Conference Title

The Third International Conference on

Computing Technology and Information
Management (ICCTIM2017)

Conference Dates
December 8-10, 2017

Conference Venue
Metropolitan College, Thessaloniki, Greece

ISBN:
978-1-941968-46-8 ©2017 SDIWC

Published by
The Society of Digital Information and Wireless
Communications (SDIWC)
Wilmington, New Castle, DE 19801, USA
www.sdiwc.net
 Table of Contents

C-Learning: Learning on Cloud ……………………………………………………………………………………………………. 1

Enterprise System Maturity – Past, Present and Future: A Case Study of Botswana …………………… 6

FAD Platforms: Proprietary Solutions …………………………………………………………………………………………. 15

Mobile Learning: A Case of Study ……………………………………………………………………………………………….. 21

Message Hiding with Pseudo-Random Binary Sequence Utilization …………………………………………... 25

Optimising the Innovation Capacity of Public Research Organisations - An Agent-Based

Simulation Approach ………………………………………………………………………………………………………………….. 32

Text Classification Using Time Windows Applied to Stock Exchange …………………………………….……..39

Towards Specification Formalisms for Data Warehousing Requirements Elicitation Techniques … 45

Using Dense Subgraphs to Optimize Ego-centric Aggregate Queries in Graph Databases …………… 59

A Secure Method for the Global Medical Information in Cloud Storage based on the Encryption
and Data Embedding ………………………………………………………………………………………………………………….. 68

Virtual Local Area Network (VLAN): Segmentation and Security ……………………………………………….… 78

Ontology-Based Data Mining Approach for Judo Technical Tactical Analysis ……………………….………90

Celestial Spectra Classification Based on Support Vector Machine …………………………………………….. 99

The Agent-Based Model of The Dynamic Spectrum Access Networks with Network Switching
Mechanism ………………………………………………………………………………………………………………………………… 106
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

C-Learning: Learning on Cloud

Pasquina Campanella
Department of Computer Science
University of Bari “Aldo Moro”
via Orabona,4 – 70126 Bari – Italy
pasqua13.cp@libero.it

ABSTRACT

Current trends in the digital age lead us to learn
that in the near future the access tools will become
more contextual. It deduces that it no longer has
any sense to identify an e-learning system in a
single monolithic platform but as a set consisting
of several interoperable components and
subcomponents to rationally manage the various
heterogeneous activities a training process can
undergo. In this scenario, cloud learning is born,
“cloud formation”, which combines the ability to
draw resources distributed with contextual
information. But the problems that arise in the era
of distributed computing or fragmentation of
workload into an arbitrary number of sub-tasks to
be distributed to an unknown number of
heterogeneous machines spread around the world
are that there is no the absolute certainty that
network machines are always available (latency,
unpredictable network crashes) then continuous
monitoring is essential. In this context, the e-
learning platform Docebo Cloud has been studied
to analyze the different response times.
KEYWORDS
platform, cloud computing, services, architecture,
test.
1 INTRODUCTION
Since the ‘90 with Grid Computing and today,
with the evolution of technologies and ways
of using the users, we are witnessing the
proliferation of an interaction between
computing systems for a computational
cooperation that moves the classical view of
ICT towards large datacenters located in the
territory [3], [6], [9], [22]. Hence, the rise of
web 2.0 and content sharing and publishing
services has led to the ability of users to have
advanced services without having to resort to
classical management of local resources [4],
[16], [20], [25]. These advances in the
performance of digital components have
resulted in a huge increase in the scope of IT
environments, and consequently, the need to
manage them uniformly in a single “cloud”
was born [6], [26], [30]. The need for such
environments is particularly felt in the
exponential growth of network connected
equipment and real time data streaming
processes, as well as in the spread of service
oriented architectures and applications,
collaborative and research projects [3], [19].
Cloud architecture has been the best candidate
for solving some of the problems generated
by large scale data processing for many
computer giants [27], [29]. In this context, a
new hybrid model of resource utilization
offered by computer networks was named,
which was named Cloud Computing [20],
[28], [29]. Cloud computing is therefore a
new approach to the provision of continuous
ICT resources that enables easy access to the
on demand network to a configurable
computational resource pool [1], [7], [17],
[19]. Although the cloud landscape is still
extremely young, in recent years it has
become increasingly important in Information
and Communication Technology (ICT) and is
the new technology that will enable the entire
education system to be changed in the near
future, high-tech e-learning services that bring
high economic savings [10], [18], [19]. The
prospect leads to the development of
CLearning (Cloud-learning) and CMobile-
Learning (CloudMobile-Learning) where the
user will have access to data that will be
shared in the cloud based on his request. The
expression “cloud learning” can be translated
as “cloud formation”, pointing to a virtual

ISBN: 978-1-941968-45-1 ©2017 SDIWC 1

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
space where you can store, share and consult
training data (work documents, formats
training formats, meeting records) on a
remote server [6], [23]. There are currently
several suppliers on the market that have
enhanced their data center hosting
applications in the cloud: from giants such as
amazon, ibm, google, microsoft, sun
microsystems to cloud realties offered by
smaller companies such as goGrid [24]. You
can access applications through a browser and
you can use any device that accesses the
network (pc, notebook, tablet, cell phones)
(Fig.1). In a cloud computing environment
(cloud or cloud computing) three distinct
actors are configured [2], [5], [12], [15], [17]:
Infrastructure Provider: provides
platforms by providing services (storage,
applications, computing capabilities)
generally following the pay-per-use model;
Service Provider / Cloud User: choose
and configure the services offered by the
vendor. Implement a service that uses the
resources provided by the infrastructure
provider and offers it to the end user.
Cliente Final Client: uses services
configured by the service provider. In certain
cases the administrator and the end customer
may coincide.
Figure 1. Representation of Cloud connections
Cloud computing, therefore, is a new way of
conceiving the supply and use of IT services
utilizing the convergence of three key
elements [8], [15], [17], [27]:
utility computing;
virtualization of computing resources;
ISBN: 978-1-941968-45-1 ©2017 SDIWC
software as a Service.
Cloud computing defines several delivery
service models among which the main ones
are [6], [12], [17]:
Software as a Service (SaaS) – the
provider's applications are accessible to the
consumer from various client devices through
a thin client interface (eg. gmail, google docs,
salesforce.com CRM 8 solutions, zoho docs);
Platform as a Service (PaaS) –
consumers have control over deployed
applications and hosting environment
configurations (eg. google app engine,
Force.com);
Infrastructure as a Service (IaaS) –
consumers can, depending on their needs,
provision storage, processing and use of
resource based networks, eg. Amazon S3
(Simple Storage Service), Amazon EC2
(Elastic Compute Cloud), GoGrid.
This article discusses the Docebo Cloud
platform in the first section, then experimental
results conducted at various learning times
and finally conclusions and future
developments.
2 DOCEBO CLOUD PLATFORM
The spread of virtualization and cloud
computing technologies, coupled with an
increasing need to cut down on application
and system management costs in the IT
world, led to the spread of IT services
delivery policies in on demand mode to allow
the diffusion of new models, or the extension
of existing ones, related to software
distribution and access to software
applications [11], [14], [21], [27]. So here are
the software, platforms, or infrastructures that
are made available as services. These services
can be considered as core components for the
development of cloud computing [6], [24],
[29]. Based on this, the docebo cloud
platform, the e-learning platform as a service
designed to allow teachers and educators to
create and manage online courses with ample
interaction possibilities, is being studied.
Docebo was born as an evolution of spaghetti
learning, an LMS, developed in 2003 by the
same team of developers [7], [13]. Today it
2
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
has a widespread use and in this the
foundation that has introduced it has migrated
to develop its e-learning platform in the
Cloud/ As a Service direction and this has
come to fruition in the docebo 7.0 release,
accessible directly online “as service” without
any necessary installation. Docebo provides
e-learning and tutor creators with a host of
tools, such as forum and chat, user and group
management, test and surveys, document
archives, and the ability to connect to some
videoconferencing services such as Teleskill
[14], [29].
3 EXPERIMENTAL RESULTS
Many problems in online training depend, in
most cases, on a management structure that is
unable to adequately manage the didactic
organizational complexity given by the
introduction of new teaching technology
solutions [14], [16]. Based on this they
wanted to conduct tests on the e-learning
docebo cloud platform, migrating to cloud/ As
a service solutions to allow teachers and
educators to create and manage online courses
with ample interaction possibilities. Tests
were conducted in the community with 150
students, students and teachers, aged 20 to 42,
willing to evaluate the new cloud impact. The
Docebo Cloud Platform is accessed by
connecting to an account as shown in Fig. 2
and exploring the various learning features
ranging from distance learning, lectures in
videoconferencing, archiving and making
available teaching files, monographic courses,
evaluation test, alerts, discussion forums.
ISBN: 978-1-941968-45-1 ©2017 SDIWC
Figure 2. Docebo Cloud platform access and
monitoring
Docebo Cloud regarding continuous
monitoring of organizational-educational
complexity, fragmentation of workload into
an arbitrary number of sub-tasks to be
deployed to an unknown number of machines,
heterogeneous among them, scattered all over
the world, or that there is no the absolute
assurance that network machines are always
available (latency times, unpredictable
network crashes) has provided variable cloud
response times in managing various tasks as
shown in Fig.2, a reduction occurred with a
number of users less than 100, much
dependent on the skills of the interlocutors.
4 CONCLUSIONS and FUTURE
DEVELOPMENTS
Concluding C-Learning sets the groundwork
to be considered as an innovative form of web
2.0, where the needs of the user go beyond
voice communication and demand strong
demand for services that need connectivity
and transmission bandwidth, which no longer
has to be rigidly fixed by contracts, but must
be flexible and must be readily adapted to
users requests and also as the latest form of
globalization, leading to a number of end user
benefits, in particular for the resources that
may be required and obtained on-demand. It
is arguing as a model that changes the way
resources are made, because they are
decoupled from technology and
“encapsulated” in IT services. Such services
are dynamic and flexible and can be used
3
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
individually or in larger business contexts,
allowing optimal use of resources because
they are shared among multiple users
(multitenant model). The article analyzed the
didactic-organizational complexity of the e-
learning docebo cloud platform and provided
variable response times in managing various
tasks. The experimentation carried out in the
community by a number of students trying to
explore the new platform for content delivery
has achieved an optimum result with a
number of users less than 100, much
dependent on the skills of the interviewees.
REFERENCES
[1] M.M. Alabbadi, “Cloud computing for
education and learning: Education and learning
as a service (eLaaS)”, Proc. of the 14th
International Conference on Interactive
Collaborative Learning (ICL2011), IEEE, pp.
589-594.
[2] M. Al-Zoube, “E-learning on the Cloud”,
International Arab Journal of e-
Technology,vol.1, n. 2, 2009, pp. 58-64.
[3] M. Armbrust, A. Fox, R. Griffith, A.D. Joseph,
R. Katz, A. Konwinski, G. Lee, D. Patterson, A.
Rabkin, I. Stoica, M. Zaharia, “A View of
Cloud Computing”, Commun. Acm, vol. 53, n.
4, 2010, pp. 50-58.
[4] D. Bein , W. Bein, P. Madiraju, “The Impact of
Cloud Computing on Web 2.0”, eij, vol. 9, n. 1,
2009, pp. 5-12, USA.
[5] P. Campanella, “Cloud Computing: un nuovo
paradigma”, Atti VIII convegno nazionale Sie-l,
connessi! Scenari di innovazione nella
formazione e nella comunicazione, 14-15-
16/09/2011, Ledizioni, Reggio Emilia, Italy.
[6] P. Campanella, “Cloud E-learning: un nuovo
binomio”, Atti Didamatica 2016 – Innovazione:
sfida comune di scuola, università, ricerca e
impresa, 30° edizione Aica, 19-20-21/04/2016,
pp. 128-137, Udine, Italy.
[7] P. Campanella, “Method of experimental
evaluation of ICT in teaching”, Atti del
convegno Elearn 2011, world conference on e-
learning in corporate governement, healthcare e
higher education organized by AACE,
Honolulu, Hawaii, USA, 17-18-19-20-
21/10/2011.
[8] D. Chandran, S. Kempegowda, “Hybrid E-
learning Platform based on Cloud Architecture
Model: a proposal”, Proc. International
conference on signal and image processing
(ICSIP), 2010, pp. 534-537, IEEE.
ISBN: 978-1-941968-45-1 ©2017 SDIWC
[9] J. Cappos, I. Beschastnikh, A. Krishnamurthy,
T. Anderson, “Seattle: a platform for
educational cloud computing”, Acm sigcse
bulletin, vol. 41, n. 1, 2009, pp. 111-115, March
3-7, New York, USA.
[10] T. Ercan, “Effective use of cloud computing in
educational institutions”, Procedia-Social and
Behavioral Sciences, vol. 2, pp. 938-942,
05/01/2010.
[11] R. Katz, P. Goldstein, R. Yanosky, “Cloud
computing in higher education”, 2010,
Educause.
[12] A. Khajeh-Hosseini, D. Greenwood, and I.
Sommerville, “Cloud Migration: A Case Study
of Migrating an Enterprise IT system to IaaS”,
IEEE, 3rd International Conference on Cloud
Computing, 2010, pp. 450-457.
[13] S. Impedovo, P. Campanella, G. Facchini, G.
Pirlo, R. Modugno, L. Sarcinella, “Learning
Management Systems: un’analisi comparativa
delle piattaforme open-source e proprietarie”,
Atti Didamatica 2011 - Informatica per la
didattica, Torino 04-05-06/05/2011.
[14] S. Impedovo, IAPR FELLOW, IEEE S. M., P.
Campanella, “Docebo Cloud: apprendimento e
nuove tecnologie”, Atti Didamatica 2012 -
Informatica per la didattica, Taranto 14-15-
16/05/2012.
[15] G. Lin, D. Fu, J. Zhu, “Cloud computing: IT as
a Service”, 2009, vol. 11, n.2, pp. 10-13, ACM.
[16] G. Lin, G. Dasmalchi, J. Zhu, “Cloud
computing and IT as a Service: Opportunities
and Challenges”, Proc. of the IEEE
International Conference on Web Services,
2008.
[17] A. Manzalini, C. Moiso, E. Morandin, “Cloud
computing: stato dell’arte e opportunità”,
Notiziario Tecnico Telecom Italia, n. 2, 2009.
[18] M. Miller, “Cloud Computing: Web-Based
Applications that Change the Way you Work
and Collaborate Online”, 2008.
[19] E. Martins Morgado, “An exploratory essay on
Cloud Computing and its Impact on the use of
Information and Communication Technologies
in Education”, in education in a technological
world: communicating current and emerging
research and technological efforts, ed. Mendez-
vilas, formatex, 2011.
[20] S. Ouf, M. Nasr, Y. Helmy, “An enhanced e-
learning eco system based on an integration
between cloud computing and Web 2.0”, Proc.
of the 10th IEEE International Symposium on
Signal Processing and Information Technology
(ISSPIT), pp. 48-55, 15-18/12/2010.
[21] P. Pocatilu, F. Alecu, M. Vetrici, “Using Cloudth
Computing for Elearning Systems”, Proc. 8
4
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

Wseas international conference on data

networks, communications, computers,
(dncoco'09), Baltimore USA, 2009, pp. 54-59.

[22] L. Qian, Z. Luo, Y. Du, L. Guo, “Cloud

computing: an overview”, 2009, pp. 626-631,
springer.

[23] N. M. Rao, C. Sasidhar, V. S. Kumar, “Cloud

computing Through Mobile Learning”, (ijacsa),
International journal of Advanced computer
science and applications, vol 1, n. 6, 2010, pp.
42-46.

[24] N. Sultan, “Cloud computing for education: a

new dawn ?”, International journal of
Information Management, elsevier, issue 2, vol.
30, 2010, pp. 109-116.

[25] Velev G. Dimiter, “Challenges and

Opportunities of Cloud-based Mobile
Learning”, ijiet, iacsit, vol. 4, n. 1, 2014, pp. 49-
53.

[26] M. A. Vouk, “Cloud computing: issues,

research and implementations”, Department of
computer science, North Carolina State
University, Raleigh, North Carolina, USA,
Journal of computing and information
technology, vol. 4, 2008, pp. 235–246.

[27] B. Wang, H.Y. Xing, “The application of cloud

computing in education informatization”, in
International Conference on computer science
and service system, IEEE, pp. 2673-2676, 27-
29/06/2011.

[28] S. Yanfei, “Cloud Computing applications in

the network learning”, in China outside school
basic education edition, 2010.

[29] Q. Zhang, I. Cheng, R. Boutaba, “Cloud

Computing: state-of-the-art and research
challenges”, Journal of internet services and
applications, vol.1, n.1, 2010, pp. 7-18, springer.

[30] L. Wang, G. von Laszewski, A. Younge, X. He,

“Cloud Computing: a perspective study”, vol.
28, n. 2, 2010, pp. 137-146, springer.

ISBN: 978-1-941968-45-1 ©2017 SDIWC 5

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
Enterprise System Maturity – Past, Present and Future:
A Case Study of Botswana
Oduronke T. Eyitayo
Department of Computer Science
University of Botswana
Private Bag 00704
Gaborone, Botswana
eyitayoo@mopipi.ub.bw
ABSTRACT
The early stages of computerisation were pre-
occupied with managing activities such as
operations, programming, data collection etc.
This then developed into information systems.
Such systems include: customer management
systems, financial systems, human resource
systems, business intelligence systems, asset
management systems, waste management
systems, document management systems,
workflow management systems, and hundreds
of other systems. In recent years, these
systems are being integrated with each other to
such an extent that it is oftentimes necessary to
view them, not as hundreds of different
systems, but one single system comprising of
several systems, often referred to as enterprise-
wide information system. Many organizations
however have encountered difficulties when
attempting to modify enterprise resource
planning systems to their business operations.
These factors led to the growth of Enterprise
Architecture. This paper focuses on studying
the trend in the development of the Botswana
Enterprise system. In order to trace the
Botswana maturity story through literature the
question the researcher set out to answer was:
How have Enterprise systems evolved over
time, and where is it now? These questions
will be answered with insights from the
academic literature using online databases.
The study uses Gartner model to analyse the
Botswana story. The planning and
implementation approaches adopted by the
nation explicitly explain how the development
agenda of a country if followed through can
impact positively on the country's
development.
ISBN: 978-1-941968-45-1 ©2017 SDIWC
Keywords: Botswana; Enterprise Resource
Planning; Enterprise Architecture; Enterprise
Maturity; Gartner Maturity Model
1. INTRODUCTION
The first era of computerisation emerged
from the 1960s and was referred to as the
data processing era while the 1970
onwards was referred to as the
management information systems (MIS)
era [1]. The first area focused on efficient
transaction handling and effective resource
control, while the second focused on the
effective problem resolution and support
for decision making. The two eras
overlapped as data processing continued to
mature. Ward and Peppard [1] noted that
as more data were stored in computer
systems, its use matured into using
information to increase the effectiveness of
decision making. The legacy of data
processing applications was often at best
fragmented data resource. This led to
organisations re-organising data and
applications into integrated data-based
systems to enable MIS to be developed.
As the business value of personal
computers became clear to enterprises, the
next logical step was to link them together.
The US Defense Advanced Research
Projects Agency's ARPANET initiative in
the sixties and early seventies was the
dawn of efficient and large-scale
networking. In the early nineties it reached
a global impact with the Internet. As the
interconnected networks grew, the power
and number of applications using them
6
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
increased, and consequently their
importance grew. With the evolution of
extended computer networks, the computer
itself lost its position as the main focus,
which instead shifted to the availability
and capacity of the networks. Ward and
Peppard [1] in their own analysis
described the third era as strategic
information systems era, characterised by
improved competitiveness.
An offshoot of the third era is the
enterprise systems. Today, there are
information systems for most of the tasks
performed in an enterprise. Such systems
are: customer management systems,
contract management systems, financial
systems, human resource systems,
business intelligence systems, asset
management systems, document
management systems, workflow
management systems, and hundreds of
other systems. In recent years, these
systems have been integrated with each
other to such an extent that it is oftentimes
necessary to view them, not as hundreds of
different systems, but one single system of
systems resulting enterprise-wide
information system.
As the information systems grew in scope
and number, the information they stored
and the information they would benefit
from having access to also increased [2].
This situation led to demands for
automated exchange of information
between the systems. Technological
development responded rapidly to these
demands by devising methods for
connecting two systems to each other.
However, the number of connections
between systems has become
overwhelming which has led to
complicated systems with new challenges
when attempting to modify enterprise
resource planning systems to meet
business operations. This has led to the
growth of Enterprise Architecture.
ISBN: 978-1-941968-45-1 ©2017 SDIWC
Enterprise Architecture began getting
attraction in the mid to late 1980s after
John Zachman published an article
describing a framework for information
systems architectures in the IBM Systems
Journal [3]. The field initially began to
address two problems: System
complexity—Organizations were spending
more and more money building IT
systems; and Poor business alignment—
Organizations were finding it more and
more difficult to keep those increasingly
expensive IT systems aligned with
business need. The cost and complexity of
IT systems have exponentially increased,
while the chances of deriving real value
from those systems have dramatically
decreased [4].
Chen et. al. [5] noted that Enterprise
architecture is a “challenging but still
confusing” concept. It is defined in many
different ways. Engelsman et al. [6]
describe Enterprise Architecture (EA) as
“a design or a description that makes clear
the relationships between products,
processes, organisation, information
services and technological infrastructure; it
is based on a vision and on certain
assumptions, principles and preferences;
consists of models and underlying
principles; provides frameworks and
guidelines for the design and realisation of
products, processes, organisation,
information services, and technological
infrastructure.” Another definition states
that it comprises a collection of simplified
representations of the organisation, from
different viewpoints and according to the
needs of different stakeholders [7]. EA
helps describe and manage changes in
enterprises so as to enhance their
consistency and agility [8].
Enterprise architecture has evolved as a
discipline since the 1990s. Early work
focused on architecture models, principles,
and standards that comprise the content of
the enterprise architecture. As companies
have gained practical experience
7
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
implementing EA concepts, they have
become more concerned with enterprise
architecture management (EAM) and its
success factors [9]. This is the focus of the
fifth era.
The fifth era is what Ward and Peppard [1]
classified as their own fourth era. This is
where sustainability from an IS
perspective can be defined as an
organisations ability to continually deliver
explicit business value through IS/IT thus
leading to advantage. Winter, Legner and
Fischbach [9] recognised that architecture
a key concern of information management
today. They noted that there is now
general acknowledgment that only
continuous and holistic management of the
Enterprise Architecture (EA) can ensure
the sustainability, agility, and strategic
alignment of corporate IT environments.
The purpose of this paper is to look at
Botswana Enterprise System and how this
has matured over the years, as well as have
a view into the future.
2. EA MATURITY MODELS
There are several models used in
measuring maturity levels within an
organisation. This section considers three
of such models.
One of the popular maturity models in
Enterprise Architecture is Gartner’s
maturity model. The maturity model has
five different levels of maturity within
enterprise architecture practices. These
five levels are also similar to the five eras
discussed earlier. In Level 1, that is the
Non-existent phase, enterprise architecture
either has not existed until that point, or is
at it’s very early stages of implementation.
Level 2 is reactive, at this phase, enterprise
architecture is officially recognized by the
organization. However, the nature of the
practice is one of response to needs, rather
than pre-planned. At level 3, Enterprise
ISBN: 978-1-941968-45-1 ©2017 SDIWC
Architecture is at the functioning phase. At
level 4 an integrated system is needed in
order to deliver value consistently. At the
final phase, level 5, the enterprise
architectures becomes a natural way of
working; the effect trickles down across
the organization [10]
Another model was one used by Feurer
[11]. He described enterprise architecture
as a continuous journey. He classified the
maturity levels into four: the business silo
architecture, the standardized technology
stage, the optimized core stage, and the
business modularity architecture. At the
business silo stage, companies primarily
use Information Technology (IT)
investments to meet local business needs.
The second maturity stage, standardized
technologies are used to move from local
applications to a shared infrastructure. At
the third stage, optimized core stage,
companies view data and applications
from an enterprise perspective and so
mainly focus on shared data and enterprise
systems. IT extracts crucial data from
enterprise systems and makes it reusable
for business processes and other IT
applications. Business processes optimized
in the third stage are refined and
modularized in the fourth stage. The
outcomes from these activities are reusable
modules (mostly based on Web service
technology) connected to core data and
backend processes. IT ensures that these
modules can be integrated into business
process.
The third model considered here is
Capability Maturity Model (CMM). This
is a methodology used to develop and
refine an organization's software
development process. The model describes
a five-level evolutionary path of
increasingly organized and systematically
more mature processes. At the initial level,
organisation does not provide a stable
environment for developing and
maintaining software. At level 2, the
repeatable level, policies for managing
8
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

software project and procedures to Google Scholar Found

implement them are established. The Information System and Botswana 149000
defined level, which is level 3, the IT/ICT and Botswana 153000
standard process for developing and Enterprise System and Botswana 32700
maintaining software across the Enterprise Architecture and Botswana 6660
organisation is documented. At level 4, the Date: 23rd August 2016
managed level, organisation sets
quantitative quality goals for both software 4. FINDINGS - THE BOTSWANA
products and processes. Productivity and STORY
quality are measured at this phase. At the There were a few literature found that gave
optimising level which is level 5, the a good picture of the story and maturity of
organisation is focused on continuous Information Systems and Enterprise
process improvement [12] Architecture in Botswana. This is shown in
Table 3.
CMM is more focused on software
development process. Feurer and Gartner Table 3: Relevant Literature on Maturity of
models are quite similar. However, the Information Systems and Enterprise
five distinct stages in Gartner make it a Architecture in Botswana
Author Year Title
more detailed model. In this paper, due to
Ayoku 2006 Information and
Gartner's being a trusted source of Ojedokun Communication
analysis, the model is used as a focus for and Technology (ICT) Systems
this study. Kgomotso in Botswana Government
Moahi Departments
[13]
3. METHODOLOGY Stephen 2006 E-readiness of SMEs in the
M. ICT sector in Botswana
In order to trace the Botswana maturity Mutula, with
story, through literature the question to be Pieter van respect to information
answered was - How have EA evolved Brakel access
[14]
over time, and where is it now? Answering Faith- 2010 A CMM Assessment of
these questions with insights from the Michael Information Systems
academic literature, the information and E. [15] Maturity
background on Information Systems, Levels in Botswana
Information Technology and Enterprise Nugi 2011 State of Information
Nkwe Technology Auditing in
Architecture as related to Botswana was [16] Botswana
searched using Ebscohost (Table 1) and Nugi 2012 E-Government: Challenges
Google Scholar (Table 2) Nkwe[17] and Opportunities in
Botswana
Table 1: Ebscohost Search Oduronke 2014 Enterprise resource
Temitope planning (ERP) systems –
EBSCO Host Found Eyitayo Is Botswana winning? A
Information System and Botswana 465 [18] Question on culture effects.
IT/ICT and Botswana 17132
Enterprise System and Botswana 11
Enterprise Architecture and Botswana 1022 According to Ojedokun and Moahi [13],
Date: 23rd August 2016 the use of information technology was
introduced by government in 1966 in the
Table 2: Google Search
Office of the Accountant General, for the
processing of payroll and accounts using
accounting machines which were later
replaced by ICT Punched Card Tabulators
in 1969. This corresponds to the first era

ISBN: 978-1-941968-45-1 ©2017 SDIWC 9

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

that was referred to as the data processing National What was achieved
by Ward and Peppard [1]. After this era, a Development
Plan (NDP)
number of critical ICT systems that have Phase
been implemented in Botswana implemented the Master
government departments. The details are Information Systems Plan
described using the government National (MISP); Teacher Management
Development Plan (NDP). Since System (TMS) and the Student
Selection System (SSS) [13]
independence in 1966, Botswana’s
development process has been guided by Automation of some post office
successive National Development Plans counters, and the installation of
(NDPs). These have provided a medium Performance Management
term planning and budgeting framework System by Botswana Post [13]
(typically 5-6 years) for capital and
The Ministry of Trade, Industry,
recurrent expenditure, and have been a key Wildlife and Tourism
feature of Botswana’s system of (MTIWAT), developed the
development management. The plans Ministry's website, automated the
outline the government’s development office of Wildlife and National
Parks [13]
priorities for the plan period as well as the
NDP 9 (2003/4 The Ministry of Trade, Industry,
policies, programmes and projects required - 2008/09) Wildlife and Tourism
to achieve those priorities [19]. (MTIWAT) : Automate the
process of company registration
The Information System development over and business names, trade and
the years is shown in Table 4 following the industrial licenses and tourism
information management,
NDP plan. develop document management
and workflow management
Table 4: Information Technology Development process; review and develop all
According to the National Development Plan Department of Wildlife and
National Parks systems [13]
National What was achieved
Development Ministry of Local Government
Plan (NDP) (MLG): Tribal Land
Phase Administration System for land
NDP 7 Voters' registers, a new payroll use planning and management,
(1991/92- system and vehicle registration installation of Human Resources
1996/97) [13] and Payroll package to all the
NDP 8 (1997- National Registration System, Local Authorities, Financial
2002) Payroll System, and Management Computer System,
Computerised Personnel and Project Management.
Management System (CPMS) Document Management System,
[13] a website; database for recording
tribal ceremonies, and Social
Livestock Identification and Benefits System[13]
Trace back System (LITS),
Automated System for Customs Ministry of Works, Transport and
Data (ASYCUDA), Vehicle Communications (MWTC):
Registration System [20,21] Development of communications
Department of Supplies, network infrastructure. A
Government Data Network I, Tax comprehensive review and
Payer Management System I, and computerisation of the Road
Trade Statistics [13] Transport Permits system
earmarked to be executed as an
Water Utilities Corporation also integral part of the Vehicle
implemented the Master Registration and Licensing [13]
Information Systems Plan (MISP
Water Utilities Corporation also

ISBN: 978-1-941968-45-1 ©2017 SDIWC 10

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

National What was achieved National What was achieved

Development Development
Plan (NDP) Plan (NDP)
Phase Phase
The Ministry of Mineral, Energy Offices, some embassies and the
and Water affairs (MMEWA): Government of Botswana. [18]
Development of Energy
Information System; expansion The beginning of the
of the Water Resources Government Enterprise
Information System; installation Architecture [22]
of a computerised supplies
system, inventory management 5. DISCUSSION
system and a library information
system; expansion of its
computerised billing system, and The different National Development plans
the implementation of a addressed the key process areas of the
Geosciences Information System Gartner’s Maturity model. Prior to NDP 7,
[13]
it was at the Non-existent phase. A few
The Ministry of Health: data processing applications were
Integrated Health available. The existence of information
Communication System and systems began during the NDP 7 and 8
software for Princess Marina and phases. This is the reactive phase where
Nyangabwe hospitals [13]
the practice is one of response from every
The Customs department: ministry was to develop based on needs,
Taxpayer Management System rather than pre-emptive planning and
(TMS) [13] action. During the NDP 9 phase, business
NDP10 Growth e-learning, E-passport; outcome became part of the core value.
(2009/10- E-license, E-health and E-
2016/17 Business[16]
This phase saw the introduction of systems
with whole ministries in mind rather than
Computerized Case just tasks. There now exist what can be
Management; Computerization called functioning systems such as billing
of Lobatse High Court; DPSM systems, library systems and tribal land
Computerized Personnel
Management System;
administration system. This development
Computerization of Students now led to the integration phase in NDP
Records and Grant Loan Scheme; 10 with the development of integrated
Computerization of Teaching system. This is where the nation is
Service Management; presently. This era has brought about
Government Accounting and
Budget System; Integrated
systems that work together. As the
Patient Management System; economy of the country is improving,
National Archives and Records organisations across Botswana are
Management Systems; increasing their investments in Enterprise
Computerization of Civil and Resource Planning (ERP) technology as a
National Registration;
Computerization of Labor and
way of helping in the growth of their
Social Security; Local business. Some institutions that have
Government Computerization of implemented ERP systems are banks,
Human Resource Management; Water Utilities Corporation, Botswana
Computerization of Social Post Offices, some embassies, University
Benefit and Reconciliation
System [17]
of Botswana and the Government of
Botswana [18].
Implementation of ERP systems:
Some implemented ERP systems Still during the NDP10, In an official
are banks, Water Utilities gazette released on September 15, 2015 by
Corporation, Botswana Post
the Public Procurement and Asset Disposal

ISBN: 978-1-941968-45-1 ©2017 SDIWC 11

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
Board (PPADB), the Botswana
government announced that Korea IT
Consulting was selected as the winner of
the “Professional Services for the
Development of an e-Government Service
Oriented Enterprise Architecture” project,
initiated by the Ministry of Transport and
Communications of Botswana [23].
Integration seem to be never-ending
processes, because the enterprise continues
to evolve in its ever-changing environment
as a result of adaptation to external forces,
advances in technology, emerging business
models, new regulations and/or
optimisation of internal solutions; making
what is today a fully integrated system, the
partly integrated system of tomorrow. In
the present state, the enterprise
architecture practice is delivering value.
The country is now in a phase of defining
integration, standards etc. and how all this
come together within the various sectors of
the nation. This is being done in the
development of a Government Enterprise
Architecture [24].
The NDP 11 (2017-2023) hopes to achieve
what is highlighted in the e-Government
master plan Government [25]. There are
fifteen areas of focus are: Upgrade e-
Government Strategy; Advancement of e-
Governance; Project Evaluation System; e-
Document System; Network Optimization;
e-Education System; National Health
Information System; Business Activity
Support System; e-Procurement System;
Local Government Informatization; Civil
Affairs Single Portal; Administrative
Information Sharing Centre; Job Portal; e-
Agriculture and Government Enterprise
Architecture (GEA) [25].
Looking at the well laid out government
plan, if all goes as planned, it can be
predicted that the end of NDP 11 should
lead to a Ubiquitous state (level 5) where
enterprise architectures success has a
trickle-down effect across the nation. At
that point, Enterprise Architecture
ISBN: 978-1-941968-45-1 ©2017 SDIWC
becomes a natural way of working and the
principles behind the discipline are widely
adopted. This is depicted in Figure 1.
6. CONCLUSION
This paper reveals significant progress
made by the Botswana Government in
attaining maturity. The achievements of
the Botswana government have improved
service delivery and increased accessibility
to government services. One lesson to be
learnt from the Botswana government
approach is the incorporation of major ICT
projects into its national development
plans, the phased implementation of the
ICT projects, and the governments'
commitment to achieving the strategies
developed. This approach has enabled a
great deal of achievements. National
development plans, among others,
particularly serve as major statements of
government's development policies and
strategies.
Enterprise plans in many countries are not
yet very clear; however Botswana has
taken a step forward to plan its national
architecture. It is therefore recommended
that other countries in Africa emulate the
example of the Botswana government by
embarking on a strategised plan of an
integrated enterprise. The planning and
implementation approaches (i.e.
incorporation into national development
plans and phased implementation) adopted
are particularly worthy of emulation. It
explicitly explains how the development
agenda of a country if followed through
can impact positively on the country's
development.
Defining maturity model is not really
where a nation will find its value. The
knowledge gained from current position
and insight from the model gives the
model its worth. This information can be
used to influence development roadmaps
for enterprise architecture practices.
12
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

Figure 1: Botswana’s Maturity Chart

REFERENCES
[1] J Ward and J Peppard, Strategic Planning [5] D Chen, G Doumeingts, and F Vernadat,
for Information System, 3rd ed. UK: John "Architectures for enterpriseintegration
Wiley, 2002. and interoperability: Past, present and
future.," Computers in Industry, vol. 59,
[2] P Johnson, R Lagerström, and M Ekstedt, pp. 647-647, 2008.
IT Management with Enterprise
Architecture.: ePub, [6] W Engelsmana, D Quartelc, H Jonkersa
www.ics.kth.se/MAP.pdf, 2012. and M van Sinderen, "Extending
enterprise architecture modelling with
[3] The Open Group (2015) A Historical business goals and requirements,"
Look at Enterprise Architecture with John Enterprise Information Systems, vol. 5,
Zachman. [Online]. HYPERLINK no. 1, pp. 9-36, 2011.
"https://blog.opengroup.org/2015/01/23/a-
historical-look-at-enterprise-architecture- [7] M Lankhorst, Enterprise Architecture at
with-john-zachman/ Work: Modelling, Communication, and
Analysis. New York: Springer-Verlag,
[4] R Sessions (2007, May) A Comparison of 2005.
the Top Four Enterprise-Architecture
Methodologies. [Online]. HYPERLINK [8] O Noran, A Meta-Methodology for
"https://msdn.microsoft.com/en- Collaborative Networked Organisations.
us/library/bb466232.aspx" School. Brisbane, Griffith University:
School of Computing and Information
Technology, 2005.

