聚类集成方法研究.PDFVIP

38 2 Vol.38 No.2 20112 Computer Science Feb 2011 ( 吉林大学计算机科学与技术学院 长春130012) ( 吉林大学符号计算与知识工程教育部重点实验室 长春130012) 聚类集成通过对原始数据集的多 个聚类结果进行学习和集成, 得到 一个能较好地反映数据集内 结构的数 据划分聚类集成能够较好地检测和处理孤立点,提高聚类结果质量综述了聚类集成的相关知识,介绍了聚类集成 的相关概念和优点;根据使用的聚类算法介绍了3 种产生聚类成员方法,分析了各自的优缺点及适用条件; 介绍了目 前已有的一致性函数, 阐述了其基本原理,并指出了其局限; 最后讨论了未来的研究方向 聚类集成,聚类成员, 一致性函数,聚类算法 TP391 Research on Cluster Aggregation Approaches Y NG Caoyuan LIU Dayou Y NG Bo CHI Shuzhen JIN Di (College of Computer Science and T echnology,Jilin U niversity, Changchun 130012, China) ( Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education,Jilin University, Changchun 130012, China) Abstract Clustering aggregation can offer a partition that could better reflect the inherent structure of the data set by studying and integrating many clustering results of the original data set . Clustering aggregation could detect and deal with the isolated points preferably,which improves the quality of clustering.T his paper made an overview of the rele vant knowledge of the clustering aggregation, presented the concepts and advantages of clustering aggregation.It pre sented three approaches to get clustering members according to the uesd cluster algorithms; analysed their respective advantages, disadvantages and application conditions; presented the existing consensus functions; explained the basic principles and pointed out their limitations.Finally, it discussed the future research directions. Keywords Clustering aggregation, Clustering member, Consensus function,Clustering algorithms [7]