Solving software module clustering problem by evolutionary algorithms

Well organized software is easy to maintain but software modularization is complicated because of the number of modules. Automated software module clustering is transformed to a search-based problem. This paper describes the experiments on real-world problems of software module clustering by metaheuristic search methods such as genetic algorithms. This paper introduces the Grouping Genetic Algorithm (GGA) to the benchmarks. The fitness function measures a module granularity which is cohesion and coupling. Empirical result reports that the GGA outperforms a genetic algorithm with string representation.