ISBN: 978-1-941968-45-1 ©2017 SDIWC 13

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
[9] R Winter, C Legner and K Fischbach,
"Introduction to the special issue on
enterprise architecture management,"
Information Systems and e-Business
Management, vol. 12, no. 1, pp. 1-4, 2014.
[10] B Burke and M Blosch. (2015) ITScore
Overview for Enterprise Architecture.
[Online]. HYPERLINK
"https://www.gartner.com/doc/3092223/its
core-overview-enterprise-architecture"
[11] S Feurer (2007) Enterprise Architechture -
Maturity Stages. [Online]. HYPERLINK
"https://archive.sap.com/kmuuid2/70edb7
ce-5fd8-2910-1583-
db2b2cd98298/Enterprise%20Architecture
%20-%20Maturity%20Stages.pdf"
[12] M C Paulk, B Curtis, M B Chrissis, and C
V Weber, "Capability Maturity Model for
Software, Version 1.1," Carnegie Mellon
University, Pittsburgh, Pennsylvania,
1993. [Online]. HYPERLINK
"https://www.sei.cmu.edu/reports/93tr024.
pdf"
[13] A A Ojedokun and K IL Moahi,
"Information and Communication
Technology (ICT) Systems in Botswana
Government Departments," African Jornal
Library Archival & Information Science,
vol. 16, no. 2, pp. 79-88, 2006.
[14] S M Mutula and P Brakel , "E-readiness of
SMEs in the ICT sector in Botswana with
respect to information access," The
Electronic Library, vol. 24, no. 3, pp. 402-
417, 2006.
[15] F E Uzoka, "A CMM Assessment of
Information Systems Maturity," MIS
Review, vol. 16, no. 1, pp. 53-84, 2010.
[16] N Nkwe, "State of Information
Technology Auditing in Botswana," Asian
Journal of Finance & Accounting, vol. 3,
no. 1, pp. 125-137, 2011.
[17] N Nkwe, "E-Government: Challenges and
Opportunities in Botswana," International
Journal of Humanities and Social Science,
vol. 2, no. 17, pp. 39-48, 2012.
ISBN: 978-1-941968-45-1 ©2017 SDIWC
[18] O T Eyitayo, "Enterprise Resource
Planning (ERP) Systems – Is Botswana
Winning? A Question on Culture Effects,"
Issues in Informing Science and
Information Technology, vol. 11, pp. 47-
55, 2014.
[19] A M Land, "Strengthening of National
Capacities for Nationa lDevelopment
Strategies and Their Management: An
Evaluation of UNDP's Contribution:
Country Study – Botswana," 2010.
[20] S Serero and M Moreri, E-government in
Botswana: From Digital Divide to E-
Economy- Issues and Strategies for Public
Policy, July 22nd-25th, 2002.
[21] A Bose, G Dick, O F Seitei, and S
Obuseng, "Botswana Human
Development Report 2002:Contributions
of Science and Technology and ICT for
Healthy Governance," Draft Outline,
UNDP 2002.
[22] J Sejabosigo (2016, February) Daily
News. [Online]. HYPERLINK
"http://allafrica.com/stories/20160218019
6.html"
[23] Y Choul-Woong, "Korea IT Consulting
Wins E-Government EA Deal from
Botswana," Korea IT Times, September
2015. [Online]. HYPERLINK
"http://www.ifg.cc/aktuelles/nachrichten/r
egionen/163-kr-suedkorea-south-
korea/52120-korea-it-consulting-wins-e-
government-ea-deal-from-botswana"
[24] Government of Botswana. Botswana’s
National e-Government Strategy 2011-
2016. [Online]. HYPERLINK
"http://workspace.unpan.org/sites/internet/
documents/unpan048687.pdf"
[25] Government Modernization Office.
Botswana e-Government Master Plan
2015-2021. [Online]. HYPERLINK
"http://www.cit.co.bw/downloads/e-
govenment%20master%20plan%20presen
tation%20to%2014th%20e-
government%20board%20version2.pdf"
14
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

FAD Platforms: Proprietary Solutions

Pasquina Campanella
Department of Computer Science
University of Bari “Aldo Moro”
via Orabona, 4 – 70126 Bari – Italy
pasqua13.cp@libero.it

ABSTRACT

In the face of development continuous of internet,
the training resources in the network have grown
informally. For this reason, the need for tools to
extract their knowledge is constantly widening.
Starting with the evolution created by social web,
this article explored potential of current
technology tools, with particular reference to the
features of e-learning platforms, and related to
proprietary solutions regarding content delivery
modes, user-based monitoring tests, as well as
evaluation techniques in order to better manage
interactive online courses, which make the web
user active in the production process. In this
direction, what is called lifelong learning is what
follows.
KEYWORDS
monitoring, platforms, performance, collaborative
learning, features.
1 INTRODUCTION
The panoramic of FAD platforms has seen a
continuous evolution over the years. The term
“platform” means the technological
infrastructure that allows e-learning activities
or online course management, integrating
teaching modules, evaluations within learning
groups [1], [10], [11], [12]. In order to
promote use of more advanced and interactive
platforms, a proprietary solution analysis is
proposed that would serve as a useful
contribution to the development of different
forms of collaborative learning and require
new capabilities for integrated management of
formative components of social networks.
The analysis was determined by the fact that
literature allows only partially to obtain an
objective evaluation of platforms and how
they support processes learning, considering
their peculiarities, needs and problems [1],
[10], [11], [14]. From early 1990, models for
evaluating learning management systems
have been developed [4], [7]. The
Commonwealth of Learning model, which
examines various features such as usability,
accessibility, collaborative functionality,
manualization, installation, technical support,
standard compliance, interoperability and
content reusability, tracking [15]. Below are
the different sections on proprietary platforms
analyzed in their respective studies and
simulations as well as evaluation sessions and
finally conclusions and future developments.
2 PROPRIETARY PLATFORMS
Some of the major proprietary platforms are
listed in different ways, considering the
sharing, participation and collaboration of
web 2.0 and in particular blog, feedback, chat,
forum, podcasting and wiki (Tab.1) [2], [4],
[9], [13]:
Table 1 - Ownership proprietary platforms
In particular, the study was aimed at
communication between learning objects,
tracking activities and the results obtained.
On-line questionnaires, forum interventions to
highlight the different polarities of expression
[2], [13]. In delivering the courses, the

ISBN: 978-1-941968-45-1 ©2017 SDIWC 15

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

observations and comparisons made have

been used to understand the user friendly role
of online platform's navigability, how to use
multimodal lessons, the content, the materials,
the verification and evaluation forms used,
interactive support during the course phases.
The following results show a simplified link
between teaching collaborative and user
satisfaction, while a varying influence on
degree of socialization gained. In the analysis
below, the duration of courses, course
management, quality of distributed material,
quality of transmitted theoretical lectures and
exercises were evaluated by providing the Figure 2. Monitoring Centra
following self-assessment questionnaire
(Fig.1): Elluminate Live
Hybrid platform, virtual classroom and e-
conferencing. Multiplatform on windows,
linux and mac os systems [4], [7], [12]. The
performance assessment of a 150 student (20
to 30 year old) students in the community
reported a 30% average duration, distributed
material quality and lessons, exercises and
test averaging 40% (Fig.3). Application
sharing issues were improved by continuing
to chat, monitoring was balanced.

Figure 1. Platforms monitoring questionnaire

Centra
Web-based collaborative platform with
features such as web conferencing, virtual
classroom, web seminar, net meetings [2],
[4],[12]. The performance evaluation of 100
students (ages 18 to 30) in the community
reported the duration of courses and course
management with 54% average, quality of
lessons, exercises and tests averaging 56% Figure 3. Monitoring Elluminate Live
(Fig.2). Communication issues have improved
in videoconferencing. e/pop
E-learning tool for content sharing,
multiplatform windows, mac os [3], [4], [12].
The performance evaluation of a 100
contingent (aged between 20 and 30 years) in
the community reported the duration of
courses and quality of distributed material
with an average of 35%, quality of lessons,
exercises and test with an average percentage

ISBN: 978-1-941968-45-1 ©2017 SDIWC 16

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
of 45% (Fig.4). The monitoring was balanced
in the various tests conducted.
Figure 4. Monitoring e/pop
Groove
E-learning tool for collaborative learning,
synchronous and asynchronous solutions [4],
[6], [14]. The performance assessment of a
100 students (18 to 30 year old) in the
community reported a 35% lifetime of
courses, 45% of distributed media, quality of
lessons, exercises, and test with an average of
60% (Fig.5). Monitoring balanced in the
various test.
Figure 5. Monitoring Groove
HotConference
E-learning tool for collaborative sharing,
synchronous and asynchronous solutions [5],
[7], [12]. The performance evaluation of a
150 year old students (aged 19 to 25) in the
community reported a 45% lifetime of
courses, 50% of distributed media, quality of
ISBN: 978-1-941968-45-1 ©2017 SDIWC
lessons, exercises, and test with an average of
48% (Fig.6). Well balanced test monitoring.
Figure 6. Monitoring HotConference
LearnLinc
E-learning tool for collaborative learning,
synchronous and asynchronous solutions [4],
[8], [12]. The performance evaluation of a
100 degree students (between 20 and 25 years
old) in the community reported a 60%
lifetime of courses, 40% of distributed
material quality, quality of lessons, exercises,
and test with an average of 50% (Fig.7).
Balanced monitoring in the various test
conducted.
Figure 7. Monitoring LearnLinc
Lotus Learning Space
Groupware platform for online learning, with
two modules learning space core and
learning space collaboration [4], [8], [12].
The performance evaluation of a 100
contingent (between 20 and 30 years old)
17
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
students in the community reported a lifetime
of 50%, material quality distributed by an
average of 45%, quality of lessons, exercises
and test with an average of 60% (Fig.8).
Testing carried out somewhat balanced.
Figure 8. Monitoring Lotus Learning Space
Netlearning
E-learning platform for synchronous and
asynchronous collaborative learning [8], [4],
[12]. The performance evaluation of a 100
contingent (between 25 and 35 years) students
in the community averaged 55% of course
time, material quality distributed by an
average of 50%, quality of lessons, exercises
and test with an average of 56% (Fig.9). Well
balanced test monitoring.
Figure 9. Monitoring Netlearning
Saba Learning Enterprice
Modular e-learning platform consists of Saba
Publisher and Saba Content for mixed and
customizable learning [4], [6], [12]. The
ISBN: 978-1-941968-45-1 ©2017 SDIWC
performance assessment of a 100 contingent
(20 to 30 year old) students in the community
averaged 60% of course time, material quality
distributed by an average of 50%, quality of
lessons, exercises and test with an average of
53% (Fig.10). Almost perfectly balanced
monitoring in the various test conducted.
Figure 10. Monitoring Saba Learning Enterprice
T-learn
E-learning platform for collaborative learning
[4], [7], [12]. The performance evaluation of a
100 contingent (between 20 and 30 years old)
students in the community averaged 30%,
material quality distributed by an average of
45%, quality of lessons, exercises and test
with an average of 53% (Fig.11). Almost
perfectly balanced monitoring in the various
test conducted.
Figure 11. Monitoring T-learn
WebCT
18
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
Standalone e-learning platform for
synchronous and asynchronous learning [3],
[5], [12]. The performance evaluation of a
100 contingent (between 20 and 30 years old)
students in the community averaged 40% of
course time, material quality distributed by an
average of 50%, quality of lessons, exercises
and test with an average of 43% (Fig.12).
Quite balanced monitoring in the various test
conducted.
Figure 12. Monitoring WebCT
WebConference
E-learning platform with the possibility of e-
meeting, allows sharing of applications [2],
[4], [12]. The performance assessment of a
total of 100 students (aged 25 to 35) in the
community has averaged 41% of course time,
material quality distributed by an average of
45%, quality of lessons, exercises and test
with an average of 53% (Fig.13). Monitoring
in the various test carried out fairly balanced.
Figure 13. Monitoring WebConference
ISBN: 978-1-941968-45-1 ©2017 SDIWC
The platforms that reported the best results for
the duration of courses are Saba Learning
Enterprice and LearnLinc; for distributed
material quality are WebCT, Saba Learning
Enterprice, Netlearning and Hotconference;
for quality theoretical lectures transmitted and
quality exercises performed are Groove and
Lotus Learning Space.
3 CONCLUSIONS and FUTURE
DEVELOPMENTS
To conclude, the potential of existing
proprietary solutions was analyzed. In
particular, the study focuses on monitoring
content training activities in order to identify
those features considered fundamental as
evolution developed by increasingly
innovative technology, which puts the user at
the center of the use of new content. In this
scenario, social collaborative learning has
been considered that transforms the system
from a learning material container into a
knowledge sharing and management tool.
Practice trials involving course planning,
tracking activities, evaluation reports, forums,
discussions, application sharing nd
communication between learning objects
showed a positive trend considering
simulations in terms of traceability and results
to appreciate the questionnaires given to the
trainees in terms of acquiring knowledge and
variations in performance by those who
participated in it for flexible learning. In
delivering the courses, the observations and
comparisons made have been used to
understand the user friendly role of online
platforms navigability, how to use multimodal
lessons, the content, the materials, the
verification and evaluation forms used, the
interactive support during the course phases.
There are still some improvements that hit the
interface of most platforms and the lack of
flexibility. The analysis has proved to be
decisive considering peculiarities of the
literature regarding in-depth studies in the
field. At the moment, further developments
tend to favor interoperability.
REFERENCES
19
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

[1] M. Banzato, D. Corcione, Piattaforme per la [11]
didattica in rete, TD-tecnologie didattiche n. 33,
2004, pp. 22-31, edizioni menabò, Ortona.
[12]
[2] P. Campanella, Piattaforme per l’uso integrato
di risorse formative nei processi di e-learning,
Atti Didamatica 2015 – Studio ergo lavoro –
dalla società della conoscenza alla società delle
competenze, Aica, 15-16-17/04/2015, Genova,
Italy.
D. Colombo, Formazione a distanza, ambienti e
piattaforme telematiche a confronto, 2001.
D. F. Garcia, C. Uria, J. C. Granda, F. J. Suarez ,
F. Gonzalez, A functional evaluation of the
commercial platforms and tools forrd synchronous
distance e-learning, Proc. of the 3 wseas/iasme
international conference on educational
technologies, Arcachon, France, 2007, pp. 330-
335.

[3] P. Campanella, Piattaforme Proprietarie: [13] S. Impedovo, P. Campanella, G. Facchini, G.

Un’analisi metodologica, Atti Didamatica 2015
– Studio ergo lavoro – dalla società della
conoscenza alla società delle competenze, Aica,
15-16-17/04/2015, Genova, Italy.
[4] P. Campanella, Platforms and methods for the
integrated use of educational resources in the
processes of e-learning, Ed-Media 2011, world
conference on educational multimedia,
hypermedia and telecommunications 2011,
chesapeake aace, in t. Bastiaens & m. Ebner
(eds.), Proc. of the world conference on
educational multimedia, hypermedia and
telecommunications 2011, chesapeake, va:
AACE, 27-28-29-30/06-01/07/2011, pp. 2375-
2384, Lisbona, Portogallo.
[5] P. Campanella, Functional Comparison of the
Tools and Commercial Platforms in distance e-
learning, Proc. of the IADIS International
conference e-learning 2011, 20-21-22-
23/07/2011, Roma, Italy.
Pirlo, R. Modugno, L. Sarcinella, Learning
Management Systems: Un’analisi comparativa
delle piattaforme open-source e proprietarie,
Atti Didamatica 2011 - Informatica per la
didattica, Aica, 04-05-06/05/2011, Torino, Italy.
[14] S. Luciani, Caratteristiche tecniche e
funzionalità didattiche delle piattaforme per
l’apprendimento on-line, in Formazione e
cambiamento, web magazine sulla Formazione,
2004.
[15] M. Pedroni, Dall’interoperabilità delle
piattaforme all’integrabilità dei moduli
interattivi, Omniacom editore, pp. 731-735, Atti
Didamatica 2004, Ferrara, 06-08/05/2004.

[6] P. Campanella, Platforms for use integrated

resources formative processes in e-learning,
Proc. the 2nd international conference on digital
information processing, data mining, and
wireless communications (DIPDMWC2015),
16-18/12/2015, Islamic azad University, uae
branch, pp. 181-186, Dubai.

[7] P. Campanella, Learning Management Systems:

A Comparative Analysis of Open-source and
Proprietary Platforms, Proc. the 2nd International
conference on digital information processing,
data mining, and wireless communications
(DIPDMWC2015), 16-18/12/2015, Islamic azad
University, uae branch, pp. 187-192, Dubai.

[8] P. Campanella, A Comparative nd Assessment of

E-learning Platforms, Proc. the 2 International
conference on digital information processing,
data mining, and wireless communications
(DIPDMWC2015), 16-18/12/2015, Islamic azad
University, uae branch, pp. 193-198, Dubai.

[9] S. Campanella, G. Dimauro, A. Ferrante, D.

Impedovo, S. Impedovo, M. G. Lucchese, R.
Modugno, G. Pirlo, L. Sarcinella, E. Stasolla, C.
A. Trullo, E-learning platforms in the Italian
Universities: The technological solutions at the
University of Bari, WSEAS Transactions on
advances in engineering education, issue 1, vol.
5, 2008, pp. 12-19.

[10] F. Colace, M. De Santo, M. Vento, Evaluating

on-line learning platforms: A case study, Proc.
of the 36th Hawaii International conference on
system sciences (hicss’03), Hawaii, IEEE press.

ISBN: 978-1-941968-45-1 ©2017 SDIWC 20

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

Mobile Learning: A Case of Study

Pasquina Campanella
Department of Computer Science
University of Bari “Aldo Moro”
via Orabona,4 – 70126 Bari – Italy
pasqua13.cp@libero.it

ABSTRACT

Digital media technologies and broadband well as reported experimental results and
communications networks are undergoing conclusions and future developments.
profound transformations and new trends in
content development for training are emerging. In
2 CASE STUDY
this scenario with the rapid introduction of mobile
devices among which the smartphone prevails, it
was possible to meet the needs of users with a Within the framework of mobile learning, a
paradigm that involves the “learning in mobility” study was launched that consider analysis as
process. This article provides a case study for well as modular integration with Oracle i-
mobile oracle i-learning and claroline platforms Learning and Claroline platform plug-in on
considering its content delivery, user-based four different mobile operating systems such
monitoring test as well as evaluation techniques to as android, iPhone OS, symbian, windows
better manage interactive online courses that make mobile, in order to promote communication
the user active web participant in the production by means of services [2], [16], [17].
process. Accessing the mobile learning area through a
specially created application that bring
KEYWORDS
important features such as viewing content
platforms, questionnaires, learning, monitoring, and other content, proven on emulators and
interoperability. real life devices. The prototype examined is
the mobile oracle i-learning and claroline
1 INTRODUCTION platform of which the screenshots are shown
(Fig.1):
In the last few years, there has been a large
scale, large scale distribution of mobile
devices such as cell phones, handhelds,
pocketPC, ebook, tablet pc, smartphones, tv-
phonics, ipod, ipad and other portable
devices, personal communication devices are
becoming devices suitable for displaying
multimedia contents. A new communication
tool, a new frontier for e-learning: mobile
learning [1], [2], [3], [4], [5], [7], [8], [9],
[10], [12], [13], [14], [18], [19], [21], [23],
[26], [27], [28]. Users pass from simple users
to content creators, designed, modified, or
simply shared [10], [11], [20], [22]. Today
there is a strong demand for immediate use,
easily assimilable and immediate oriented
applications [16], [17]. In this context, a Figure 1. Oracle iLearning - Claroline mobile
follow-up study case has been launched, as platforms screening

ISBN: 978-1-941968-45-1 ©2017 SDIWC 21

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
The user interface was very intuitive. The
issues they raised concern: What didactic and
communicative models are most effective?
What technological solutions can encourage a
broad participation of users? Based on these
questions, evaluation sessions were performed
as well as simulations with actual users of
prototypes created. Interest has focused on a
particular subset of applications in which the
mobile device has been used to increase and
improve communication.
3 EXPERIMENTAL RESULTS
During the simulations, the students involved
were subjected to observations and then
interviewed through questionnaires (Fig.2) to
ascertain their views.
Figure 2. Questionnaire on mobile learning
technologies
The research was carried out at the Intelligent
Systems Lab and involved 80 students aged
19 to 21 in Computer Science, who were
asked to comment on how m-learning has
them supported during the learning phase as
well as to arouse their interest. The students
have proved to be in favor of
experimentation, even when technical
problems have prevented them from working
properly, although only some of the features
have been found to be responsive to the
collaborative level access and consultation.
Following is a graphical representation of the
conducted questionnaire (Fig.3):
ISBN: 978-1-941968-45-1 ©2017 SDIWC
Figure 3. Users questionnaire results in mobile
learning
The results outlined above, for the first
question, have found that 71% of students
claim that mobile technologies have
supported it adequately during the learning
phase. For 24%, the technologies did little to
support the learner during the learning phases.
While 5% said they did not receive any
support from the use of new technologies. For
the second question, 76% claim that mobile
devices make the learning phase more
interesting. For 21% use is of little interest,
while 3% does not involve involvement.
Asked what he wants to receive from his
mobile device, the most common response
was to get new information. Asked what a
mobile device can offer, 90% of users
answered more and learn better. Ultimately,
positive results were achieved in terms of
satisfaction, acquisition of knowledge and
performance variations by those who
participated in it. It is crucial that the learner
has access to a flexible learning strategy and
that all teaching resources are available at any
time and in various types of support to allow
users access to information based on their
preferences, attitudes and the needs [6], [16],
[24], [25]. The data shows a wide availability
of IT and a significant predisposition to the
use of mobile devices. The mobile device
learning experience finds that around 95%
rated the teaching method positively and
about 90% wanted to continue studying
through the mobile phone. The aim was to
create a flexible learning model, which makes
it possible to access information with every
22
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
type of device and produce flexible materials
taken from different situations [4], [15], [17].
The critical issues are the small size of the
screen, which does not allow for a large
amount of content but only essential concepts,
the difficulties of interoperability between the
various devices and connectivity has been
somewhat fragmented.
4 CONCLUSIONS and FUTURE
DEVELOPMENTS
Concluding the rapid diffusion of large scale
mobile devices such as cell phones,
handhelds, pocketPC, ebook, tablet pc,
smartphones, tv-phonics, ipod, ipad, and other
portable devices bring new trends in
developing content for training or so-called
“learning in mobility”. A new frontier for e-
learning: mobile learning. In this context, a
study was conducted between students on
mobile oracle and claroline platforms tested
on four different mobile operating systems
such as android, iPhone OS, symbian,
windows mobile considering its content
delivery, user-based monitoring tests and
evaluation in order to better manage
interactive online courses, which make the
web user active in the production process.
The results obtained were positive in terms of
satisfaction, acquisition of knowledge and
performance variations by those who
participated in it. Ultimately, it is crucial that
the learner has access to a flexible learning
strategy and that all teaching resources are
available at any time and in different types of
support. The minor criticisms are related to
interoperability between the different devices
and further studies are being carried out.
Ultimately, we can say that m-learning is
aimed at bridging the emerging needs of
digital natives and training outcomes.
REFERENCES
[1] B. Alexander, “Going Nomadic: Mobile
Learning in Higher Education”, Educause
review, vol. 39, n. 5, 2004, pp. 28-35.
[2] M. Alier, J. Casany, P. Casado, “Mobile
extension of a Web based Moodle Virtual
Classroom”, in P. Cunningham, M. Cunningham
(ed.), expanding the knowledge economy: issue,
ISBN: 978-1-941968-45-1 ©2017 SDIWC
applications, case studies, vol. 4, 2007, pp. 1169-
1176, IOS press.
[3] S. Al-khamayseh, A. Zmijewska, E. Lawrence,
G. Culjak, “Mobile Learning Systems for Digital
Natives”, Proc. of 6th Iasted International
Conference on web based education, 2007, pp.
252-257, Chamonix, France.
[4] J. Attewell, “Mobile Technologies and Learning:
a Technology Update and M-learning project
summary”, learning and skills development
agency, United Kingdom, 2005.
[5] P. Campanella, “Mobile Learning: New forms of
education”, Proc. of 10th International
Conference on Emerging e-learning technologies
and applications, ICETA 2012, IEEE, 08-
09/11/2012, pp. 51-56, Stará Lesná, the high
tatras, Slovakia.
[6] P. Campanella, “Mobile learning application for
Android using web service”, in p. Resta (ed.),
Proc. of Society for information technology &
teacher education international conference, SITE
2012, AACE publish, 05/03/2012, pp. 1677-
1682, Austin, Texas, USA.
[7] Y. Y. Chan, S. C. Chan, C. H. Leung, A. K. W.
Wu, “Mobilp: a mobile learning platform for
enhancing lifewide learning”, Proc. of the 3rd
IEEE international conference on advanced
learning technologies, Athens, Greece, pp.457-
457, 09-11/07/2003.
[8] S. J. Geddes, “Mobile learning in the 21st
century: benefit for learners”, knowledge tree E-
journal, vol. 30, n.3, 2004, pp. 214-228.
[9] G. Guazzaroni, “Fare esperienze di
apprendimento con tecnologie di mobile
learning”, tratto da giornata di studio sul mobile
learning organizzato dal collaborative knowledge
building group (CKBG), Genova, 2010.
[10] J. Herrington, A. Herrington, J. Mantei, I. Olney,
B. Ferry, “New technologies, new pedagogies:
mobile learning in higher education”, faculty of
education, University of Wollongong, Australia,
2009.
[11] J. Herrington, J. Mantei, A. Herrington, I. W.
Olney, B. Ferry, “New technologies, new
pedagogies: mobile technologies and new ways
of teaching and learning”, ascilite of teaching
and learning, in atkinson, r & mcbeath, c (eds),
proc. Asclite, 2008, pp. 419-427, Melbourne,
Australia.
[12] D. Keegan, “Mobile learning - theth next
generation of learning”, Proc. of the 18 asian
association of open universities annual
conference, Shanghai, China, 28-30/11/2004.
[13] J. Kossen, “Mobile e-learning: when e-learning
becomes m-learning”, Palmpower magazine,
2005.
23
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

[14] A. Kukulska-hulme, J. Traxler, “Mobile
learning: a handbook for educators and trainers”,
vol. 8, n. 2, Rroutledge, London, 2005.
[15] S. Impedovo, IAPR Fellow, IEEE S. M., P.
Campanella, “Mobile recommended system on
Android platform”, Proc. of the 18th
International conference on distributed
multimedia systems, DMS 2012, dblp, pp. 33-38,
09-10-11/08/2012, eden roc renaissance Miami
beach, Florida,USA.
[16] S. Impedovo, IAPR Fellow, S. M. IEEE, P.
Campanella, “Mobile Computing: sviluppo
applicazione Voip su Symbian OS”, Atti
Didamatica 2012 - Informatica per la didattica,
Aica, Taranto 14-15-16/05/2012.
[27] K. Yordanova, “Mobile Learning and integration
of advanced technologies in education”, Proc. of
the international conference on computer
systems and technologies, compsystech’07, pp.
1-5, Acm, 14/06/2007.
[28] B. Zuga, I. Slaidins, A. Kapenieks, A. K.
Strazds, “M-learning and mobile knowledge
management: similarities and differences”,
International journal of computing &
information sciences, vol. 4, n. 2, 2006, pp. 58-
62.

[17] S. Impedovo, P. Campanella, G. Facchini, G.

Pirlo, “Mobile Platforms: un’analisi
comparativa”, Atti VIII convegno nazionale sie-
l, “connessi! Scenari di innovazione nella
formazione e nella comunicazione”, Ledizioni,
pp. 499-504, Reggio Emilia, Italy, 14-15-
16/09/2011.

[18] C. Leung, Y. Chan, “Mobile Learning: A new

paradigm in electronic learning”, Proc. of the 3rd
IEEE International conference on advanced
learning technologies (ICALT’03), 09-
11/07/2003, pp. 76-80.

[19] L. Milrad, “Mobile Learning: challenges,

perspectives, and reality”, in mobile learning:
essays on philosophy, psychology and education,
passagen verlag Vienna, Austria, 2003, pp.151-
164.

[20] A. Mulliah, E. Stroulia, “Mobile devices for

collaborative learning in practicum courses”,
International Journal of mobile learning and
organisation, vol. 3, n.1, 2009, pp. 44-59.

[21] M. Pieri, D. Diamantini, “Il mobile learning”, ed.

Angelo Guerini e associati, Milano, 2009.

[22] H. Ryu, D. Parsons, “Innovative mobile learning:

techniques and technologies”, information
science reference, IGI publishing, Hershey, New
York, 2009, pp. 21-46.

[23] P. Seppala, H. Alamaki, “Mobile learning in

teacher training”, Journal of computer assisted
learning, vol. 19, 2003, pp. 330-335.

[24] M. Sharples, D. Corlett, O. Westmancott, “The

design and implementation of a Mobile Learning
resource”, Journal of personal and ubiquitous
computing, vol. 6, n. 3, 2002, pp. 220-234.

[25] M. Sharples, “The design of personal mobile

technologies for lifelong learning”, in Computers
and Education, vol. 34, 2000, pp. 177-193.

[26] J. Traxler, “Defining mobile learning”, Proc.

Iadis international conference Mobile Learning
2005, IADIS press, pp 261-266, Lisbona,
Portogallo.

ISBN: 978-1-941968-45-1 ©2017 SDIWC 24

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
Message Hiding with Pseudo-Random Binary Sequence Utilization
Gabriel Bugár, Juraj Gazda, Dušan Levický
Faculty of Electrical Engineering and Informatics,
Technical University of Košice, Letná 9, 042 00 Košice, Slovak Republic
gabriel.bugar@tuke.sk
ABSTRACT
This paper presents a novel computational secret
message hiding technique, also called Code-
booked steganography technique based on Code
Division Multiple Access (CDMA) technology
with having regard to security and perceptibility
of hidden message. This CDMA technique and
their variations have mainly exploitation in
mobile phone communication standards, the
world's most widely used cell phone standard. We
model the cover object transmission over two
interacting social networks using a user based
model in which we apply the fundamental
principles of the steganography. We show that the
imposed features of CDMA techniques agree with
imperatives claimed upon steganography systems
and it facilitates the creation of an unsuspicious
communication channel.
KEYWORDS
Steganography; SSIM; DCT; DWT; Social
Network.
1 INTRODUCTION
Steganography provides techniques of hidden
communication that is mainly used in
countries where usually the freedom of
speech is forbidden. The fundamental
principle lies in embedding the secret
message into a camouflage inconspicuous
cover data; where one of the indispensable
requirement is to ensure that an unwanted
party will not be aware of the existence of the
embedded information in a cover object.
From that place, its goal is to hide the
presence of communication [1], [2].
Nowadays, the most used unsuspicious digital
communication cover objects are digital
document files, image and audio files,
programs or protocols [3]. Mainly the large
ISBN: 978-1-941968-45-1 ©2017 SDIWC
file size media files are excellent for
stenographic transmission because of their
large part of redundancy [4]. Overleaf, in
some situation, the knowledge and possibility
of detecting a hidden communication and
messages may be needed. This part of science
is often called steganalysis, and its primary
objective is to break a steganography system
and uncover a hidden communication [5].
This steganalysis has also led to several very
sophisticated image stenographic algorithm
invention. Just a brief review of these: the
most popular one is the JSteg method, which
embedding a secret message in sequential
order of the quantized DCT coefficients LSBs
while skipping 0s and 1s [6]. The next, also
prevalent one is the OutGuess method, this
disseminates data into the LSB of quantized
DCT coefficients. The coefficients
modification is followed by a correction
process to ensure consistent distributions of
any related pair of the DCTs. Another well-
known method algorithm applies the
technique of matrix encoding to hold secret
message using LSB DCTs in F5 [7].
Steganography channel expressed in this
paper can represent a conventional system by
hiding a message in other regular
inconspicuousness information - a secret
image in a cover image. There, in the system
verification process, we tried to improve the
algorithm substantiality, and that's why we
used as the transport media the user social
network account information wall. The
embedding process relies upon hiding a
modified secret image in each area of DWT
transformation coefficients of quantized DCT
coefficients LSB plane. It means the
execution of two separate transformations of a
cover object before the embedding of the
secret message takes place. The acquisition of
25
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

proposed method based on pseudo-random

SSIM  A, B  
2  B  C1   2 cov AB  C2 
A
(4)
binary sequences holds the meaningful    B2  C1   A2   B2  C2 
2
A
improvement of confidence level in
steganography systems. Cover object Here, MAXI is the maximum possible value
degradation is satisfactorily small, in this of the object.
manner; the stego image quality distortion is
more imperceptible to the human eye. 2.2 2D - Discrete Cosines Transformation
Secondly, adding the cryptographic element
to the stenographic algorithm in the process of Discrete Cosines Transformation (DCT)
embedding represented by the CDMA performs an object transformation from
behavior is also enormously increasing the spatial to the frequency domain. We decided
security and imperceptibility of the to use the 2D modification of DCT. 2D-DCT
steganography system. provides upper computational convergence;
therefore, it offers better usability in the field
2 GENERAL SUPPOSITIONS of steganography. The high computational
convergence means that by multiplying the
In this section, we briefly describe some horizontally oriented 1D basis functions with
needed necessary knowledge to understand a vertically oriented set of the same features,
the proposed model. A fundamental of the we make 2D basis functions. Consequently,
process of secrecy a secret message relies on the 2D-DCT is a direct extension of the 1D
the differences of stego object. The stego case and is given by the following way:
object represents the original cover object N 1 N 1
 2i  1u   2 j  1v 
Fu ,v  Cu Cv  f (i, j ) cos  cos  (5)
with an embedded hidden message. A i 0 j  0  2N   2N 
function of conformity can express its  
1 1
measure. Hence, the objective conformity  u0  v0
 N  N
measures, which stem from a statistical Cu   Cv   (6)
approach, are based on measuring the  2  2

 N otherwise 
 N otherwise
distortion between an original cover object
and modified cover object = the stego object.
2.3 2D - Discrete Haar Wavelet
2.1 Peak Signal to Noise Ratio (PSNR) Transformation

The static multi-level images use several Discrete Wavelet Transform based on Haar
metrics of the objective quality evaluation. In wavelet (HT) is the simplest useful energy
the field of steganography, those usually compression process which can effectively
measure the distortion between stego and serve very useful and fast object
cover objects. One of the most common is the decomposition. The Haar transform, like all
PSNR, based on the mean squared error wavelet transforms, decomposes a discrete
(MSE). Given a noise-free M×N monochrome object into sublevels {a1|d1} of half its length,
image object I and its noisy approximation K, where a1 = [a1, a2 …aN/2] represents the
then MSE is defined as [8]: approximation (average) coefficients and d1 =
[d1, d2 …dN/2] the detailed (difference)
2

 I (i, j )  K (i, j )
M N

MSE 
1 coefficients. The first value of approximation
(1)
M .N i 1 j 1 coefficient a1 is computed by taking the
PSNR  20 log10 MAX I   10 log10 MSEdB
average of the first pair of values of and then
(2)
multiplying it by the square root of 2.
255 2 Application of 2D-HT to an object will
PSNRCSF  10 log10 dB
(3) retrieve transformation coefficients that are
 I  I  g 
N1 N2
1 W 2

N1 .N 2 i 1 j 1
CSF defined as a decomposition of an input object,
also known as approximation component LL
and detailed components LH, HL and HH [9]:

ISBN: 978-1-941968-45-1 ©2017 SDIWC 26

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
LL (m, n)  a(2m  1, n)  a(2m, n) 2
1 2
(7)
LH (m, n)  2 a(2m  1, n)  a(2m, n)
1 2
(8)
HL(m, n)  d (2m  1, n)  d (2m, n) 2
1 2
(9)
HH (m, n)  2 d (2m  1, n)  d (2m, n) 
1 2
(10)
2.4 Spread Spectrum Image
Steganography (SSIS)
The Code-division multiple access is one of
the numerous access techniques, where
several transmitters can send information
simultaneously over a single communication
channel. This method stands on mathematical
apparatus of orthogonal pseudorandom
number sequences (PNS) which allow
multiple accesses on the same communication
without interference. This experience is
achieved by orthogonality and the
autocorrelation aspect of PNS. The
autocorrelation controls the impact of noise
on the stenographic communication. Direct
spreading technique (DSS) by PNS stimulates
to use properties applicable in steganography.
Firstly, PN sequences have an almost
subequal quantity of -1 and 1. It is why a
spread data (secret data) do not directly
appear in bulks or other periodical sequences,
which could be a stimulus to an intruder to
uncover the communication. Secondly, the
longer length of PNS provides better
correlation results; therefore, a higher number
of error/broken bits during the extraction
process of the secret message can be admitted
[2010].
3 PROPOSED ALGORITHM
IMPLEMENTATION
The main idea of proposed method utilizes a
combination of DCT and DWT transform in
the process of message embedding.
Subliminal channel creation processes grey
scale images as the cover object, and it
employs the technique of CDMA for great
secret messages spreading into transform
coefficients of 2D-DCT and 2D-DWT during
the procedure of embedding. This method
works with free hidden messages, which
means that hidden message objects could
ISBN: 978-1-941968-45-1 ©2017 SDIWC
consist of any binary data file. The
embedding process is starting with electing of
the proper cover object from the group of
appropriated objects. These were prepared for
this usage based on steganography expected
properties. The elected one is in the first step
decomposed into single color components
where each one is divided into blocks with
size eight by eight. Every block is
transformed into the frequency domain by
force of the DCT transformation. Thus we get
matrices of decimal numbers according to the
DCT behavior (compress the energy to DC
and low spatial frequencies, i.e., highest
numbers). We can say that the energy packing
properties of the DCT are excellent.
Afterward, we decomposed every transform
coefficient matrices by DWT transform.
There, in this DWT perform we use just
coefficients without fractions; i.e., we use
only the integer part of the decimal numbers.
DWT based on Haar wavelets has a unique
advantage of accrued transform coefficients.
According to the chapter 2.3: 2D-HT, the
DWT provides one block of approximation
(HH) and three blocks of detailed coefficients
(horizontal LH, vertical HL, and diagonal LL)
on each level of decomposition. The main aim
was to create a stealth communication, where
the secret message will account a very high
level of imperceptibility in the cover object.
Therefore, the detail coefficients are the most
convenient area for secret message
embedding. Also, a critical requirement was
the system capacity. There, the amount of
embedded data (capacity) and total distortion
of the cover object are considerably and
reciprocally related. Sometimes the system
capacity is represented as the payload, i.e., it
can be expressed as some embedded bits per
cover image picture element (bit/pixel). Our
proposed algorithm uses the 1-level
decomposition of discrete 2D Haar wavelet
transformation (2D-HT). In general, the
approximation coefficients are not very
suitable for modification because there is the
most information content of the whole cover
image. Vice-versa, detailed coefficients are
the most convenient area for doing some
operations (for hiding a secret data here). We
27
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
tried to design the algorithm in the way of
minimal modification of transformation
coefficients as the imperceptibility of secret
message would then be better. We build our
work on the fascinating feature of 2D-HT
coefficients; therefore, there is a high
correlation between the transformation
coefficients on the same position in different
decomposition areas. It means, if coefficient
in LL on a position (i, j) is an even-numbered
integer then the coefficients in LH, HL, and
HH on the same location (i, j) are also even-
numbered integers. Our next finding during
the investigations was that if we change one
transform coefficient value in one
decomposition area, then the inverse 2D-HT
(2D-IHT) changes this value back to the
original one. Otherwise, when we do the same
change also in other three areas of
decomposition at the same position, then this
change will be preserved. Another and
curious problem is a due behavior of 2D-IHT.
Sometimes, the reconstructed image pixel
values may not be integer numbers. After this,
the image would not be possible to
reconstruct and illustrate. From these
findings, we concluded, that there exist only
four suitable changes of transform
coefficients. Each modification affects a
different set of image pixel values after 2D-
IHT is performed. Table 1. shows all possible
coefficient changes.
pixel 2D-HT decomposition areas
position HH HL
f(2m-1,2n-1) increase (+) increase (+)
f(2m-1,2n) increase (+) decrease (–)
f(2m,2n-1) decrease (–) increase (+)
f(2m,2n) decrease (–) decrease (–)
Table 1. Transformation coefficients suitable changes.
The first column determines a set of specific
pixel values positions. Variables m and n are
pixel indexes, where m = 1, 2, 3… M and n =
1, 2, 3... N and image size are intended as
M×N. The "increase" label on the table 1.
represents a modification of the value by
increasing it according to the value that has
been modified by the embedding of secret
message in LL area on the same punctual
ISBN: 978-1-941968-45-1 ©2017 SDIWC
position. Contrariwise, the "decrease" label in
the table 1. represents a modification of the
coefficient value by increasing it according to
the value that has been modified by the
embedding of secret message in LL area in
the same position.
3.1 Process of embedding
We designed the proposed method to be
usable for all kind of secret digital messages,
i.e., it can be any numeric binary data file.
Firstly, the elected cover image is
decomposed by 2D-HT. Based on the
previous chapter; this transform provides an
approximation HH and three detail coefficient
areas (horizontal HL, vertical LH, and
diagonal LL) on each level of decomposition.
It was necessary to accomplish the property of
private communication as well as the secret
message imperceptibility in the cover image.
Therefore, as it was explained, the detail
coefficients are the most convenient area for
secret message embedding. Another, not a
negligible requirement is the system capacity.
The capacity or payload can be defined as
some embedded bits per pixel (bit/pixel). The
secret message embedding process can be
achieved in three embedding approaches,
where every method provides a specific level
of robustness. Differences between
approaches are mainly in the fact that the
same data bits are replicated into other
LH coefficients of detailed areas. In individual
increase (+) strategies, we try to embed secret data bit
decrease (–)
from one area of detail coefficients until all
three areas HL, LH, LL respectively.
decrease (–)
Introduction of higher method approaches
increase (+) introduces more details coefficients
modifications without contribution to the
system payload. However, higher approach
guarantees the higher secret message
robustness.
We defined a set of applicable transform
coefficients C' suitable for embedding
process, which was elected from detail areas
HL, LH, LL (5). It was also needed to prepare
and define a set of hidden message bits S
before embedding itself (6).
C   cij | 1  i  M c ,1  j  N c  (11)
28
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
S  si | 1  i  K , si  0,1
The preparation consists of two parts. First
part is dealing with the determination of the
length of Code-word, which determines the
size of secret data stream segmentation.
Analogically, the length of Code word {bi | i =
1, 2... N} directly determines the number of
pseudorandom sequences (PNSK). This
PNSK are used in the process of spreading,
ergo the SSIS approach, shown in Fig. 1.
Figure 1. Secrete message spreading before
embedding itself.
Secondly, the secret key can be used in the
process of PNSK alignment, with which the
secret message is spread. This key can denote
positions of bits during the extraction process.
However, one significant problem with this
approach is that this secret key needs to be
transported and present on the receiver side. It
represents the weakness of this approach, and
it can lead to uncovering of the covert
communication.
Subsequently, when the secret message is
correctly spread, they are embedded in the
transformation domain of 2D-HT coefficient
regarding attributes of steganography. In this
embedding process, we firstly perform
modulo calculations, and comparisons of the
secret message summarized spread bits Si
with the detail coefficients in the form of (13–
16).
if Si= 0:
C  , if mod(| C  |,1)  0 (13)
C 
ij
   
E
ij
 sgn C  
ij
| C  |
ij
 1 2 if mod(| C  |,1)  0 (14)
if Si= 1:
Cij , if
CijE  
 sgn Cij   | Cij | 1 2 if
ISBN: 978-1-941968-45-1 ©2017 SDIWC
(12) The most unusual behavior in here is the case
when no changes are needed to achieve the
secret message bit be embedded. In other
words, the secret message bit is embedded
without any modification of transform
coefficient.
Nevertheless, independently of minimal
modifications, we encountered an unpleasant
problem on the receiver side. The issue we
found was related to stego object distortion,
caused by the implementation of multiple
transformations. Inverse operations mislead
decimal numbers into spatial domain and
rounding to integers introduces this issue.
This distortion is responsible for incorrect
extraction of secret message. It was the reason
for PNS implementation. Autocorrelation
attributes of PNS significantly improve
resultful extraction. In general, a more
extended PNS are accounted for uplifting
autocorrelation characteristics, hence better-
gained results. Moreover, an additional
approach termed Extraction with Error
Correction (EEC) was applied during the
extraction to improve error correction. The all
possible states calculations (2n) on the output
of defined code-word could be held as
reference data cell. However, incrementing of
code-word length requires higher computer
power, i.e., more time-consuming.
4 Results and Ascertainments
The total system capacity of the proposed
method is a quarter of used cover image size
in the 1-level decomposition of 2D-HT. The
maximum useful payload is 0.25 bit/pixel in
case of using the total capacity, and it also
varies depending on some used detail
coefficient during the process of embedding
and from the applied code-word length. The
total number of coefficient elected for
embedding determines the capacity in binary
form, and it does not vary with growing
ij
number of PNS. Unfortunately, the system
ij
capacity is decreasing with increasing length
of PNS. Considerable differences can be
mod(| Cij |,1)  0 (15) observed in the amount of added or removed
mod(| Cij |,1)  0 (16) energy to or from the cover object.
Subsequently, if more PNS are used in the
29
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

process of spreading, then more the is

embedded to cover object. This unwanted
property is measured by objective quality
measures (PSNR, PSNRCSF, SSIM
imperceptibility of human perception).

