Multi agent paradigm used to complexity measure for perfective software maintenance

Abstract

Agent is a computing entity which mimics the behavior of a human being in problem solving strategy. Multi agents systems are the group of agents which have different tasks and work in cooperation, coordination and in communication to each other. Apart from wide applications in software engineering tasks, design, development and testing, it has a vital role in software maintenance also. Few effective attempts have been made in this direction. We made an attempt to develop multi agent system for perfective maintenance using JADE-architecture. These results will helps to assess and estimate the complexity of change in a particular method, particular class and particular file with their effect to other set of methods, classes, and files (clusters). This knowledge can be deployed for adaptive maintenance in which a method or class or file are being changed or replaced. We are in process of developing and adding two more agents one adaptive and another evaluating in the context of software maintenance. © 2014 IEEE.