www.elsevier.com/locate/advengsoft
Abstract
The selection and use of an appropriate procurement system are fundamental to the success of a construction project. However, the
procurement selection process involves the analysis of complex and dynamic criteria such as cost certainty, time certainty, speed, flexibility,
etc. Procurement selection is, therefore, plagued with uncertainty and vagueness that is difficult to be represented by a generalized set of
rules. In reality, decisions in procurement selection are usually derived from intuition and past experience. Case-based reasoning (CBR)
appears to be an appropriate approach to meet the requirements of the procurement selection process because of the value of experiential
knowledge. This paper reviews the practicality and suitability of a CBR approach for procurement selection through the development of a
prototype case-based procurement advisory system. In this prototype system, procurement selection cases are represented by a set of
attributes elicited from experienced procurement experts. The system is powered by a fuzzy similarity retrieval mechanism, which gives a
greater accuracy than the normal similarity retrieval process. The results indicate that the CBR approach can suitably model the
characteristics of construction procurement selection, and provide an indication of potential outcomes to any apparently suitable procurement
methods.
q 2003 Elsevier Science Ltd. All rights reserved.
Keywords: Construction procurement; Procurement selection; Case-based reasoning; Fuzzy similarity retrieval
established by the users. The assignment of appropriate the degree of success of that project (outcome) are provided
linguistic classifications calls on the experience of the users. to the users for consideration. When the users are convinced
Predefined values that replicate the linguistic classifications that the case selected is comparable to the current case, and
of fuzzy PSC, such as low certainty, high speed, etc. are that the solution and outcome are acceptable, they can apply
presented to the users in accordance with the recommen- the previous solution to the new project. Otherwise, the users
dation of Ng et al. [27]. might have to adapt the case to suit the distinctive
Depending on the weightings and linguistic classifications characteristics of the new project.
entered, similar cases are retrieved by the selection module. The hypothetically ideal solution is reported to the users
The retrieval process is performed through a fuzzy similarity through the output module. All the data entered by the users
retrieval mechanism. Such a mechanism can best encapsulate including the PSC, linguistic classifications and importance
the interrelationships between various linguistic PSC, and weightings are also presented to the users for checking. The
hence ensure the retrieval of the most appropriate procure- details of the new case and the adopted or adapted
ment system for a new project. The procurement system solution(s) are stored in the case-base for future reference
adopted in the most similar historic case (solution) and and retrieval.
432 D.T. Luu et al. / Advances in Engineering Software 34 (2003) 429–438
4. An overview of the case-based procurement advisor costs, greater cost and time certainty, shorter procurement
duration, better quality, more effective and efficient
As an initial step to establish the suitability of CBR decision-making and communication, and minimization of
approaches for procurement selection, the conceptual disputes [4]. Therefore, these factors constitute the outcome
framework was developed into a prototype CPAS using a part of a case. Both successful and unsuccessful construc-
CBR shell—ARTpEnterprisee version 10. tion projects are recorded in the case-base of CPAS.
The case attributes in CPAS include those of the
4.1. Case representation symbolic and linguistic types. Symbolic attributes are
expressed by terms with unambiguous meanings. There
Aamodt and Plaza [26] pointed out that a good CBR are no implied logical relationships among the values of
system lies with a clear representation of the cases and an these attributes. An example of this is the attribute ‘time
appropriate structure for describing their contents. Like certainty’, which can be represented by Boolean values (i.e.
other CBR systems, cases in the procurement advisory yes or no). Linguistic attributes, on the other hand, are fuzzy
system consist of three main components: problem, solution linguistic variables with implied logical relationships
and outcome (Fig. 3). The problem part is represented by a
Table 1
collection of PSC. Table 1 highlights the details of the nine
Characteristics of case attributes
most commonly used PSC, namely time certainty, cost
certainty, speed, flexibility, responsibility, complexity, price Attributes Data type Domain values
competition, risk allocation and quality.