Decomp. PSNR PSNRCSF Extract. Capacity

n SSIM
area [dB] [dB] [%] utiliz. [%]
LL 59,654 233,267 0,9917 49,1
1 LL, HL 56,164 69,630 0,8714 65,7 95,2
LL, HL, HH 53,946 77,038 0,6121 78,5
LL 60,653 257,846 0,9939 74,1 Figure 2. Difference values of PSNR and system
2 LL, HL 57,875 72,009 0,8745 87,0 46,1 capacity.
LL, HL, HH 54,247 82,975 0,6178 97,9
LL 61,887 278,246 0,9951 90,1
4 LL, HL 58,367 86,451 0,8785 95,5 23,8 The differences are even more exposed to
LL, HL, HH 55,947 87,462 0,6221 100
LL 62,384 293,533 0,9975 95,7
longer code words, where better cross-
8 LL, HL 59,942 95,865 0,8812 97,9 11,8 correlation attribute of PNS stands its place.
LL, HL, HH 56,142 93,672 0,6274 100
LL 63,174 311,520 0,9998 98,4
16 LL, HL 60,285 106,647 0,8864 100 5,95 5 Conclusion
LL, HL, HH 57,056 98,975 0,6376 100

Table 2. Extraction rate results and PSNR values for
different cover objects.
In case of LL and n=1 (if one decomposition
area is utilized and no code word applied), the
satisfactory reconstruction rate is around
49%. This rate is unacceptable and not
applicable for our purposes. However, the
stego image visual quality values (PSNR-
SSIM) are incredibly satisfying. In case of
used LL, HL and HH coefficients for
embedding, the reconstruction rate has
increased to around 78% what is much better,
but still not enough for our purposes. It was
the reason why we implemented code words.
In Tab. 2 are all combinations of reached
results. It is evident if the length of the code
word is increasing then the extraction rate is
getting better, but at the expense of the
amount of payload to be embedded (amount
of secrete message bits). The Fig. 2 depicts
the difference values of PSNR between
applied decomposition areas and achieved
capacity utilization.
In this paper, we prepare steganography
method that uses properties of the
combination of the transformation domain of
2D-HT, 2D-DCT and Direct spreading
technique of CDMA. The objectives and
embedding algorithm solved the problem of
secret message reconstruction issue formed in
research [4], where the transformation domain
of DCT creates a wrong place for secret
message embedding. As it has been shown,
the capacity depends on length, but no longer
on number of PNS. Moreover, the error
correction algorithm increases the fruitfulness
of secret message extraction. Another
critically considered benchmark according to
steganography is imperceptibility of a human
observer to degradation of the cover object.
These measured values are decreasing with
employment of more decomposition areas of
2D-HT. This handicap is solved by increasing
the length of PNS because of cross-
correlation improvement. The autocorrelation
allows proper identification of PNS, thus the
secret message bit.

Acknowledgment

This work was supported by the Slovak

Research and Development Agency, project
number APVV-15-0055, and by the Scientific
Grant Agency of the Ministry of Education,
Science, Research and Sport of the Slovak

ISBN: 978-1-941968-45-1 ©2017 SDIWC 30

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

Republic under the contract No. 1/0075/15

and University Science Park TECHNICOM
for Innovation Applications Supported by
Knowledge Technology, ITMS:
26220220182, supported by the Research &
Development Operational Programme funded
by the ERDF. This project is being co-
financed by the European Union.

REFERENCES
[1] Gowda, S. N. (2016, July). Dual layered secure
algorithm for image steganography. In Applied
and Theoretical Computing and Communication
Technology (iCATccT), 2016 2nd International
Conference on (pp. 22-24). IEEE.

[2] Shih, F. Y. (2017). Digital watermarking and

steganography: fundamentals and techniques. CRC
Press.

[3] Kishor, S. N., Ramaiah, G. K., & Jilani, S. A. K.

(2016, May). A review on steganography through
multimedia. In Research Advances in Integrated
Navigation Systems (RAINS), International
Conference on (pp. 1-6). IEEE.

[4] Bugár, G., Levický, D., Broda, M., & Hajduk, V.

(2016, April). A novel block-based data hiding
scheme in SVD-DCT composite domain.
In Radioelektronika (RADIOELEKTRONIKA),
2016 26th International Conference (pp. 336-339).
IEEE.

[5] Banoci, V., Broda, M., Bugar, G., & Levický, D.

(2014). Universal image steganalytic
method. Radioengineering, 23(4), 1213-1220.

[6] Trivedi, M. C., Sharma, S., & Yadav, V. K. (2016,

March). Analysis of Several Image Steganography
Techniques in Spatial Domain: A Survey.
In Proceedings of the Second International
Conference on Information and Communication
Technology for Competitive Strategies (p. 84).
ACM.

[7] Westfeld, A. (2001). F5—a steganographic

algorithm. In Information hiding (pp. 289-302).
Springer Berlin/Heidelberg.

[8] Bánoci, V., Bugár, G., & Levický, D. (2010,

April). Information hiding using pseudo-random
number sequences with error correction.
In Radioelektronika (RADIOELEKTRONIKA),
2010 20th International Conference (pp. 1-4).
IEEE.

[9] Bugar, G., Banoci, V., Broda, M., Levicky, D., &
Miko, E. (2013, September). Blind steganography
based on 2D Haar transform. In ELMAR, 2013
55th International Symposium (pp. 31-35). IEEE.

ISBN: 978-1-941968-45-1 ©2017 SDIWC 31

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
Optimising the Innovation Capacity of Public Research Organisations – An
Agent-Based Simulation Approach
Davy van Doren
EA European Academy of Technology and Innovation Assessment GmbH
Wilhelmstr. 56, 53474 Bad Neuenahr-Ahrweiler, Germany
davy.vandoren@ea-aw.de
ABSTRACT
Although there is considerable attention for the role
of public research organisations in driving
innovation, methodologies that can be used to
analyse and optimise these roles are limited. An
agent-based model could potentially assess ex-post
the operation of public research organisations
within innovation systems, as well as allows ex-
ante analysis of potential intervention effects for
optimising their innovation capacity. Such concept
needs to entail two perspectives. From the
perspective of innovation, it should represent not
only organisational capacities and capabilities
required for research and development, but also
integrate antecedent and subsequent processes
related to the act of innovation. From the
perspective of innovators, it should integrate the
organisational metabolism that enables innovation,
as well as contain agglomerate effects of
organisational operation on the overall dynamic of
innovation networks. Based on such requirements,
some options for designing an agent-based model
are suggested.
KEYWORDS
agent-based model, public research organisations,
innovation network, innovation policy,
organisational learning
1 INTRODUCTION
For the governance of technological
trajectories, it has been argued that public
research organisations can be important for
enabling innovation [1]. As political directions
over time have moved towards increased
embracement of technology
ISBN: 978-1-941968-45-1 ©2017 SDIWC
commercialisation, the traditional independent
and upstream-position of public research has
been replaced by one that contains richer and
more interactive actor relationships [2][3]. Due
to the evolving roles and functions of public
research organisations, there is need to
understand the conditions that enable
innovation within public research
organisations and optimise their effectiveness.
One way to assess innovation within public
research organisations is through developing a
computational model. In the following
sections, some general features of such
modelling concept are described. Also, to
allow analytical distinction regarding the
various dimensions and levels that underlie the
innovation process, a theoretical framework is
proposed.
2 THEORETICAL FRAMEWORK
2.1 Dimensions of Innovation
In the context of public research organisations,
innovation trajectories could be considered as
consisting of four dimensions: (1) the
production of science, (2) the processing of
science to enable innovation, (3) the initiation
of innovation, and (4) the transfer or
conversion of innovation through
entrepreneurship:
• 1 - Science: With innovation defined as
a form of applied knowledge, theories
on innovation have generally been
based on the discovery and
management of scientific knowledge
by actors [4].
• 2 - Enabling innovation: Previous
research has highlighted the need for
32
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
organisational antecedents, capacities
and installed processes that can enable
innovation [5][6][7].
• 3 - Innovation: Organisations require
capabilities to actually realise the
production of technological inventions,
as well as skills to manage acquired
technological capital [8][9].
• 4 - Entrepreneurship: Entities require
adaptability and flexibility to handle
changing operational environments
resulting from technological change,
for example through driving
organisational diversification or
manage institutional spin-offs
[10][11][12][13].1
2.2 Levels of innovation
De facto representations of these dimensions
of innovation are likely to differ between
research organisations. To create analytical
distinction within and between institutes, three
conceptual levels of innovation could be
distinguished: (I) within entities, (II) between
entities, (III) and outside entities.
I - Within entities – entity level
Much has been written how organisational
structures relate to innovation. For this, it
seems helpful to classify entities based on
organisational capacity and organisational
processes, as well as how these relate to
organisational output [14][15]:
• Capacity: It has been shown how
knowledge management can be
considered from a capacity and
capability perspective [16], as well as
how firms need to recognise and
assimilate external knowledge for
survival [17].
• Processes: There has been attention
towards optimisation of distinct
processes, for example the process of
knowledge generation through
learning or the communication of
1
In recent years, various initiatives have started to
elevate the potential of innovation within regional
ISBN: 978-1-941968-45-1 ©2017 SDIWC
information [18][19]. Also, various
studies are conducted on how
processes could be optimally integrated
and adapted to improve system
flexibility [20]. Furthermore, research
has been conducted at the level of
employees, for example how human
relations and human resource
management influences organisational
performance [21][22][23].
• Output: Based on the architecture of
capacity and processes, the output of an
organisation in terms of innovation is
influenced. For example, it has been
shown how an organisation’s
performance in innovation relates to
different modes of strategy formulation
[24].
II - Between entities - network level
Between entities, research on networks and
innovation systems have highlighted the
relevance of inter-organisational activity for
explaining the dynamic of innovation [25],
also in order to create economic regions of
specialisation that can induce national and
international competitive advantages [26].
Furthermore, innovation and entrepreneurship
are known to occur within diverse actor
constellations and at different levels of scale
[27][28][29].
III - Outside entities - environment level
The dynamic of innovation is also strongly
influenced by effects external to entities and
entity relations – i.e. variables that are not
directly associated to the relationships existing
between entities. The environment constitutes
conditions that strongly influence the character
and context of (national) innovation systems
[25][30]. Such conditions are generally outside
the direct sphere of organisational control.
Based on the made categorisation of
innovation, the following theoretical
framework is composed (Figure 1).
entrepreneurship programs (e.g. REAP -
http://reap.mit.edu/).
33
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
Environment
Network
Entity
Science
Capacity
Processes
Performance
Enabling Innovation
Capacity
Processes
Performance
Figure 1. Theoretical framework to analyse innovation within and between organizations
3 METHODOLOGICAL APPROACH
The composed theoretical framework can be
operationalised by means of agent-based
modelling.
In the context of evolving innovation
networks, agent-based models have been
designed to simulate actor behaviour, as well
as to specify the inter-relationships between
new knowledge, knowledge transfer, market
selection and reward structures [31]. As these
types of simulations allow for ex-ante
evaluation and the analysis of non-linear
relationships, they have been adapted to
contexts ranging from university-industry
collaborations to research policy effectiveness
[32][33].
There are some requirements to be considered
in the development of simulation models. First,
for simulation models to be useful in real-
world settings, they should be able to make
past representations of innovation systems
reliably. In addition, they should be able to
make potential future projections and allow the
ISBN: 978-1-941968-45-1 ©2017 SDIWC
Entrepreneurship
Capacity
Processes
Performance
Innovation
Capacity
Processes
Performance
exploration of successful strategies within the
space that is being modelled. As such, models
should be flexible enough to explore multiple
scenarios and alternative actor-configurations,
in order to provide insights how overall agent
fitness can be improved.
To achieve such functionality within models,
there is need for legitimate data-generation and
data-management procedures. For this, three
methodological features are proposed: (1)
multi-level integration through ego-networks,
(2) empirical- and simulation-data embedment
through scores, and (3) complexity reduction
through model modularisation.
3.1 Methodological feature 1 – Multi-level
integration through ego-networks
Multi-level integration through ego-networks
relates to what and how data is used in setting-
up and calibrating the operational settings of
the model. Based on principles of evolutionary
economics, approaches have often been based
on large aggregate datasets to describe
innovation diffusion processes in relation to
industrial structures [34] or have integrated
qualitative observations on historic sector
34
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
development [35]. Also, in the tradition of
Monte Carlo simulation and NK modelling,
developed methodologies have included levels
of uncertainty for calculating actor fitness
under different scenarios, as well as to obtain
parametric probability distributions in the
context of nonlinear evolution [36][37][38].
However, a predominant use of network-based
data might obscure structures that are more
relevant at micro-level. To introduce a multi-
level dynamic in the set-up of models, ego-
network data could be used to integrate
information regarding the intensity, quality
and nature of actor connections. Through
analytical embedment of preferential
attachment, strength of ties, collaboration
intensity and variation, number of co-
participations, or the profile of partnerships
[39], empirically observed ego-networks of
entities could be used in calibrating and
validating model output.
3.2 Methodological feature 2 – Connecting
empirical- and simulation-data through
indicators and scores
The connection of empirical- and simulation-
data relates to the observability versus
unobservability of various innovation
phenomena. This is an important consideration
within modelling approaches that aim to
integrate both types of phenomena through
either collected empirical observations or in
silico produced data. The inclusion of both
data-types represents a certain duality and
challenge in translating multi-level based
innovation theory into an operating agent-
based model. To align innovation models with
innovation theories, it is therefore important to
obtain some form of data integration between
levels. For this, a system based on indicators
and scores is proposed to improve the
validation of innovation models.
Indicators
Indicators can enable measurement of
innovation trajectories dimensions (i.e.
science, enabling innovation, innovation,
ISBN: 978-1-941968-45-1 ©2017 SDIWC
entrepreneurship), as well as measurement of
the different levels at which innovation occurs
(i.e. within entities, between entities, and the
environment). Data used for the calculation of
indicators can be based on either empirical data
(quantitative or qualitative) or simulation data.
To come up with a comprehensive set of
indicators, previously developed criteria that
highlight organisational aspects influencing
the ability to innovate were identified as a
potential starting basis [40]. Such criteria can
then be used as a basis for the identification or
formulation of additional indicators (Table 1).
Scores
After indicators have been identified or
formulated, scores can be developed that
represent a certain overall dynamic or
performance of innovation. Scores should be
considered as a unit that aggregates multiple
indicators and that is constructed as rooted
within theoretical logic or stylised facts. As
scores allow the integration of both empirical
and simulation data, they function as links
between the collected empirical- and produced
simulation-data of simulations.
3.3 Methodological feature 3 –
Modularisation of model architecture
The modularisation of model architecture
relates to a need to methodologically deal with
system complexity. Popular concepts of
evolutionary economics have often been
translated into the architecture of models
explaining innovation dynamics. Concepts like
the bounded rationality of actors or the
transformation of produced knowledge have
often resulted in complex model architecture.
Complex model architecture decreases
operational understanding due to multiple
parameter interaction and the emergence of
feedback- and feedforward-processes. Also,
complex model architecture reduces model
flexibility, which hampers the adaptation of
models towards specific research questions or
contexts.
35
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

Table 1. Analytical grid for the categorisation of individual indicators and aggregated indicators. For illustrative
purpose, some example indicators are provided.
Dimensions of innovation
Levels of Science Enabling Innovation Innovation Entrepreneurship
innovation
Within entities Capacity - Revealed Scientific - Budget for applied - Revealed - Potential for spin-
Advantage (RSA) research Technological off
- Scientific personal - Science- Advantage - Entrepreneurship
innovation (RTA) programs
connectors - Engineers
Processes - Science focussed - Market research / - Innovation - Transfer of IPR
incentives identification focussed - Resource allocation
- Internal research exploitation incentives / management
projects opportunities - Identification
licensees
Output - Publications (basic - Publications - Patents - Spin-offs
research) (applied research) - Technologies
- Market assessment
Between Capacity - Basic research - Applied research - Innovation - Shared
entities collaborations collaborations networks infrastructures
- Organisational
embedment of
institutes
Processes - Interorganisational - Applied research - Contract - Identification
coordination partner projects research potential clients
- Basic research partner
projects
Output - Co-publications (basic - Co-publications - Product - Market niche
research) (applied research) development
Environment - Systemic / national - Systemic / national - Funding and - Institutional support
need for science need for transfer for market niches
innovation instruments - Price of innovation

In order for models to explain more complex
patterns of innovation but still be coherent and
flexible, models could be designed in a
modular fashion [41]. Additional to modular
models that are being developed in technology
development to overcome measurement
challenges and understand complex
interactions – for example biotechnology
[42][43] – some attempts have also been made
in the context of innovation dynamic [44].2
The construction of modules should depend on
several aspects. Based on the availability of
data and analysis of indicators and scores,
modules can be composed based on principle
components that relate to the applied analytical
framework and align with the chosen scope of
analysis. Second, the creation of modules is
also likely to depend on end-user
requirements. Based on articulated needs
regarding what output a model should produce
– both in terms of intelligence and interface
possibilities – innovation phenomena could be
translated into either central or peripheral
modules.
CONCLUSION
To understand the conditions that enable
innovation within public research
organisations, a preliminary methodological
concept was proposed that can be applied
within the construction of innovation models.
This concept – rooted in theories of
organisational structure, innovation systems
and evolutionary economics – can be
operationalised to include both antecedents
and impacts of innovation, as well as to
integrate processes that occur within and
between involved agents. For the further
development and operationalisation of this

2
The operational implications of multiple perspectives https://www.imagwiki.nibib.nih.gov/sites/default/files/
have also previously been discussed in research; for Ropella,%20Glen05Aug14cah.pdf
example, see

ISBN: 978-1-941968-45-1 ©2017 SDIWC 36

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
concept, some potential methodological
features were suggested to improve the
simulation of agent-based models.
REFERENCES
[1] W. M. Cohen, R. R. Nelson, and J. P. Walsh,
“Links and Impacts : The Influence of Public
Research on Industrial R & D,” Manage. Sci.,
vol. 48, no. 1, pp. 1–23, 2002.
[2] E. von Hippel, The sources of innovation, vol.
53, no. 9. 2013.
[3] M. Gibbons and R. Johnston, “The roles of
science in technological innovation,” Res.
Policy, vol. 3, no. 3, pp. 220–242, 1974.
[4] J. A. Johannessen, B. Olsen, and J. Olaisen,
“Aspects of innovation theory based on
knowledge-management,” Int. J. Inf. Manage.,
vol. 19, no. 2, pp. 121–139, 1999.
[5] C. E. Helfat et al., Dynamic capabilities:
Understanding strategic change in
organizations. Oxford, Blackwell Publishing,
2007.
[6] J. J. P. Jansen, F. a J. van den Bosch, and H. W.
Volberda, “Managing Potential and Realised
Absorptive Capacity: How do Organisational
Antecedents Matter?,” Acad. Manag., vol. 48,
no. 6, p. 16, 2005.
[7] J. Woodhill, “Capacities for institutional
innovation: A complexity perspective,” IDS
Bull., vol. 41, no. 3, pp. 47–59, 2010.
[8] F. E. García-Muiña and E. Pelechano-Barahona,
“The complexity of technological capital and
legal protection mechanisms,” J. Intellect. Cap.,
vol. 9, no. 1, pp. 86–104, 2008.
[9] H. M. Grimm, “The diffusion of Bayh-Dole to
Germany: Did New public policy facilitate
university patenting and commercialisation?,”
Int. J. Entrep. Small Bus., vol. 12, no. 4, pp.
459–478, 2011.
[10] H. Chesbrough and R. S. Rosenbloom, “The
role of the business model in capturing value
from innovation: evidence from Xerox
Corporation’s technology spin-off companies,”
Ind. Corp. Chang., vol. 11, no. 3, pp. 529–555,
2002.
[11] K. Pavitt, M. Robson, and J. Townsend,
“Technological Accumulation, Diversification
and Organisation in UK Companies, 1945-
1983,” Manage. Sci., vol. 35, no. 1, pp. 81–99,
1989.
[12] A. Walter, M. Auer, and T. Ritter, “The impact
of network capabilities and entrepreneurial
orientation on university spin-off performance,”
J. Bus. Ventur., vol. 21, no. 4, pp. 541–567,
2006.
[13] M. G. Jacobides and S. G. Winter,
“Entrepreneurship and firm boundaries: The
ISBN: 978-1-941968-45-1 ©2017 SDIWC
theory of a firm,” J. Manag. Stud., vol. 44, no.
7, pp. 1213–1241, 2007.
[14] R. G. M. Kemp, M. Folkeringa, J. P. J. de Jong,
and E. F. M. Wubben, Innovation and firm
performance. 2003.
[15] J. Birkinshaw, G. Hamel, and M. J. Mol,
“Management Innovation,” Acad. Manag. Rev.,
vol. 33, no. 4, pp. 825–845, Oct. 2008.
[16] A. H. Gold, A. Malhotra, and A. H. Segars,
“Knowledge Management: An Organizational
Capabilities Perspective,” J. Manag. Inf. Syst.,
vol. 18, no. 1, pp. 185–214, 2001.
[17] W. M. Cohen and D. A. Levinthal, “Absorptive
Capacity: A New Perspective on Learning and
Innovation.,” Adm. Sci. Q., vol. 35, no. 1, pp.
128–152, 1990.
[18] B. Levitt and J. G. March, “Organizational
Learning,” Annu. Rev. Sociol., vol. 14, no. 1, pp.
319–338, 1988.
[19] T. J. Allen, “Managing the Flow of
Technology,” MIT Press Cambridge MA, p.
320, 1977.
[20] J. E. Ettlie and E. M. Reza, “Organizational
Integration and Process Innovation,” Acad.
Manag., vol. 35, no. 4, pp. 795–827, 1992.
[21] C. Truss, A. Shantz, E. Soane, K. Alfes, and R.
Delbridge, “Employee engagement,
organisational performance and individual well-
being: exploring the evidence, developing the
theory.,” Int. J. Hum. Resour. Manag., vol. 24,
no. 14, pp. 2657–2669, 2013.
[22] V. H. Hailey, E. Farndale, and C. Truss, “The
HR department’s role in organisational
performance,” Hum. Resour. Manag. J., vol. 15,
no. 3, pp. 49–66, 2005.
[23] J.-M. Hiltrop, “The impact of human resource
management on organisational performance:
Theory and research,” Eur. Manag. J., vol. 14,
no. 6, pp. 628–637, 1996.
[24] R. E. Miles, C. C. Snow, A. D. Meyer, and H. J.
Coleman, “Organizational strategy, structure,
and process,” Acad. Manag. Rev., vol. 3, no. 3,
pp. 546–562, 1978.
[25] B.-Å. Lundvall, National Systems of
Innovation: Towards a Theory of Innovation
and Interactive learning. London, 1992.
[26] M. E. Porter, “The Competitive Advantage of
Nations. (cover story),” Harv. Bus. Rev., vol. 68,
no. 2, pp. 73–93, 1990.
[27] W. Vandekerckhove and N. a. Dentchev, “A
Network Perspective on Stakeholder
Management: Facilitating Entrepreneurs in the
Discovery of Opportunities,” J. Bus. Ethics, vol.
60, no. 3, pp. 221–232, Sep. 2005.
[28] H. Choi, S.-H. Kim, and J. Lee, “Role of
network structure and network effects in
diffusion of innovations,” Ind. Mark. Manag.,
vol. 39, no. 1, pp. 170–177, 2010.
[29] F. W. Geels, “Ontologies, socio-technical
transitions (to sustainability), and the multi-
37
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
level perspective,” Res. Policy, vol. 39, no. 4,
pp. 495–510, 2010.
[30] M. Steiner, Clusters and regional
specialisation : on geography, technology and
networks, no. 8. 1998.
[31] N. Gilbert, A. Pyka, and P. Ahrweiler,
“Innovation Networks - A Simulation
Approach,” J. Artif. Soc. Soc. Simul., vol. 4, no.
3, pp. 1–14, 2001.
[32] P. Ahrweiler, A. Pyka, and N. Gilbert, “A New
Model for University-Industry Links in
Knowledge-Based Economies,” Soc. Sci., pp.
218–235, 2011.
[33] P. Ahrweiler, M. Schilperoord, A. Pyka, and N.
Gilbert, “Modelling research policy: Ex-ante
evaluation of complex policy instruments,”
Jasss, vol. 18, no. 4, 2015.
[34] R. R. Nelson and S. G. Winter, An evolutionary
theory of economic change. 1982.
[35] F. Malerba, R. Nelson, L. Orsenigo, and S.
Winter, “History-friendly models: An overview
of the case of the computer industry,” JASSS,
vol. 4, no. 3, 2001.
[36] V. N. Kolokoltsov, Nonlinear Markov
processes and kinetic equations. Cambridge
Tracts in Mathematics, 2010.
[37] R. Y. Rubinstein and D. P. Kroese, “Simulation
and the Monte Carlo Method,” Wiley, p. 377,
2008.
[38] S. A. Kauffman and E. D. Weinberger, “The NK
model of rugged fitness landscapes and its
ISBN: 978-1-941968-45-1 ©2017 SDIWC
application to maturation of the immune
response,” J. Theor. Biol., vol. 141, no. 2, pp.
211–245, 1989.
[39] A. L. Barabási, H. Jeong, Z. Néda, E. Ravasz,
A. Schubert, and T. Vicsek, “Evolution of the
social network of scientific collaborations,”
Phys. A Stat. Mech. its Appl., vol. 311, no. 3–4,
pp. 590–614, 2002.
[40] Helmholtz Gemeinschaft, “BMBF-
Förderprojekt Enabling Innovation –Erprobung
des Management-Tools,” 2014.
[41] D. Scerri, S. Hickmott, A. Drogoul, and L.
Padgham, “An Architecture for Modular
Distributed Simulation with Agent-Based
Models,” Proc. 9th Int. Conf. Auton. Agents
Multiagent Syst. (AAMAS 2010), pp. 541–548,
2010.
[42] B. K. Petersen, G. E. P. Ropella, and C. A. Hunt,
“Toward modular biological models: defining
analog modules based on referent physiological
mechanisms.,” BMC Syst. Biol., vol. 8, p. 95,
2014.
[43] G. Sunwoo Park, E. P. Ropella, and C. A. Hu,
“PISL: A Large-Scale In Silico Experimental
Framework for Agent-Directed Physiological
Models,” 2005.
[44] S. H. Chen and B. T. Chie, “A functional
modularity approach to agent-based modeling
of the evolution of technology,” in Lecture
Notes in Economics and Mathematical Systems,
2006, vol. 567, pp. 165–178.
38
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

Text Classification Using Time Windows Applied to Stock Exchange

Pavel Netolický, Jonáš Petrovský, František Dařena, Jan Žižka

Department of Informatics, Faculty of Business and Economics, Mendel University in Brno,

Zemědělská 1, 613 00 Brno, Czech Republic

pavel.netolicky@mendelu.cz, jonas.petrovsky@mendelu.cz, frantisek.darena@mendelu.cz,

jan.zizka@mendelu.cz

ABSTRACT different application areas (computer

Each day, a lot of text data is generated. This data
comes from various sources and may contain
valuable information. In this article, we use text
classification to discover if there is a connection
between textual documents (specifically Facebook
posts) and changes of the S&P 500 stock index.
The index values and documents were divided
into time windows according to the direction of
the index value changes. In the first experiment,
we used a batch processing approach to put the
documents from all windows into one data set and
a classification accuracy of 62% was achieved. In
the second experiment, we used a data stream
approach to divide documents into twelve data
sets created from two neighboring windows and
we achieved an accuracy of 68%. This indicates
that posts, which companies write on their
Facebook pages, are partially related to the
performance of the stock index. Taking the
concept change into account also enables better
quantification of this relationship.
KEYWORDS
Machine Learning, Classification, Text Mining,
Stock Exchange, Time Windows, Data Streams
1 INTRODUCTION
A huge amount of data is constantly being
generated by people and organizations. The
speed of data creation is rapidly growing and
we use the term “data stream” for the constant
flow of new data [1].
Data streams may be of various data types
(text, image, numeric) and come from
networks monitoring, scientific experiments,
internet search, social networks etc.). In
comparison to batch processing (for which we
have all data available at once), data streams
processing needs a different approach,
because classical approaches are not effective
or even feasible [2].
In this article, we will focus on the connection
between text documents published on the
Internet and movements of stock prices
(represented by a composite value of stock
index). Some research in this area uses
structured (quantitative) data to analyze the
impact of data on stock prices [3].
Unstructured data (like text) may provide us
with another complementary information with
additional hard-to-quantify knowledge [4].
Behavioral finance theory says that emotions
may deeply influence behavior and decision
making of individuals as well as whole
human societies [5]. This means that the
prices on capital markets are (more or less)
influenced by emotions, moods and opinions
of market participants [6]. These attributes are
often contained in text documents and
therefore we decided to use text data for our
research.
[7] examined the connection between the
content of messages posted to a discussion
board and movements of the Czech stock
index. We will expand this approach further
by focusing on the US stock market, using a
larger number and another type (Facebook

ISBN: 978-1-941968-45-1 ©2017 SDIWC 39

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
posts) of text data and treating stock prices
and related text documents as data streams
divided into time windows as we suppose that
the reasons of stock price changes evolve in
time.
2 CURRENTLY USED METHODS
To model the behavior of a stock price with a
relation to the content of text data we can use
classification in a way that we examine the
direction of the change of the stock price to
create classes. This approach was used for
example by [6]. The problem can be seen as
text classification – given a text, decide its
class (direction of the price movement).
However, we must overcome two problems.
The first problem is the definition of classes.
[8] used a threshold value of 1% price change
for the class determination. The second
problem lies in choosing correct features.
Many studies used just single words and this
simple unigram bag-of-words model provided
good results in [8].
There exist a wide range of supervised
learning algorithm that can be uses for the
text classification. An interesting approach is
described in [9] – it focuses on sentence-level
sentiment analysis of movie reviews. They
used the cosine normalization, Term
Presence, and Smoothed delta IDF as
weighting schemes and the Recursive Neutral
Tensor Network algorithm to achieve an
accuracy of 87.60%. [10] used Naïve Bayes
and SVM as algorithms and unigrams,
bigrams, unigrams with bigrams, and
unigrams with POS (Parts-of-speech) as
features. The bigrams showed a lower
accuracy then unigrams – the reason is that
the resulting vectors were very sparse. All in
all, the type of features used in the bag-of-
words model has a little (maximal 2–3%)
impact on the accuracy.
3 DATA AND METHODOLOGY
The goal of the work was to examine whether
the content of text documents published on
the Internet has any connection with stock
price movements. We decided to use
ISBN: 978-1-941968-45-1 ©2017 SDIWC
Facebook posts from company pages as the
text data, because it has been a very rarely
used data source for this area of research, we
have lots of available data, and it might bring
new interesting insights.
3.1 Stock prices
In our research, the values of the S&P 500
Index were used to represent stock prices. The
index values reflect stock prices of the
selected blue chip (large and famous)
companies on the US stock market. The
historical values of the index were
downloaded from the website investing.com.
For each trading day, we have a closing (end-
of-day) numeric value of the S&P 500 Index
available.
3.2 Text data
As the text data, posts from Facebook pages
of the companies from the S&P 500 Index
were used. In total, we examined 431
company pages. The company’s Facebook
page contains a sequence of documents
arranged according to their publication time.
These short postings are created by the
company representatives. Figure 1 shows an
example of a post on the Intel’s page. A post
may be commented by Facebook users.
However, the comments were not used in the
analysis.
In total, 138,713 Facebook posts published
between 1. 1. 2015 and 15. 10. 2016 were
used.
Figure 1. Example of a Facebook post
40
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
3.3 Classification methodology
We used text classification to predict whether
the given document is connected with an
upward or downward movement of the S&P
500 Index.
We examined the time series of the S&P 500
Index values between 1. 1. 2015 and 15. 10.
2016 and found time intervals (windows) in
which the change (either positive or negative)
of the index value between the first and the
last day of the interval was at least 5%. In
total, 24 such windows were found. In 12 of
them, the index value grew and in 12 it
declined. The length (a number of days) of the
time windows varied between 4 and 30. Then,
each document was, based on the time
window in which it was published, assigned
a class: 1 (up) for the positive index value
change, 2 (down) for the negative one. Figure
2 shows an example of time windows
between 1. 1. 2016 and 1. 4. 2016 with the
assigned classes.
Figure 2. Classification classes identified in the time
series of the stock index values
We decided to perform two types of
experiments with different data sets used for
the classification. In the first experiment,
documents from all 24 windows were put into
one data set. The results for this experiment
are presented in section “Batch approach”.
In the second experiment, we divided the
documents into 12 data sets. Each data set
consisted of the documents from two
neighboring windows: one with an upward
movement and one with a downward
movement. The windows represented two
classes for the classification. The results for
ISBN: 978-1-941968-45-1 ©2017 SDIWC
this experiment are presented in section
“Neighboring windows approach”.
Text pre-processing and conversion
The raw text of each document was processed
by a Python script as follows:
1. Remove all whitespace.
2. Lowercase all letters.
3. Tokenize the document – get words (using
TreebankWordTokenizer).
4. Filter words – minimal length of three
letters, exclude numbers.
The edited text was converted into
a structured format by using a Python library
called scikit-learn and its Vectorizer class.
Only words that occurred at least 5 times in
the whole document collection were included
in the resulting vector representation.
The documents were converted to a bag-of-
words representation using three different
weighting schemes for the term-document
matrix [11, p. 21–26]:
• Term Presence (TP): 1 if a term was present
in a document, 0 if not.
• Term Frequency (TF): number of times
a term was present in a document.
• TF-IDF: TF (local weight) multiplied by the
IDF (global weight).
Classification
The converted data was split into the training
(60%) and testing (40%) set. Each bag-of-
words representation was processed by 10
classifiers (with default settings – no
parameter optimization was made) in scikit-
learn. The classifier’s performance was
evaluated by the achieved accuracy
(proportion of the correctly classified
instances on all examined instances [10, p.
268]) on the test set.
4 RESULTS AND DISCUSSION
One set of the text data (Facebook posts)
together with the S&P 500 Index values was
used to prepare the data for classification. The
class-labelled data set was processed using
the three weighting schemes (TP, TF, TF-
41
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

IDF) by 10 classification algorithms. In total, Table 4. The comparison of average accuracies

30 classification results were obtained. achieved by each classifier.
Average
Classifier
accuracy
4.1 Batch approach
ExtraTreesClassifier 0.613
RandomForestClassifier 0.609
Tables 1, 2, 3 and 4 show the results for
MultinomialNB 0.602
classification of the data set including all
windows. LogisticRegression 0.601
BernoulliNB 0.598
Table 1. Facebook posts – data statistics. LinearSVC 0.597
Total Class 1 Class 2 Number of MLPClassifier 0.596
samples samples samples words DecisionTreeClassifier 0.564
138,713 85,286 53,427 33,397 NearestCentroid 0.533

4.2 Neighboring windows approach

Table 2. Facebook posts – classification results.
Accuracy Precision Recall F1 score Tables 5, 6, 7, and 8 show the results
achieved for the 12 data sets consisting of two
0.623 0.605 0.623 0.614
neighboring windows.
Table 1 shows the statistics about the data Table 5. Facebook posts – neighboring windows:
used for the classification. It is obvious that data statistics.
the data set was quite unbalanced (with more Data Data set Number
Class 1 Class 2 Total
documents marked with index value going set
samples samples samples
balance of
up). Table 2 shows the best classification no. ratio words
results. The highest accuracy (62%) was 1 477 520 997 0.917 691
achieved with the TF-IDF weighting scheme 2 1,156 719 1,875 1.608 1,523
and the Multinomial Naïve Bayes 3 4,293 953 5,246 4.505 3,627
classification algorithm. 4 1,187 2,008 3,195 0.591 2,350
5 8,971 2,207 11,178 4.065 6,523
Table 3 tells us that the used weighting 6 3,052 4,802 7,854 0.636 4,995
scheme was not very important. However, we 7 3,454 3,998 7,452 0.864 4,684
can see that the highest average accuracy was 8 8,399 15,292 23,691 0.549 10,441
achieved by TF-IDF. 9 5,198 2,211 7,409 2.351 4,646
10 1,918 2,828 4,746 0.678 3,191
Table 3. The comparison of average accuracies 11 26,111 8,139 34,250 3.208 13,269
achieved for each weighting scheme. 21,070 9,750 30,820 2.161 12,116
12
Average accuracy from
Weighting scheme
all experiments
Table 5 shows the statistics about the data
TF-IDF 0.598
used for the classification. Because the length
TP 0.587
of the windows was variable, the numbers of
TF 0.586
documents greatly vary. It is also visible that
most of the data sets are imbalanced. This
Table 4 shows for each classifier the average should be taken into account when evaluating
accuracy from all experiments. We can see the results.
that the decision tree classifiers
“ExtraTreesClassifier” and
“RandomForestClassifier“ performed the best
with the accuracy around 61%.

