关键词: PAM算法, 粒子群优化算法, 聚类分析, 有效性函数

Abstract: PAM is the first K-medoids algorithm proposed by one of the algorithm is relatively robust, but PAM sensitive to initial value, easily falling into local convergence. PSO algorithm is used to optimize the PAM, a approach of clustering based on PSO and PAM algorithm is proposed, making full use of both PAM and the PSO for the advantages of different issues, to continuously update the PAM clustering center. Through the establishment of cluster validity function based on entropy, the performance of the hybrid clustering algorithm is evaluated. the UCI data test results show that the hybrid clustering method has high accuracy of clustering.

Key words: Partitioning Around Medoid（PAM）, Particle Swarm Optimization（PSO）, cluster analysis, validity function