Fitness for Solving SMCP Using Evolutionary Algorithm

Abstract

Search based software engineering is a paradigm with prime focus to apply search techniques and meta-heuristics for solving various software engineering open NP hard problems. Evolutionary meta-heuristics techniques are already proven to provide optimized solutions to other software engineering problems like automated software test data generation, project estimation, class responsibility assignment to name few. Software module clustering is such an open problem of software engineering that cannot be solved in definite manner. For a given module dependency graph of software system, there exists large number of possible partitions. Identifying a good partition to cluster all modules of software system is an exhaustive search that cannot be carried out in finite manner. With search based techniques applying evolutionary algorithms, an optimized solution can be identified with evaluating goodness of a given partitions. Identifying a fitness function that can direct the search towards optimal solution is very critical. This paper discusses various types of fitness applicable to solve software module clustering problem.