ISBN: 978-1-941968-45-1 ©2017 SDIWC 42

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

Table 6. Facebook posts – neighboring windows: Table 8. The comparison of average accuracies
classification results. achieved by different classifiers applied to the
Data set neighboring windows of the Facebook posts.
Accuracy Precision Recall F1 score
no.
Data set no. Classifier Avg. accuracy
1 0.584 0.602 0.584 0.593
2 0.615 0.598 0.615 0.606 1 NearestCentroid 0.587
3 0.808 0.653 0.808 0.722 2 LogisticRegressionCV 0.578
4 0.639 0.733 0.639 0.683 3 LogisticRegression 0.748
5 0.803 0.807 0.803 0.805 4 LogisticRegressionCV 0.633
6 0.646 0.653 0.646 0.650 5 LogisticRegression 0.784
7 0.554 0.553 0.554 0.553 6 MultinomialNB 0.630
8 0.674 0.669 0.674 0.672 7 ExtraTreesClassifier 0.553
9 0.721 0.727 0.721 0.724 8 SGDClassifier 0.656
10 0.618 0.614 0.618 0.616 9 MultinomialNB 0.701
11 0.800 0.782 0.800 0.791 10 LogisticRegressionCV 0.595
12 0.698 0.702 0.698 0.700 11 ExtraTreesClassifier 0.788
Average 0.680 0.674 0.680 0.676 12 SGDClassifier 0.673

According to Table 6, the average accuracy Table 8 shows the classifier that achieved the
(as well as the F1 score) was 68%. The best highest accuracy for each data set. We can see
accuracy (as well as F1 score) was achieved that most of the times the Logistic Regression
for data sets 3 (72%), 5 (80%), and 11 (79%). (5 times) achieved the best result. Among the
The reason for this might be that they have a other classifiers, the Multinomial Naïve
balance ratio around 4 (with more documents Bayes classifier, Extra Trees Classifier, and
marked with index value going up). Stochastic Gradient Descent (SGD) Classifier
were the most successful twice and the
Table 7. The comparison of average accuracies Nearest Centroid was the best only once.
achieved with different weighting schemes applied
to the neighboring windows of the Facebook posts. 5 CONCLUSION
Data set no. TP TF TF-IDF
The goal of the work was to examine whether
1 0.539 0.534 0.534
the content of text documents published on
2 0.555 0.556 0.570
the Internet (specifically Facebook posts) has
3 0.744 0.753 0.769
any connection with stock price movements.
4 0.626 0.625 0.639
We used the values of the S&P 500 Index and
5 0.735 0.736 0.765
divided them into 24 time windows with
6 0.610 0.609 0.620
either growing or decreasing index value
7 0.534 0.534 0.536
trend. Subsequently, we examined (using the
8 0.633 0.629 0.646 classification accuracy) the connection
9 0.663 0.661 0.675 between the documents’ content and the trend
10 0.568 0.562 0.578 of the index value in the time window in
11 0.754 0.747 0.766 which was the document published.
12 0.637 0.644 0.658
Average 0.633 0.633 0.646 Two types of experiments were performed. In
the first one, the documents from all 24
From Table 7 can be seen that the highest windows were put into one data set and we
average accuracy provided the TF-IDF achieved an accuracy of 62%. The second
weighting scheme (+1% in comparison to TP experiment, in which we divided the
and TF). documents into 12 data sets formed from two
neighboring windows, provided better results
– the average accuracy was 68%. Moreover,

ISBN: 978-1-941968-45-1 ©2017 SDIWC 43

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
for three data sets the accuracy was even
higher – 72%, 79% and 80%. This means that
classifying data from the neighboring
windows brings on average better results than
using only one data set. This might be related
to the concept drift [12] phenomenon which
requires a further investigation for this
specific domain.
The achieved accuracy around 70% tells us
that the posts which companies write on their
Facebook pages are partially related to
the performance of the whole stock index.
It must be noted that we did not optimize the
parameters of used classification algorithms.
By doing this, we might achieve a slightly
higher accuracy.
This area could be further researched in
various directions. Firstly, the analysis may
be performed on more types of documents
(e.g., newspaper articles). Secondly, the class
assigning method may be enriched by using
various thresholds of the index value changes
(not only 5%). Thirdly, it might be interesting
to examine not the whole stock index, but the
stock prices of the individual companies
instead.
ACKNOWLEDGEMENT
This research was supported by the Czech
Science Foundation [grant No. 16-26353S
”Sentiment and its Impact on Stock Markets”]
and Internal Grant Agency of Mendel
University [No. PEF_DP_2017001
“Searching for semantic information and
gaining knowledge from text data streams
with new machine learning methods”] and
Internal Grant Agency of Mendel University
[No. PEF_DP_2017022 “Acquiring, filtering
and analyzing of texts for stock markets”].
ISBN: 978-1-941968-45-1 ©2017 SDIWC
REFERENCES
[1] Aggarwal, C. C. Data Streams: Models and
Algorithms. 2007. Springer
[2] Gama, J. Knowledge discovery from data
streams. CRC Press, 2010.
[3] Petrovský, J., Netolický, P. and Dařena, F.
Examining Stock Price Movements on Prague
Stock Exchange Using Text Classification.
International Journal of New Computer
Architectures and their Applications (IJNCAA).
Vol. 7 No. 1. (2017). pp. 8-13. ISSN 2412-3587.
[4] Sven S. Groth, Jan Muntermann. An intraday
market risk management approach based on
textual analysis. Decision Support Systems .
Volume 50. Issue 4. March 2011. Pages 680-691
[5] Colm Kearney, Sha Liu. Textual sentiment in
finance: A survey of methods and models.
International Review of Financial Analysis.
Volume 33. May 2014. Pages 171–185.
[6] Bollen, J., Mao, H. and Zeng, X. Twitter mood
predicts the stock market. Journal of
Computational Science. 2011. vol. 2. no. 1. p. 1–
8.
[7] Kaplanski, G. and Levy, H. Sentiment and stock
prices: The case of aviation disasters. Journal of
Financial Economics. 2010. vol. 95. no. 2. p.
174–201.
[8] Lee, H., Surdeanu, M., MacCartney, B. and
Jurafsky, D. On the Importance of Text Analysis
for Stock Price Prediction. In: LREC. 2014. p.
1170-1175
[9] Maas, A. L., Daly, R. E., Pham, P. T., Huang, D.
Learning word vectors for sentiment analysis. In:
Proceedings of the 49th Annual Meeting of the
Association for Computational Linguistics:
Human Language Technologies. 2011. vol. 1. p.
142–150.
[10] Go, A., Bhayani, R. and Huang, L. Twitter
sentiment classification using distant
supervision. CS224N Project Report. Stanford.
2009. vol. 1. p. 12.
[11] Weiss, S. M., Indurkhya, N. and Zhang, T.
Fundamentals of Predictive Text Mining.
London: Springer. 2010. ISBN 978-1-84996-
225-4.
[12] Lindstrom, P, Delany, S. J., Mac Namee, B.
(2010) Handling Concept Drift in Text Data
Stream Constrained by High Labelling Cost.
Florida Artificial Intelligence Research Society
Conference (FLAIRS). Florida, 19-21, May.
44
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

Towards Specification Formalisms for Data Warehousing Requirements

Elicitation Techniques
Isaac N. Mbala (1) and John A. van der Poll (2)
(1)
School of Computing, Florida, University of South Africa
(2)
Graduate School of Business Leadership (SBL), Midrand, University of South Africa
(1)
isaacnkongolo64@gmail.com, (2) vdpolja@unisa.ac.za

phase of requirements analysis for DW

ABSTRACT
Various studies have demonstrated that, often
data warehouse projects miss to meet with the
business objectives and business requirements of
the target company. Many of these projects fail to
recognize the importance of comprehensive
requirements elicitation and subsequent
specification phases. The requirements executed
during these phases of definition are incomplete
or incoherent, leading to incorrect specifications.
In this paper we present the challenge of data
warehousing projects failure and propose a
methodology of requirements engineering for the
specification of data warehousing requirements.
Our approach leans towards a hybrid-driven
technique for requirements analysis for data
warehouse systems. Formalisms for requirements
elicitation are proposed, followed by formal
specification in Z.
KEYWORDS
Data warehouse design, Requirements
engineering, Formal methods, Formal
specification, Z.
1 INTRODUCTION
Nowadays, data warehousing (DWH) is the
technology more used in large industries [1],
[2]. The usage of data warehousing involves
the creation of a data warehouse (DW) made
up of data marts (DMs) or operational
databases with business intelligence (BI)
embedded in the resultant structure. The goal
of this paper is upon a methodology of
requirements engineering (RE) for the
specification of requirements used during the
systems design. Numerous authors have
researched requirements analysis for DW
systems, notably [1], [2], [3], [4], [5], [6], [7],
[8], [9], [10], and [11] to name but a few, and
rather few research works have attempted to
address the requirements engineering aspect.
[5], [7] and [12] have defined a data
warehouse as “a subject oriented, integrated,
time-variant, and nonvolatile collection of
data in support of management’s decisions”.
A data warehouse is recognized as one of the
most complex information systems and its
maintenance and design are described by
numerous complexity coefficients [2], [5],
[7]. A data warehouse is thought of as the
cornerstone of business intelligence systems
[12], [13].
As observed by [2], the objective of the data
warehouse is to yield concise analysis in
order to assist decision makers and also to
increase corporeal organizational
performance. To build a conventional
operational system, it requires to take in
consideration not only demands on how to
automatically perform operations of the
company, but to also develop a DW system,
the analytical requirements sustaining the
process of decision-making need to be
captured [11]. The data warehouse
development and maintenance is a complex
and tedious task which requires solving many
different problem types [14]. The design of
DW systems is instead different from the
design of traditional operational systems that

ISBN: 978-1-941968-45-1 ©2017 SDIWC 45

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
provide data to the warehouse since the
objective of data warehouse projects is
fundamentally relied on the support of the
process of decision making of the enterprise
in order to facilitate the analysis processes
[2], [5], [8], and [11].
It is well documented that a prominent reason
why many DW projects have failed in the
past is not only because they attempted to
supply strategic information from operational
systems while those operational systems
were not intended to provide strategic
information [5], but also because the
requirements analysis phase were often
overlooked [1], [2], and [3] during the design
process and because of these reasons, [2] and
[14] have declared that over 80% of DW
projects miss to meet with the users and
stakeholders’ requirements. The analysis
phase of requirements can be executed
informally based on simple requirements
glossaries instead of formal diagrams but
such an informal (or maybe semi-formal)
approach may be inappropriate for a
requirements-driven framework that requires
more organized and comprehensible
techniques [3].
Data warehouse projects are similar in
several phases to any software development
project and claims a definition of different
activities which ought to be executed related
to demands collection, design and
implementation within an operational
platform, amongst other activities [1], [14].
Despite the similarity to general software
development, the effective development of a
DW relies upon the quality of its models
(design and specification) [15]. However, the
system success under the development may
be strongly affected by the discovering
process of involved stakeholders demands
and sustaining those demands while
ISBN: 978-1-941968-45-1 ©2017 SDIWC
transforming and documenting them under a
form which will be analyzable and
communicable. This process is generally
known as Requirements Engineering (RE)
[11], which is a necessary and vital phase in
the software development life cycle (SDLC)
[16].
According to [16] and [17], RE is the process
which is intended to collect, document,
analyze and manage requirements for
systems and software product throughout the
SDLC. According to [5], RE in the data
warehouse arena has acquired increased
importance and it has the goal of identifying
the information demands of the decision
makers. However, researchers are actually
attempting to utilize numerous techniques of
requirements engineering to analyze the
specification of data warehouse systems in
order to avoid the risk of failure. Several
techniques and methods are used in the
requirements engineering activities and in
this paper we are more interested on formal
methods [16], [18], and [19] for analyzing
system behavior, factors of risk and problems
related to its implementation [20] during the
design of the system.
The use of Formal Methods (FMs) in the
construction of reliable software has been
controversial for a number of decades.
Advocates of such techniques point to the
advantages to be gained in constructing
provably correct systems, especially in the
arena of mission/safety-critical systems, e.g.
nuclear power plants and aviation systems.
Critics of FMs object to the steep learning
curve involved in mastering the underlying
discrete mathematics and formal logic
needed for the effective use of the
methodology. Yet, the literature suggest that
using FMs in the design of data warehouse
systems ought to be useful in improving the
46
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

reliability of such systems and other directions for future work are presented in
functionality [19]. Section 6.
Formal methods are mathematical 1.1 Research questions
approaches sustained by tools and techniques
In this paper we aim to find answers to the
for the verification of the desired and
following questions:
necessary properties of software or hardware
systems. FMs are necessary for the control of RQ1: What are the requirements elicitation
the quality parameters such as completeness; approaches for data warehouse development?
correctness; and consistency and verification
of requirements of a system [13] and they are RQ2: To what extent may formal
based on (often discrete) mathematical specification facilitate data warehousing?
notations and logic to clearly and accurately
RQ3: How may the two (2) prominent
express requirements specification [21].
elicitation techniques be combined?
In the research work published by [16],
Formal methods are more likely to be used at 2 DATA WAREHOUSE
the levels of design and verification of the SYSTEMS DESIGN
software development. As observed by [13],
Building a DW system is unlike transactional
formal methods are attached with the three
systems with respect to the development; the
techniques which are formal specification,
structures are not only ones to be thought of
formal checking (discharging proof
as in those kind of source systems, but
obligations), and refinements. Formal
cognizance should also be given about the
specifications aim to provide for an
purposes and strategies of the organization
unambiguous and coherent complement to
[8]. Data warehouse systems have the
natural language descriptions [21], [22] and
purpose of supporting the process of decision
are rigorously validated and verified
making of an enterprise. The development of
conducting to the early detection of
a DW requires that the analytical
specification errors [22]. One of the broadly
requirements supporting the decision making
used formal specification languages amongst
process be captured and such requirements
many different formal specification
are usually not easy to extract and specify
languages is the Z, selected in this paper
[11]. A DW is generally defined as the
owing to its simplicity and widely-usedness
linking of a number of operational databases
in the formal methods arena [22].
with the aforementioned intelligence (e.g.
This paper is structured as follows. Following decision-making) added to the resultant
our research questions below, we introduce in structure. Since a DM is viewed as a subset
Section 2 the fundamental concepts of data of a DW, we view a data mart as being one of
warehouse systems design by discussing the operational databases in the DW.
various design approaches. Section 3
Subsequently, we formalize a DW as follows:
presents a Z specification of a data warehouse
star schema and Section 4 addresses related 𝑛
work in this area. In Section 5 we address our 𝐿𝑖𝑛𝑘 𝐷𝐵𝑖 , where
methodology and finally, conclusions and i=1

ISBN: 978-1-941968-45-1 ©2017 SDIWC 47

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
(∀i)(∀j) (1 ≤ i, j ≤ n ⦁ i ≠ j ⇒ 𝐷𝐵𝑖 ∩ 𝐷𝐵𝑗 =
∅)
The above definition assumes that different
databases, when correctly normalized do not
contain common elements, except of course,
for foreign key matches.
The design of data warehouse systems is
unlike from the design of transactional
systems that provides data to the warehouse
[8]. There are two well-known authors in the
world of data warehousing; Bill Inmon and
Ralph Kimball, advocating complementary,
yet different techniques to the design of data
warehouses. The technique applied by Bill
Inmon is the familiar top-down design which
begins with the Extraction-Transformation-
Loading (ETL) process working from
external data sources in order to build a data
warehouse, whilst Ralph Kimball applies the
equally well-established bottom-up
technique which begins with an ETL process
for one or more data marts separately. Most
of proponents of data warehouse design
subscribe to either of the two techniques [12].
2.1 Data warehouse systems design
Approaches
The DW systems design is based on two
approaches that are alternative and inverses
of each other viz: the Data-driven approach
also known as the Supply-driven approach
and the Requirement-driven approach also
known as the Demand-driven approach [1],
[3], [7], and [10]. The process of
development of a DW starts with the
identification and collection of requirements.
The design of the multidimensional model is
next, followed then by testing and
maintenance [9]. To develop a data
warehouse requires a set of steps to
accomplish throughout the process, namely,
the requirements analysis phase, conceptual
ISBN: 978-1-941968-45-1 ©2017 SDIWC
phase, logical phase and physical phase [6],
[8]. According to [9], the design stage is the
most significant operation in the successful
construction of a DW.
The following sections elucidate the context
of requirements analysis and conceptual
design which are considered as the two main
phases within the data warehouse systems
design process [8].
2.1.1. Requirements Analysis
Requirements analysis has as its aim
detecting which knowledge is useful for
decision making by investigating the user’s
demands and expectations in user-driven and
goal-driven approaches, or by verifying the
validity of operational data sources in a data-
driven approach [8]. Requirements analysis
of users plays a crucial role in data warehouse
systems design. It has a major influence upon
the taking of decisions throughout the data
warehouse systems implementation [2], [23].
The requirements analysis phase leads the
designer to unveil the multidimensional
schema necessary elements (facts, measures
and dimensions) which are claimed to assist
future data manipulations and calculations.
The multidimensional schema has a
significant impact on the success of DW
projects [2], [3], and [14].
Several research works have published on the
various approaches used during the
requirements analysis phase of DW systems
design, leaning on the two techniques
mentioned above – the Top-down technique
and Bottom-up technique. Implementations
of these are: Data-driven approach, Goal-
driven approach, User-driven approach, and
Mixed-driven approach [1], [2], [3], [4], [5],
[6], [7], [8], [9], [10], and [11]:
48
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
 The Data-driven approach also known as
the supply-driven approach. It utilizes
the bottom-up technique and yields
subject-oriented business data schemas
by only leaning on the operational data
sources and ignoring business goals and
stakeholder needs.
 The Goal-driven approach applies the
top-down technique rather than the
bottom-up technique. It allows for the
generation of information like Key
Performance Indicators (KPIs) of
principal business areas by relying on
business objectives and granted business
processes only, and essentially overlooks
data sources and user demands.
 The User-driven approach is similar to
the goal-driven approach, and it applies
the top-down technique. It permits to
produce analytical requirements
interpreted by the dimensions and
measures of each subject by ignoring
business goals and data sources.
A user-driven approach starts with a detailed
agreement of the needs and expectations of
the users and this brings about numerous
advantages like increasing of the production,
enhancing of the work quality, support and
training costs reductions and improvement of
general user satisfaction [24]. User
requirements analysis does not prescribe to a
standard approach which designers may rely
on for designing their data warehousing
projects [16]. As declared by [6] the data-
driven approach yields a conceptual schema
through a re-engineering process of the data
sources by ignoring the end users and
stakeholders’ contribution, whilst the
requirements-driven approach aims at
generating the conceptual schema by only
ISBN: 978-1-941968-45-1 ©2017 SDIWC
considering the needs expressed by end users
and stakeholders. The correct elicitation of
user requirements remains a fine challenge
and many techniques, e.g. the use of JAD
(Joint Application Design) sessions [25] have
been put forward.
These three primary aforementioned
approaches have their merits and demerits.
However, in an attempt at overcoming this
problem, numerous authors suggested the
mixed-driven approach which consists of
combining two or even all three the primary
approaches (either user-driven and data-
driven approaches or goal-driven and data-
driven approaches or a combination of user-
driven, goal-driven and data-driven
approaches) as detailed in [2], all aimed at
getting a “best result” that will meet users and
stakeholders’ demands and expectations.
According to [14] and [26], the requirements-
driven approach is also called the analysis-
driven approach; the supply-driven approach
is also called the source-driven approach and
the requirement/supply-driven approach is
known as an analysis/source-driven
approach, but is also known by the name of a
hybrid-driven approach.
The various approached discussed above are
elicited in Figure 1 below.
Top-down
Requirements- DW
driven Source-driven
User-driven
Supply-driven
Demand-
Data-driven
driven
Goal-driven
Analysis-driven Bottom-up
DW
Figure 1: Complementary top-down & bottom-up
49
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

Table 1 below presents the advantages and Figure 2 below depicts the analysis-driven
disadvantages of approaches grouped by approach framework with all the steps
technique. considered:

Table 1: Advantages and disadvantages of techniques. Identify users

Technique Approach Advantages Disadvantages Determine analysis demands

Top- User- A coherent It’s not
data flexible to the Define, refine and prioritize goals
down driven
dimensional requirements
Goal- views change during
through data the
driven marts is implement-
provided ation. Model business processes
Demand- from the Detail
Determine
driven DW. user needs processes Specify
It’s easy It’s highly for services
Analysis
-driven from a data expose to the accomplish or
warehouse to risk of failure. ment of activities
Require reproduce a goals
data mart.
ments-
driven
Bottom- Data- It’s less The data view
up driven expose to the for each data Document requirements specification
risk of mart is
Source- failure. narrowed.
driven Figure 2: Requirement-driven approach framework [25]
It facilitates The
the return on penetration of
Supply- investment redundant
driven and leads to data within 2.1.1.2. Supply-driven Approach
concrete each data
results in mart. In the supply-driven approach, the
short time. development of the conceptual schema leans
on the data available in the operational
The above discussion presents an answer to systems. The objective of this approach is to
our RQ1 above. identify multidimensional schemas which
may be conveniently implemented over the
2.1.1.1. Requirements-driven Approach legacy operational databases (data marts –
see above). An exhaustive analysis is made
The conceptual schema development within
over these databases to elicit the essential
the requirements-driven approach is based on
elements which can depict facts with attached
business- or user requirements. The
measures, dimensions and hierarchies and the
organizational purposes and demands that the
discovering of these elements induced to an
DW systems are expected to present, support
initial DW schema which can correspond to
the process of decision making which include
many different analysis goals [14], [26].
the requirements needed for the conceptual
schema. The information collected serves as Figure 3 below represents the supply-driven
a foundation for the initial data warehouse approach framework with the all considered
schema development [14], [26]. steps.

ISBN: 978-1-941968-45-1 ©2017 SDIWC 50

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

Document before the final DW schema is developed, the

Identify
operation Apply requireme essential step in all the aforementioned
al derivation nts approaches that we concentrate on in this
systems process specificati paper, is the document requirements
on
specification step. This enables the
documentation of all the information
Figure 3: Supply-driven approach framework [25] obtained from the previous step. This step
will contain the business requirements and
2.1.1.3. Hybrid-driven approach business objectives expressed in more detail,
e.g. what operations may be done, who has
The hybrid-driven approach combines the
the right to access to the data, what measures
two abovementioned approaches that can be
and dimensions are represented, etc.
used in parallel in order to obtain a best
design [14], [26]. The operation of 2.1.2. Conceptual design
requirements mapping takes place while fact
Although most of the research on the design
tables, measures and dimensions are
identified during the decisional modeling, of DWs concerns their logical and physical
and are mapped over entities into the source models, the essential foundation to build a
schema [3]. data warehouse on is an accurate conceptual
design that is well-documented and
The hybrid-driven approach framework is thoroughly fulfills the requirements [27]. We
illustrated below in Figure 4.
addressed some works that discussed the
Supply-driven Requirement-driven conceptual design of data warehouse systems
above.
Identify Identify
operational users
In [28] the authors proposed a hybrid
systems
technique which has as objective to
summarize techniques while defining user
Determine requirements for DWs in order to get a
Apply analysis multidimensional conceptual model using an
derivation demands
ensemble of rules Query-View-
process
Transformation (QVT) based on the
multidimensional normal forms for the
Matching
process preciseness. In the paper published by [15] an
approach of the multidimensional star
schema validation supported through
Document requirements specification reparation solutions has been proposed while
assisting designers within the detection of
Figure 4: Hybrid-driven approach framework constraint violations and by suggesting
reparation solutions based on a number of
According to [14] and [26] for all the
mistake-based rules formalized in Prolog.
previously described approaches, showing
the different iterations which can be claimed

ISBN: 978-1-941968-45-1 ©2017 SDIWC 51

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

However, the main purpose of this stage is to
develop a conceptual diagram that will meet
the functional requirements from the
requirements analysis stage (requirement-
driven approach) and the data model
designed from the legacy operational systems
[8] in order to fulfill the users and
stakeholders’ demands and expectations.
This phase would be useful for the
representation of the necessary elements into
the multidimensional schema after the
specification of requirements.
According to [29] a schema is defined by the
relation between facts and dimensions. Fact
is the subject of analysis or the focus of
interest in the process of decision making
[27] and dimensions are different
perspectives used for the analysis of facts.
Fact contains numerical attributes commonly
called measures [29].
Figure 5 below depicts a multidimensional
schema for data warehouse systems,
following the suggestions in [15].
Fact – table Dimensions –
1.. n table 1.. n
Measures Attributes
Figure 5: A multidimensional schema
Next we present a formal specification of the
star-based structure in Figure 5 in Z.
3 A FORMAL SPECIFICATION
consists of measures and additional
attributes, therefore we have as basic types:
[Measure, Attribute]
A Fact table consists of attributes, but given
the structure of the star schema [15] we
distinguish between measures and ordinary
attributes as follows.
Fact_table
measures : ℙ Measure
attributes : ℙ Attribute
measures ∩ attributes = ∅
While measures of fact tables in a star schema
are also attributes, we give them special
status to reflect the structure defined in [15].
A Dimension as per Figure 5 consists of a
number of fact tables.
Dimension
dimension : ℙ Fact_table
As per the Established Strategy for
constructing a Z specification, a specifier has
to define an initial state, and discharge a
proof obligation (PO) arises, namely, show
that such an initial state may be realized (i.e.
it exists).
Subsequently, the initial state of the star state
is given by:
Init_Dimension
Dimension′

As per the Established Strategy for dimension′ = ∅

constructing a Z specification [18], we define
the basic types of the system. Our state The PO that arises is:

⊢ ∃ Dimension′ ⦁ Init_Dimension

ISBN: 978-1-941968-45-1 ©2017 SDIWC 52

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

Therefore, ⊢ ∃ dimension′ | dimension′ = ∅

from which it follows that the initial state of Dim - Cars
the system contains no fact tables. IdCar
Subsequently, measures = ∅ ∧ attributes = ∅. Model
Therefore the initial state Init_Dimension above Type
Brand
can be realized.
The above Z specification provides an
answer to our RQ2 above.
Fact - Renting Dim - Date
Figure 5 suggests that in the IdCar Year
multidimensional schema, we can have idAgency Month
Day Week
several facts link to various dimensions. We
Amount Day
choose to design the DW conceptual schema
with the model of star.

In Figure 6 we show how a multidimensional Dim - Agency

star schema may appear for a specific case. idAgency
Country
Dimension Dimension City
table 1 table 3
(attributes) (attributes)
Figure 7: A multidimensional star schema

Fact table
(Measures) 4 RELATED WORK

Dimension Several works in the literature have

table 2
(attributes)
Figure 6: A multidimensional star schema [12]
Example 1
Consider an example of a car rental company
represented in Figure 7 to illustrate the
generic multidimensional star schema
concepts in Figure 6.
An example where a decision maker is
interested in analyzing the fact Renting in
terms of amount measure is described in [15].
addressed requirements analysis for data
warehouse systems and just few works try to
link requirements engineering to DW
systems [30]. Requirements analysis is
thought of as one of the significant tasks to
facilitate the success of a data warehouse
project [2]. However, we discuss below some
work that examine the requirements analysis
phase in more detail within the data
warehouse systems design arena.
[6] Proposed a hybrid approach based on the
two main approaches used for data
warehouse systems design (requirements-
driven approach and supply-driven approach)
which helps to produce the conceptual

ISBN: 978-1-941968-45-1 ©2017 SDIWC 53

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

schema based on a graph-oriented Various authors, e.g. [14], [28] have

representation, using in turn an automatic discussed different approaches and
operation of data sources reengineering, methodologies used for the DW systems
based on a set of constraints derived from design. Following that, we have chosen to use
requirements of users. the hybrid approach to create our framework
as a goal of this paper.
In [5] the authors introduced various
requirements elicitation methodologies by 5 OUR METHODOLOGY
addressing advantages of each of them and
Our methodology is depicted Figure 8
they investigated the role of requirements
below. It is organized as a sequence of steps
engineering within the DW development and
where each step follows in depth the
made a comparative study of different
application abstraction level, as the
requirements elicitation methodologies (GDI
requirements of the project are collected to
model, AGDI model, Use case approach,
form the requirements basis. Such
DFM, CADWA, Tropos) for the DW
requirements are ultimately formally
development.
specified, using e.g. Z.
[10] Discussed requirements analysis
As is evident from the diagram, we base our
approaches for the design of data warehouse
approach on both the hybrid-driven approach
systems and suggested a requirements
as well as requirements-driven approach,
analysis framework based on business
aimed at assisting designers of data
objects for data warehouse systems to
warehouse systems to design a conceptual
identify the analytical requirements and
schema that will meet the users and
further refined those requirements to map
stakeholders’ needs and expectations.
onto the conceptual level of the model of data
Therefore, we are not concerned with all
warehouse design, using either requirements-
these steps of requirements elicitation for
driven or supply-driven approach for DW
documenting requirements – details may be
requirements analysis. The author declared
observed in [14].
that the multidimensional data model of the
conceptual level is the primary deliverable of The two sets of requirements stem from
the data warehouse requirements analysis. bottom-up and top-down analysis
respectively. Furthermore, for the
In [9] the author proposed an object oriented
requirements analysis, requirements are
framework for DW systems conceptual
collected from dissimilar users and on the
design and used UML (Unified Modeling
supply side from different legacy operational
Language) within the design process of
systems. Finally the requirements are
software system development as this latter
combined as shown and discussed below.
has become a standard for object oriented
modelling during design and analysis phases.
He also made a comparative study of various
data warehouse design approaches and
schemas used, according to different authors.

ISBN: 978-1-941968-45-1 ©2017 SDIWC 54

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

Supply-driven Requirement-driven Algorithm 1: Enhanced SRE – Merge

structured and unstructured data
sets obtained from bottom-up and top-
Identify Identify down requirements elicitation.
operational users
systems Begin
(* Elicit unstructured- &
structured data through
Determine requirements- and supply driven
elicitations. *)
Apply analysis
derivation demands (* Assume m structured data sets *)
process for i := 1 to m step 1 do
Si := Structured data set i of
supply-driven requirements;
Matching (* Assume n unstructured data sets *)
process
for j := 1 to n step 1 do
Uj := Unstructured data set j of
user requirements;
Document requirements specification
(* Amalgamate the two data sets
into two separate sets S and U. ⋃
denotes set-theoretic arbitrary
Formal specification union. *)
m n

Conceptual schema
S := ⋃ Si ; U := ⋃ Uj

Figure 8: Proposed framework

i := 1 j := 1

(*Combine outcome of elicitations

The crucial step in Figure 8 above is the
matching of the two sets of requirements
obtained through the top-down processes and
the bottom-up processes. It is plausible that
these requirements sets may be non-
homogenous and in different formats, e.g.
one may contain structured data (bottom-up
data), while the other set may contain
unstructured data obtained through
incomplete and often inconsistent user
requirements.
To merge the two data sets we enhance an
SRE (Software Requirements Elicitation)
algorithm defined in [25] as follows:
as per Figure 8 into set C.*)
C := S ∩ U
End.
Following the construction of set C
containing both structured and unstructured
data in Algorithm 1, the next step is to
construct a requirements definition [31], [17]
as per Figure 8. From the definition a formal
specification [22] would be constructed in
line with the ideas presented in Section 3.
Algorithm 1 provides an answer to our RQ3
above.

ISBN: 978-1-941968-45-1 ©2017 SDIWC 55

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
6 CONCLUSIONS AND FUTURE
WORK
In this paper we introduced several works
which discuss about approaches used for data
warehouse systems design. Various authors
in the literature review [1], [2], [3], [4], [5],
[6], [7], [8], [9], [10], [11] and [32] have
proposed different techniques and/or
approaches and frameworks to address the
issue of DW systems design. We also
investigated the extent to which the use of
formal methods for data warehouse systems
may alleviate such failures within the design
process.
We proposed a requirements engineering
methodology which represents our proposed
framework which is created from one of the
existing requirements analysis frameworks
for the specification of system requirements
to requirements analysis for the design of
DW systems at the conceptual level. The use
of a formal specification as a last step in the
process is proposed.
We anticipate our proposed framework to
facilitate with removing vagueness or
inconsistencies, at least to some extent. Such
ambiguity is often present in natural language
specifications used to specify data
warehouses.
As for future work, we intend to test our
methodology on a case study using the CZT
toolset, aimed at facilitating the building of a
formal specification.
The combined set C in Algorithm 1 above
contains both structured and unstructured
data, hence the set is non-homogeneous. Z,
however, is a strongly typed language,
containing only “homogeneous” sets.
Therefore specifying the above processes in
Z will require us to devise a mechanism to
ISBN: 978-1-941968-45-1 ©2017 SDIWC
deal with structuredness as well as
unstructuredness in the same set.
REFERENCES
[1] N. Jukic and J. Nicholas. (2010). A
Framework for Collecting and Defining
Requirements for Data Warehousing
Projects. Journal of Computing and
Information Technology, 18(4), pp. 377–384.
https://doi.org/doi:10.2498
[2] N. H. Z. Abai, J. H. Yahaya and A. Deraman.
(2013). User Requirement Analysis in Data
Warehouse Design: A Review. Procedia
Technology, 11, pp. 801–806.
https://doi.org/10.1016/j.protcy.2013.12.261
[3] P. Giorgini, S. Rizzi and M. Garzetti. (2008).
GRAnD: A goal-oriented approach to
requirement analysis in data warehouses.
Decision Support Systems, 45(1), pp. 4–21.
https://doi.org/10.1016/j.dss.2006.12.001
[4] D. T. A. Hoang. (2011). Impact Analysis for
On-Demand Data Warehousing Evolution. In
ADBIS (2) (pp. 280–285). Retrieved from
https://pdfs.semanticscholar.org/ae5c/a847a
8afc046951e34653fcbd3ade06322cb.pdf
[5] S. Mathur, G. Sharma and A. K. Soni. (2012).
Requirement elicitation techniques for data
warehouse review paper. International
Journal of Emerging Technology and
Advanced Engineering, 2(11), pp. 456–459.
Retrieved from
https://www.researchgate.net/profile/Girish_
Sharma4/publication/259827907_IJETAE_1
212_84/links/02e7e52e0dc392211f000000.p
df
[6] F. Di Tria, E. Lefons and F. Tangorra. (2011).
GrHyMM: A graph-oriented hybrid
multidimensional model. In Lecture Notes in
Computer Science (including subseries
Lecture Notes in Artificial Intelligence and
Lecture Notes in Bioinformatics) (Vol. 6999
LNCS, pp. 86–97). Springer.
https://doi.org/10.1007/978-3-642-24574-
9_12
[7] M. Golfarelli. (2010). From User
Requirements to Conceptual Design in Data
Warehouse Design. Data Warehousing
Design and Advanced Engineering …, 15.
https://doi.org/10.4018/978-1-60566-756-
0.ch001
56
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

[8] M. El Mohajir and I. Jellouli. (2014). 6-

Towards a Framework Incorporating KbWAhUrJsAKHateCJQQ6AEIOTAF#v=o
Functional and Non-Functional nepage&q=incomplete%20requirements%2
Requirements for Datawarehouse 0for%20data%20warehouse&f=false
Conceptual Design. IADIS International
Journal on Computer Science and [15] A. Salem and H. Ben-Abdallah (2015). The
Information Systems, 9(1). Retrieved from design of valid multidimensional star
http://citeseerx.ist.psu.edu/viewdoc/downloa schemas assisted by repair solutions. Vietnam
d?doi=10.1.1.640.5590&rep=rep1&type=pd Journal of Computer Science, 2(3), pp. 169–
f 179. https://doi.org/10.1007/s40595-015-
0041-1
[9] R. Jindal. (2012). Comparative Study of Data
Warehouse Design Approaches : A Survey. [16] M. Dos Santos Soares and D. S. Cioquetta.
International Journal of Database (2012). Analysis of techniques for
Management Systems, 4(1), pp. 33–45. documenting user requirements. In
https://doi.org/10.5121/ijdms.2012.4104 International Conference on Computational
Science and Its Applications (pp. 16–28).
[10] A. Sarkar. (2012). Data Warehouse Springer. Retrieved from
Requirements Analysis Framework: http://link.springer.com/chapter/10.1007/978
Business-Object Based Approach. -3-642-31128-4_2
International Journal of Advanced Computer
Science and Applications, 3(1), pp. 25–34. [17] I. Sommerville. (2015). Software
https://doi.org/10.14569/IJACSA.2012.0301 Engineering (10th ed). Pearson.
04 [18] B. Potter, J. Sinclair and D. Till. (1996). An
[11] A. Nasiri, E. Zimányi and R. Wrembel. Introduction to Formal Specification and Z,
(2015). Requirements Engineering for Data 2nd edition, Prentice Hall International
Warehouses. Retrieved from Series.
http://code.ulb.ac.be/dbfiles/NasZimWre201 [19] J. Q. Zhao. (2007). Formal Design of Data
5incollection.pdf Warehouse and OLAP Systems.
[12] T. Oketunji and O. Omodara. (2011). Design [20] S. A. Han and H. Jamshed. (2016). Analysis
of Data Warehouse and Business Intelligence of Formal Methods for Specification of E-
System. Master Thesis. Retrieved from Commerce Applications, 35(1), pp. 19–28.
http://www.diva-
portal.org/smash/record.jsf?pid=diva2:8310 [21] S. H. Bakri, H. Harun, A. Alzoubi and R.
50 Ibrahim. (2013). the Formal Specification for
the Inventory System Using Z Language. The
[13] T. Pandey and S. Srivastava. (2015). 4th International Conference on Cloud
Comparative Analysis of Formal Computing and Informatics, (64), pp. 419–
Specification Languages Z, {VDM} and B. 425.
International Journal of Current
Engineering and Technology, 5(3), pp. 2277– [22] M. Gulati and M. Singh. (2012). Analysis of
4106. Retrieved from Three Formal Methods-Z, B and VDM.
http://inpressco.com/wp- International Journal of Engineering, 1(4),
content/uploads/2015/06/Paper1082086- pp. 1–5. Retrieved from
2091.pdf http://www.ijert.org/browse/june-2012-
edition?download=297:analysis-of-three-
[14] E. Malinowski and E. Zimányi. (2008). formal-methods-z-b-and-vdm&start=120
Advanced data warehouse design: from
conventional to spatial and temporal [23] J. Schiefer, B. List and R. Bruckner. (2002).
applications. Retrieved from A holistic approach for managing
https://books.google.co.za/books?id=XPMV requirements of data warehouse systems, 13.
W3PtGtEC&pg=PA259&lpg=PA259&dq=i Retrieved from
ncomplete+requirements+for+data+warehou http://aisel.aisnet.org/cgi/viewcontent.cgi?ar
se&source=bl&ots=pbxOHI231W&sig=Zuf ticle=1372&context=amcis2002
rX0byf75gvHMUeMOg3-
9z_v4&hl=en&sa=X&ved=0ahUKEwjF9df [24] M. Maguire and N. Bevan. (2002). User

ISBN: 978-1-941968-45-1 ©2017 SDIWC 57

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

requirements analysis. In Usability (pp. 133–

148). Springer. Retrieved from
http://link.springer.com/chapter/10.1007/978
-0-387-35610-5_9
[25] W. F. Friedrich and J.A. van der Poll. (2007).
Towards a Methodology to Elicit Tacit
Domain Knowledge from Users.
Interdisciplinary Journal of Information,
Knowledge and Management (IJIKM),
Volume 2, 2007, pp. 179 - 193. ISSN: Print
1555-1229. URL: www.ijikm.org
[26] E. Zimanyi. (2006). Requirements
Specification and Conceptual Modeling for
Spatial Data Warehouses, 4278(October
2017). https://doi.org/10.1007/11915072
[27] M. Golfarelli, D. Maio and S. Rizzi. (1998).
Conceptual design of data warehouses from
E/R schemes. Proceedings of the Thirty-First
Hawaii International Conference on System
Sciences, 7, pp. 334–343.
https://doi.org/10.1109/HICSS.1998.649228
[28] J. N. Mazón, J. Trujillo and J. Lechtenbörger.
(2007). Reconciling requirement-driven data
warehouses with data sources via
multidimensional normal forms. Data and
Knowledge Engineering, 63(3), pp. 699–725.
https://doi.org/10.1016/j.datak.2007.04.005
[29] A. Vaisman and E. Zimányi. (2014). Data
Warehouse Systems.
https://doi.org/10.1007/978-3-642-54655-6
[30] F. Rilston, S. Paim, A. E. Carvalho and J. B.
De. Castro. (2002). Towards a Methodology
for Requirements Analysis of Data
Warehouse Systems, pp. 146–161.
[31] I. Sommerville. (2004). Software
Engineering. A Brief History of Computing
(9th ed). Pearson.
https://doi.org/10.1111/j.1365-
2362.2005.01463.x
[32] W. H. Inmon. (2005). Building the Data
Warehouse (4th ed). Indianapolis, Ind:
Wiley.

