一种基于PSO&PAM的聚类算法  

Approach of clustering based on PSO & PAM algorithm

作　　者：黄翔 蔡碧野[1] 孟颖[1] 

机构地区：[1]长沙理工大学计算机与通信工程学院,长沙410114 [2]湖南大众传媒职业技术学院网络传媒系,长沙410100

出　　处：《计算机工程与应用》2013年第4期149-152,共4页Computer Engineering and Applications

基　　金：国家自然科学基金(No.10926189;No.10871031);湖南省教育厅重点项目(No.10A015)

摘　　要：PAM是最早提出的k-medoids算法之一,该算法比较健壮,比k-means算法鲁棒性更强,但是PAM对初始值敏感,易陷入局部收敛。利用PSO算法对PAM进行优化,提出一种基于PSO和PAM的聚类方法,充分利用PAM和PSO两者对于不同问题的优势,来不断地更新PAM的聚类中心。通过建立基于熵的聚类有效性函数,对混合聚类算法的性能进行客观评价。从来自UCI的数据的测试结果表明,这种混合聚类的方法有较高的聚类正确率。PAM is the first K-medoids algorithm proposed by one of the algorithm is relatively robust, but PAM sensitive to ini- tial 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 contin- uously update the PAM clustering center. Through the establishment of cluster validity function based on entropy, the perfor- mance of the hybrid clustering algorithm is evaluated, the UCI data test results show that the hybrid clustering method has high accuracy of clustering.

关 键 词：PAM算法 粒子群优化算法 聚类分析 有效性函数 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]