An Improvement of Fuzzy C-Means Clustering Using Adaptive Particle Swarm Optimization

Abstract

Fuzzy C-Means (FCM) algorithm is one of the most popular fuzzy clustering techniques. However, it is easily trapped in local optima. Particle swarm optimization (PSO) is a stochastic global optimization model, which is used in many optimization problems. In this paper, a hybrid clustering algorithm, called HAPF, based on adaptive PSO (APSO) and FCM is proposed, in order to take advantage of the merits of both APSO and FCM. In HAPF the state of swarm aggregation is divided into three situations: strong, loose, and medium, respectively representing swarm's exploitation phrase, exploration phrase, and a balance between the two phrases. In addition, the interval of population diversity measured by the variance of population fitness is partitioned into three sections. After mapping the relationship between the swarm aggregation situations and the value of population diversity, three inertia weight groups are dynamically adjusted accordingly to the real-time state of population diversity. Experimental results show that the proposed HAPF is able to escape local optima and find better optima than other seven well-known clustering algorithms.