基于重叠度增量的模糊聚类有效性函数  

Fuzzy Clustering Validity Function Based on the Increment of Overlap

作　　者：欧卫华[1] 

机构地区：[1]怀化学院数学系,湖南怀化418008

出　　处：《计算技术与自动化》2009年第4期115-118,共4页Computing Technology and Automation

基　　金：湖南省教育厅资助科研项目(07C507);湖南省教改资助项目(湘教通263号191)

摘　　要：提出用重叠度来刻画模糊类间的距离,在此基础上针对模糊划分总重叠度有随类数增加而单调递增的趋势,提出基于重叠度增量的聚类有效性函数。该算法由重叠度增量最大值来确定最佳聚类数,不但克服了传统有效性函数的单调问题,而且计算简单。基于模糊C-均值聚类算法(FCM),应用多组测试数据对其进行性能分析,并与当前广泛应用且具代表性的有效性函数进行深入比较。仿真结果表明,该函数的有效性和优越性。The method was proposed for measuring distance of fuzzy cluster by overlap, a new validity function based on the incremental of overlap was proposed against the overlap of fuzzy patition increased monotonous with the increasing number of categories. The best number of cluster determined by the maximun increment of overlap, the algorithm not only overcome the effectiveness of the tradtional validity function ,but is simple. Extensive tests of the index in a conventional model selection （FCM）algorithm have been performed using generated data sets and public domain data sets, and eomparision with several existing and important indices has been made. The results show clearly the effficiency and advantages of the new index.

关 键 词：聚类有效性 重叠度增量 模糊C-均值算法 聚类数 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]