APPLICATION OF SBSE TECHNIQUES FOR HIERARCHICAL SOFTWARE CLUSTERING

Abstract

Software systems evolve and change with time due to change in business needs. At some stage the available architectural description may not best represent the current software system. Accurate understanding of software architecture is very important because it helps in estimating where and how much change is required in the software system to fulfill changing business needs. It also helps in making decisions related to reusability of software components. The understanding of software architecture also plays vital role in estimating cost and risk of change in software system. In some cases, especially for legacy systems such a description does not readily exist. For such cases, we can use source code to extract architecture of the software system. Software Clustering is an approach to decompose large software system into smaller manageable sub systems to get system architecture. Software clustering, however, is an NP-hard problem. Search Based Software Engineering (SBSE) provides optimization algorithms which are search based and can be applied to Software Engineering problems. Particle Swarm Optimization (PSO) is a metaheuristic search technique based on biological behaviors and can be used to solve NP-hard problems. This thesis provides a framework for solving software clustering problem with PSO. Experimental results show fast convergence and stable results. In this thesis, software clustering process is presented in detail. Different Search Based Software Engineering (SBSE) techniques are discussed but focus is on Particle Swarm Optimization (PSO). The thesis focuses on design, implementation and analysis of PSO algorithm applied to software clustering problem. The objective of this paper is to solve software clustering problem using PSO and examine the effectiveness of PSO comparative to Genetic Algorithms (GA). Simulation results show that the PSO approach has stable results and it requires smaller computational effort as compared to GA.