The solution part contains information relevant to the Time certainty Symbolic Yes; no
Cost certainty Symbolic Yes; no
procurement system used in a past construction project and
Speed Linguistic High; medium; low
its sub-managerial systems, including tendering method and Flexibility Linguistic High; medium; low
contractual arrangement. As for the outcome part of a case, Responsibility Linguistic High; medium; low
feedback detailing the degree of success of the past project Complexity Linguistic High; medium; low
is provided. The degree of success of a construction project Price competition Linguistic High; medium; low
Risk allocation Linguistic High; medium; low
is measured by the level of client’s satisfaction on critical
Quality Linguistic Prestige; good; basic
success factors such as reduction in capital and lifecycle
D.T. Luu et al. / Advances in Engineering Software 34 (2003) 429–438 433
analysis and consideration. Fig. 5 shows the outcome Further case details can be made available should the
generated by the matching and scoring processes. users click on the view details button. As shown in Fig. 6,
In the matching result box (Fig. 5), five most similar cases these additional details include project’s PSC and their
along with their similarity scores are retrieved and presented linguistic classifications, the procurement system used in
to the decision-makers. By selecting a particular retrieved the stored case and its full details (i.e. mechanism,
case in the matching result box, the users can review the case advantages and disadvantages), sub-managerial systems
details, and the client’s evaluation on the project performance, (i.e. tendering method and contractual arrangement), and
such as communication efficiency, team performance, future the extent to which the retrieved case differs from the
working relationship between client and the main contractor. current one.
4.3. Adaptation case to better suit expectations button, the system displays
an adaptation screen to guide the users through the
A combination of different adaptation strategies was adaptation process (Fig. 7).
adopted for CPAS. For instance, if the users are satisfied that The upper part of the adaptation screen (Fig. 7) displays
the retrieved case closely resembles the current one, they the adaptation information, i.e. the full extent of differ-
can employ a null adaptation strategy by simply adopting ences between the retrieved and the present cases as well
the solution of the matching case to the new case without as advices on case attributes that might require subsequent
any modification. However, when the intrinsic character- adaptation. Built-in adaptation rules (Table 3) determine
istics and requirements between the two cases differ, the extent to which adaptation is required for the retrieved
modifications to the historic solutions might be desirable. case. By selecting any particular case attribute highlighted
Critic-based adaptation [29] and parameterized adaptation for potential further modifications, the users are provided
[23] strategies are provided to help decision-makers come with a list of activities that might be implemented to
up with a more suitable solution. Once the users have improve the performance of this attribute to the expected
decided to adapt the retrieved solution and chosen the adapt level. This list was compiled according to the advantages
436 D.T. Luu et al. / Advances in Engineering Software 34 (2003) 429–438
Table 4 selection models, the CBR approach caters for the fuzzy
Case study procurement solutions generated by experts and CPAS characteristics and the complicated intrinsic interrelation-
Decision- Solution ship of PSC. Since CBR is an experience-based approach, the
maker experiences of previous cases can be made available to the
Procurement Tendering method Contract type users to provide an early indication of the likely future
systems outcomes of the prospective project. This feature is not
addressed in the current analytical approaches for procure-
Expert A Design and Selective competition AS4300
build using a list of pre-
ment selection.
qualified contractors Based on the information collected from the experts, a
Expert B Design and Negotiation with the Design and construct conceptual framework for a CPAS for construction was
build mentioned local contracts–Master devised. The framework was subsequently developed into a
contractor with a Builders Association computer prototype using a CBR shell—ARTpEnterprise.
view to go into a of New South Wales
long term partnering
The prototype, using trial data, has demonstrated that CBR
relationship approach can provide appropriate recommendations for the
Expert C Design and Negotiation with the AS4300 procurement of a hardware retail facility. Further validation
build mentioned local on the performance of the system also yielded satisfactory
contractor with a results.
view to go into a
long term partnering
As the process of identifying the PSC, their conditions
relationship (i.e. linguistic classifications), and impacts (i.e. importance
Expert D Design, Selective competition C21 contract weightings) is largely dependant on previous experiences of
construct & using a list of pre- decision-makers, it seems sensible to model this process
maintenance qualified contractors
using the CBR approach too. The development of such a
CPAS Design and Negotiation Design and construct
build contracts—Master sub-system for procurement selection is being investigated.