ISBN: 978-1-941968-45-1 ©2017 SDIWC 58

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
Using Dense Subgraphs to Optimize Ego-centric Aggregate Queries in Graph
Databases
Ali Ben Ammar
Higher Institute of Computer Science and Management
Kairouan University
Kairouan, Tunisia
ali.benammar@isigk.rnu.tn
ABSTRACT
In this paper, we present an approach to optimize
ego-centric aggregate queries, in graph databases,
by precomputing (materializing) some of their
results. Ego-centric aggregate queries allow to
graph nodes, called consumers, to aggregate events
from others nodes, called producers. Our
contribution consists of discovering the densest
subgraph, that represents the tightly coupled nodes.
We consider that these nodes are the most actives
i.e. they have two main features: the high access
frequency and the strong correlation between them.
Then, we precompute the results of the ego-centric
aggregate queries that are implemented on the
active nodes. Our experimentation has shown that,
when the graph database size is less voluminous,
our approach is able to improve the response time
of the ego-centric aggregate queries. However, for
large graph databases, we should adjust the
selection way of active nodes to improve the
management load of ego-centric aggregate queries.
KEYWORDS
Query Optimization; Ego-centric Aggregate
queries; Graph Databases; Dense Subgraphs.
1 INTRODUCTION
Graph database (GDB) systems are
particularly used to store highly interconnected
data and to query them. They were arisen,
during the last decade, to satisfy the
requirements of applications with graph-like
data structure, such as social networks,
telecommunication networks, linked
webpages, …. The early surveys on GDB and
their tools are presented in [1] [2] [3] [4] [5]
[6]. Recently, there is a trend towards temporal
ISBN: 978-1-941968-45-1 ©2017 SDIWC
graph databases [7]. A formal definition of
GDB using graph terminology is provided in
[8] as follow. A GDB is considered as a finite
edge-labeled graph. i.e. let ∑ be a finite
alphabet, and V a countably infinite set of node
ids, then a GDB over ∑ is a pair 𝐺 = (𝑁, 𝐸),
where 𝑁 is the set of nodes (a finite subset of
𝑉), and 𝐸 is the set of edges, i.e., 𝐸 ⊆ 𝑁 ×
∑ × 𝑁. That is, we view each edge as a triple
(𝑛, 𝑎, 𝑛′), whose interpretation, of course, is an
𝑎-labeled edge from 𝑛 to 𝑛′. Therefore,
in the rest of the paper, we will use the terms
graph and graph database interchangeably.
In parallel to the development of GDB systems
and languages, some techniques have been
invented to optimize query response time. A
survey of these main techniques is presented in
[9]. In this paper, we are interested in the
optimization of a class of these queries, called
ego-centric aggregate queries.
According to [10], in an ego-centric aggregate
query, the querier corresponds to a node in the
graph, and is interested in an aggregate over
the current state or the recent history of a local
neighborhood of the node in the graph. Some
examples of ego-centric aggregate queries are
presented in [10] such as trend analysis in
social networks where the goal is to find, for
each user, the trends (e.g., popular topics of
discussion, news items) in his or her local
neighborhood. Similarly, in a phone-call
network or an analogous communication
network, we may be interested in identifying
interesting events or anomalies (e.g., higher
than normal communication activity among a
group of nodes). In [11], the authors use the
terms consumer and producer to present ego-
59
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
centric aggregate queries. According to [11],
an ego-centric query allows to a graph node,
called consumer, to aggregate events from
others nodes, called producers.
In this paper, we present an approach to
optimize ego-centric aggregate queries in
GDB by materializing some of their results.
The main contribution of our approach is to
reduce the cost of affecting materialization
decisions to the results of ego-centric
aggregate query. Firstly, we discover the
densest subgraph from the underline graph.
Dense subgraphs represent the tightly coupled
nodes. Then, we materialize all the results of
the ego-centric aggregate queries that are
implemented on the nodes of the densest
subgraph.
The rest of this paper is structured as follows.
The next section discusses related works.
Section 3 presents our approach. It contains the
details of the node classification and the
decision affectation, which constitute the two
main steps of our approach. Section 4 presents
the evaluation of the proposed approach. The
section 5 is the conclusion.
2 RELATED WORKS
Ego-centric aggregate queries are special case
of aggregate queries which are widely treated
in relational databases and data warehouse [12]
[13] [14] [15], in data streams [16], and in
sensor networks and distributed databases [17]
[18]. The materialization of query results is the
main technique, which is used to optimize
aggregate queries, in such domains. It allows
pre-computing and storing their results, called
views, to avoid computing/recomputing them
whenever they are asked. The topics of
materialization are mainly: What data do we
materialize? Where do we materializing it?
And when updating materialized data?
Aggregated results are selected based on the
following criteria: they serve frequent queries
or they are shared by some queries. The works
in [12], [13], [14], [15] have proposed
approaches to best select materialized views.
The approach proposed in [19], optimizes the
update load of aggregate data (materialized
views). In the context of data streams, [16],
ISBN: 978-1-941968-45-1 ©2017 SDIWC
[20] and [21] use aggregate data to share work
across different queries with different sliding
windows.
Aggregate queries over graphs are
fundamentally different and have not been
widely studied in previous works. [22] has
proposed a language for querying and
analyzing GDB by using aggregates and
ranking. [10] and [11] have addressed the
management of ego-centric aggregate queries
in large graphs.. Both [11] and [10] have
addressed the issue of when the events should
be transmitted from the producers to the
consumer. The two possible ways are: either at
query time or precomputed on the consumer.
The former way consists of traversing the
producers at each read of the consumer. In
[10], this way is called pull task and it
corresponds to an on-demand update of
aggregate data. However, the later one consists
of pre-computing aggregate query answer at
each new write in the producers. In [10], this
way is called push task and it corresponds to
online update. The work in [11] proposes to
retrieve events from high-rate producers at
query time and materialize, in aggregation
nodes, events that come from low-rate
producers. The work in [10] proposes detailed
solution that begins by constructing an
aggregation overlay graph and then makes a
decision for each node of this graph whether to
aggregate events on it (push decision) or not
(pull decision). The aggregation overlay graph
is constructed to encode the computations to be
performed when an update or a query is
received. The main advantage of the
aggregation overlay graph is that it allows
sharing partial aggregates across different ego-
centric aggregate queries. The decisions of
materializing events on nodes are made based
on the cost of push and pull tasks. As we have
seen, what distinguishes [10] from [11] is the
answer to the issue of where do we store
materialized data? For this reason, the solution
of [10] has integrated intermediate aggregation
nodes. These two approaches are most closely
related to our work, which consists of deciding
whether the result of an ego-centric aggregate
query should be materialized (push decision)
or not (pull decision). We have not evoked the
60
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
issue of where materialized data should be
stored. Simply, we have supposed that
materialized results are stored on the ego- node
i.e. the consumer. The main contribution of our
approach is to reduce the cost of affecting
decisions (pull/push) to ego-centric aggregate
queries. Instead of continuously measuring the
access and update frequencies of nodes to
affect decisions, we propose to intervene
periodically to identify the most active nodes
(producer or consumer), which are the nodes
of the densest subgraph. Then, we materialize
the result of every ego-centric aggregate query
that is contained on an identified active node,
i.e. the producers of this query will receive the
decision push. To the best of our knowledge,
there is no approach that uses dense subgraphs
as a way to classify and optimize ego-centric
aggregate queries.
3 OPTIMIZING EGO-CENTRIC
AGGREGATE QUERIES
In our approach, each ego-centric aggregate
query is executed in one of the following two
ways. The first way consists of querying the
inputs from the neighborhood only when the
user requests the ego node. However, in the
second way, inputs are pre-computed and kept
up-to-date; then, when the user requests it, the
ego-centric aggregate query is executed on the
precomputed data which reduce latency.
Consequently, our hybrid approach affects the
pull decision to the ego-centric aggregate
queries whose results should be pre-computed,
and the push decision to the rest of the queries.
Therefore, there are two steps in our approach:
1. Classify ego-centric aggregate queries
by identifying the nodes whose query
results should be pre-computed. This
task is performed periodically, i.e. at the
expiration of any time interval with a
predetermined duration;
2. Affect and apply update decision. This
task exploits the result of the previous
classification to optimize the server load.
It applies pull or push decisions for
executing ego-centric aggregate queries,
as we have explained here above.
ISBN: 978-1-941968-45-1 ©2017 SDIWC
These two steps follow an iteratively process
in order to make it possible to adapt the node
classes to the recent changes in the graph
databases.
In the rest of this section, we will develop these
two steps.
3.1 Classification of graph nodes
We distinguish two types of graph nodes: (i)
the most active nodes; and (ii) the less active
ones. The first set contains the nodes that
participate, as sources/targets, in the most parts
of the graph edges (represent events in real
world). The second set contains the nodes with
low participation in the graph growth. In our
approach, the ego-centric aggregate queries, to
which we affect the push (online) policy, are
implemented on a part of the first set, since
their access/update frequencies will be
important. However. The ego-centric
aggregate queries, to which we affect the pull
(on demand) policy, belong to the second class
of nodes, where the access/update are less
frequent.
To specify the first set of nodes, we have
chosen to look for the tightly coupled nodes of
the graph in the recent time. We suppose that
these nodes represent the main interests of
users, i.e. topics that capture popular attention
and in which a group of nodes has participate
as sources or targets. In literature, tightly
coupled nodes are called dense subgraph.
Dense subgraphs may correspond to emerging
stories in social media, hot topic of discussion
in a forum …. We use dense subgraphs, to
discover the more active nodes, because:
 Dense subgraphs group the nodes that are
frequently updated/accessed and that
capture the most part of interactions.
 Dense subgraphs indicate the
trends/interests of users in the current
and probably in the future time.
For example, in figure 1, the densest subgraph
is composed of vertices {a, b} and the edges
between them. These two nodes attract the
major part of the graph edges. However, {c} is
less used in the interactions between nodes. So,
if the vertex a or b contains an ego-centric
61
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
aggregate query, then its result will be
precomputed (push decision) else we apply the
pull decision for it.
Figure 1. Example of graph having a densest subgraph.
Dense subgraphs may be identified from
directed or undirected underline graphs. In this
approach we consider the directed graph since
the recent applications of GDB need directed
edges like social networks, communication
networks, email networks, financial
transaction ….. According to [23], where the
underline graph is directed, the problem of
identification of dense subgraphs is formulated
as follow:
Let G(V,E) be a directed graph, where
V is the set of vertices and E is the set
of directed edges between vertices of V.
To identify the densest subgraph
𝑀(𝑉𝑚 , 𝑇𝑚 ) of G, we search the subsets
𝑆 ⊆ 𝑉 and 𝑇 ⊆ 𝑉 so that:
- All the edges from S to T are
included in E, i.e. 𝐸(𝑆, 𝑇) = {𝑒𝑖,𝑗 ∈
𝐸, 𝑣𝑖 ∈ 𝑆, 𝑣𝑗 ∈ 𝑇};
- The subgraph composed of 𝐸(𝑆, 𝑇)
has the maximum density from all
the subgraphs, i.e
𝑚𝑎𝑥𝑆,𝑇 ⊆ 𝑉 {𝑑(𝑆, 𝑇)} where 𝑑(𝑆, 𝑇)
represents the density. This density
of directed graphs was introduced in
|𝐸(𝑆,𝑇)|
[24] as follow: 𝑑(𝑆, 𝑇) =
√|𝑆|.|𝑇|
- 𝑇𝑚 = 𝐸(𝑆, 𝑇); 𝑉𝑚 = 𝑆 ⋃ 𝑇;
In literature, the algorithms, of identification
of dense subgraphs, are either with or without
overlap. The approach without overlap
discover a set of dense subgraphs so that the
intersection between each couple of dense
subgraphs is null [23]. However, the approach
with overlap authorizes the intersection
between dense subgraphs [25]. In this paper,
since our aim is limited to identify the most
active vertices, the overlap of subgraphs
ISBN: 978-1-941968-45-1 ©2017 SDIWC
becomes not interesting. So, we will use the
greedy approximation algorithm [23] to
discover dense subgraphs.
The greedy approximation algorithm identify
the densest subgraph through several
iterations. In each iteration, it removes the
node that has the minimum degree vertex
according to a certain rule.
3.2 Affectation of push/pull decisions
In our approach, we intervene periodically to
decide whether the result of an ego-centric
aggregate query should be precomputed (push
decision) or not (pull decision). In other words,
every time when a pre-specified interval of
time (called intervention period 𝑝𝑖 ) is expired,
we search the densest subgraph to classify the
nodes used in the last period 𝑝𝑖 . The vertices of
the densest subgraph, issue from the
classification task, are added to a set 𝑁 that
regroups the vertices of the densest subgraphs
of the previous periods. In other words, if 𝑀𝑘
is the set of vertices of the densest subgraph of
the period 𝑝𝑘 , then 𝑁 = ⋃𝑘=0,…,𝑖 𝑀𝑘 .The
decision for an ego-centric query 𝑞𝑧 which is
implemented on a vertex 𝑣𝑗 , is made according
to the following rules:
 If 𝑣𝑗 ∈ 𝑁 then the decision for 𝑞𝑧 is push
 Else the decision for 𝑞𝑧 is pull
𝑁 is incrementally constructed because:
 The identification of the densest
subgraph from the wall underline graph
is highly complex. In our approach, only
the last increment to the graph is used to
search the dense subgraph;
 New active nodes may arise over time
and we should adapt the update policy of
the ego-centric aggregate queries they
contains.
(a) state at 𝑡0 (b) state at 𝑡1 (c) state at 𝑡2
Figure 2. Example of underline graph with increments.
62
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
For example in figure 2, we have three states
of the underline graph in three different times.
The dotted edges and vertices represent the
changes (the increment) on the underline graph
from 𝑡𝑖 to 𝑡𝑖+1 , where 𝑡𝑖+1 = 𝑡𝑖 + 𝑙 and 𝑙 is
a specified time interval, i.e. the duration of 𝑝𝑖 .
The table 1 presents the densest subgraphs of
the periods 𝑝0 , 𝑝1, 𝑝2 corresponding to the
three states of figure 2; and the evolution of the
content of the set 𝑁 across time. The 3 last
rows of the table present the decisions for three
ego-centric aggregate queries 𝑞1 , 𝑞2 , 𝑞3
implemented respectively on the vertices a, c
and f.
Table 1. Example of densest subgraphs and
decisions.
Period 𝒑𝟎 𝒑𝟏
Vertices of the densest {a, b} {a, d}
subgraph
N {a, b} {a, b, d}
Decision for 𝒒𝟏 push push
Decision for 𝒒𝟐 Pull pull
Decision for 𝒒𝟏 Pull pull
3.3 Experimentation
We evaluate our proposed approach using the
following questions:
1. Do the nodes, which are considered the
most active ones and whose query
results have been precomputed, have
really optimized the server load and the
response time of ego-centric aggregate
queries?
2. How relevant is the choice of the
duration of the period to reclassify nodes
and begin a new iteration?
3. Do the size of the dataset has an impact
on the result of our approach?
In order to answer these questions, we have
evaluated our approach against the two
datasets: CollegeMsg temporal network
[http://snap.stanford.edu/data/CollegeMsg.ht
ml] and Super User temporal network
[http://snap.stanford.edu/data/sx-
superuser.html]. The first dataset is comprised
of private messages sent on an online social
network at the University of California, Irvine.
Users could search the network for others and
then initiate conversation based on profile
information. An edge (u, v, t) means that user
ISBN: 978-1-941968-45-1 ©2017 SDIWC
u sent a private message to user v at time t. The
graph of this dataset contains 1899 nodes and
59835 temporal edges. The time span of this
dataset is 193 days. The second dataset is a
temporal network of interactions on the stack
exchange web site https://superuser.com.
There are three different types of interactions
represented by a directed edge (u, v, t):
 User u answered user v's question at time
t;
 User u commented on user v's question at
time t;
 User u commented on user v's answer at
time t.
This second dataset is comprised of 194085
𝒑𝟐 nodes and more than 1 million edges (1443339
{b, f} temporal edges). Its time span is 2773 days.
{a, b, d, f}
Push
In this experimentation, each vertex 𝑣𝑖 , which
Pull corresponds to a user, is considered having an
push ego-centric aggregate query that summarizes
the reactions of neighbors to the messages
(emails, questions, answers, comments) of 𝑣𝑖 .
In order to answer the first question, we ran our
system on the dataset CollegeMsg, here above
described. We measured the update load,
which is the required time to update the results
of all the ego-centric aggregate queries, and the
average of query response time. We have split
the time span to periods. The duration of each
period is 7 days i.e. one week. The table 2
presents the result of this test, in milliseconds,
for the three policies of running queries:
 Push decision for all the vertices, called
precomputation policy;
 Pull decision from all vertices, called on
demand policy;
 Hybrid policy according to the principle
of our approach.
From table 2, we can see that our approach
allows the optimization of the update load of
the precomputation policy by more than 46%
and the query response time of the on-demand
policy by more than 15%. It is obvious that the
on-demand policy produces the minimum
update load, since query results are computed
only when the query is asked. The
63
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

precomputation policy produces the optimal Table 3. Selectivity and rate of participation.
response time, since the results of all queries Of all vertices Of active
vertices
are precomputed. We conclude, from this Average number by 334 5
experiment, that our approach gives the best period
compromise between an acceptable quality of Rate of the active vertices 5/334 = 1.5%
with respect to the total
service (QoS) and a low update load i.e. we number of vertices
optimize the query response time using a low Average number of edges 2209 806
update load. (events) by period
Rate of participation of 806/2206 = 36.49%
active vertices in the
Table 2. Results of the first experiment. edges
Policy Precomputation On Hybrid
demand In order to answer the second question we have
Update load in ms 2369445 1065968 1271694 executed our system on the dataset
Average of query 6.08 23.95 20.21
CollegeMsg with varying the period duration
response time in from 3 to 150 days. We have measured how
ms much our approach is capable to optimize the
Rate of - - 46.32%
optimization of total load i.e. the access and update loads. We
precomputation calculated the optimization rate. The
policy load by our
approach optimization rate is the percentage of
Rate of - - 15.60% decreased/increased time that resulted by
optimization of
query response applying our approach with rapport to the two
time of On demand other policies. For example, the optimization
policy by our
approach rate of the precomputation policy is calculated
as follow:
We explain, in table 3, why this first
experiment has produced good results. The 𝑡𝑜𝑡𝑎𝑙 𝑙𝑜𝑎𝑑 𝑜𝑓 𝑡ℎ𝑒 𝑝𝑟𝑒𝑐𝑜𝑚𝑢𝑝𝑡𝑎𝑡𝑖𝑜𝑛 𝑝𝑜𝑙𝑖𝑦 − 𝑡𝑜𝑡𝑎𝑙 𝑙𝑜𝑎𝑑 𝑜𝑓 𝑜𝑢𝑟 𝑎𝑝𝑝𝑟𝑜𝑎𝑐ℎ

selectivity, of the algorithm of classification of 𝑡𝑜𝑡𝑎𝑙 𝑙𝑜𝑎𝑑 𝑜𝑓 𝑡ℎ𝑒 𝑝𝑟𝑒𝑐𝑜𝑚𝑢𝑝𝑡𝑎𝑡𝑖𝑜𝑛 𝑝𝑜𝑙𝑖𝑦

nodes at each period 𝑝𝑖 , was 1.5% i.e. one or

The figure 3 presents the results of this
two from the nodes of the underline graph, are
experiment.
considered active vertices. However, although
this low selectivity, the active vertices have Optimization rates
participate in about 37% of edges in the next
20.00%
period 𝑝𝑖+1 . This indicates that our
classification approach select the vertices that 0.00%
150 90 days 60 days 30 days 15 days 7 days 3 days
have the high impact on the access/update days -20.00%
load. Also, the precomputation, of results of
-40.00%
the ego-centric aggregate queries which are
implemented on the selected vertices, was -60.00%
profitable i.e. compared to their update -80.00%
frequency, the access frequency of these
optimization rate of precomputation policy
queries is very high. In addition, the other ego-
centric aggregate queries, which are optimization rate of on demand policy

implemented on the passive vertices, have a

low impact on the server load, even if their
results weren’t precomputed.
Figure 3. Optimization rates in collegeMsg dataset.
From the figure 3, we conclude that, regardless
of the period duration, our hybrid approach
give less load than the on demand policy but
the best rate was for a duration of 60 days (≃
8%). However, in rapport with the
precomputation policy, there is an
optimization of the total load only for short

ISBN: 978-1-941968-45-1 ©2017 SDIWC 64

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
periods (less than 30 days), where the best rate
is for duration of 3 days (≃ 15%). For long
periods, the selectivity, of our classification
algorithm, is high. For example, for the period
duration 90 and 150 days, the selectivity was
27.45% and 30%, respectively. In addition, the
refreshment of the list of selected vertices will
be slow, i.e. it may not follow the change in the
user interests. Consequently, some vertices
that are expected to be active will consume a
lot of update time but they will never be
accessed. For example, the participation of the
expected active vertices, for period duration 90
and 150 days, was 59% and 47%, respectively.
In addition to that, the possibility, to select new
active vertices to reduce the total load, is
delayed. This makes the precomputation
policy, which precomputes the results of all
queries, more profitable than our hybrid
approach for long periods. On the other hand,
this failure, of classifying vertices, did not
prohibit our approach to produce a profit in
rapport with the on demand policy. The main
conclusion of this experiment is that, to
optimize the two policies, the period duration
shall not exceed 30 days.
In order to answer the third question we have
executed our system on the dataset Super User,
here above described. We have measured the
optimization rate of our approach with rapport
to the two other policies. The figure 4 presents
the results of this experiment.
What we conclude, from the figure 4, is that
our hybrid approach has never produced a
profit with rapport to the on demand policy for
all the period durations of the test. The rate of
optimization with rapport to the on demand
policy was around 0% . In most cases, it was
negative. The second conclusion is that the
precomputation policy was the worst and we
have optimized it by nearly 20%, in all cases.
The main reason, that has led to these results is
the high variety of users and of their interests
in the case of large dataset. In other words, the
correlation between users, which is the basis to
select active vertices and affect the push
decision for them, is low in such large datasets.
Moreover, even if there is a correlation
between a group of users (in our dataset, users
ISBN: 978-1-941968-45-1 ©2017 SDIWC
may be correlated because they discuss an
issue), it does not last in time. This makes the
results, of the ego-centric aggregate queries,
that are precomputed (push decision), are less
or never requested; that's why the
precomputation policy was costly i.e. it
precomputes all the results for nothing.
However, in the first dataset (CollegeMsg),
where users know each other and are in
permanent contact, the results of our hybrid
approach were good.
Optimization rates
30.00%
20.00%
10.00%
0.00%
90 days 60 days 30 days 7 days 5 days
-10.00%
optimization rate of precomputation policy
optimization rate of on demand policy
Figure 4. Optimization rates in super user dataset.
Our main conclusion from these experiments
is that our approach may be used to optimize
ego-centric aggregate queries in graph
databases with limited number of users and
highly connected vertices across time.
However, our approach is not recommended
for large graph databases with less correlated
nodes across time.
4 CONCLUSION
In this paper, we have proposed an approach to
optimize ego-centric aggregate queries in
graph databases. Ego-centric aggregate queries
allow to a graph node, called consumer, to
aggregate events from others nodes, called
producers. The most used technique to
optimize such queries is the materialization of
their results either in the consumer node or in
the producer ones. We have developed a policy
that materializes only the results of ego-centric
aggregate queries which are implemented on
active nodes. A node is considered active if it
is an element of the densest subgraph. For this
reason, we have begun by discovering the
densest subgraph. We have supposed that the
densest subgraph regroups the nodes whose
65
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
access/update frequency is high and the
correlation between them is strong. The results
of our experimentation have demonstrated
that, in case of small graphs, our approach
produces a low management load with rapport
to the scenarios of precomputing all query
results or computing query result at query time.
However, for large graphs, computing query
result at query time is the best scenario. The
failure of our approach, in case of large graphs,
comes from the level of correlation between
the selected active nodes, which has not held
strong. Consequently, access frequencies are
decreased and the expected optimization has
not been realized. Therefore, our future works
will focus on how improving the way of
selecting active nodes in order to be profitable
across time.
REFERENCES
[1] R. Angles and C. Gutierrez, "Survey of graph
database models," ACM Comput. Surv., Vol. 40,
No. 1, pp. pp. 1-39, 2008.
[2] K. N. Satone, "Modern Graph Databases
Models," International Journal of Engineering
Research and Applications , 2014.
[3] P. . T. Wood, "Query languages for graph
databases," SIGMOD Rec Vol. 41, No. 1, pp. pp.
50-60, 2012.
[4] P. Macko, D. W. Margo and M. I. Seltzer,
"Performance introspection of graph databases,"
in 6th Annual International Systems and Storage
Conference, Haifa, Israel , June 30 - July 02,
2013.
[5] P. Jadhav and R. Oberoi, "Comparative Analysis
of Graph Database Models using Classification
and Clustering by using Weka Tool,"
International Journal of Advanced Research in
Computer Science and Software Engineering, pp.
438-445, Volume 5, Issue 2, February 2015.
[6] H. R. Vyawahare and P. P. Karde, "An Overview
on Graph Database Model," International Journal
of Innovative Research in Computer and
Communication Engineering, Vol. 3, Issue 8,
August 2015.
[7] A. Campos, J. Mozzino and A. A. Vaisman,
"Towards Temporal Graph Databases," CoRR
abs/1604.08568 , 2016.
[8] J. Reutter, Graph Patterns: Structure, Query
Answering and Applications in Schema
Mappings and Formal Language Theory, PhD.
Dissertation, University of Edinburgh, 2013.
ISBN: 978-1-941968-45-1 ©2017 SDIWC
[9] Ben Ammar, Ali, "Query Optimization
Techniques In Graph Databases," in CoRR
abs/1609.01893 , 2016.
[10] J. Mondal and . A. Deshpande, "EAGr:
Supporting Continuous Ego-centric Aggregate
Queries over Large Dynamic Graphs," CoRR/
abs/1404.6570, 2014.
[11] A. Silberstein, J. Terrace, B. . F. Cooper and R.
Ramakrishnan, "Feeding Frenzy: Selectively
Materializing Users’ Event Feeds," in SIGMOD,
2010.
[12] H. Gupta and . I. S. Mumick, "Selection of
Views to Materialize in a Data Warehouse,"
IEEE Trans. on Knowl. and Data Eng. 17(1), pp.
pages 24-43. , Jan. 2005..
[13] I. Mami, Z. Bellahsene and R. Coletta, "A
Declarative Approach to View Selection
Modeling.," T. Large-Scale Data- and
Knowledge-Centered Systems, pp. 115-145,
2013.
[14] P. P. Karde and V. M. Thakare, "Selection &
Maintenance of Materialized View and It’s
Application for Fast Query Processing A
survey," International Journal of Computer
Science & Engineering Survey (IJCSES), Vol.1,
No.2, November 2010.
[15] Y. . D. Choudhari and S. K. Shrivastava, "Cluster
Based Approach for Selection of Materialized
Views," International Journal of Advanced
Research in Computer Science and Software
Engineering, Volume 2, Issue 7, July 2012.
[16] S. Krishnamurthy, C. Wu and M. J. Franklin,
"On-the-fly sharing for streamed aggregation," in
SIGMOD, 2016.
[17] S. Madden, M. J. Franklin, J. M. Hellerstein and
W. Hong, "TAG: a Tiny Aggregation service for
Ad-Hoc sensor networks," in OSDI, 2002.
[18] A. Silberstein and . J. Yang, "Many-to-Many
Aggregation for Sensor Networks," in ICDE,
2007.
[19] X. Zhang, L. Yang and D. Wang, "Incremental
view maintenance based on data source
compensation in data warehouses," International
Conference on Computer Application and
System Modeling (ICCASM), pp. 287-291, 22-
24 Oct. 2010.
[20] S. Wang, E. A. Rundensteiner, S. Ganguly and S.
Bhatnagar, "State-Slice: New Paradigm of Multi-
query Optimization of Window-based Stream
Queries," in VLDB, 2006.
[21] L. Al Moakar, "Class-Based Continuous Query
Scheduling in Data Stream Management
Systems," in PhD thesis, Univ. of Pittsburgh,
2013.
[22] A. Dries and S. Nijssen, "Analyzing graph
databases by aggregate queries," MLG@KDD,
pp. 37-45, 2010.
66
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

[23] M. Charikar, "Greedy approximation algorithms

for finding dense components in a graph,"
APPROX , pp. 84-95, 2000.
[24] R. Kannan and V. Vinay, "Analyzing the
Structure of Large Graphs," Manuscript, , August
2009.
[25] O. D. Balalau, F. Bonchi, T. H. Hubert Chan, F.
Gullo and M. Sozio, "Finding Subgraphs with
Maximum Total Density and Limited Overlap,"
WSDM , pp. 379-388, 2015.

ISBN: 978-1-941968-45-1 ©2017 SDIWC 67

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

A Secure Method for the Global Medical Information in Cloud Storage based on the
Encryption and Data Embedding
1 Soheil Nezaket, Islamic Azad University, IAU, UAE branch, Dubai, UAE
2 Dr. Mohammad V. Malakooti, Islamic Azad University, IAU, UAE branch, Dubai, UAE
3 Dr. Navid Hashemitaba, Islamic Azad University, IAU, Tehran Central Branch, Tehran, Iran
1 soheilnezakat@yahoo.com, 2malakooti@iau.ae 3 nhtaba@yahoo.com

Abstract: videos with high resolution from hospitals and

The cloud computing has become the most popular
platform and environment for both data computing
and information storage. The big corporations,
businesses, and enterprise companies moved their
infrastructures on the cloud for the cheap storage
facilities and processing. Though, most the cloud
service providers have a highly secure system for
information storage and computing but the
malicious cloud users or other attackers could
access our data on the cloud during the transmission
or while information are stored. We have suggested
a technique that can be used to encrypt the
information prior to transmission over the cloud
system. Our technique is for the global medical
information in cloud storage using the orthogonal
transform algorithms for the encryption and data
embedding processes. Our technique is fast,
reliable, robust, and reversible and can be used to
increase the security and integrity of the
information stored on the cloud systems.
Keywords: Cloud, Computing, Encryption,
Decryption, hiding, data Security, scramble.
Research Objective:
Picture Archiving and Communication System
(PACS) as well as Hospital Information System
(HIS) are one of the most sophisticated systems
which have been used in healthcare, generally in
hospital and health systems. PACS traditionally
were associated with the Radiology department
where the radio graphical images were captured for
the Diagnosis and treatment of patients with broken
legs, arms, hands, nose, head, and so on. In recent
year, the invention of advanced technology used in
hospital and healthcare department helped to
incorporate PACS with the images obtained from
other departments such as, cardiology, pathology,
oncology, and dermatology. Thus, a massive
amount of images captured in hospital and
healthcare centers need to be stored and achieved as
the patient medical files. Nowadays PACS has
improved a lot and it has captured images and
ISBN: 978-1-941968-45-1 ©2017 SDIWC
healthcare centers. The captures information must
be compressed as well as be encrypted before it has
been transferred over the internet to highly secured
cloud storage facilities for the easy access, low cost,
and high security.
HIS is a comprehensive, integrated information
system designed to manage all aspects of the
hospital’s operation, such as medical, financial,
administrative and legal issues. It provides a
common source of information about the patient’s
health information and his illness history. The
system should have capability to store data in a
secure storage facility that can be accessed easily
and information can be retrieved by authorized user
through some type of authentication and
classification which define the user’s access level.
HIS has integrated with most of the medical devices
and controls the information processing and
handling of huge amount of data which should be
compressed and secured prior to transmission over
the networks. We have proposed a new method
which has used the patient information and applied
image segmentation and DCT as well as scrambling
techniques on the DCT coefficients. We also added
the IRIS information to increase the level of
security and processed it through cellular automata
to reduce the size of IRIS information where data
accuracy and confidentiality is very important. Our
innovative algorithm is based on the information
encryption which has used orthogonal transform
matrices for the encryption process, along with
scrambling algorithm, as well as using biometric
features and the IRIS information processed through
the cellular automata to obtain a high secured and
reversible algorithm for the information
compassion, encryption, scrambling and IRIS
coding..
Introduction:
The cloud computing is an infrastructure in which
several computer networks, servers, storage units,
software, hardware, services, applications, resources
68
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
are interconnected to form an information pool that
can be accessed over the internet with minimal
management efforts. The individual users,
corporations, and even enterprises can perform data
storage and computing on their own private cloud or
using the third–party cloud facilities rapidly,
efficiently, with high speed and acceptable security.
The most important feature of the cloud computing
is the resource sharing with easy access, high
reliability, minimum cost, high speed, and
acceptable security. Cloud storage is a hot topic
nowadays as the data storage capacity rates are
increasing manifold’s every year and has become a
reality that all data centers and organizations should
consider it.
The cloud computing and storage is a fast growing
technologies in which it has received the high
attention of scientists, researchers, enterprises, and
industrial communities due to its low cost, easy
access, high speed, and reliability which bring data
security and portability. Cloud Storage empowers
the real time data access over the network. Each
user with their pre-specified privilege can have a
real time and on-demand access to the online data
pool which is shared among all cloud uses but with
different data type, accessibility and performance.
Cloud Storage occurs as central data storage idea
which its main goal is avoid the data replication,
duplication, provide quick-secure-suitable data
exchange and also prevent massive paper
displacement. The cloud storage speeds up all the
basic cloud features like: collaboration, agility,
scalability, availability. In some cases, Cloud
storage brings the centralized data management in
parallel with data security. The most important issue
that need to be considered regarding to the cloud
computing is the uncertainty that exist in its security
while the information is transmitted through the
cloud facilities or during the computation or storage
process[1]. The security risk for storing the vital
information over the public, cloud storage, or any
online facilities are high because the user have no
control over the cloud storage and many people
simultaneous can access the cloud and breach the
security.
Data movement for the companies is a big problem.
In order to decrease this issue, the cloud storage
providers ensure the data owners that they can
continue with the same security and privacy
monitor their data[2]. Although the cloud
computing has solved the storage problem of
corporations and enterprises in which they have
ISBN: 978-1-941968-45-1 ©2017 SDIWC
saved millions of dollars using the cloud storage,
but the security issues are still under consideration.
The security threats are major challenges for the
cloud users and many businesses and they are not
comfortable to use cloud storage facility and prefer
to use their own infrastructure and storage systems.
We have proposed a new method that can be used to
scramble and encrypt the information prior to the
transmission process. While the information is
encrypted then the hacker no longer can retrieve the
original data during the transmission or while the
data is stored on the cloud network.
Cloud storage brings scalability, economies, and
flexibility all together. But still security in the cloud
storage has its own proportional worries. The cloud
storage will solve security problems occurred in the
cloud computing. problems related to compliance,
privacy and legal matters[3]. Cloud storage will
covers most of the security eras considered for the
shared data. The study has shown that most of the
cloud storage providers are aware of the extreme
important of the data integrity and security and they
have launched the new firewalls and software for
their network security as well as the data
integrity[4].
Study of the Cryptography:
The cryptography system brings latest and modern
security protocol. Cryptography can be used in a
correct way and wrong way and also it has the
possibility to protect the wrong things. Transferring
message in a secure way is the goal of
cryptography. Cryptography has been used to
prevent the insecurity in a way that information is
encrypted by the predefined algorithm and those
who have knowledge of the algorithm as well as the
secret keys can decrypt the information and retrieve
the original data from the encrypted message. Once
the message is encrypted, the meaning will change
and without key no one can find out what is that.
This meaning will be revealed only after the correct
recipient tries to access it. The encrypted message
will be obtained through some reversible algorithm
in which the original data can be obtained by
inverse option through the decryption algorithm. If
the algorithm is not reversible then it cannot be used
for the encryption and decryption processes.
The accrued inconsistency between the security and
cryptography is now over 20 years. The people who
works and study about the security don’t find the
cryptography tools much useful. A more complex
method of information security is the lossless data
69
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
hiding or information embedding in which vital
data, such as social security or credit card numbers,
are hidden inside the coded information rather than
the original message. We have optional
transformation such a s DCT to convert the original
message into coded message. Once the code
message is obtain then the vital information is
inserted into mid-frequency area of the coded
message. This process is based on the mathematical
algorithm that can be used to retrieve the data from
the decode information in reversible technique[5].
The mathematical function which is using the
encryption named as cryptographic algorithm. The
used method in that algorithm is working with key
and a record that can be number or phrase. Keys are
generally giant numerical figures. You can store
keys in encoded forms. The keys can be stored in
two files on your hard disc using PGP.
Review on biometrics:
Nowadays, the importance of exigency to identify
users of convenience and services has improved and
became more on the significant issues not only for
controlling accesses a system & service, but also to
determine who will have the rights. Biometrics is
currently being applied all over the world in various
ways. All these systems are generally computer-
based solutions, where the procedure of validation
is running in at the server’s side or workstations.
Biometrics relies on two major categories: Physical
and Behavioral as mentioned in below.
Physical Modalities:
Biometric systems that can be used for the human
identification are based on the following physical
elements, such as face shape, fingerprint, retinal
scans, and Iris scans. The DNA information also
can be used in the high security area such as Data
Centers, military sites, or the aerospace launch
controller. We have not used the DNA test in our
research due to lack of available information and
only used face shape, fingerprint, eyes(retinal scans,
and Iris scans) as the most important elements of
the physical modalities.
Face Shape:
The face recognition technology is one of the
advanced technics that can be used to measures and
match the unique characteristics of the human
biometric features for the identification and
authentication. The digital camera connected to the
face recognition software can obtain the detected
image and extract their features which then match
ISBN: 978-1-941968-45-1 ©2017 SDIWC
them against the stored featured in the database. In
addition, we can obtain extra features by measuring
the distance between the eyes and mouth, as well as
the distance between the nose and mouth. The total
biometric features collected from the human face
can be in the wide of potential applications
Fingerprint:
The fingerprint is made up of pattern of ridges,
valley and furrows on the surface of the fingertip.
Fingerprints have been used for the criminal
investigation by the law enforcement officers to
identify the people for more than 100 years. The
finger pattern is created in the first step of growing
fetus in the uterus. This feature is a unique feature
of each human. The fingerprint information can be
obtained by using a fingerprint reader or fingerprint
scanner. The scanner is a biometric device that can
be used to identify a person based on the acquisition
and recognition of those unique patters in a
fingerprint. The fingerprint scanners are the most
popular types of the biometric security and used
with a variety of the systems on the market for the
general and mask market usage.
Eyes
In spite of the reduced size of this organ, it provides
two reliable modalities: The Retina and Iris.
3.1 Retina Scans
The retina is one of the modalities which provide
better performance results; however, this technique
is not well accepted due to the eye invasion during
the acquisition process.
3.2 Iris Scans:
The Irises are the most important biometric features
that can be used for the authentication. The patterns
in our irises are unique and hardly can be replicated,
meaning those irises authentications are safe and
secure. The Irises authentications mostly have been
used in the immigration, international police, airport
security, and criminal justice systems. The Iris
scanners have provider the high resolution images
that can be used to identify the person with the high
reliability and accuracy inside the airports, highly
secured data centers, and even in the criminal
justice systems. The development of highly
resolution camera and scanner as well as the fast
and robust software make it possible to retrieve the
high resolution iris features from the captured
image and quickly compare it with the existence Iris
features inside the database. The high resolution
70
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
iris scanner along with the underline software for
feature extraction has made this modality cost
effective for commercial applications.
Behavioral modalities:
Behavioral modalities are based on the data derived
from an action performed by the user.
1- Voice recognition:
Voice recognition is a combination of a physical
and a behavioral characteristic. Voice is based on
two factors: first, on language and the way of
speaking which helps in the mouth disposition for
modulation and secondly, in physical traits such as
vocal chords or the mouth itself. Voice recognition
is the identification of individuals from the
characteristics of their voices and often referred to
as voice biometrics. The voice characteristics can be
used for both authentication and identification. The
characteristics of the voice or its features can be
obtained by applying the Discrete Fourier
Transform and other feature extraction techniques.
These features can be saved for the speaker
identification as well as of the voice recognition.
The acoustic features obtained from voice analysis
reflect both anatomy (size and shape of throats and
mouth) as well as learning the behavioral patterns.
2-Signature
Signature systems can observe only the result of the
action, i.e. the signature. These systems do not
require that the signature is made at the time of user
identification. Thus, it can be used in forensic
Biometrics. Signature recognition is a type of
behavioral biometric in which users can write their
signatures on the paper or in digitizing tablet. When
the signature is typed on paper it can be digitized by
using a scanner or camera for offline recognition. In
contrast when users write their signature in a
digitized tablet, the online recognition can be
applied to analyses the signature based on their
some features such as pressure, spatial coordinates x
and y, azimuth, inclination, and pen position The
most popular signature recognition techniques are
dynamic time warping, hidden Markov Model, and
vector Quantization.
3-Gait:
The gait is the person’s manner of walking. It is
referred to the style of stepping or the locomotion
achieved through the movement of human limbs.
The variety of gait patterns are characterized by
differences in limb movement patterns and it can
ISBN: 978-1-941968-45-1 ©2017 SDIWC
recognized by overall velocity, forces, kinetic and
potential energy cycles, and changes in the contact
with the surface. This trait although not very
distinctive among users, provides an extra
advantage: it can be measured at a distance,
avoiding contact with the user. This modality,
therefore, is interesting for surveillance
applications.
Security:
High security is needed when we are talking about
the biometrics. Because the biometrics features are
grafting with the individuals features and data. This
security must be occurring in two phases. The first
phase of protection is in the storage, where they
store. And the second phase is where the
information exchange happens.
Global Medical Cloud Storage
By using our proposed Image embedding algorithm,
one can easily store the patient’s personal
information at the specific location of the image.
There are several data embedding techniques that
have been proposed by different authors. The one
which is close to our proposed algorithm is the act
of hiding the secret information into the image data
during the encoding process. The encoded image
along with embedded secret information will be
delivered to the decoder[6].
We proposed a new approach that can hide data in
general images, including personal information,
medical information and any type of data that can
be converted to numbers (e.g. ASCII Code). In
contrast to traditional techniques of the data hiding,
the new embedding process of lossless data hiding
methods must be invertible and user be able to
completely restore the original image after the
extracting the embedded secret Information[7].
Our proposed model is based on the a novel
Lossless, Secured data embedding algorithm in
which the vital information can be embedded into
the personal or radiology image while preserving
the quality of cover image and maintaining the
security of the data to be embedded[8]. In many
embedding algorithm, a huge amount of data will be
embedded into the cover image with a high security
and high resolution while the extraction of the
embedded data are required to be lossless and
robust[9]. We have applied the Hilbert curve as
well as encryption with the Iris Code on the original
data and hide the data inside the mid frequency
areas of the DCT coefficients. The general equation
71
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

for a 2D (N by M) image DCT is defined by the following equation.

