最適クラスター探索における粒子群最適化ファジーC-Means法のXie Beni尺度を用いた新しい手法の提案

抄録

Fuzzy C-Means (FCM) is a popular method used in clustering. FCM is enhanced　with an evolutionary algorithm called Particle Swarm Optimization (PSO) to avoid local optimal problem. On the other hand, Xie Beni Index (XBI) is used for evaluating the quality of clustering by measuring both of cluster compactness by intra-cluster deviation and separateness by inter-cluster pairwise distance of centers. In this paper we propose a new method in FCMPSO by minimizing XBI to evaluate the temporary solution. The result shows that our new method can improve the quality of clustering.