Modelling Virtual Enterprises Using A Multiagent Systems Approach: Case Of Construction Industry In Kenya

ABSTRACT

Disparate enterprises can pool together their core competencies to form a temporary organization in order to exploit a market opportunity. This inter-organizational collaboration of enterprises is commonly referred to as a virtual enterprise (VE). The success of any VE is dependent on the partner members’ performance and influence of the partner attributes on its performance. These members and their attributes need to be carefully evaluated. The competitive advantage of any VE is jeopardized by the time it takes to set it up when the information available about the partners is insufficient. Extensive research on the evaluation and selection of partners has been done, but only a few studies have considered evaluation and selection of partners for VEs in the construction industry. Little research has been done in evaluating the factors that affect the VE performance. This research evaluates and selects partners representing ten construction companies to carry out project tasks for a large building in Nairobi. Each partner has business, technical and management expertise. Qualitative and quantitative research methods were used in this study. The qualitative method comprised interviews (with the stakeholders). Subsequently, quantitative methods, namely, Fuzzy Analytical Hierarchy Process (FAHP) and Reduced Group Fuzzy Analytical Hierarchy Process (RGFAHP), Multi-Criteria Decision Making (MCDM) algorithms, were applied. A technique called Partners Selection and Performance Evaluation Technique (PaSPET) is proposed. The technique combines fuzzy approximate reasoning with conventional Analytical Hierarchy Process algorithm, designed to deal with imprecise evaluators' judgement. A Multi-Agent Systems (MAS) approach was chosen to simulate VEs. A MAS is a computerized system composed of multiple interacting intelligent agents within an environment. Prior evidence of MAS to facilitate formation of VEs is lacking. Results, however, show that the chosen techniques are both efficient and effective. In particular, RGFAHP reduces the number of pairwise comparisons required when a large number of attributes are to be compared. Validation of the system, carried out by stakeholder evaluation, show that the approach is approximately 87% accurate in evaluation and selection of partners and partners' performance evaluation.