𝑁−1
2 1⁄ 2 1⁄ 𝜋.𝑢 𝜋.𝑣
𝑓(𝑢, 𝑣) = (𝑁) 2( ) 2 ∑ ∑𝑀−1
𝑗=0 𝐴(𝑖). 𝐴(𝑗). 𝑐𝑜𝑠 [ (2𝑖 + 1)] cos [ (2𝑗 + 1)] . 𝑓(𝑖, 𝑗) (EQ. 1-1)
𝑀 𝑖=0 2.𝑁 2.𝑀

Where f(i,j) is the sample image point and f(u,v) is
the sampled DCT coefficients, M, and N are the
image rows and columns of image, respectively.
GMCS characteristics:
GMCS (Global Medical Cloud Storage) Algorithm
consists of two major security levels that can be
used to hide information of many Patients inside
their images (Personal and radiology). The first
security level is applied on Data itself, which
contain multiple security levels (conversion,
encryption, scrambling).
All the data at the begging stage will be converted
to the ASCII codes.
Data getting scrambled (Peano Hilbert).
The last stage is encryption.
The second security level is at the image
processing.
Transforming image by Discrete Cosine Transform
formula. Merging the image pixels with the data.
Archiving is the last stage.
Once the personal information or patient’s vital data
has been obtained from the client file then they will
be embedded inside the image file using our highly
secured algorithm without losing even one bit of the
information.
In this paper we focused on using part of
(embedding information's into the LSB and
modifies DCT values). There are 3 main concerns
that are clearly showed in Figure 1.

Getting Resizing the Convert Applying Keeping

patient’s image if image into 2D DCT on each required digit
images, either needed to array of blue 8*8 block of in the decimal
personal or 320*248 pixels the blue part
medical pixels

Gathering Convert the Scrambling Merging all the information and data
patient data data into 2D array of together;
(medical or ASCII code data by Peano
personal) and place it in Hilbert curve
to 2D array 1- Adding converted scrambled
and encrypted data to the end
of the transformed value
Generating Using MD-5 Encrypt the Dara 2- Encrypt all data once again for
512-digit or cellular with Generated double secure purpose
IRIS code automata to Code
from eye scan reduce 512-
digit code

Figure 1-Block Diagram of DCT along with scrambling on the Patient’s Information, and IRIS code

How can we select DCT coefficients to embed the
data?
How can we embed the data in each block of image
by using the Hilbert-space-filling curve scrambling
and encryption?
How can we embed the data into image in secure
manner and what technique do we use to enhance
the efficiency.
ISBN: 978-1-941968-45-1 ©2017 SDIWC
A Hilbert curve (also known as a Hilbert space-
filling curve) is a continuous fractal space-filling
curve first described by the German mathematician
David Hilbert in 1891. The Hilbert curve (Figure 2)
In particular appears to have useful characteristics
of the cells after 4 subdivision steps[10].
72
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

Hilbert rules:
Rules:
L  +RF-LFL-FR+
R  -LF+RFR+FL-
(-) Turn right
(+) Turn left
Ln+1 = +RnF − LnF Ln − F Rn+
Rn+1 = −LnF + RnF Rn + F Ln−

Figure 2-Peano Hilbert Scrambling

Authentication and Encryption key: Cellular Automata:

Encryption will use a 32-digit code that will be
XOR with the scrambled data matrix but for the
encryption another data is required. Iris code is the
3rd type input. This input is neither image nor text
data that can be simply fetched from person itself.
This input is neither image nor text data that can be
simply fetched from person itself. With utilizing iris
scanner, a single 512-digit unique code will be
generated.
As the 512-digit code is a big code, there are couple
choices for the compression. Depends on the
advantages and disadvantages of the following
method, we can choose either MD-5 or Reversal
Cellular automata.
MD-5:
MD5 is an encryption method that widely used as a
hash function encryption. It accepts a message as
input and generates a fixed-length output, which is
generally less than the length of the input message.
The output is called a hash value, a fingerprint or a
message digest. Here, in our algorithm the MD-5
will generate a 32-digit code which is 4X smaller
than iris code. 9*1031 unique code can be generated
with MD-5 algorithm.
The general use of cellular automata is to expand a
rul. But in our model we have utilized cellular
automata to reduce the size of 512-digit Iris code.
The 512-Digit iris code has 512*4=2048 bits. The
goal of using cellular automata is to reduce 4096
bits to 128 bits which is 32-digit code
Functional GMCS:
The process of global medical cloud storage
contains the below steps:
Image processing
First we divide the image into 3 layers, Red, Green,
and Blue. Blue layer is selected to be transformed
by DCT. Here we apply the DCT transform by
multiplying each 8X8 blocks as you see in below,
the value of the transformed image is Boolean
which contains 5 to 8 decimal digits.
DCT Matrix:

ISBN: 978-1-941968-45-1 ©2017 SDIWC 73

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

0.3536 0.4904 0.4619 0.4175 0.3536 0.2778 0.1913 0.0975

0.3536 0.4157 0.1613 -0.0975 -0.3536 -0.4904 -0.4619 -0.2778
0.3536 0.2778 -0.1913 -0.4904 -0.3536 0.0975 0.4619 0.4175
0.3536 0.0975 -0.4619 -0.2778 0.3536 0.4175 -0.1913 -0.4904
0.3536 -0.0975 -0.4619 0.2778 0.3536 -0.4175 -0.1913 0.4904
0.3536 -0.3778 -0.1913 0.4904 -0.3536 -0.0975 0.4619 -0.4175
0.3536 -0.4175 0.1913 0.0975 -0.3536 0.4904 -0.4609 0.2778
0.3536 -0.4904 0.4619 -0.4175 0.3536 -0.2778 0.1913 0.0975
Table 1: DCT Coefficients
Image Block:
255 255 255 255 255 255 255 255
255 255 255 255 255 255 255 255
255 255 255 255 255 255 255 255
255 255 255 255 255 255 255 255
255 255 255 255 255 255 255 255
255 255 255 255 255 255 255 255
255 255 255 255 255 255 255 255
255 255 255 255 255 255 255 255
As you see in below, the value of the transformed image is Boolean which contains 5 to 8 decimal digits.

721.24891 721.24891 721.24891 721.24891 721.24891 721.24891 721.24891 721.24891

57.0000 57.0000 57.0000 57.0000 57.0000 57.0000 57.0000 57.0000
56.9999 56.9999 56.9999 56.9999 56.9999 56.9999 56.9999 56.9999
57.0000 57.0000 57.0000 57.0000 57.0000 57.0000 57.0000 57.0000
56.9999 56.9999 56.9999 56.9999 56.9999 56.9999 56.9999 56.9999
57.0000 57.0000 57.0000 57.0000 57.0000 57.0000 57.0000 57.0000
56.9999 56.9999 56.9999 56.9999 56.9999 56.9999 56.9999 56.9999
57.0000 57.0000 57.0000 57.0000 57.0000 57.0000 57.0000 57.0000

Table 2: The Transformed Image Values

Data Processing transferred into 2D matrix. The data will be

At the second phase the text data are getting
processed. It can be any kind of data that is related
to the patient. Personal information, medical records
such as lab result, radiology reports and etc. are part
of the Electronic Health Record (EHR) Data being
transferred to matrix in the ASCII Code formats.
The matrix dimension depends on the Hilbert-curve
round that we choose. Here we have used Hilbert
L4 (3rd round), so the matrix is 8X8.

S O H E I L S H
A H R Y A R N E
Z A K A T 2 2 0
2 1 9 9 0 0 0 9
8 9 1 2 1 2 3 4
5 6 7 T E H R A
N T E H R A N S
H E M I R A N I
Table 3: Patients’ Information

ISBN: 978-1-941968-45-1 ©2017 SDIWC 74

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

83 79 72 69 73 76 83 72
65 72 82 89 65 82 78 69
90 65 75 65 84 50 50 48
50 49 57 57 48 48 48 57
56 57 49 50 49 50 51 52
53 53 52 53 52 53 53 53
78 84 69 72 82 65 78 83
55 56 54 55 56 54 55 56
Table 4: Patient’s Information Converted to ASCII

After transferring into the matrix and convert the

characters into the ASCII codes, the scrambling
take place as part of security.

83 79 72 65 90 520 49 65
75 57 57 65 89 82 72 69
73 65 82 76 83 72 69 78
50 48 57 48 48 50 84 48
49 52 53 50 51 52 53 53
78 83 56 55 54 65 82 56
55 54 69 72 53 50 49 52
53 58 56 53 78 84 56 55
Table 5: Scrambled ASCII code of Patient’s Information

Before merging them we have to encrypt our whereas you can keep more digits to increase the
scrambled data. As I mentioned earlier, for the reversed picture quality.
encryption key we are using the 512-Digit iris code. We add the second matrix values at end of the first
We are not storing the key, because the key is matrix diagonal values.
always carried by the patient (IRIS). In order to
reduce size of iris code, we can use reversible
cellular automata or the MD-5 hash algorithm.
With the MD-5 or reverse cellular automata, the
512-digit iris code reduce to 32-digit. With this 32-
digit code we encrypt the data matrix by XOR each
ASCII code with each digit of the code.
Merging
Merging step is the most critical step in the Global
Medical Cloud Storage. Hence it is important that if
you do not merge them exactly the way that it must
be done, you cannot unmerge them later when you
want to retrieve the file.
Here we have two matrices. One is the converted,
transformed blue pixel layer with the Boolean
value. And the other one is data which are
converted, scrambled and encrypted with the iris
code.
We are merging the encrypted asci codes in to the
decimal part. The picture accuracy that we spoke
about is the decision that we make here about the
values of the transformed pixels. In this paper we
have decided to keep 5 digits in decimal part,
ISBN: 978-1-941968-45-1 ©2017 SDIWC 75
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

721.24891 721.24891 721.24891 721.24891 721.24891 721.24891 721.24891 721.24891083

57.0000 57.0000 57.0000 57.0000 57.0000 57.0000 57.0000079 57.0000
56.9999 56.9999 56.9999 56.9999 56.9999 56.9999072 56.9999 56.9999
57.0000 57.0000 57.0000 57.0000 57.0000065 57.0000 57.0000 57.0000
56.9999 56.9999 56.9999 56.9999090 56.9999 56.9999 56.9999 56.9999
57.0000 57.0000 57.0000050 57.0000 57.0000 57.0000 57.0000 57.0000
56.9999 56.9999049 56.9999 56.9999 56.9999 56.9999 56.9999 56.9999
57.0000065 57.0000 57.0000 57.0000 57.0000 57.0000 57.0000 57.0000
Table 6- The Coded Patient’s Information after merging the IRIS coding and DCT, Scrambling Values

Double compression Reversible data hiding technique enables the

The matrix which is in the size of picture
320X248, is saved in a text file. The text files
size is usually bigger than the image size, so we
compressed the text file with using the
Microsoft visual studio archiving method
(*.gz). The format includes a cyclic redundancy
check value for detecting data corruption.
Compressed GZip-Stream objects written to a
file with an extension of .gz can be
decompressed using many common
compression tools; however, this class does not
inherently provide functionality for adding files
to or extracting files from zip archives
[Microsoft].
Data retrieving method in GMCS
exact recovery of original contents upon
extraction of embedding data[9]. The retrieving
method is critical, because minor mistake can
ruin the retrieving process. The retrieving
process is exactly stepping backward the data
embedding process. For example, if the
diagonal value is 721.24891, and the encrypted
ascii code is 48, the merged value will be like:
721.24891048. Each row of the data matrix has
been merged with each diagonal values of each
block. I.e. if you have two 8X8 data, then 16
8X8 matrix blocks is needed. Because we have
16 row and each row must be merged with each
diagonal values.
The entire process of retrieving data is
illustrated in Figure 3.

Present the Removed the

data padding
characters

Present the Create and Transfer the Apply IDCT Descramble

data save the driven values the data
image matrix

Archived file Extract the Un-merge the Compress the Decrypt with Extract the
values data iris using the iris code vital values
MD5 or
cellular
automata

Figure 3-Date Retrieving Block Diagrams

Conclusions and future works: scrambling algorithm, as well as using biometric

features and the IRIS information processed through
Our algorithm is based on the information the cellular automata to obtain a high secured and
encryption which has used orthogonal transform reversible algorithm for the information
matrices for the encryption process, along with compassion, encryption, scrambling and IRIS
ISBN: 978-1-941968-45-1 ©2017 SDIWC 76
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
coding. In this paper we proposed a new Data
Encryption, Compression, along with new data
embedding algorithms that no one has ever used in
PACS and HIS .We have examined the feasibility
of Encryption and Data Embedding in the medical
cloud storage which can be used to provide the data
portability, compression, as well as the security. We
also discovered the existence of several factors
about the security and compression in the cloud
storage around the Hospital Information System.
In this paper we have shown that our model is
robust and secure and predetermined goals and
objectives have been achieved. Our main goals are
listed as following:
Being able to store all the medical records
independently but integrated.
Making the Health Records to be accessible from all
over the world, without any data loss.
Making the Electronics Health Record (EHR) to be
highly confidentiality by applying the Data
References:
1. S. Ajoudanian, M.R.A., A Novel Data Security
Model for Cloud Computing. IACSIT International
Journal of Engineering and Technology, 2012, 4.
2.A.Ross, A.K.J., Human Recognition Using
Biometrics. Appeared in Annals of
Telecommunications, Feb 2007. 62: p. 11-35.
3. A.Bessani, M.C., B.Quaresma F.Andr´e Paulo
Sousa, DEPSKY: Dependable and Secure Storage
in a Cloud-of-Clouds.
4.Moritz Borgmann, T.H., Michael Herfert, Thomas
Kunz,, and Marcel Richter, U.V., Sven Vowe, On
the Security of Cloud Storage Services. Fraunhofer
Institute for Secure Information Technology,SIT,
2012.
5. Ohyama, S., Niimi, M., Yamawaki, K., and
Noda, H., Lossless Data Hiding Using Bit-Depth
ISBN: 978-1-941968-45-1 ©2017 SDIWC
Compression, Encryption, and biometric feature
extraction along with data hiding. We can achieve
the highest level of security even though the patient
is swoon by using biometric information along with
a 512 digit of iris code.
This study has opened a wide range of research
about the electronic health record with security,
encryption, and compression while maintaining the
portability and accessibility. If we think deeply we
could easily find out that the Global Medical Cloud
Storage is the best world wide solution, but the
model need to be fitted into the different regional
requirements, because there are certain rules and
regulations that restrict some countries to exchange
patient’s information and health records. The health
integration and record exchange with worldwide
EHR solution, is the next research step which will
empower the GMCS worldwide.
Embedding for JPEG2000 Compressed Bit-Stream.
2008: p. 151-154.
6. Huang, H.-C., Lai, W.-H., and Chang, F.-C.,
Content-Adaptive Multi-level Data Embedding for
Lossless Data Hiding. 2011: p. 29-32.
7.Wu, J.-H.L.a.a.M.-Y., An Iterative Method for
Lossless Data Embedding in BMP Images.
8. Malakooti,M.V., . Khederzadeh, M., A Lossless
Secure data embedding in image using DCT and
Randomize key generator. DICTAP 2012.
9.Naheed, T., I. Usman, and A. Dar, Lossless Data
Hiding Using Optimized Interpolation Error
Expansion. 2011: p. 281-286.
10.Mishra, S., AN INTUITIVE METHOD FOR
HILBERT CURVE CODING. International Journal
of Computing and Corporate Research, 2011,1.
77
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
Virtual Local Area Network (VLAN): Segmentation and Security
Abbas Mehdizadeha Kevin Suinggia Mojtaba Mohammadpoorb Harlina Haruna
a
Dept. of Computing, Nilai University, 71800 Putra Nilai, Negeri Sembilan, Malaysia.
b
Dept. of Computer & Electrical Eng., University of Gonabad, Gonabad, Iran.
mehdiizadeh@ieee.org, kevingizxc@gmail.com, mohammadpur@gonabad.cac.ir, harlina@nilai.edu.my
ABSTRACT
A virtual local area networks (VLANs) have
recently expanded into an integral property of
switched LAN solutions from every main LAN ma-
teriel seller. One of the motives for the attention
placed on VLAN functionality now is the fast im-
plementation of LAN switching that commenced
two decades ago. Many more anxious organiza-
tions and companies are moving rapidly into net-
works featuring private port LAN switching de-
signs. VLANs demonstrate an alternate solution to
routers for broadcasting containment, since VLANs
permit switches to also possess broadcast traffic.
With the implementation of switches in continu-
ity with VLANs, each network segment can have
as few as one user, while broadcast domains can
be as big as 1,000 users or probably even more.
This paper presents what exactly a VLAN is and
how VLAN memberships are implemented in a
switched network. Membership in a VLAN can
be based on MAC addresses, port members, IP ad-
dresses, IP multicast addresses and/or a combina-
tion of these aspects. VLANs are cost effective as
well as time effective, can decrease the traffic of the
network, and give an extra security. VLANs give
upgraded system security. In a VLAN system en-
vironment, with various communicating areas, sys-
tem administrator can have control over every port
and client. A malevolent client can no more simply
connect their station to any port of switch and snif-
fer the system movement utilizing a bundle sniffer.
The system overseer controls every port and what-
ever assets it is permitted to utilize. VLANs confine
delicate movement beginning from inside an under-
taking department itself.
KEYWORDS
Local Area Network, Virtual LAN, Security, Seg-
mentation, VXLAN.
ISBN: 978-1-941968-45-1 ©2017 SDIWC
1 INTRODUCTION
The world today relies heavily on technology
to do daily work. Important information is al-
ways being transferred quickly from within a
company. The advancement of technology has
helped to enhance the way we transfer informa-
tion. The use of Virtual Local Area Network
(VLAN) is more popular now than ever. But
what is a VLAN? VLANs are simple, yet they
offer a wide variety of capabilities and options
to improve the network. VLAN is a technology
that provide dividing a physical network into
logical at Layer two. Functionally, VLANs al-
low a network administrator to partition a local
network into separate, independent networks.
VLANs are often implemented in large net-
works as well as small VLANs. In larg net-
works, VLANs are sometimes implemented to
combine physically separate LAN segments or
LANs into one logical LAN.
In this paper, we will discuss about the seg-
mentation of a VLAN, including why VLANs
should be considered in a smaller network.
We will go through the segmentation process
of the LAN to the VLAN, and the configura-
tion of a switch to form separate LAN seg-
ments. It is imperial for the company to se-
cure the network from any attackers that tries
to steal company information. Network secu-
rity technologies provide protection of the net-
work against the theft and misuse of confiden-
tial data and secure against malicious attacks
from viruses/worms. Without a security so-
lution, the company risks unauthorized intru-
sions, network downtime, service disruption,
regulatory noncompliance and even legal ac-
tion. Companies uses VLAN as a way to con-
nect the networks in their company [1][2].
An expansive scope of themes identified with
system security have been discussed in the re-
78
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
cent researches, and a decent rundown of the
system security issue have been given. Great
systems ought to work easily with different
systems, be straightforwardly to clients, give
remote get to, and keep up crest execution.
Then again, secure systems ensure private data,
keep system execution solid, and underscore
information respectability. The two measure-
ments are regularly at chances [3].
In this paper, we will discuss more on the secu-
rity of the VLAN. We will indulge further into
the discussion of the different types of attack
such as ARP poisoning and VLAN hopping on
the VLAN and how we can use different ways
of preventing such attacks such as the use of
static ARP entries. In section 2, we will discuss
the related work that focuses on the segmenta-
tion and the security of the VLAN and the ben-
efits of the VLAN. In the next subsection 2.1,
we will examine the benefits of using VLANs.
We will be viewing the benefit in terms of the
scalability, cost, ease of use, integrity, virtual
work group and security. In subsection 2.2, we
will be going more in depth about the segmen-
tation of the VLAN and see how it differs from
the traditional LAN. In subsection 2.3, we will
examine about VXLAN, how it differs from
VLAN and its benefits. Then in subsection 2.4,
we will cover more about the security of the
VLAN where we will examine the network at-
tacks on a VLAN and the existing methods that
exist for protection against them. In section 3,
we will examine about the future works that
are in stored for the improvement of VLAN us-
age such as VXLAN. Lastly, we will conclude
about the VLAN segmentation and security.
2 RELATED WORK
In this section, we will be focusing on the seg-
mentation and the security of the Virtual Local
Area Network (VLAN)[4][5].
The Internet, the largest network, is the global
system of interconnected computer networks
that use the Internet protocol suite (TCP/IP) to
connect billions of computers and electronics
devices worldwide. It consists of millions of
private, public, academic, business, and gov-
ernment sector networks of local to global
ISBN: 978-1-941968-45-1 ©2017 SDIWC
scope. It carries an extensive and wide range
of information resources and services, such as
the World Wide Web (WWW), electronic mail,
telephony, voice and video over IP [6], and file
sharing using peer-to-peer networks [7].
As the internet is the network of networks.
We then look into the different type of net-
works. The LAN interconnects computers and
devices within a limited area such as a resi-
dence, campus, school, laboratory, or office.
On the other hand, there is a Wide Area Net-
work (WAN), which covers a larger geographic
distance compared to LAN. A network larger
than LAN and smaller than WAN could be con-
sidered as Metropolitan Area Network (MAN),
covering an area of a few blocks of a city to the
area of an entire city [8].
If LAN uses physical administration to create
a network, VLAN was created by using logical
networks to divide a physical switch and sep-
arate hosts that are not supposed to access to
each other. A VLAN allows the creation of dif-
ferent networks on a single physical switch at
the data link layer [9]. To subdivide a network
into VLANs, one configures a network switch
or router. This allows different departments
in the company to have different networks on
a single physical switch. This saves cost as
only a single physical switch is needed, and it
can absolutely simplify network implementa-
tion and design, as it can be configured through
software rather than hardware. VLANs allow
network administrator to group hosts together
even if the hosts are not on the same network
switch. Without VLANs, grouping hosts will
need to relocate the nodes or rewire the data
links [10]. VLAN allows the flexibility for
changes should there be a need to reconfigure
the network [11]. However, VLANs are still
vulnerable to network attacks.
From protecting user data against the grow-
ing number of threats to ensuring the conti-
nuity of the business, IT Security is an es-
sential element in any organization IT infras-
tructure. As IT professionals being able to
benchmark against our peers, assess a threat, or
just having some understanding of why a secu-
rity project is important to the business is key
79
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
[12]. Many techniques to address this issue has
been discovered. However, these techniques
will require improvement over the years as net-
work attackers are getting better in attacking a
VLAN [1][2].
2.1 Benefits of VLAN
The benefits of VLAN will be dsicussed in this
section [13][14].
Scalability The ability to add, move and
change networks are attained with low cost and
less effort by just configuring a switch port into
the segmented VLANs and assigning clients to
the different VLANs.
Security VLAN is able to provide a more se-
cure environment, as network administrators
can have control over each port. A malicious
user can no longer just plug their device into
any switch port and sniff or steal the network
traffic using a sniffer software without getting
detected. The port and whatever resources it is
allowed to use could be controlled by network
administrator. The sensitive traffic originating
from an enterprise department within itself can
be restricted by VLANs.
Cost Saving The cost of creating a network
or expanding a network can be decreased by
eliminating the need for additional expensive
network equipment like longer cables or extra
routers. VLANs will provide the ability to use
better bandwidth and resources and therefore
the network can work more efficiently.
Easy Troubleshooting Network administra-
tors are able to observe the activities of the
VLANs more easily. Therefore network prob-
lems can be easily traced and identified and
rectified.
Integrity The network is segmented logi-
cally which divide a physical switch and sep-
arate hosts that are not supposed to access to
each other. This ensure that data is not com-
promised when handled.
ISBN: 978-1-941968-45-1 ©2017 SDIWC
Virtual Work Group Another benefits of
VLAN is to create virtual workgroups. For in-
stance, co-workers from different departments
that is working on a part of big project or pos-
sibly same project can send message from one
another without having to be in the same de-
partment. This can help in reducing traffic in
the network.
2.2 Segmentation
Larger network is broken into smaller sections
by implementing VLANs. This is called as
segmentation of the network which allows eas-
ier management of the network. A LAN is
a combination of computers and devices con-
nected to each other in small area to interoper-
ate and share resources.
Routers look at the network address of packets
and use different routing protocols to send the
packet to its destination efficiently. Switching
could reduce the number of nodes by using the
same network segment, resulting in lower con-
gestion on each segment. In switched hubs or
bridges, each node can have its own network
segment, and therefore have access to all of the
network bandwidth of the segment. Switching
bridges can look deep into a packet and use
protocol information and the like to provide a
level of filtering and prioritization [15][16].
For the segmentation section, we also have
divided it into subsections. In subsection
2.2.1, we have the types of VLAN member-
ship where it explains how the VLAN are as-
signed. In subsection 2.2.2, we view the types
of VLANs and how they differ from each other
by the specific function they are used for. We
then move to the VLAN operation in subsec-
tion 2.2.3, this is where we show how the op-
eration work so that each switch port can be
assigned to a different VLAN. In subsection
2.2.4, we examine how we can identify the
VLAN and what ports are used. We then go
more in depth to view the methods used by
VLAN for identification. Lastly for this sec-
tion, we view the routing traffic between the
VLANs.
80
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
2.2.1 VLAN Membership
There are two types of VLAN memberships:
Static VLANs Static VLANs are configured
by network administrator, mainly because of
security reason . Since a is assigned manually
to a VLAN will be always finxed and main-
tained. This type of mebership is of course
easy to set up and configure but the manual
update is required if any changes in the host.
Static VLANs are n ot feasible to be imple-
mented for a large network which required fre-
quent updates. In this case a dynamic solution
is suggested.
Dynamic VLANs VLANs can be assigned
automatically using software, based on hard-
ware address (MAC), protocols and applica-
tions. For instance, assume that a MAC ad-
dresses has been listed into centralized VLAN
software. If it is attached to an unas-
signed switch port, the management database
of VLAN can look up for the hardware ad-
dress and assign and configure the switch port
into the correct VLAN. The difficulty of this
method is to setup the database at the initial
level [17].
2.2.2 Types of VLANs
In modern networks, there are a number of dif-
ferent types of VLANs. Some of them can be
explained and classified based on their traffic
classes. The other VLAN types could be de-
fined by the particular function that they serve.
Data VLAN A data VLAN could be used
and configured to bear with the traffic gener-
ated by a user. It would not include a VLAN
carrying voice or management traffic. It is
common practice to distinguish voice traffic
and management traffic from data traffic. It is
sometimes referred to as a user VLAN. These
VLANs are developed to separate the network
into groups of users or groups of devices.
Default VLAN When the default configura-
tion is loaded at the initial bootup, all the ports
ISBN: 978-1-941968-45-1 ©2017 SDIWC
of a switch become a part of the default VLAN.
These switch ports that are now part of de-
fault VLAN, in fact they are part of the same
broadcast domain. It means any device con-
nected to any switch port is allowed to com-
municate with other devices on other switch
ports. VLAN 1 is considered as default VLAN
for Cisco switches.
Native VLAN A native VLAN is defined
to an 802.1Q trunk port which considered as
the links between switches to provide the traf-
fic transmission associated with more than a
VLAN. It supports traffic coming from many
VLANs, as well as traffic that does not come
from a VLAN, which considered as tagged and
untagged traffic, respectively. The trunk port
(802.1Q) places untagged traffic on the default
VLAN 1 which is known as native VLAN. Na-
tive VLANs are specified to maintain back-
ward compatibility with untagged traffic com-
mon to legacy LAN scenarios. It is a best prac-
tice to configure the native VLAN as an un-
used VLAN, different from VLAN 1 and other
VLANs. It can be dedicated as a fixed VLAN
to serve the role of the native VLAN for all
trunk ports in the switched domain.
Management VLAN Any VLAN that is
configured to access the management capabil-
ities of a switch is considered as management
VLAN which by default is VLAN 1. By as-
signing an IP address and subnet mask to the
Switch Virtual Interface (SVI), the manage-
ment VLAN can be created and managed via
HTTP, Telnet, SSH, or SNMP.
It is not a good practice to choose VLAN 1
as management VLAN (which is by default)
because the out-of-the-box configuration of a
Cisco switch. In the past, the management
VLAN for a 2960 series switch was the only
active SVI. On 15.x versions of the Cisco IOS
for Catalyst 2960 switches, more than one ac-
tive SVI will be possible. However, having
more than one management VLAN which the-
oretically a switch can have, will give oppor-
tunity to network attackers. There is a risk if
the native VLAN is the same as the manage-
81
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
ment one. Therefore, the native VLAN should
be distinct from any othe VLANs if it is used.
Voice VLANs To support Voice over IP
(VoIP), a different VLAN is required which
can be called as a voice VLAN. VoIP traffic
needs the followings:
• Assured bandwidth to provide acceptable
voice quality
• Priority of transmission over other types
of network traffics
• Routing capability around congested ar-
eas on the network
• Lesser delay (less than 150ms) across the
network
These requirements need to be achieved to sup-
port VoIP. However, the configuration of these
requirements is beyond the scope of this paper,
but it is useful to briefly discuss on how a voice
VLAN works between a switch, a computer,
and a Cisco IP phone.
2.2.3 VLAN Operation
Figure 1. VLAN Operation.
While each switch port could be associated to
a separate VLAN, the ports associated to the
same VLAN share broadcasts. Once a device
enters the network, it automatically considers
the VLAN membership of the port it is at-
tached to. For a host to be a part of any VLAN,
it must be given an IP address that belongs to
the appropriate subnet.
ISBN: 978-1-941968-45-1 ©2017 SDIWC
2.2.4 Identifying VLANs
A port on a switch could be associated to only
one VLAN or to all VLANs. A port could be
configured manually as an access or trunk port.
Let the Dynamic Trunking Protocol (DTP) op-
erates on a per-port basis to set the switch port
mode. It can be done by negotiating with the
port on the other end of the link [17]. There
are two different types of links in the switched
network:
i) Access Ports: An access port normally car-
ries the traffic of only one VLAN. In this case,
traffic is both sent and received in native for-
mats without VLAN tagging. Anything arriv-
ing on an access port is simply considered to
belong to the VLAN assigned to the port. Any
device connected to an access link is not aware
of a VLAN membership; the device just as-
sumes its part of the same broadcast domain
and doesn’t recognize the physical network
topology. Access-link devices cannot send and
receive data to and from devices outside their
VLAN unless the routing is configured. It can
only make a switch port to be either an access
port or a trunk port but not both. It must be
noted that the access port can only be attached
to one VLAN only [17].
ii) Trunk Ports: Trunk ports on the other hand
is able to carry multiple VLANs at a time. A
trunk link is 100 or 1000 Mbps point-to-point
link between two switches, switch and router,
or even between a switch and server, and it car-
ries the traffic of multiple VLANs from 1 to
4094 at a time. This is a great functionality
because ports can be set up to have a server
in two separate broadcast domains at the same
time, so the users will not have to cross a net-
work layer (layer 3) device to log in and access
it. The other benefit is that trunk links are able
to carry various amounts of VLAN data across
the link [17].
2.2.5 VLAN Identification Method
VLAN identification is where switches can
keep track of all frames as they are traveling
through a switched network. It defines how
switches can identify which frames belong to
82
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
which VLANs where there is more than one
trunking method.
i) Inter-Switch Link (ISL): Inter-Switch Link
(ISL) is a method that tag VLAN data onto
an Ethernet frame. This data tagging per-
mits VLANs to be multiplexed over a trunk
through an external encapsulation method
(ISL). In fact, it allows the switch to recog-
nize the VLAN membership of a frame over
the trunked link. By implementing ISL, inter-
connection of multiple switches can be done
and still maintain VLAN information as traffic
travels ; between switches on trunk links. ISL
operates at layer two by encapsulating a data
frame with a new header and Cyclic Redun-
dancy Check (CRC). It is used for Fast Ether-
net and Gigabit Ethernet links only. ISL rout-
ing is versatile and can be used on a switch
port, router interface and server interface cards
to trunk a server [17].
ii) IEEE 802.1Q: it is a standard method cre-
ated by IEEE for frame tagging, IEEE 802.1Q
inserts a field into the frame to identify the
VLAN. If trunking between a Cisco switched
link and a different brand of switch are need
to be done, 802.1Q must be used for the
trunk to work. The basic purpose of ISL and
802.1Q frame-tagging methods is to provide
inter-switch VLAN communication. Also, it
should be noted that any ISL or 802.1Q frame
tagging is removed if a frame is forwarded out
an access link; tagging is used across trunk
links only [17].
2.2.6 Routing between VLANs
Routing between VLANs (Inter-VLAN rout-
ing) is based on forwarding network traffic
from one VLAN to another VLAN using a
router. It allows devices connected to the dif-
ferent VLANs to communicate with each other
by using a router. Nodes in a VLAN stay in
their own broadcast domain and will be able
to communicate freely. VLANs can make net-
work partitioning and traffic separation at layer
two, data link layer. Therefore, if hosts or
any other IP addressable device want to com-
municate between VLANs, a layer 3 device is
needed to provide routing services. Using vir-
ISBN: 978-1-941968-45-1 ©2017 SDIWC
tual technology and protocols to segment a net-
work can be useful to control broadcast traf-
fic and implement security boundaries. How-
ever, allowing absolutely NO access between
VLANs is never very beneficial. To solve this
problem, implementation of inter-VLAN rout-
ing is suggested. At; the CCNA level, there are
two ways we can make this happen namely:
a) Connect a unique router port to each VLAN
b) Create a router on a stick.
2.3 Virtual Extended Local Area Network
(VXLAN)
Now that we have learned that traditional
network segmentation has been provided by
VLANs that are standardized under the IEEE
802.1Q group. VLANs provide logical seg-
mentation of Layer 2 boundaries or broadcast
domains. However, due to the inefficient use of
available network links with VLAN use, rigid
requirements on device placements in the data
centre network, and the limited scalability to a
maximum of 4094 VLANs, using VLANs has
become a limiting factor to IT departments and
cloud providers as they build large multitenant
data centres.
In this section, we will discuss about the
VXLAN standard which Cisco, in partnership
with other leading vendors proposed IETF as
a solution to the data centre network chal-
lenges posed by traditional VLAN technology.
The VXLAN standard provides for the elastic
workload placement and higher scalability of
Layer 2 segmentation that is required by to-
day’s larger application demands.
2.3.1 VXLAN Benefits
VXLAN is proposed to provide the same Eth-
ernet Layer two network services as VLAN
does today, but with greater flexibility and ex-
tensibility. The following benefits are offered
by VXLAN as compared to VLAN [18]:
1. Flexible placement of multitenant segments
throughout the data centre gives a way to ex-
tend the Layer two segments over the underly-
ing shared network infrastructure. Therefore,
tenant workload can be put among physical
83
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
pods in the data centre.
2. Higher scalability to address more Layer
two segments compared to VLANs, which
results in limiting scalability of only 4094
VLANs. VXLAN uses a 24-bit segment
ID known as the VXLAN Network Identi-
fier (VNID). This enables up to 16 million
VXLAN segments to coexist in the same ad-
min domain.
3. Better utilization of network paths available
in the underlying infrastructure where VLAN
uses the Spanning Tree Protocol for loop pre-
vention. It should noted that it uses half of
the network links in a network by blocking re-
dundant paths. In contrast, VXLAN packets
are transferred through the underlying network
based on its Layer 3 header and can take com-
plete advantage of Layer 3 routing, Equal Cost
Multi Path (ECMP) routing, and link aggrega-
tion protocols to be able to use all available
paths.
2.4 Security
A broad range of topics related to network se-
curity has been discussed, and a good sum-
mary of the network security problem has
been provided. Good networks should operate
smoothly with other networks, be transparently
to users, provide remote access, and maintain
peak performance. On the other hand, secure
networks protect confidential information, pro-
vide reliability, and give data integrity [19].
The two dimensions are often at odds [3].
2.4.1 Network Attacks on a VLAN
With the increasing advancement in technol-
ogy, the need for network security is also in-
creasing. This is because as technology ad-
vance, it also increases the wide range of hack-
ing tools availability. These hacking tools are
used on different type of networks where we
have the Wide Area Network and the Local
Area Network [20].
While VLAN is important in maintaining a
flexible networking [11], it also raises se-
curity issues as corporations keep important
data that is being transferred around using the
ISBN: 978-1-941968-45-1 ©2017 SDIWC
VLAN. Network security issues on the VLAN
[21][22][23] is very important and should be
considered, discussed and analyzed. Here we
consider some of the more common network
attacks on a VLAN.
Address Resolution Protocol (ARP) Spoofing
Attacks:
ARP spoofing, ARP cache poisoning, or ARP
poison routing, is a method that an attacker
sends (spoofed) Address Resolution Protocol
(ARP)[24][25] messages onto a local area net-
work. The main aim of an attacker is to asso-
ciate the its MAC address with the IP address
of another node, such as the default gateway.
Therefore, the attacker is able to cause any traf-
fic meant for that particular IP address to be
sent to the attacker instead [20]. As shown in
figure 3, the attacker sends fake packets which
has similar IP address to the original IP address
to the server claiming to be the genuine host.
When the server receives the packet and con-
siders the MAC address of the attacker to be
the intended destination for the packets since
the attacker is using a similar IP address as the
original host, it starts sending the data to the
attacker instead. The attacker can receive data
intended for the genuine recipient. ARP spoof-
ing may allow an attacker to intercept data
frames on a network, modify the traffic, or stop
all traffic [26].
Figure 2. Routing under normal operation.
LAN Hopping:
VLAN hopping is considered as a computer se-
curity exploit, where an attacker can attack net-
worked resources on a VLAN. The basic con-
cept behind all these kind of attacks is for at-
tack host on a VLAN to gain access to traffic
on other VLANs that are not supposed to be
84
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
Figure 3. ARP attack example with the malicious user
as the man-in-the-middle.
accessible. In a VLAN hopping attack, the at-
tacker will send tagged packets with the VLAN
ID of the targeted VLAN. The VLAN ID will
then be checked and pass the packet to the tar-
geted VLAN [20]. In figure 4, the attacker at-
tacks the VLAN through the trunk of the net-
work. The packet sent by the attacker will con-
tain the VLAN ID of the targeted VLAN.
Figure 4. Attacking the VLANs through the trunk of the
network.
MAC Flooding:
MAC flooding is a method implemented to
collude the security of network switches. In
this attack, the attacker will send many dif-
ferent fake source MAC addresses [20]. The
purpose of attacker is to consume the limited
memory set aside in the switch to store the
MAC address table. Traffic is then flooded
and becomes accessible as the switch can no
longer keep the specific destination address in
the memory. Since the traffic is flooded, the at-
tacker can then simply see the traffic generated
and sniff out the information. Figure 5 shows
the attacker sending various MAC addresses to
the switch with the intention to limit the MAC
address table of the switch.
ISBN: 978-1-941968-45-1 ©2017 SDIWC
Figure 5. MAC address flooding.
2.4.2 Existing Methods for Protection
against Network Attacks on the
VLANs
In this section, we will discuss about the dif-
ferent types of method used to protect against
network attacks on the VLAN.
Static ARP Entries:
One of the methods to prevent network attacks
is the use of static ARP entries into the ARP ta-
ble on the switch. IP address-to-MAC address
mappings in the local ARP cache might be stat-
ically entered so that nodes decline ARP reply
packets. While static entries provide some se-
curity against ARP spoofing attacks, this re-
sults in high maintenance efforts as address-
mappings of all systems in the network have to
be done statically one by one. A static Address
Resolution Protocol (ARP) entry is a constant
entry in the ARP cache. A static ARP entry
can be managed from a node/device or a work-
station. It is not regularly used under normal
circumstances, however, it is used when add
or delete of an entry from the cache is needed.
The following command could be used to add
a static ARP cache entry:
C:\>arp -s 192.168.1.17 6c-fc-03-a3-7f-81
The arp command makes a static entry in ARP
cache, to launch a communication session with
the node that has a 192.168.1.17 IP address,
it is not needed to start the process with an
ARP request as the target node’s MAC address
is already known. If a similar ARP entry has
not been added to the target node/host, the tar-
get host needs to send an ARP request to the
computer to find out the MAC address. After
85
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
adding the static ARP entry, the ARP cache on
the computer may look like the following:
C:\>arp -a
Interface: 192.168.1.137 --- 0x50006
Internet Address Physical Address Type
192.168.1.17 6c-fc-03-a3-7f-81 static
192.168.1.254 77-d8-e5-f2-43-6d dynamic
The connection is up with the node at
192.168.1.17 until the MAC address of the tar-
get computer changes, which could be because
of a change of network card, or probably an
operation changes the MAC address. When
this happens, it is needed to delete the invalid
ARP entry with an arp -dcommand, such as
arp -d 192.168.1.17.
Cisco router keeps ARP entries in the cache
for four hours (240 minutes), while Windows
workstations can only keep for a maximum of
ten minutes. This is common on routers be-
cause they tend to spend most of their time
dealing with the same nodes. A router is nor-
mally configured as a default gateway for de-
vices of the network, which is why they see the
same nodes communicating with that for most
of the time in a day, and as long as those nodes
keep sending data through the router, they will
remain in the ARP cache. For a router con-
nected to large network segments, this would
result in a rather large ARP listing or ARP ta-
ble. More of the router’s memory will be con-
sumed with a large ARP table, so the caching
time that Cisco has chosen was the result of
memory consumed by the ARP cache against
the ARP’s need for fresh MAC information. To
create a static ARP entry for a router, it can be
done by entering Global Configuration mode,
which the arp command looks like this:
#arp 192.168.1.17 6cfc.03a3.7f81 arpa
After entering this command, the ARP cache
contains the IP-MAC address pair, which
would not age-out of the cache. This can be
seen by the dash in the Age column. Static
ARP entries are not usually recognized to an
interface like the dynamic entries are.
Router\#show arp
Prot. Address
Int. 192.168.1.1
Age(min) Hardware Addr
- 0050.43bf.7c82
Int. 192.168.1.17 - 6cfc.03a3.7f81
ISBN: 978-1-941968-45-1 ©2017 SDIWC
If the entry is no longer needed, or if it is
needed to be changed to something else, the
no arp command removes the original entry:
Router(config)\#no arp 192.168.1.17
If two nodes communicate with each other
constantly throughout the day, the static ARP
would be added. By adding static ARP entries
for both systems in each other’s ARP cache,
some network overhead will be reduced, in the
form of ARP requests and ARP replies. It is
also good for prevention against flooding at-
tacks where the ARP cache is being flooded
with random entries. Static ARP could help to
find out which entries are allowed and which
should be dropped.
Ingress Filtering:
The use of ingress filtering is a method of en-
suring that incoming packets are actually from
the networks which they claim to originate.
The switch is configured with ingress filter-
ing to accept only the allowed packets. Any
router that deploys ingress filtering, checks the
source IP field of IP packets it receives, and if
the packets do not have an IP address in the
IP address block, the packets will be dropped.
However, addresses can be faked and ingress
filtering will still accept the packets should the
same allowed address by the ingress filtering
be used by the attacker [20]. Ingress filtering is
to confirm whether inbound packets arriving at
a network are from the source that they claim
to be from before entry (or ingress) is given or
not.
It takes the advantage of the Layer two IP-
address filtering capability of a router at the
network’s edge and if it has a high probability
of being malicious, the traffic will be blocked.
At its simplest, ingress filtering involves estab-
lishing an access control list. This list contains
the IP addresses of permitted source addresses.
Conversely, the access control list may also be
used to block prohibited source addresses. The
following source IP addresses will be blocked
by ingress filtering:
Type
ARPA
• Already in use IP address, IP address
ARPA within the internal network. By blocking
86
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
the source IP, attacker from spoofing an
internal IP address to take advantage of a
poorly written firewall rule could be pre-
vented.
• Private IP addresses. By blocking this
addresses, malicious traffic coming in
from an improperly configured Internet-
based host or an attacker’s spoofed ad-
dress could be prevented.
• loopbacks IP addresses. If the loopback is
spoofed, this helps to prevent this type of
traffic.
• multicast addresses[27]. Blocking multi-
cast addresses could help to prevent un-
desired multicast traffic that seems to be
spam.
• Service or management network IP ad-
dresses. The attacker is not bale yo use
the public Internet to gain unauthorized
access to network services running at the
network application layer and above.
The traffic from specific regions of the world
could be whitelisted by network admin, and
can be blacklisted to not allow a specific region
to have access to its environment. Some free
subscription-based services could be found to
create access control lists for network border
routers.
3 VXLAN ENHANCEMENTS
VXLAN has a higher scalability to address
more Layer 2 segments. VLANs uses a 12-
bit VLAN ID to address Layer 2 segments,
which results in limiting scalability of only
4094 VLANs. VXLAN uses a 24-bit seg-
ment ID known as the VXLAN Network Iden-
tifier (VNID), which enables up to 16 million
VXLAN segments to coexist in the same ad-
ministrative domain.
We now discuss the VXLAN encapsulation
and packet format. VXLAN is a Layer 2
overlay scheme over a Layer 3 network. It
uses MAC Address-in-User Datagram Proto-
col (MAC-in-UDP) encapsulation to provide a
means to extend Layer 2 segments across the
data centre network. VXLAN is a solution to
support a flexible, large-scale multitenant en-
vironment over a shared common physical in-
ISBN: 978-1-941968-45-1 ©2017 SDIWC
frastructure. The transport protocol over the
physical data centre network is IP plus UDP.
VXLAN defines a MAC-in-UDP encapsula-
tion scheme where the original Layer 2 frame
has a VXLAN header added and is then placed
in a UDP-IP packet. With this MAC-in-UDP
encapsulation, VXLAN tunnels Layer 2 net-
work over Layer 3 network. VXLAN intro-
duces an 8-byte VXLAN header that consists
of a 24-bit VNID and a few reserved bits. The
VXLAN header together with the original Eth-
ernet frame goes in the UDP payload. The 24-
bit VNID is used to identify Layer 2 segments
and to maintain Layer 2 isolation between the
segments. With all 24 bits in VNID, VXLAN
can support 16 million LAN segments.
VXLAN uses VXLAN Tunnel EndPoint
(VTEP) devices to map tenants’ end devices to
VXLAN segments and to perform VXLAN en-
capsulation and de-encapsulation. Each VTEP
function has two interfaces: One is a switch in-
terface on the local LAN segment to support
local endpoint communication through bridg-
ing, and the other is an IP interface to the
transport IP network. The IP interface has a
unique IP address that identifies the VTEP de-
vice on the transport IP network known as the
infrastructure VLAN. The VTEP device uses
this IP address to encapsulate Ethernet frames
and transmits the encapsulated packets to the
transport network through the IP interface. A
VTEP device also discovers the remote VTEPs
for its VXLAN segments and learns remote
MAC Address-to-VTEP mappings through its
IP interface. The functional components of
VTEPs and the logical topology that is created
for Layer 2 connectivity across the transport IP
network is shown in Figure 6.
The VXLAN segments are independent of the
underlying network topology; conversely, the
underlying IP network between VTEPs is in-
dependent of the VXLAN overlay. It routes
the encapsulated packets based on the outer IP
address header, which has the initiating VTEP
as the source IP address and the terminating
VTEP as the destination IP address [28][29].
87
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
Figure 6. The functional components of VTEPs.
4 CONCLUSION
As technology is advancing and improving at
a high rate on a daily basis, more methods of
managing the network of these technologies
are being developed. Since there are millions
of network all around the globe, one of the
special methods to manage these networks is
the creation of logical addressing. One way to
manage the networks is the creation of phys-
ical way of addressing which is called Local
Area Networking (LAN). To address the issue
with handling many networks, logical address-
ing was created where component only needed
to be in the same sub network to interact with
each other.
With the wide usage of VLAN, there is a con-
cern of security of the network, as well as scal-
ability and network management, which have
been discussed in this paper. As sensitive data
is broadcasted on a network, there are sev-
eral risks and threats to the network. VLANs
can minimize this threat by placing only those
users on the network data on a VLAN with ac-
cess. This will reduce chances of an intruder
gaining access. With the implementation of
VLAN we can also have control of broadcast
domains, setting up firewalls, prohibition of ac-
cess and alerting a network manager in case of
an attack by an outsider. In this paper, we can
conclude that utilization of virtual local area
networks can surely simplify network manage-
ment and also provide networks with improved
security.
ISBN: 978-1-941968-45-1 ©2017 SDIWC
REFERENCES
[1] Cisco, “What Is Network Security? - Cisco Sys-
tems”, 2016.
[2] Cisco, “Solutions - Cisco Systems, What You
Need to Know about Network Security”, 2016.
[3] Baker, Richard H., “Network Security: How to
Plan for it and Achieve it”, McGraw-Hill, Inc.,
1995.
[4] Catalyst 4500 Series Switch Cisco IOS Software
Configuration Guide, 12.2(25)EW, “Understand-
ing and Configuring VLANs [Cisco Catalyst 4500
Series Switches]”, 2016.
[5] Surabhi Surendra Tambe, “Understanding Vir-
tual Local Area Networks”, International Journal
of Engineering Trends and Technology (IJETT),
Vol.25 No.4, 174-176, 2015.
[6] Sun, Lingfen, Is-Haka Mkwawa, Emmanuel Jam-
meh, and Emmanuel Ifeachor, “Guide to voice and
video over IP: for fixed and mobile networks”,
Springer Science & Business Media, 2013.
[7] Tsimonis, G., & Dimitriadis, S., “Brand strategies
in social media. Marketing Intelligence & Plan-
ning”, 32(3), 328-344, 2014.
[8] Van Heddeghem, W., Lambert, S., Lannoo, B.,
Colle, D., Pickavet, M., & Demeester, P., “Trends
in worldwide ICT electricity consumption from
2007 to 2012”, Computer Communications, 50,
64-76, 2014.
[9] Altunbasak, Hayriye, and Henry Owen, “An archi-
tectural framework for data link layer security with
security inter-layering”, Proceedings IEEE South-
eastCon, 2007.
[10] Wilkins, S., “Virtual vs. Physical LANs: Device
Functionalities”, Pearson IT Certification, CCNA
Routing and Switching 200-120 Network Simula-
tor, 1st Edition, 2015.
[11] Nishino, H., Nagatomo, Y., Kagawa, T., & Hara-
maki, T., “A Mobile AR Assistant for Cam-
pus Area Network Management”, In IEEE 2014
Eighth International Conference on Complex, In-
telligent and Software Intensive Systems (CISIS),
pp.643-648, 2014.
[12] Shepard, D., “84 Fascinating & Scary IT Secu-
rity Statistics”, 2015 CYBERTHREAT DEFENSE
REPORT, 2015.
88
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

