加权共协矩阵聚类融合研究  

Study on Weighted Cluster Ensemble of Co-association Matrix

作　　者：李丹丹[1] 褚丽丽[1] 

机构地区：[1]辽宁工程技术大学系统工程研究所,辽宁葫芦岛125105

出　　处：《科技和产业》2010年第1期89-91,118,共4页Science Technology and Industry

基　　金：国家自然科学基金项目(70971059)

摘　　要：当前的聚类融合算法大多不考虑进行融合的成员的质量,当聚类成员存在聚类质量差或者有噪声干扰时,融合结果将受到影响。本文提出了一种对聚类成员进行加权的融合算法。该方法对聚类成员的质量进行简单评价,并据此对其赋予权重,生成加权共协矩阵,能够有效的消减聚类成员质量差异和噪声对融合的影响,进而得到更好的融合结果。Resent research on cluster ensemble mostly pays little attention to the qualities of cluster members, poor qualities and noise in cluster members may affect the ensemble result. This paper proposed a weighted cluster ensemble approach. This approach made a simple evaluation of the quality of clustering members, and set weights to all cluster members according to their qualities. Then weighted co-association matrix was generated. This approach can effectively reduce the effect of poor qualities and noise in cluster members, therefore it can afford better ensemble result.

关 键 词：聚类融合 权重 共协矩阵 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]