Optimization of multi-way clustering and retrieval using genetic algorithms in reusable class library

Abstract

In order to improve code reliability and development productivity, software reuse is a clear solution and a reuse library based on object-oriented technology is essential. It is also very important to classify components elaborately and retrieve them accurately in the reuse library. In this paper, we present genetic algorithms for multi-way clustering, in which the number of clusters, similarity in a cluster and similarity between clusters are taken into consideration with the aim of finding optimized clusters into which components are classified, and for cluster-based linear retrieval with the aim of finding an optimal query which retrieves clusters containing components similar to a given query. We compare genetic algorithms with simulated annealing algorithms for multi-way clustering and cluster-based retrieval. The results of our experiments demonstrate that generic algorithms produce better solutions than those obtained by simulated annealing algorithms. We implemented a Reusable Class Library (RCL) using these methods, which is based on CORBA.