Software Module Clustering Using Hybrid SocioEvolutionary Algorithms

Design of the software system plays a crucial role in the effective and efficient maintenance of the software system. In the absence of original design structure it might be required to re-identify the design by using the source code of the concerned software. Software clustering is one of the powerful techniques which could be used to cluster large software systems into smaller manageable subsystems containing modules of similar features. This paper examines the use of novel evolutionary imperialist competitive algorithms, genetic algorithms and their combinations for software clustering. Apparently, recursive application of these algorithms result in the best performance in terms of quality of clusters, number of epochs required for convergence and standard deviation obtained by repeated application of these algorithms. Index Terms—Genetic algorithm, Imperialist competitive algorithm, Module dependency graph, Reverse engineering Software clustering, Software maintenance.