基于实数值链接分析的ESSC融合算法  

Enhanced soft subspace clustering ensemble based on real valued link analysis

作　　者：王丽娟[1] 郝志峰[1,2] 蔡瑞初[2] 温雯[2] 

机构地区：[1]华南理工大学计算机科学与工程学院,广州510006 [2]广东工业大学计算机学院,广州510006

出　　处：《计算机应用研究》2014年第5期1366-1369,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(61070033,61100148,61202269);广东省自然科学基金资助项目(S2011040004804);广东省科技计划项目(2010B050400011);软件新技术国家重点实验室开放项目(KFKT2011B19);广东高校优秀青年创新人才培育项目(LYM11060);广州市科技计划项目(12C42111607,201200000031);番禺区科技计划项目(2012-Z-03-67)

摘　　要：为了进一步提升ESSC聚类融合性能,采用实数值链接分析(real valued link analysis)计算聚类融合中模糊数据类的相似性。根据模糊决策及其相似性定义优化的融合信息,从而达到改进聚类性能的目的。实验选用了两个仿真数据库和五个UCI数据库。实验结果表明,基于实数值链接分析的ESSC聚类融合算法(RLA-ESSCE)的性能优于K-means聚类算法(KMC)、ESSC、ESSCE。In order to further improve the performance of ESSCE, real-valued link analysis had been proposed to compute the similarity between fuzzy clusters in ESSC clustering ensemble （RLA-ESSCE）. The clustering ensemble information was refined according to fuzzy decision and its similarity. Therefore the performance according to refined clustering ensemble information had been improved. Experiments were conducted on two synthetic datasets and five UCI datasets. Experimental results show that RLA-ESSCE is better than K-means clustering （KMC） ,ESSC and ESSCE.

关 键 词：增强的软子空间聚类 聚类融合 实数值链接分析 聚类融合信息 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TP301.6[自动化与计算机技术—控制科学与工程]