[13] Haq, Ul, Syed Ehtesham, and Suraiyaa Parveen,
“Implementation of network architecture, its secu-
rity and performance analysis of VLAN”, Interna-
tional Journal of Advanced Research in Computer
Science, 8, no. 7, 2017.
[14] Nguyen, Van-Giang, and Young-Han Kim, “SDN-
Based Enterprise and Campus Networks: A Case
of VLAN Management”, Journal of Information
Processing Systems, 12, no. 3, 2016.
[15] Derfler FJ, Freed L, Douglas P, Robbins L, Adams
S., “Illustrator-Troller M. How networks work”,
Que Corp, 2000.
[16] Henry, Paul David, “Strategic networking: From
LAN and WAN to information superhighways”,
Coriolis Group, 1996.
col”, 19th IEEE Annual Computer Security Appli-
cations Conference, 2003.
[25] Cisco, “Configuring the Address Resolution Proto-
col (ARP), Cisco Content Services Switch Routing
and Bridging Configuration”, 2004.
[26] Zargar, S. T., Joshi, J., & Tipper, D., “A survey of
defense mechanisms against distributed denial of
service (DDoS) flooding attacks”, IEEE commu-
nications surveys & tutorials, 15(4), 2046-2069,
2013.
[27] Mehdizadeh, A., Abdullah, R. S. A. R., Hashim,
F., Ali, B. M., Othman, M., & Khatun, S., “Reli-
able key management and data delivery method in
multicast over wireless IPv6 networks”, Wireless
personal communications, 73(3), 967-991, 2013.

[17] Pal, G. Prakash, and Gyan Prakash Pal, “Virtual [28] Cisco, “VXLAN Overview: Cisco Nexus 9000 Se-
Local Area Network (VLAN)”, International Jour- ries Switches”, 2016.
nal of Scientific Research Engineering & Technol-
ogy (IJSRET), 1: 006-010, 2013. [29] Arista, “VXLAN: Scaling Data Center Capacity”,
2016.
[18] Kapadia, Shyam, Puto H. Subagio, Yibin Yang,
Nilesh Shah, Vipin Jain, and Ashutosh Agrawal,
“Implementation of virtual extensible local area
network (VXLAN) in top-of-rack switches in a
network environment”, U.S. Patent 9,565,105,
2017.

[19] Mehdizadeh, A., Abdullah, R. S. A. R., & Hashim,

F., “Secure group communication scheme in wire-
less IPv6 networks: An experimental test-bed”, In
IEEE International Symposium on Communica-
tions and Information Technologies (ISCIT), (pp.
724-729), 2012.

[20] Riddell, A., Darrell, A., Drewett, C., Sangster, W.,

“LAN Security Guide”, First Edition, ISBN: 978-
0-473-19020-0, Allied Telesis Inc, 2012.

[21] Rouiller A.S., “VLAN security weaknesses and

countermeasures”, GIAC Security Essentials Prac-
tical, SANS Institute, 2016.

[22] Heron, S., “Ten top threats to VLAN security”,

Redscan, 2014.

[23] Akram, J., Akram, N., Mamoon, S., Ali, S.,

& Naseer, N., “Future and Techniques of Im-
plementing Security in VLAN”, Journal of Net-
work Communications and Emerging Technolo-
gies (JNCET), 7(5), 2017.

[24] Bruschi, Danilo, Alberto Ornaghi, and Emilia

Rosti, “S-ARP: a secure address resolution proto-

ISBN: 978-1-941968-45-1 ©2017 SDIWC 89

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
Ontology-Based Data Mining Approach for Judo Technical Tactical Analysis
Ivo La Puma1 and Fernando Antonio de Castro Giorno1,2
1
Institute of Technological Research, São Paulo, Brazil
2
Pontifical Catholic University, São Paulo, Brazil
ivolapuma@gmail.com and fgiorno@gmail.com
ABSTRACT
Coaches and other judo experts conduct technical
tactical analysis of judokas to understand the
techniques they use to win fights, and to identify
defensive strategies to counter an opponent’s
actions. Computer systems based on artificial
intelligence (AI) techniques are used in the
technical tactical analysis and in predicting
results, injury prevention, talent discovery, and
game strategy evaluation in various collective and
individual sports. However, there are no studies
related to the use of AI in judo. This paper
proposes a data mining approach using an
ontology for the technical tactical analysis of
judo. As a proof of concept, a data mining tool
was developed to identify sequential patterns in
judo combat actions and assist in strategic
decision-making. An ontology of judo fight was
also developed and used to model the database.
The approach was found to be valid as the tool
yielded the information needed to satisfy the
desired performance analysis requirement. As
contribution, it is expected this paper enables the
flourishing of new researches or applications
through the ontology of judo fight as well as
validates the model of mapping requirements
performance analysis and data mining methods
used in this study.
KEYWORDS
Data mining, Sequential pattern mining,
Ontology, Judo, Technical tactical analysis.
1 INTRODUCTION
Technical tactical analysis or performance
analysis describes the method of analysis used
to understand how sporting skills are applied,
and serves as a source of information for
performance enhancement in the sport [1].
From the perspective of sports science, such
ISBN: 978-1-941968-45-1 ©2017 SDIWC
analysis is carried out with one or a
combination of the following aims: finding
patterns of performance that may increase the
chances of winning a competition; predicting
the performances of an athlete or a team; real-
time decision-making on actions/reactions or
strategies needed in an event; and identifying
sports demands and selecting athletes who
can best address the demands [2].
Judo is characterized as an intermittent high-
intensity combat sport consisting of a variety
of techniques and actions taken during a
match [3]. Similar to other sporting
modalities, technical tactical analysis is
carried out in judo to determine the
performance development in the sport.
Moreover, the performance of judokas is
analyzed to understand how they use their
techniques to win a contest [4].
The volume of data produced in the sports
area is increasing. However, the potential for
analyzing and understanding these large
datasets is still below the potential for
collecting and storing them. As experts and
statisticians fail to explain the relationships
between data, approaches based on AI are
being used to support them in explaining or
predicting events or in making strategic
decisions [5]. Several studies have reported
the use of data mining based on AI techniques
in predicting results, evaluating performance
and preventing injuries to athletes, finding
talent, and evaluating game strategies in
various sports modalities [2], [5], [6].
However, there are no studies related to the
use of AI-based computer systems in the
technical tactical analysis of judo competitors
or in supporting strategic decision-making in
the sport.
90
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

which highlight its positive and negative

The main objective of this study is to propose aspects, contributions, and suggestions for
an ontology-based data mining approach in future work.
the technical tactical analysis of judo
competitors. Furthermore, the following are 2 REVIEW OF RELATED LITERATURE
secondary objectives of this study:
1. Develop an ontology of judo fight, In this section, the significant studies that
which is inspired by the technical influenced the development of the present
tactical model of the judo fight [7]. study are highlighted.
2. Build a prototype of a data mining
tool for technical tactical judo analysis The first study is the technical tactical
based on the model of mapping analysis model of judo fight [7], which uses a
requirements of sports performance statistical method based on Markov processes
analysis and data mining methods [2]. to calculate combinations of probabilities
between each phase of combat. This model
A contribution expected by this paper is to characterizes judo fight in six distinct phases
enable the flourishing of new researches or (approach, gripping, attack, defense, ground
applications through the ontology of judo working, and pause) and uses 46 groups of
fight. Other contribution is to validate the technical tactical variables to represent
model of mapping requirements performance combat actions.
analysis and data mining methods used in the
approach proposed in this study. This is followed by the model of data mining
tools for performance analysis in high-
The remainder of the paper is organized as performance sports [2]. This model suggests a
follows. In Section 2, significant studies categorization of sports according to
related to this paper are presented. In Section performance analysis requirements where,
3, the ontology of judo fight is introduced. through a mapping of data mining methods
The steps for the ontology-based data mining and techniques, it is possible to combine it
approach for technical tactical judo analysis with the characteristics of available
are defined in Section 4. The main aspects in techniques and suggest the most appropriate
building a prototype of the data mining tool approach to build the tool (Table 1). This
for the technical tactical analysis of judo, model was proposed based on the hypothesis
which serves as a proof of concept, are that the combination of certain methods and
presented in Section 5. In Section 6, analysis techniques of data mining is more appropriate
of the results obtained in the simulations of for certain sports performance analysis
technical tactical analysis of judo using the requirements.
prototype is described. Finally, Section 7
presents the final considerations on this study,
Table 1. Mapping between sports performance analysis requirements, data mining methods, and data mining technique
characteristics.
Sports performance analysis DM methods DM technique characteristics
requirements Clust. Class. R. Mod. R. Min. Interp. Prec. Flex.
Performance pattern discovery X X High Moderate Moderate
Performance prediction X X Low High High
Real-time decision-making X X Very high High Very low
Demand analysis X X Moderate Moderate Moderate
Note. DM = data mining; Clust. = clustering; Class. = classification; R. Mod. = relationship modelling;
R. Min. = rule mining; Interp. = interpretability; Prec. = precision; Flex. = flexibility.
A cell with “X” indicates that the specific performance analysis is mainly performed using the method in that column.
A blank cell shows that the specific method in that column is not frequently and effectively used for the specific
performance analysis in that row.

ISBN: 978-1-941968-45-1 ©2017 SDIWC 91

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

Finally, the last study is an ontology-based
data mining approach using clustering and
mining association rules to identify patterns in
schools in India [8]. The approach was
composed of ontology development and
attributes categorization, data mining, and
evaluation. Data from Indian schools
consisted of 242 attributes, which make it
difficult for the data mining tool to extract
useful knowledge through random grouping.
The ontology developed in this study was
based on the school data attributes, which was
used in the application to select the input data
of mining algorithms, and served as reference
in the evaluation and interpretation of results.
This approach served as an inspiration for the
data mining approach to technical tactical
analysis of judo in the present study.
data source for the data mining tool in the
technical tactical analysis of judo.
Step 1.1. Defining the ontology of judo fight.
The ontology of judo fight is
recommended, but it is possible to use
another ontology.
Step 1.2. Modelling the database. The
database must consist of entities and
attributes found in the ontology.
Step 1.3. Creating the physical database.
Step 1.4. Loading the judo fight notations in
the database.

3 ONTOLOGY OF JUDO FIGHT

The ontology of judo fight was strongly

inspired by the judo fight phases (approach,
gripping, attack, defense, ground working,
and pause) and 46 combat actions that
characterize the technical tactical model of
judo fight [7]. The development of this onto
logy was carried out using of the Noy and
McGuinness's method [9] and the ontology
pitfall detection tool (OOPS!) [10]. This
ontology was written in OWL through the
ontology editor Protégé 5.0.0 [11]. Figure 1
shows the classes of ontology.

The latest version of ontology is available at

GitHub repository
(https://github.com/ivolapuma/ontology-judo-
fight).
4 PROPOSED APPROACH
This section presents all steps comprising the
ontology-based data mining approach for the
technical tactical analysis of judo.
4.1 Building the Judo Fight Notation
Database
Phase 1 deals with the building of the judo
fight notation database, which will serve as a
Figure 1. Classes of ontology of judo fight.
4.2 Creation of the Data Mining Tool for
Technical Tactical Analysis
Phase 2 deals with the creation of the
prototype of the data mining tool, which must
discover patterns from data of fight notations
and allows the technical tactical analysis of a
judoka.
Step 2.1. Defining the performance analysis
requirements. Based on Table 1, the
requirements can be performance pattern
discovery, performance prediction, and
real-time decision-making or demand
analysis.

ISBN: 978-1-941968-45-1 ©2017 SDIWC 92

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

Step 2.2. Defining the data mining method.
The data mining method must obey the
mapping of performance analysis
requirements and data mining methods
(Table 1).
Step 2.3. Defining the data mining technique.
The technique must correspond to the
values defined by the characteristics of
interpretability, precision, and flexibility
(Table 1).
Step 2.4. Developing the prototype of the data
mining tool. This prototype must use the
selected data mining technique to process
the judo fight notations and return the
patterns discovered and desired to satisfy
the performance analysis requirement.
4.3 Simulation of Technical Tactical
Analysis of Judo
database tables according to Step 1.2 was
done based on the entities and attributes of
this ontology. In Step 1.3, PostgreSQL
version 9.6.1 for x86-64 bits Windows
platform was used to create the database. In
Step 1.4, fight notation data were imported
directly into the database. These notes came
from Professor Emerson Franchini's archive
[12] and the 120 judo matches in official
contests organized by the International Judo
Federation between Beijing 2008 and London
2012 Olympic Games, which comprise the
fighting actions of 23 athletes from among the
best of men's -81 kg category. Table 2 shows
the number of fights and actions noted by
judoka – the names are preserved. These judo
matches were under the rules applicable at
that time, and did not consider the
modifications introduced in 2017.

The simulation of the technical tactical Table 2. Number of fights and actions noted by judoka.
analysis of judo comprises the third and last
Judo fighter Total fights Total actions
phase of the proposed approach. Brazilian judoka 1 10 753
Step 3.1. Discovering patterns from judo fight German judoka 1 10 527
notations. This step should be performed in Canadian judoka 1 9 486
simulations using the prototype. Dutch judoka 1 8 555
Step 3.2. Performing the technical tactical French judoka 1 7 428
Montenegrin judoka 1 7 369
analysis of judo. The analysis must be South Korean judoka 1 6 417
based on the patterns discovered by the American judoka 1 6 390
prototype. Currently, the ontology can be Azerbaijani judoka 1 6 368
used in interpreting the results. Emirati judoka 1 6 299
Step 3.3. Evaluating the characteristics of the Moroccan judoka 1 6 263
Belgian judoka 1 5 345
data mining technique. The characteristics Italian judoka 1 5 268
of interpretability, precision, and flexibility Argentinian judoka 1 5 199
of the technique used should be compared Ukrainian judoka 1 4 170
with the values defined for these Kazakhstani judoka 1 4 167
characteristics (Table 1). Japanese judoka 1 3 210
Dutch judoka 2 3 153
Slovene judoka 1 3 94
5 PROOF OF CONCEPT Russian judoka 1 3 88
Croatian judoka 1 2 134
To validate the approach proposed in this British judoka 1 1 60
study, a proof of concept was performed that South Korean judoka 2 1 54
involved the fulfillment of the steps defined in
each phase. In this section, the most critical Following Step 2.1, the performance analysis
aspects defined in Phases 1 and 2 steps are requirements were aimed at real-time
highlighted. decision-making because it was intended to
extract information from judo matches that
The ontology of judo fight, Step 1.1 in support decision-making on actions/reactions
Section 3, was defined for this proof of or strategies needed by a judoka to win in a
concept. The creation of a model of the competition. However, the term "real-time"

ISBN: 978-1-941968-45-1 ©2017 SDIWC 93

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
must be relativized given that during a fight, a
judoka would not be able to redefine his
strategies from an information system. In Step
2.2, mining rules was the data mining method
defined because the discovery of rules or
patterns in the fighting actions of a judoka can
fulfill the performance analysis requirement.
In Step 2.3, the sequencing pattern mining
technique [13] was defined because
identifying the most frequent sequences of
actions performed by a judoka can
accomplish the goal defined in Step 2.1.
Step 2.4 comprised the development of the
prototype of the data mining tool based on the
sequential pattern mining technique. The CM-
SPAM algorithm [14], provided by SPMF
library [15] written in Java, was used in the
prototype development. The CM-SPAM
algorithm was selected because it has a very
satisfactory performance compared to other
algorithms [16]. Furthermore, its
implementation in SPMF allows the use of
optional parameters (such as maximum and
minimum pattern size and mandatory items),
which are not available in the implementation
of other algorithms (such as GSP, PrefixSpan,
SPADE, SPAM, and CM-SPADE).
Moreover, to assemble the sequence base to
enable execution of the sequential pattern
mining algorithm, the following adaptations
were considered:
Item. All fight actions performed by the same
judoka.
Element. A set of consecutive fighting
actions carried out by the same judoka
between a referee's start fight command
(hajime or yoshi) and a pause fight
command (mate, sonomama or soremade).
Eventually, if a judo player repeats some
action since hajime, that action delimits the
set of previous actions, which marks the
beginning of a new set of actions until the
next referee's command or until another
action is repeated. The break in a set of
actions is necessary because it is an
algorithm premise that an element does not
contain repeated items.
ISBN: 978-1-941968-45-1 ©2017 SDIWC
Sequence. All set of actions (or elements)
performed by the same judoka in the same
fight.
The constraint imposed by the algorithm's
premise of not allowing an element to contain
repeated items can have implications on the
results obtained with data mining. Another
important consideration concerns the attack
actions. In the judo fight notation database
model, an attack action is characterized by the
applied judo technique, the attack direction
and, eventually, the score assessed by the
referee. To allow techniques and directions of
attacks to be considered in the process of
finding sequential patterns, the set "attack +
technique + direction" will be considered as a
single action. Similarly, an eventual score will
also be considered as a single action in
addition to the element following the item
related to the attack action.
An example of an element could be
represented by the sequence of judo fight
actions “Left Anteroposterior Approach,
Trying to Grasp, Left Sleeve and Right
Sleeve, Attack + Deashi-harai + Left, Waza-
Ari.”
The prototype built in this study was called
JudoDataMining and was developed using
Java version JavaSE-1.8 (jdk1.8.0_121).
6 RESULTS ANALYSIS
This section highlights the most critical
aspects defined in the implementation of
Steps 3.1, 3.2, and 3.3 of Phase 3.
6.1 Discovering Patterns from Judo Fight
Notations
In Step 3.1, the equipment used to perform
the simulations was a Dell Vostro 3500 with
Intel® Core ™ i7 M processor 680 @
2.80GHz, 6.00GB RAM using Microsoft
Windows 10 Pro 64-bit operating system. The
simulations were carried out through a script
program to run the data mining tool for each
selected judo fighter with the following steps:
94
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
1. Performing a loop by running the tool
and initially passing the value 100% in
minimum support parameter. For each
repetition, 10% is decreased in the value
of the minimum support parameter until
it reaches 30% or the execution time
exceeds 2 seconds.
2. Performing the previous loop by running
the tool and passing the value 4 in the
minimum pattern size parameter. With
this parameter, the tool will only return
sequential patterns with at least 4 items.
3. Performing the previous loop again by
passing the value corresponding to the
IPPON item in the required item
parameter, instead of passing value in
the minimum pattern size parameter.
With this parameter, the tool will only
return sequential patterns containing
IPPON.
There are some significant numbers in this
script program.
The first number is the value 4 defined for the
minimum pattern size parameter used in the
second loop. This is because a judoka must
perform at least 3 actions to obtain a score: an
approach action, a grip action, and the attack
action. Given that a score is also considered
an item in the sequential database, the
hypothesis is that any sequential pattern
involving the achievement of the score must
have at least 4 items.
The second number is the value 10% that is
decreased at the minimum support parameter
for each repetition in the loops. This is
because the largest number of fights noted by
a judoka is 10. Thus, the most adequate
minimum value of support would be 10%,
which would represent a single pattern
sustained in at least one fight.
The third number is the minimum value of
30% for the minimum support parameter.
During the running tests in the development
phase, the execution time had degraded
excessively with the support of 20% or less,
taking hours to finish in some simulations.
ISBN: 978-1-941968-45-1 ©2017 SDIWC
The degradation is probably due to the
number of fights of each judoka, the average
size of items and elements contained in each
sequence, and the number of possible items in
the system.
The last significant number is 2 seconds as
limit for the execution time of the tool. This is
because in the running tests lasting close to 2
seconds, the number of sequential patterns
found was around 300,000. Thus, the limit
was imposed so that huge files with much
information would not be generated, which
would cost too much to analyze.
6.2 Technical Tactical Analysis of Judo
In Step 3.2, the task of a tool user (or a
technical tactical analyst) is to extract useful
information to make winning strategic
decisions from the sequential patterns found.
For this task, the tool offers two features. The
first resource is a CSV file containing the
columns: found sequential pattern, support
value, number of elements, and number of
fighting actions (example in Table 3). The
second feature is also a CSV file that present
some simple statistics extracted from the
sequential patterns found file with the
columns: item/element, type, size, and
number of occurrences (example in Table 4).
Table 3. Sample records from the sequential patterns
file of “Brazilian judoka 1”.
Sequential pattern Support Elements Actions
(Left anteroposterior 3 4 5
approach) (Trying to
grasp) (Right
anteroposterior
approach, Left collar)
(Ippon)
(Left anteroposterior 3 4 4
approach) (Left sleeve
and right sleeve) (Left
collar) (Ippon)
Table 5 presents some simple statistics on the
quantities of sequential patterns found in the
proof of concept simulations. On the other
hand, Table 6 shows the quantities of patterns
95
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

found and the execution times from the the technique used by the prototype with the
simulations performed for Brazilian judoka 1. values defined in the model proposed by
Ofoghi et al. [2]. As defined in Step 2.3, the
The analysis of the results supported the sequential pattern mining technique should
conclusion that the sequential pattern mining present the following characteristics values:
tool was able to extract useful information for Interpretability: Very high – Owing to the
tactical and strategic decision-making in judo. large amount of response information
produced.
Table 4. Sample records from the simple statistics file Precision: High – Owing to the high degree
of the sequential patterns of “Brazilian judoka 1”. of dependence between the results
Item/element Type Size Occurrences
generated and the important decisions
Right anteroposterior Item 1 21 taken.
approach Flexibility: Very low – Owing to the very
Left collar and right Item 1 9 short time limit to use the results.
sleeve
(Right anteroposterior Element 2 8 6.3.1 Interpretability
approach, left collar)

Table 5. Statistics of the sequential patterns found in In this proof of concept, the results are
the proof of concept simulations. sequential patterns consisting of a series of
consecutive elements - or sets of items -
Number of Minimum Maximum Average representing judo fight actions. Given that
simulations number per number per these results are not numerical values and do
simulation simulation
not require interpretation in reading, their
173 0 5757309 168815.8
interpretation can be considered as an easy
Table 6. Quantities of patterns found and execution task. Moreover, a judo specialist is widely
times from the simulations performed for “Brazilian familiar with the fighting action names, which
judoka 1”. further enhances the interpretability. If there
is any difficulty or doubt in some fighting
Support Min. Required Number of Execution
pattern item patterns time (s)
action, the ontology of judo fight can be used
size discovered as support. Thus, it is possible to justify that
100% 21 0 the interpretability of the sequential pattern
90% 325 0.016 mining is very high because having
80% 10626 0.125 understood what a sequential pattern
70% 1280831 9.656
100% 4 1 0 represents allows for an interpretation without
90% 4 249 0 difficulties.
80% 4 10475 0.109
70% 4 1280453 9.786 6.3.2 Precision
100% Ippon 0 0
90% Ippon 0 0
80% Ippon 0 0 In this study, the results are sequential
70% Ippon 0 0 patterns, which represent a series of fight
60% Ippon 0 0.015 action sets that occur with certain frequency.
50% Ippon 0 0 Typically, a sequential pattern reveals similar
40% Ippon 0 0
information as a judo coach would find by
30% Ippon 41 0.016
watching a judo match once or more.
Therefore, any combat strategy decision will
6.3 Evaluation of the Characteristics of the
be made based on the information found by
Data Mining Technique
the judo coach.
In Step 3.3, the evaluation should be made by
By regarding the judo fight notations as
comparing the values of each characteristic
accurate and reliable, it is possible to consider
(interpretability, precision, and flexibility) of

ISBN: 978-1-941968-45-1 ©2017 SDIWC 96

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
that the results will maintain the same level of
accuracy and reliability. However, to confirm
that accuracy, it is necessary to carry out in
practice the combat strategies defined based
on the information extracted from the
sequential patterns found. Given that practice
is not part of the scope of this study, the
precision of the sequential pattern mining
could not be verified to be high.
6.3.3 Flexibility
Given that the performance analysis
requirement is real-time decision-making, and
the data mining method is mining rules, the
flexibility characteristic must be very low
(Table 1). However, the term "real-time" must
be relativized. Sequential patterns are found
on a sequence base assembled from the judo
fight notations of a given judoka. In judo, the
time limit for using the found sequential
patterns is until the beginning of the match
when the strategies have already been decided
between the athlete and his coach.
The sequential pattern mining technique does
not allow variations in parameters and data
format. The minimum support parameter is
obligatory. The processed data must be in the
form of a sequence database. This proof of
concept shows that the judo fight notation
data have to be adapted to use the data mining
technique.
In this study, the number of sequential
patterns found per time unit (average of
150,000 patterns per second) is highly
satisfactory when the support parameter
and/or the number of judo fights processed
are not very low. On the other hand, the
execution time degrades when the number of
sequential patterns found is very large.
Considering the time it would take for an
analyst to work on the results, which can be
large according to the minimum support
parameter, sequential pattern mining can
support the technical tactical analysis. Thus,
the low flexibility of the sequential pattern
mining can be justified.
ISBN: 978-1-941968-45-1 ©2017 SDIWC
7 CONCLUSION
This paper described a prototype that was
developed using a sequential pattern mining
technique to find patterns from a judo fight
notation database, to extract useful
information that would enable decision-
making on actions/reactions or strategies
needed to defeat a particular opponent in a
contest. The analysis of the simulation results
with the prototype shows that the use of the
sequential pattern mining technique produces
the needed information to satisfy the desired
performance analysis requirement. Thus, it is
confirmed that the main objective of this
study was achieved as the proof of concept
was able to validate the proposed data mining
approach.
This study has the following contributions.
The first is the novelty of using AI in judo. To
the best of our knowledge, no work has been
found on that subject during out search.
Second further research on the subject is
enabled through the development of an
ontology of judo fight and the proposed data
mining approach. Finally, this study validated
the mapping of performance analysis
requirements and data mining methods [2].
A limitation of this study is that when there is
a small amount of fight notation data for a
judoka, the use of the tool becomes infeasible
owing to the degraded execution time and/or
the large number of resultant sequential
patterns. Another limitation is that it is not
possible to verify the level of accuracy and
reliability of the sequential patterns found. To
do so, it would be necessary to practice the
judo strategy decisions based on the
information extracted using the tool, which
was outside the scope of this study.
In future research, the information extracted
from the sequential patterns is expected to be
evaluated in terms of making strategic judo
combat decisions, thereby verifying the
accuracy of the sequential pattern mining. A
future study may also apply this approach to
the technical tactical analysis of judo using
97
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

other data mining techniques or methods.