Builders Association Once developed, this sub-system should overcome the
of New South Wales inherent weaknesses of the CPAS proposed in this paper, as
Client Design and Negotiation with the In-house written it will no longer have to rely rigidly on a list of nine PSC
build mentioned local contract incorporating
alone for case representation and retrieval.
contractor clauses for partnering
[11] Chan APC, Ho DCK, Tam CM. Design and build project success factors: [21] Ward SC, Curis B, Chapman CB. Objectives and performance in
multivariate analysis. J Construct Engng Mgmt 2001;127(2):93–100. construction projects. Construct Mgmt Econ 1991;9:343 –53.
[12] Cheung SO, Lam TI, Wan YW, Lam KC. Improving objectivity in [22] Masterman JWE. An introduction to building procurement systems.
procurement selection. J Mgmt Engng 2001;17(3):132–9. London: E & FN Spon; 1992.
[13] Alhazmi T, McCaffer R. Project procurement system selection model. [23] Kolodner JL. Case based reasoning. San Mateo, CA: Morgan
J Construct Engng Mgmt 2000;126(3):176 –84. Kaufmann; 1993.
[14] Brandon ES, Basden A, Hamilton IW. Expert system: the strategic [24] Schank RC, Kass A, Riesbeck CK. In: Erlbaum L, editor. Inside case-
planning of construction projects. UK: Royal Institution of Chartered based reasoning. Hillsdale, NJ: Associates Publishers; 1994.
Surveyors and University of Salford; 1988. [25] Barletta R. An introduction to case-based reasoning. AI Expert 1991;
[15] Ng ST, Luu DT, Chen SE. Decision criteria and their subjectivity in 43– 9.
construction procurement selection. Aust J Construct Econ Bldg 2002; [26] Aamodt A, Plaza E. Case based reasoning: foundational issues,
2(1):70–80. methodological variations and system approaches. Artif Intell
[16] Chang CY, Ive G. Rethinking the multi-attribute utility approach Commun 1994;7:39–59.
based procurement route selection technique. Construct Mgmt Econ [27] Ng ST, Luu DT, Chen SE, Lam KC. Fuzzy membership functions of
2002;20:275 –84. procurement selection criteria. Construct Mgmt Econ 2002;20(3):
[17] Ireland V. The role of managerial actions in cost, time and quality 285 –96.
performance of high rise commercial building projects. Construct [28] Liao TW, Zhang Z, Mount C. Similarity measures for retrieval in
Mgmt Econ 1985;3(1):59–87. case-based reasoning systems. Appl Artif Intell 1998;12:267–88.
[18] Nahapiet H, Nahapiet J. The management of construction projects: [29] Brown SJ, Lewis LM. A case-based reasoning solution to the problem
case studies from the UK & USA. Ascot: Chartered Institute of of redundant resolutions of nonconformances in large-scale manu-
Building; 1985. facturing. In: Smith R, Scott C, editors. Innovative applications of
[19] Hamilton IW. In: Brandon PS, editor. Developing expert systems for artificial intelligence 3, AAAI Press; 1993. p. 121 –33.
management applications. Building cost modeling and computers. [30] Gonzalez AJ, Xu L, Gupta UM. Validation techniques for case-based
London: E & FN Spon; 1987. reasoning systems. IEEE Trans Syst Man Cybern 1998;28(4):
[20] Tookey JE, Murray M, Hardcastle C, Langford D. Construction 465 –77.
procurement routes: re-defining the contours of construction procure- [31] O’Keefe RM, Balci O, Smith EP. Validating expert system
ment. J Engng Construct Architect Mgmt 2001;8(1):20–30. performance. IEEE Expert 1987;Winter:81–7.