关联规则的相似性度量与聚类研究

由于进行关联规则挖掘过程中会产生大量规则,给关联规则的后期分析与利用带来了巨大障碍.针对关联规则的特点,提出了一种新的规则相似性度量方法,通过相似性度量方法推出新的规则距离度量方法,运用系统聚类中的类平均法进行聚类.实验结果表明,该距离度量方法考虑了关联规则的整体信息,依据聚类谱系图和规则散点图,确定了类和类的个数,有利于规则的分类处理.

Research on similarity of association rules and clustering

In the process of mining association rules,lots of rules are gotten,which bring great obstacles to analyze and utilize the association rules later.According to the characteristics of association rules,a new similarity measurement method is proposed in cluster analysis.The new distance approach is deduced by similarity measurement method.Then average method of hierarchical clustering is used to cluster rules.The results show that the new distance is useful for rules.The class and number of class is determined based on hierarchical diagram and rules plot,and in favor of classification of rules.