最大频繁集的数据聚类方法

Maximum Frequent Itemsets－based Clustering Algorithm for Market Transaction

提出了一种新的聚类方法.针对货篮数据的特点,运用概念格获取最大频繁项目集,并以此作为初始聚类,采用适合于货篮数据的相似性测量方法,求得聚类结果.研究表明,该方法所生成的聚类比其它传统方法更优化,而且效率较高.

： A novel market basket analysis based clustering algorithm is proposed.According to the specific feature of market transaction,maximum frequent itemsets can be obtained via concept lattice.As the initial cluster,these maximum frequent itemsets can be used to acquire the clustering results,with the similarity measurement method adaptive to market transaction.Our study shows that the algorithm not only generates optimal clusters than traditional algorithms,but also exhibits good efficiency．