一种新的频繁项集精简表示方法及其挖掘算法的研究

Research on a New Concise Representation of Frequent Itemset and Its Mining Algorithm

频繁项集挖掘是数据挖掘研究领域的一个基本问题,其瓶颈在于频繁项集全集的结果过多,冗余现象严重.主要的解决思路是只挖掘全体频繁项集中有代表性的子集,使得这种子集或者可满足应用的需要或者可由它们导出其他项集.最大项集和闭项集便是这类解决方案中两种最典型的子集形式.在最大项集和闭项集的基础上,提出了元项集这一新的频繁项集精简表示方法.首先,证明了最大项集和闭项集都是元项集的特例,且元项集所包含的项集数目介于二者之间;其次,讨论了元项集的性质.最后,通过在闭项集挖掘算法DCI-Closed-Index的基础上引入剪枝策略,设计了一个元项集挖掘算法.实验结果表明,所提出的挖掘算法是有效的和高效的.

Frequent itemset mining has become an important data mining task and a focused theme in data mining research. The bottlenecks of frequent itemset mining are as follows： On the one hand, the number of all frequent itemsets is usually extremely large. On the other hand, there is often severe redundancy in the resultant itemsets. To overcome these problems, recently several proposals have been made to construct a concise representation of the frequent itemsets, instead of mining all frequent itemsets. The aim is that the resultant subset can either satisfy the requirements of applications, or can derive all the other frequent itemsets. Maximal itemset and closed itemset are two most typical representative subsets of all frequent itemsets. Based on maximal itemset and closed itemset, a new concise representation of frequent itemset, namely meta itemset, is proposed. It is proved that both maximal itemset and closed itemset are special cases of meta itemset. The cardinality of meta itemset is between those of maximal itemset and closed ire：reset. Then, property of meta itemset is discussed. Finally, by introducing pruning strategies to DCI-closed-index, which is a closed itemset mining algorithm, an algorithm for mining meta itemset is proposed. Experimental results show that the proposed algorithm is effective and efficient.