An agent-based Knowledge Management System for collaborative software maintenance environment Design and evaluation

Abstract

Software maintenance (SM) environment is highly complex, knowledge-driven and collaborative. Therefore, a Knowledge Management System (KMS) is critical to provide an environment for sharing and sustaining knowledge among members of SM community of practice (CoP). Based on earlier works on KMS framework for collaborative SM, an agent-based KMS prototype system called SM Collaborative Agent Team (SCAT) is designed and developed, to assist maintainers to share information and perform their SM activities. The aims of this paper are to discuss the design of SCAT and to evaluate SCAT's success as a KMS tool. Designing SCAT involves three major aspects: Multi Agent System (MAS) design, Combined Ontology of SM and Domain knowledge areas, and System design. For evaluation, SCAT was benchmark against another agent-based KMS prototype. A survey was conducted to gauge the responses of users and maintainers on the success of both KMS tools, based on Wu & Wang KMS Success model. The results of the survey are analyzed using Rasch rating-scale model to determine if SCAT fares better than Rodriguez. Based on t-test comparison on overall Rasch estimated measures, SCAT KMS is slightly better than Rodriguez KMS. Using Common Person Equating (CPE) plot, SCAT are better than Rodriguez KMS in several key areas. In future, SCAT could be enhanced to improve the weak areas.