Approximate frequent itemsets compression using dynamic clustering method

Frequent-itemsets mining often faces the problem of generating a large collection of frequent itemsets, which is too large to be carefully examined and understood by the users. To reduce the output size of frequent itemsets, we propose using a dynamic clustering method to compress the frequent itemsets approximately in this paper. Concretely, two frequent itemsets intra-cluster similarities, expression similarity and support similarity, are defined according to the specific requirements of frequent itemsets compression. Based on the above two similarity measures, the frequent itemsets clustering criterion and its related clustering algorithm are developed. Specially, our method has two features: 1)users neednpsilat specify the number of frequent itemsets clusters explicitly; 2)userpsilas expectation of compression ratio is incorporated. Our initial experimental results show that our approximate frequent itemsets method is feasible and the compression quality is good.