[12] Emerson Franchini B-9119-2012.
Finally, it is expected that the approach will ResearcherID.com. Available at
be adapted to other sports as well. http://www.researcherid.com/rid/B-9119-2012.
Accessed on 26Sep2017.

REFERENCES [13] R. Srikant, and R. Agrawal. Mining sequential

patterns: Generalizations and performance
[1] A. Lees, “Technique analysis in sports: a critical improvements. In The International Conference on
review.” J Sports Sci, vol. 20, issue 10, pp. 813- Extending Database Technology, 5, pp. 1-17,
828, 2002. 1996.

[2] B. Ofoghi, J. Zeleznikow, C. MacMahom, and M.
Raab. Data Mining in Elite Sports: A Review and
a Framework. Measurement in Physical Education
and Exercise Science, vol. 17, issue 3, pp. 171-
186, 2013.
[14] P. Fournier-Viger, A. Gomariz, M. Campos, and
R. Thomas. Fast Vertical Mining of Sequential
Patterns Using Co-occurrence Information. In
Pacific-Asia Conference on Knowledge Discovery
and Data Mining, 18, pp. 40-52, 2014.

[3] G. Marcon, E. Franchini, J.R. Jardim, and T.L.
Barros Neto, “Structural analysis of action and
time in sports: Judo.” J Quant Anal Sports, vol. 6,
issue 4, article 10, 2010.
[4] B. Miarka. Construção, validação e aplicação de
um programa computadorizado para análise de
ações técnicas e táticas em atletas de Judô:
diferenças entre classes, categorias e níveis
competitivos. (MSc Thesis, in Portuguese).
Universidade de São Paulo, Brazil, 2010.
[5] M. Haghighat, H. Rastegari, and N. Nourafza, “A
review of data mining techniques for result
prediction in sports.” ACSIJ Advances in
Computer Science: an International Journal, vol. 2,
issue 5, pp. 7-12, 2013.
[15] P. Fournier-Viger, J.C.W. Lin, A. Gomariz, T.
Gueniche, A. Soltani, Z. Deng, and H.T. Lam. The
SPMF open-source data mining library version 2.
In Joint European Conference on Machine
Learning and Knowledge Discovery in Databases,
23, pp. 36-40, 2016.
[16] P. Fournier-Viger, J.C.W. Lin, R.U. Kiran, Y.S.
Koh, and R. Thomas. A Survey of Sequential
Pattern Mining. Ubiquitous International - Data
Science and Pattern Recognition, vol. 1, issue 1,
pp. 54-77, 2017.

[6] H. Novatchkov, and A. Baca, “Artificial

intelligence in sports on the example of weight
training.” J Sports Sci and Med, vol. 12, issue 1,
pp. 27-37, 2013.

[7] B. Miarka. Modelagem das interações técnicas e

táticas em atletas de Judô: comparações entre
categoria, nível competitivo e resultados de
combates do Circuito Mundial de Judô e dos Jogos
Olímpicos de Londres. (PhD Thesis, in
Portuguese). Universidade de São Paulo, Brazil,
2014.

[8] P. Roy, S.S. Sathya, and N. Kumar. Ontology

Assisted Data Mining and Pattern Discovery
Approach: A Case Study on Indian School
Education System. Adv Nat Appl Sci, vol. 9, issue
6, pp. 555-560, 2015.

[9] N. Noy, and D.L. McGuinness. Ontology

Development 101: A Guide to Creating Your First
Ontology. University of Stanford, USA, 2002.

[10] M. Poveda-Villalón, A. Gómez-Pérez, and M.C.

Suárez-Figueroa. OOPS! (Ontology Pitfall
Scanner!) an on-line tool for ontology evaluation.
International Journal on Semantic Web and
Information Systems, vol. 10, issue 2, pp. 7-34,
2014.

[11] M. Musen. The Protégé project: A look back and a

look forward. AI Matters Association of
Computing Machinery Specific Interest Group in
Artificial Intelligence, vol. 1, issue 4, pp. 4-12,
2015.

ISBN: 978-1-941968-45-1 ©2017 SDIWC 98

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

Celestial Spectra Classification Based on Support Vector Machine

Jingchang Pan, Gaoyu Jiang, Yude Bu, Zhenping Yi and Xin Tan
School of Mechanical, Electrical & Information Engineering, Shandong University at Weihai,
Weihai 264209, China
pjc@sdu.edu.cn, jgyxyyxy@gmail.com, buyude001@163.com
yizhenping@sdu.edu.cn tanxin_0911@163.com

ABSTRACT preprocessed spectral data is classified by this

Spectra classification is essentially a pattern
recognition problem, so using SVM (Support
Vector Machine) to do spectral classification is
feasible. In addition, the spectra of special objects
can be found through the classification. In this
paper, we applied the SVM method to spectra
classification by Matlab simulation programs, and
analyze the results of the experiments.
Experimental results show the ideal classification
effects.
KEYWORDS
Support vector machine, Spectrum, Classification
1 INTRODUCTION
The large number of spectra obtained from
sky surveys such as the Sloan Digital Sky
Survey(SDSS) and the survey executed by the
Large sky Area Multi-Object fibre
Spectroscopic Telescope (LAMOST) provide
us with opportunities to search for peculiar or
even unknown types of spectra [1]. How to deal
with the massive spectra to obtain the necessary
scientific information in time has become a
very important issue, and spectral classification
of celestial bodies is one of the key tasks.
Support Vector Machine (SVM) is an effective
method to solve this problem [2]. In this paper,
Support Vector Machine is discussed, the
method, and the applicability of the method in
spectral classification is verified.
2 SVM
SVM is a new pattern recognition method
developed on the basis of statistical learning
theory in recent years, having many unique
advantages in solving nonlinear and high
dimensional pattern recognition problems.
Because of its excellent learning performance,
this technology has been successfully applied
in many fields such as face detection,
handwritten numeral recognition, automatic
text classification and so on.
2.1 Optimal Hyperplane
The SVM method is proposed from the
Optimal Hyperplane in the linear separable
case. As shown in Fig. 1, in two-dimensional
space, two classes of linear separable case, the
circles and squares in the figure represent two
classes of training samples. H is a line for
classification that separates the two classes
without errors. For one class, straight line H1 is
parallel to H and passes through the samples
that are closest to line H, for the another class,
the same is true of line H2. The distance
between them is called the margin. The
so-called optimal classification line is that it
can not only divide the samples into two

ISBN: 978-1-941968-45-1 ©2017 SDIWC 99

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
classes correctly, but also maximize the margin.
The former guarantees minimal empirical risk
(e.g. the training error is 0). It can be seen later
that the maximum margin is to minimize the
fiducial range in the bound of extending, so that
the real risk is minimal. When extended to high
dimensional space, the optimal classification
line becomes the optimal hyperplane.
Figure 1 illustration of optimal hyperplane
After a series of mathematical deduction,
the optimal classification function is obtained.
n transforming the original feature space into a
f ( x)  sgn{( w *  x )  b*}  sgn{  i* yi ( xi  x )  b*}
i 1
(2-1)
2.2 Generalized Optimal Hyperplane
The optimal hyperplane is discussed on
the premise of linearly separable. In linear
inseparable cases, a slack term i  0 can be
added.
yi [(w  xi )  b] 1  i  0
(2-2)
The problem of generalized optical
hyperplane can be evolved to find the minimum
values of the following functions under the
constraints of condition (3-2):
n
1
 ( w ,  )  ( w  w )  C ( i )
2 i 1
(2-3)
ISBN: 978-1-941968-45-1 ©2017 SDIWC
Where C is a specified constant which
controlls the degree of punishment for samples
that were misclassified, so that the trade-off
between the ratio of the misclassified samples
and the complexity of the algorithm is
achieved.
2.3 Support Vector Machine
The final classification discriminant
function ( w  x  b  0 ) of optimal and
generalized linear classification functions that
were discussed above contains only the inner
product ( x  xi ) of support vectors for
classified samples and training samples. Thus,
in order to solve the optimal linear
classification problem in a feature space, what
we only need to know is the inner product
operations in this space. If the inner product
K ( x, x') is used instead of the dot product in
the optimal hyperplane, it is equivalent to
new feature space, and the discriminant
function equation (3-1) is updated.
n
f ( x )  sgn{  i* yi K ( xi , x )  b*}
i 1
(2-4)
Other conditions of the algorithm are
unchanged. This is what is called Support
Vector Machine.
The basic ideas of support vector
machines can be summarized as followed.
Firstly, the input space is transformed into a
high dimensional space by nonlinear
transformation which is achieved by defining
proper inner product functions. Then, the
optimal Hyperplane is obtained in this new
space.
3 SPECTRA CLASSIFICATION
The automatic identification of celestial
100
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
Figure 2.
spectra must contain astronomical knowledge[3].
For astronomers, spectral lines are the most
important feature of spectra. The transition of
atoms or molecules between different energy
levels in a celestial body will absorb or emit
spectral lines. Different atoms and molecules
have their own specific spectral lines. The
distribution of the intensity at different
wavelengths can be used to describe the
radiation characteristics of celestial bodies.
There are all kinds of heavenly bodies in
the universe. In the field of astronomy, the
celestial bodies are first divided into normal
celestial spectral and emission-line object
spectra [4-5]. The spectrum of normal celestial
bodies includes main stars and normal galaxies,
while the emission-line object spectra includes
starburst galaxies, narrow line AGNs, wide line
AGNs and quasars.
Spectra classification can be expressed in
Fig. 2.
4 OVERALL DESIGN OF SPECTRAL
CLASSIFICATION MODEL
4.1 General Design
The work of this paper is to use SVM to
implement classifier 1, classifier 2 and
classifier 3, through which the celestial
spectrum is roughly classified.
ISBN: 978-1-941968-45-1 ©2017 SDIWC
Physical classification of celestial spectra
The training data and test data used in this
paper are derived from SDSS DR7, and the
categories involve stars (STAR), late stars
(STAR_LATE), galaxies (GALAXY), quasars
(QSO), and high redshift quasars (HIZ_QSO).
Among them, galaxies (GALAXY) are divided
into normal galaxies and emission-line
galaxies[6-8].
4.2 Experimental Environment
Experiments run in the MATLAB
R2010a mainly, and its computer environment
is configured as follows.
Processor: Intel Pentium dual-core
T2390@1.86GHz notebook processor
Motherboard: Lenovo 1GT30
Memory: 1GB (Hynix DDR2 667MHz)
Main hard drive: Hitachi
HTS542525K9SA00 (250GB)
Operating system: Windows7 Ultimate 32
LIBSVM software package developed by
Doctor Chih-Jen, Lin of National Taiwan
University was used for experiments. It’s a
simple SVM software package in common use.
4.3 Classifier Design
Training data is from DR7, downloaded
from the SDSS website. There are files of GIF
and FITS of five classes of celestial spectra
defined by Sloan including STAR,
STAR_LATE, GALAXY, QSO and HIZ_QSO.
101
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
Manual removal of both low confidence and
erroneous spectral files, dividing GALAXY
into normal galaxies and emission-line galaxies,
a total of 300 FITS files are selected as training
data, 50 files from each class. 150 spectral data
from STAR, STAR_LATE, and normal galaxies
in GALAXY are used as normal celestial
spectra, and data from QSO, HIZ_QSO, and
emission-line spectra in GALAXY are used as
emission-lines celestial spectra. A total of 100
spectral data from STAR and STAR_LATE are
used as stellar spectral data, and normal galaxy
spectra in GALAXY are used as normal galaxy
spectral data. A total of 100 spectral data of
QSO and HIZ_QSO are used as quasar spectral
data, and the emission-line galaxies in
GALAXY are used as emission-line galaxy
spectral data.
In terms of FITS files downloaded from
DR7, the abscissa of spectral is wavelength
range from 3800Å to 9200Å, and the ordinate
is the corresponding flow. Because there is
something missing in front wavelength of some
spectral, we use the wavelength range from
3810Å to 9200Å for training.
The two parameters, COEFF0 and
COEFF1, can be found in the corresponding
FITS file, representing the starting wavelength
and step size, respectively. The dimension n is
obtained from the FITS header file, that is, the
number of steps, and the wavelength of the
laboratory is obtained by the formula as
follows.
wave(n)  10COEFF 0( n1)COEFF1
(4-1)
The corresponding flow is obtained from
the table of the main block of the FITS file.
The specific experimental process is as
follows.
1) Read the training data used in the
experiment to make it can be
processed by computers.
2) Generate class labels.
ISBN: 978-1-941968-45-1 ©2017 SDIWC
3) Reduce the dimension of training data.
4) Use LIBSVM software package to
train, select parameters, generate
training templates and record training
time.
5 IMPLEMENTATION OF SPECTRAL
CLASSIFICATION MODEL
5.1 Description of Test Data
Similar to training data, test data is from
DR7, downloaded from the SDSS website.
There are files of GIF and FITS of five classes
of celestial spectra defined by Sloan including
STAR, STAR_LATE, GALAXY, QSO and
HIZ_QSO. Manual removal of both low
confidence and erroneous spectral files,
dividing GALAXY into normal galaxies and
emission-line galaxies, a total of 300 FITS files
are selected as test data, 50 files from each
class. 150 spectral data from STAR,
STAR_LATE, and normal galaxies in
GALAXY are used as normal celestial spectra,
and data from QSO, HIZ_QSO, and
emission-line spectra in GALAXY are used as
emission-lines celestial spectra. A total of 100
spectral data from STAR and STAR_LATE are
used as stellar spectral data, and normal galaxy
spectra in GALAXY are used as normal galaxy
spectral data. A total of 100 spectral data of
QSO and HIZ_QSO are used as quasar spectral
data, and the emission-line galaxies in
GALAXY are used as emission-line galaxy
spectral data.
Test data and training data are similar in
format, but differ in content.
5.2 Testing Process
1) Read the training data used in the
experiment to make it can be
processed by computers.
2) Generate class labels.
3) Use LIBSVM package to test and
record the final accuracy.
5.3 Test Results
1) The results of classifier 1 are shown
102
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

in Table 1 matrix E that obtained from 150 normal

In the row of dimensionality reduction,
E_star_43d is a matrix E that obtained from 50
star spectra which are reduced to 43 dimensions
through the PCA dimension reduction method.
Using E_star_43d to reduce dimension, the
accuracy rate is 96.6667%, and the training
time is 0.1813s. E_normal_total_43d is a
celestial spectra which are reduced to 43
dimensions through the PCA dimension
reduction method. According to the same
principle, E_celestial_total_43d is obtained
from 150 emission-line celestial spectra, and
E_total_43d is from all training spectra at a
total of 300.

Table 1. Results of classifier 1

Dimensionality E_normal_total E_celestial_tot
E_star_43d E_total_43d No
reduction _43d al_43d
Accuracy 96.6667% 97.3333% 95.3333% 96.3333% 97%
Time of
0.1813 0.1818 0.1821 0.1824 0.9378
training(s)

Table 2. Results of classifier 2

Dimensionality E_normal_tot E_star_late_43 E_normal_gala
E_star_43d No
reduction al_43d d xy_43d
Accuracy 97.3333% 99.3333% 99.3333% 96% 97.3333%
Time of
0.1349 0.1068 0.1098 0.1072 0.2800
training(s)

Table 3. Results of classifier 3

Dimensionality E_star_ E_celestial_ E_hiz_qso_ E_celestial_
E_qso_43d No
reduction 43d total_43d 43d galaxy_43d
Accuracy 96% 96% 98% 93.3333% 87.3333% 96.6667%
Time of
0.1056 0.1033 0.1097 0.1085 0.1053 0.3337
training(s)

2) The results of classifier 2 are shown in
Table 2
In the row of dimensionality reduction,
E_normal_total_43d is a matrix E that obtained
from all training spectra at a total of 150 which
are reduced to 43 dimensions through the PCA
dimension reduction method. Using
E_normal_total_43d to reduce dimension, the
accuracy rate is 97.3333%, and the training
time is 0.1349s. E_star_43d is a matrix E that
obtained from 50 star spectra which are
reduced to 43 dimensions through the PCA
dimension reduction method. According to the
same principle, E_star_late_43d is obtained
from 50 late stellar astronomical spectra, and
E_normal_galaxy_43d is from all training
spectra at a total of 50.
3) The results of classifier 3 are shown in
Table 3
In the row of dimensionality reduction,
E_star_43d is a matrix E that obtained from 50
star spectra which are reduced to 43 dimensions

ISBN: 978-1-941968-45-1 ©2017 SDIWC 103

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
through the PCA dimension reduction method.
Using E_star_43d to reduce dimension, the
accuracy rate is 96%, and the training time is
0.1056s. E_celestial_total_43d is a matrix E
that obtained from all training spectra at a total
of 150 which are reduced to 43 dimensions
through the PCA dimension reduction method.
According to the same principle,
E_hiz_qso_43d is obtained from 50 late high
redshift quasar spectra, and
E_celestial_galaxy_43d is from 50
emission-line galaxy spectra.
6. SPECTRAL CLASSIFICATION MODEL
ANALYSIS
6.1 Analysis of Experimental Result
1) Using support vector machines to
classify celestial spectra gets a good
result.
It can be seen from the above
experimental results that using support vector
machines to classify celestial spectra does an
excellent job. The accuracy is basically above
95%, and the training time is short.
2) The method of dimension reduction
has an effect on the experimental
results.
As can be seen from the above
experimental results, the experimental results
obtained by using different data for
dimensionality reduction are also quite
different, and the corresponding training time is
different.
In classifier 1, the result obtained from all
normal celestial spectra through PCA
dimension reduction method is the best. In
classifier 2, the result obtained from star or late
star spectra through PCA dimension reduction
method is the best. In classifier 3, the result
obtained from quasar spectra is the best.
From the above results, we can get a
conclusion that before dividing into two classes,
Using E matrix obtained by the PCA reduction
dimension of one of the classes to reduce the
ISBN: 978-1-941968-45-1 ©2017 SDIWC
dimension of other data, the classification result
is the best.
Moreover, it can be seen from the
experiment results that the training time is
longer without reducing dimension, and the
effect is usually better after reducing
dimension.
6.2 Next Work
1) Analyze the misclassification sample
and improve the training templates.
Misclassification samples are extracted
and analyzed separately to find out the causes
of the errors, and improve training data to
improve training templates, so as to get better
classification results.
2) Further analysis will be made on
dimension reduction.
From the above experimental results, we
can see that the process of dimension reduction
has a great impact on the experimental results,
so we can do some works from the dimension
reduction to see if we can further improve the
training template. The experiment in this paper
only involves the process of reducing to 43
dimensions, and only uses the method of PCA
dimension reduction. Spectra also can be
reduced to 3 dimension, 2 dimension and so on
to see whether the experimental results improve
or not. Other dimension reduction methods,
such as kernel entropy and component analysis
(KECA), can be used to reduce the
dimensionality, and compare the results to see
if the training templates can be further
improved.
SUMMARY
The content of this study is to use support
vector machines to classify the celestial
spectrum, and the feasibility is verified by
experiments.
After briefly introducing the LAMOST
project and the FITS file, the principle of SVM
and the classification of celestial spectrum are
introduced. The process and result of the
104
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

experiment are introduced in detail, and the ASTRONOMICAL SOCIETY, 2011,

corresponding analysis is given.
The main work in paper is as follows:
1) Study the principle of support vector
machine, understand its operation
mechanism and feasibility in reality,
and apply it into practice.
2) Understand the mechanism of
celestial spectrum classification, be
familiar with various spectra of
celestial bodies, understand the
characteristics of their spectra, and be
familiar with the corresponding FITS
documents.
3) Support vector machines are used to
classify the celestial spectra. After
repeated training, the best training
templates are obtained and then the
experimental results are obtained by
test.
We can see that using support vector
machines to classify celestial spectra is very
effective through experiments, and it is feasible
in practice.
412(4):2183-2198.
[4] Pan J C, Wang X X, Wei P, et al.
Spectroscopy and Spectral Analysis, 2012,
32(008): 2260-2263.
[5] Liu Z, Wan M J, Liu H. Electronic Test,
2016, (7): 24-25.
[6] M Marziani, M Gambaccini, Burga, G Di
Domenico, et al. Applied Radiation and
ISOTOPES, 2014, 92(6): 32-36.
[7] Yang Z Q. Application of Statistics and
Management, 2007, 26(1): 178-188.
[8] Huang L Y. Research of Automatic
Recognition of Qusar [D], 20000601, Pattern
Recognition and Intellegence Control.

ACKNOWLEDGEMENT
This work was financially supported by
the National Natural Science Foundation of
China (U1431102)

REFERENCES
[1] Cui, X., et al. The Large Sky Area
Multi-Object Fiber Spectroscopic Telescope.
Research in Astronomy and Astrophysics, 2012,
12(9): 1197-1242.
[2] Yude Bu, Fuqiang Chen, Jingchang Pan.
Stellar spectral subclasses classification based
on Isomap and SVM. New Astronomy, 2014,
28: 35-43.
[3] Daniel Thomas, Claudia Maraston, Jonas
Johansson. Flux-calibrated Stellar Population
Models of Lick Absorption-line Indices with
Variable Element Abundance Ratios.
MONTHLY NOTICES OF THE ROYAL

ISBN: 978-1-941968-45-1 ©2017 SDIWC 105

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

The Agent-Based Model of The Dynamic Spectrum Access Networks with

Network Switching Mechanism

Marcel Vološin1 and Jana Zausinova2

1
Faculty of Electrical Engineering and Informatics, 2 Faculty of Economics
Technical University of Košice, Letná 9, 040 01 Košice, Slovakia
marcel.volosin@student.tuke.sk, jana.zausinova@tuke.sk

ABSTRACT leasing [2]. The relevance of the spectrum

In this paper, we aim to introduce the initial agent-
based model, with the ability to capture two major
mechanisms (dynamic spectrum access networks
and network switching) that will be employed in
the future 5G communication systems. Dynamic
spectrum access networks aim to tackle the
traditional drawbacks of the long-term issued
frequency licenses, which results mainly in the
inefficient use of the assigned spectra. On the other
hand, the network switching mechanism allows the
user to increase its utility due to the ability to
choose the operator in the real-time. In the paper,
the two-level agent-based model with a focus on
the economic characteristics of the network is
proposed. On the first layer, the wholesale market
stipulating dynamic spectrum access network is
modeled using the MASCEM framework being
proposed initially for the grid-electricity
distribution. Second layer models the retail market
consisting of the set of user-equipment (UEs),
which is capable of the network switching. The
proposed early-stage model confirms the
theoretical expectation regarding the price
evolution and other technical indicators.
KEYWORDS
agent-based modeling, cognitive networks,
electronic markets, MASCEM
leasing raises its further importance in
conjunction with the concepts of the software-
defined radio (SDR). SDR enhances the
efficiency of the frequency spectrum
utilization through the embedded software
allowing the terminal to operate in multiple
frequency bands using numerous transmission
protocols [3]. Spectrum leasing and SDR are
technologies coined throughout the paper as
the dynamic spectrum access (DSA) strategies.
DSA strategies fundamentally change the
traditional telecom model based on the
"vertical integration". Here the single entity
delivers the service, maintains the network and
the network infrastructure [4]. Initially, the
available services were mainly limited to
telephony, radio, and television. However, we
do witness the convergence of these services
nowadays, and also the roles of the service
provider and the network owner are separated,
and the service providers get access to the
network and the end customers through
secondary spectrum trading on fair and non-
discriminatory conditions [5]. This
telecommunication concept is recognized as
the open access network.

1 INTRODUCTION Most trades today are direct trades between

Spectrum trading allows the owners of certain
spectrum licenses to transfer or lease all or part
of their rights and obligations under their
license to another party [1]. Several countries
have implemented spectrum trading, but the
trading process is often time-consuming, hence
hampering the usage. The UK regulator Ofcom
is in the forefront on the spectrum trading
arena, allowing spectrum sale and spectrum
organizations with the regulator as an
intermediate giving the final consent to
commit the trades [2]. However, in order to
facilitate spectrum trades on a shorter time
scale an organizational unit such as a band
manager could be introduced to mediate
between traders. Furthermore, organizational
units could be introduced to monitor for
compliance with committed trades and that the
spectrum is not misused. Overall, an

ISBN: 978-1-941968-45-1 ©2017 SDIWC 106

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
ecosystem is required to realize the spectrum
micro-trading [2] and to be able to analyze the
behavior of stakeholders on the both the retail
and the wholesale market. Therefore, in this
paper, the agent-based model of the dynamic
spectrum access network capable of capturing
the essential network characteristics is
proposed.
2 PROPOSED MODEL
To capture the stochastic nature of a spectrum
trading in a cognitive network an agent-based
model based on a MASCEM [6] in the
modeling tool called NetLogo was
implemented. The Multiagent Simulator of
Competitive Electricity Markets (MASCEM)
is the multiagent platform that can simulate the
variety of the different market players. The
goal of this framework is the ability to simulate
not only many types of the players but also as
many market models as possible with the long-
term and short-term decision implemented and
ready to be used. It also includes the numerous
negotiation mechanisms that can cope with the
diverse time scales. It was created to simulate
electricity markets that similarly to the DSA
networks consist of both the wholesale and the
retail market [7].
Figure 1 Model scheme
There are 3 different types of agents involved
in the proposed model as Figure 1 shows, each
interested in achieving its own goals. An
administrator, the owner of the spectrum
license, leasing its rights on the wholesale
market to the investors, who offer the services
on the retail market to the end-users willing to
use the offered services like phone calls, file
transfers, streaming etc.
ISBN: 978-1-941968-45-1 ©2017 SDIWC
2.1 Wholesale Market
In this stage, the investors acquire the chunks
of the frequency spectrum from the
administrator for the prices set via bilateral
negotiation. Following equations are used to
set the starting offers:
𝑠𝑡𝑎𝑟𝑡𝑖𝑛𝑔_𝑜𝑓𝑓𝑒𝑟𝑖+1 = 𝑠𝑡𝑎𝑟𝑡𝑖𝑛𝑔_𝑜𝑓𝑓𝑒𝑟𝑖 ± 𝛿𝑖+1
(1)
Δ𝑖
𝛿𝑖+1 = 𝑠𝑡𝑎𝑟𝑡𝑖𝑛𝑔_𝑜𝑓𝑓𝑒𝑟𝑖 × (β + )
𝐵𝑊𝑎𝑣𝑎𝑖𝑙𝑖 × 𝛼
(2)
Δ𝑖 = 𝐵𝑊𝑎𝑣𝑎𝑖𝑙𝑖 − 𝐵𝑊𝑙𝑒𝑎𝑠𝑒𝑑𝑖 .
(3)
Each new offer 𝑠𝑡𝑎𝑟𝑡𝑖𝑛𝑔_𝑜𝑓𝑓𝑒𝑟𝑖+1 calculated
by both the investor and the administrator
depends on the starting offer from the previous
iteration, the bandwidth administrator
(investor) wanted to rent (lease) and the
amount it rent (leased) and price shaping
parameters 𝛼 and β. The stakeholders set their
limit prices for each round of the negotiation
according to:
𝑙𝑖𝑚𝑖𝑡_𝑝𝑟𝑖𝑐𝑒𝑖 = 𝑠𝑡𝑎𝑟𝑡𝑖𝑛𝑔_𝑜𝑓𝑓𝑒𝑟𝑖 ± 𝜗 ,
(4)
where the 𝜗 represents the limit offer
parameter. During the process of negotiation,
agents adjust their offers simply via:
𝑜𝑓𝑓𝑒𝑟𝑖+1 = 𝑜𝑓𝑓𝑒𝑟𝑖 ± 𝜀,
(5)
the 𝜀 represents a fixed price-change
parameter, the administrator uses to lower and
the operator to increase its offer concerning the
preset 𝑙𝑖𝑚𝑖𝑡_𝑝𝑟𝑖𝑐𝑒𝑖 of the actual round.
2.2 Retail Market
Due to the different characteristics of the
frequency spectrum and the electric energy, it
is necessary to define suitable rules that will be
applied when the trading takes place on the
retail market. The acceptance probability
function was adopted from [8] with some
slight modifications. The spectrum obtained
107
Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

via bilateral negotiation is sold to the end-users

on the retail market for the prices
accommodated by the end-users’ demand. The
end-user is capable of the network switching in
the real-time, based on the instantaneous price
offered by the operator.

The following equations are used for the Figure 2 Acceptance probability according to retail
channel retail price calculation: price

𝑝𝑖 = 𝑝𝑖,𝑡 + (Ψ𝑖,𝑡−1 − 0.5) × 𝜇 Table 1 Simulation parameters

Parameter Value Description
(6)
𝑁𝑏𝑡𝑠 1 Number of base
1⁄2 (𝐵𝑊𝑎𝑣𝑎𝑖𝑙𝑖 = 0) ∧ (𝑆𝑖 = 0) stations
Ψ𝑖,𝑡−1 {0 (𝐵𝑊𝑎𝑣𝑎𝑖𝑙𝑖 > 0) ∧ (𝑆𝑖 = 0) 𝐵𝑊𝑡𝑜𝑡𝑎𝑙 200 Total number of
available channels
𝑆𝑖𝑖𝑑𝑙𝑒→𝑐𝑜𝑛𝑛 ⁄𝑆𝑖 (𝐵𝑊𝑎𝑣𝑎𝑖𝑙𝑖 > 0) ∧ (𝑆𝑖 > 0),
(7) 𝑁𝑖𝑛𝑣 2 Number of operators
𝑁𝑒𝑛𝑑−𝑢𝑠𝑒𝑟𝑠 〈300; 500〉 Number of end-users
where 𝐵𝑊𝑎𝑣𝑎𝑖𝑙𝑖 represents the amount of the 𝛼𝑎𝑑𝑚 -0.7 Price-shaping
available frequency channels, 𝑆𝑖 is the total parameter of
administrator
number of the end user connection attempts
𝛽𝑎𝑑𝑚 0.02 Price-shaping
towards an operator and 𝑆𝑖𝑖𝑑𝑙𝑒→𝑐𝑜𝑛𝑛 denotes the parameter of
number of successful connections. administrator
𝛼𝑖𝑛𝑣 -3 Price-shaping
Demand on the retail market is created by end- parameter of
operator
users, who randomly switch their states during
the simulation between the following three 𝛽𝑖𝑛𝑣 0.1 Price-shaping
parameter of
different options: IDLE, ACTIVE and operator
CONNECTED which brings stochasticity to 𝑝𝑚𝑖𝑛 0.15 Minimum wholesale
the model. In the IDLE, as the name suggests price
agents are inactive. Transition from IDLE to 𝜀 0.01 Negotiation
parameter
ACTIVE takes place randomly with the 〈0.1; 0.8〉
𝑃𝑎𝑐𝑡 End-users’
probabilities that change with the Gauss-like activation
curve on interval 〈0.1; 0.8〉 which emulates probability
different activity of users during day and night. 𝑃𝑑𝑖𝑠𝑐 1 End-users’
When ACTIVE, users evaluate each offer disconnection
probability
using following equation:
𝜗 0.3 Limit offer
𝛾 parameter
𝐴𝑃𝑖 = 1 − 𝑒 −𝑐(1−𝑝𝑖) ,
𝛾 3 End-users’ price
(8) sensitivity parameter
𝛿 0.5 End-users’ utility
acceptance probability (AP) of the best offer sensitivity parameter
represents user's tendency to make a deal with 𝜇 0.2 Price coefficient
the winning operator and switch state to 𝑐 8 Acceptance
CONNECTED. Parameters 𝑐 and 𝛾 adjust a probability
parameter
sensitivity of users towards the price 𝑝. Figure
2 acceptance probability in accordance with
3 SIMULATION RESULTS
the retail price when parameters 𝑐 and 𝛾 are set
according to the Table 1.
Multiple simulations were performed to test
the model behavior in the different scenarios.
Figure 3 shows the real-time data collected

ISBN: 978-1-941968-45-1 ©2017 SDIWC 108

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017

Figure 3 Real-time behavior of the model

Figure 4 Model behavior in accordance to the number of end-users

ISBN: 978-1-941968-45-1 ©2017 SDIWC 109

Proceedings of the Third International Conference on Computing Technology and Information Management (ICCTIM2017), Thessaloniki, Greece, 2017
from a single simulation during the time
interval of 5 days, each consisting of the 500
virtual time units so-called ticks in the
NetLogo. In Figure 3 a) the negotiated
wholesale prices between the administrator
and one of the investors can be observed. As
can be seen, price reacts to the varying end-
users' activity that follows Gauss like curve
with a slight delay. This phenomenon was
caused by equations used in MASCEM as was
found.
Figure 3 b) shows frequency channels usage.
Investors tend to rent more during peak hours
as expected and are very successful on the
retail market too. Investor's profit is also
affected by end-users' activity, but from the
Figure 3 c) it is obvious that highest profits
were gained thanks to the delay on the
wholesale market. Retail market price, Figure
3 d), does not change its mean value during
simulated days, but it's variance which is
higher during the non-peak hours.
Numerous simulations were executed to
determine the impact of the overall network
load generated by end-users. Figure 4
illustrates the stability of the model with the
different number of the end-users. The results
conclude that the increasing number of the end
users in the network, results in the higher
wholesale prices that have the higher variance
too due to the MASCEMs characteristics of the
negotiation process and the fact that activity of
end-users changes during the day. The mean
retail prices are also raised, but unlike the
wholesale price, more users result in the less
volatile thus more stable prices.
4 CONCLUSION
In the paper, we described the initial agent-
based model of spectrum trading in the
cognitive network with the variable stochastic
demand of the end users inspired by the
MASCEM. In the model, an administrator
(owner of the spectrum) and the non-
cooperating operators providing the services
were also present. Multiple simulations were
ISBN: 978-1-941968-45-1 ©2017 SDIWC
executed to verify and illustrate system's
behavior in different conditions. We can
conclude that proposed model with the
bilateral negotiation can capture the vital
characteristics of the cognitive network and
performs well regarding spectrum trading as
simulation results suggest. However, it
features the noticeable price lag as could be
seen on the plots which affect the incomes of
the operators. This phenomenon, yet being
natural in the MASCEM, would require an
additional effort to overcome when deployed
in the real environment.
REFERENCES
1. Ofcom, “Simplifying Spectrum Trading: Spectrum
leasing and other market enhancements”, 2011, p. 8
2. P. Grønsund, R. MacKenzie, P.H. Lehne, K. Briggs,
O. Grøndalen, P.E. Engelstad, and T. Tjelta,
“Towards spectrum micro-trading” [Future
Network & Mobile Summit (FutureNetw), 2012].
3. H. Arslan, ed.: “Cognitive radio, software defined
radio, and adaptive wireless systems,” Vol. 10.
Berlin: Springer, 2007, p 16.
4. N. Zhang, H. Liang, N. Cheng, Y. Tang, J.W. Mark,
and X.S.Shen, “Dynamic spectrum access in multi-
channel cognitive radio networks,” IEEE Journal on
Selected Areas in Communications 32.11, 2014, pp.
2053-2064.
5. P. Cramton and L. Doyle, “Open access wireless
markets,” Telecommunications Policy, 2017, pp.
379-390
6. I. Praça, C. Ramos, Z. Vale, and M. Cordeiro,
“MASCEM: A multiagent system that simulates
competitive electricity markets,” IEEE Intelligent
Systems 18.6, 2003, pp. 54-60.
7. Z. Vale, T. Pinto, I. Praca, and H. Morais,
“MASCEM: electricity markets simulation with
strategic agents,” IEEE Intelligent Systems, 26.2,
2011, pp. 9-17.
8. J. Pastirčák, L. Friga, V. Kováč, J. Gazda, and V.
Gazda, “An Agent-Based Economy Model of Real-
Time Secondary Market for the Cognitive Radio
Networks,” Journal of Network and Systems
Management. 12.10 2015, pp. 1-17
ACKNOWLEDGEMENT
This work was supported by the Slovak
Research and Development Agency, project
number APVV-15-0358.
110