广义可能性C均值聚类算法  

General possibilistic C-means clustering algorithm

作　　者：文传军[1] 汪庆淼[2] 

机构地区：[1]常州工学院理学院,江苏常州213002 [2]苏州大学计算机学院,江苏苏州215021

出　　处：《计算机工程与科学》2015年第5期1015-1018,共4页Computer Engineering & Science

基　　金：国家自然科学基金资助项目(61170126);常州工学院校级课题资助项目(YN1305)

摘　　要：可能性C均值聚类算法(PCM)中模糊加权指标m要求大于1,通过对PCM算法的分析讨论,将PCM算法中模糊加权指标m设置为多个独立变量,且将其取值范围进行了扩展,称之为广义可能性C均值聚类(GPCM)。GPCM从理论上分析了加权指标m的扩展取值范围,并利用粒子群算法(PSO)对样本模糊隶属度进行估计。GPCM算法突破了PCM算法对参数m的约束。仿真实验验证了所提算法的有效性。The value range of fuzzy weighting exponent m is larger than 1 in the possibilistic C-means clustering （PCM）. Through analysis and discussion on the PCM algorithm, we set the weighting expo- nent m as multiple independent variables, and extend the value ranges of the weighting exponents, thus obtaining a new clustering algorithm,named general possibilistic C-means clustering （GPCM）. The new value scope of the GPCM＇s weighting exponents is proved theoretically, and the fuzzy membership of the samples is estimated by the particle swarm optimization （PSO） algorithm. The GPCM algorithm breaks the restriction of the PCM on parameter m,and simulation results demonstrate its effectiveness.

关 键 词：模糊C均值聚类 可能性C均值聚类 加权指数 模糊判决准则 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]