Proposed Work:
To propose an multi objective optimization
algorithm to mine the intelligence rules.
Abstract
Mining of interesting relation with the objective of improving
the business plays a vital role in critical decisions in the
current scenario.
In this work, an algorithm to mine the intelligence rules from
the input data using the multi-objective optimization
algorithm is proposed.
The objectives used are profit, quality and the proposed
reliability.
Subsequently R-tree is modelled and is solved using the
improved ABC algorithm.
Improvement to the ABC algorithm is brought about in the
initial population and in the scout bee phase.
The performance is evaluated in terms of the number of rules
mined, computation time and memory usage by varying the
threshold and support value respectively.
Introduction
One of the outstanding applications of the DM is the business
intelligence and risk management as DM has a direct and
significant say on the decision-making process
The accomplishment of a BI system is not a traditional project
like the operational or transactional mechanism, indeed it
deals with multifaceted task necessitated with the suitable
infrastructure and resources.
To solve this the multi-objective optimization has emerged as
a well-acknowledged investigation
The classification rule mining may be designed as a multi-
objective optimization issue with diverse commensurable and
often contradictory objectives
In this paper, a technique to mine business intelligent rules is
proposed with the use of multi-objective improved ABC based
rule mining.
Introduction (Ctd)
Here, initially the influential patterns are found out from the
input database with the use of objectives of profit, quality and
proposed objective of reliability.
Subsequently, R-Tree is generated and then the patterns are
solved using the improved Artificial Bee Colony (ABC)
algorithm.
The rules obtained from the Improved ABC algorithm are
useful for developing business strategies
Review of Literature
C.F. Cheung and F.L. Li [20] were instrumental in effectively
launching the quantitative correlation coefficient mining
technique for business intelligence in small and medium
enterprises of the trading business.
The novel approach was competent to reveal the concealed
models of the sales and the market.
Therefore, a model business intelligence system (BIS) termed
as the correlation coefficient sales data mining system
(CCSDMS) was brought to limelight and efficiently
experimented in a specified reference site.
Review of Literature (ctd)
Hung-Pin Chiu et al. [21] spotlighted the novel clustering
based fuzzy association rule mining in the electronic
commerce (EC) scenario
A cluster-based fuzzy association rules (CBFAR) mining
architecture was brought in to concurrently tackle the issues
with three major objectives.
(a) An effective fuzzy association rule miner based on cluster-
based fuzzy-sets tables was spearheaded to locate all the
gigantic fuzzy itemsets.
(b) The novel approach needed minimal contrast to create
colossal itemsets.
(c) The fuzzy rule mining technique was employed to evaluate
the confidence values for locating the correlations between
transaction database and the browsing information database.
Review of Literature (ctd)
Alatas and Erhan Akin [22] launched the Multi-objective rule
mining employing a messy particle swarm optimization
technique.
A multi-objective messy particle swarm optimization (PSO)
approach was brought in as the investigation stratagem to
extract the classification rules within the datasets.
The optimized PSO employed the similarity measure for
neighborhood and far-neighborhood search to park the global
best particles located in multi-objective method.
For the bi-objective issue of rule mining of superior
precision/lucidity, the multi-objective approach was designed
to permit the PSO technique to return an approximation to
the higher precision/lucidity bounds, offering solutions
The captivating outcomes effectively exhibited the superlative
efficiency of the novel technique.
Review of Literature (ctd)
Jayakrushna Sahoo et al. [23] developed investigative
community when they proficiently proposed the mining the
association rules from the high utility itemsets.
In their document, they dealt with the issue of locating the
association rules employing the utility-confidence structure,
which represented a simplification of the amount-confidence
measure.
They were capable of creating the utility based non-
redundant association rules and techniques for restructuring
the entire association rules.
The fascinating upshots emerged as the ideal credentials
authenticating the superlative efficiency and competence of
the HUCI-Miner technique vis--vis popular peer methods.
Further, the upbeat test outcomes illustrated the superior
quality in the compacted illustration of the complete rule set
Review of Literature (ctd)
Thi-Thiet Pham et al. [24] introduced an effective technique
for extracting the relevant sequential rules from an attributed
prefix-tree in two phases.
In the former phase, it was able to construct a prefix-tree
which stockpiled the entire sequential patterns from a
specified sequence database.
In the later phase, it succeeded in extorting the relevant
sequential rules from the corresponding s prefix-tree.
Further, an excellent pruning system was instrumental in
scaling down the search space and the runtime in the mining
tasks.
The eye-catching test outcomes illustrated the success of the
innovative technique by authenticating its superlative
excellence over the modern approaches in extracting the
relevant sequential rules.
Review of Literature (ctd)
Wingyan Chung et al. [25] discussed the Discovering of
business intelligence from online product appraisals.
In their document, they designed a novel class of BI systems
taking cues from the rough set theory, inductive rule learning,
and data retrieval techniques.
The outcomes illustrated the novel technique was able to
usher in superior precision and exposure related to rule
quality, and brought in exciting and enlightening rules with
superior support and confidence values.
The investigations resulted in significant impacts for the
market sentiment appraisal and e-commerce reputation
administration.
Block diagram of proposed method
Objective Measures
Data Base
Improved ABC
Algorithm
Objective measures
The objective measures taken here are profit, quality index
and reliability
PROFIT
In-order to define profit, it is essential to define quantity first.
The quantity of an event set in a space Q can be described as
the average number of events in a unit of Q. It can be defined
as the total number of events of the event set divided by the
volume of the space Q. For a given space Q, the quantity of an
event set X located in each unit of Q, can be defined by,
| {x | x X ^ x insideQ } |
Quantity ( X , Q)
|Q|
Objective measures (Ctd)
For two event sets X and X and a given follow () predicate
defined by a neighbourhood function N(x) for any event x X,
the profit of X and X can be represented by,