A Hybrid Particle Swarm Optimization Algorithm for Clustering Analysis

Abstract

Clustering is a very important problem that has been addressed in many contexts and by researchers in many disciplines. This paper presents a new stochastic nature inspired methodology, which is based on the concepts of Particle Swarm Optimization (PSO) and Greedy Randomized Adaptive Search Procedure (GRASP), for optimally clustering N objects into K clusters. The proposed algorithm (Hybrid PSO-GRASP) for the solution of the clustering problem is a two phase algorithm which combines a PSO algorithm for the solution of the feature selection problem and a GRASP for the solution of the clustering problem. Due to the nature of stochastic and population-based search, the proposed algorithm can overcome the drawbacks of traditional clustering methods. Its performance is compared with other popular stochastic/metaheuristic methods like genetic algorithms and tabu search. Results from the application of the methodology to a survey data base coming from the Paris olive oil market and to data sets from the UCI Machine Learning Repository are presented.