一种基于相容关系的聚类算法

Clustering on compatible relation

聚类分析是数据挖掘中一个重要研究内容。传统的聚类算法可划分为硬聚类和模糊聚类两大类,提出一种基于对象集上的相容关系的聚类算法,该算法通过极大相容簇来对数据对象集进行分类,使得同一对象可以属于不同的簇,而每个簇又有自己独有的成员对象,从而得到既不同于硬聚类也不同于模糊聚类的聚类效果。实验进一步表明了该算法的聚类的合理性。

Cluster analysis had played a very important role in data mining. This paper proposed a new algorithm based on compatible relation. The new algorithm grouped objects by the maximum compatible clusters and permited one object belonging to several different clusters while every cluster had its exclusive members, which gained a different clustering result from the traditional cluster algorithms. The experiments get a consistent result.