自适应种群的高斯动态粒子群聚类算法  

Adaptive Population of Gaussian Dynamic PSO Clustering Algorithm

作　　者：沈亮[1] 常新功[1] 景丽荣[1] 

机构地区：[1]山西财经大学信息管理学院,山西太原030006

出　　处：《计算机系统应用》2010年第8期112-116,共5页Computer Systems & Applications

基　　金：国家自然科学基金(60873100);山西省高校科技研究开发项目(20081023)

摘　　要：聚类问题究其根本在于样本之间相似性的定义和聚类效果优劣的评价。粒子群聚类算法以其较好的聚类效果而受到广大研究者的关注。提出了一种新的衡量聚类效果的函数,并对其进行一定的分析。另外,从分析粒子群算法的拓扑结构出发,在高斯动态粒子群算法的基础上,提出了一种自适应种群的高斯动态粒子群聚类算法。实验表明,该衡量函数能够有效地评价聚类效果的优劣,其算法具有良好的聚类效果,在高维数据上表现优良。The key issue in Clustering is the definition of similarity between samples and the evaluation of pros and cons of clustering effects.PSO algorithm has drawn more attention from the majority of researchers for its preferable impact.This paper gives a new function that measures the effectiveness of the clustering algorithm and analyzes it thoroughly.In addition,from the topology of the PSO,an adaptive population of Gaussian dynamic PSO clustering algorithm is proposed based on the Gaussian dynamic algorithm.The experiment shows the measure function could effectively evaluate the pros and cons of clustering effects,and its corresponding algorithm has good clustering efficiency,better performance in the high-dimensional data.

关 键 词：聚类 粒子群算法 衡量函数 拓扑结构 自适应种群 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]