最大目标频繁模式挖掘算法研究

The Research of Maximum Target Frequent Pattern Mining Algorithm

传统的频繁模式挖掘算法往往会得到成百上千的结果模式,面对繁多的频繁模式用户通常要经过“二次挖掘”才能得到有用的目标模式。怎样根据用户需求直接挖掘用户感兴趣的目标模式是该文的研究目标。文章在FP-树的基础上设计了紧缩的、非冗余的TFP-树,它能有效过滤与目标模式无关的项和事务,而仅保留与目标模式相关的信息,缩小TFP-树的大小规模。同时根据TFP-树的规律和特点,笔者设计了最大目标频繁模式挖掘算法,算法的结果模式具有以下两个特点:(1)满足用户需求的目标模式;(2)最大模式。该实验结果验证了TFP-树算法是有效的,而且显著改善了FP-树算法的性能。

Traditional frequent pattern mining algorithms always produce hundreds of result patterns,so facing numerous frequent patterns,users have to mine second time to get the useful target patterns.How to directly mine the target patterns which is the users interested in is this paper's research motive.Basing on the frequent pattern tree(FP-tree),it designs a compressed and non-redundant target frequent pattern tree(TFP-tree).TFP-tree can filter the items and transactions which doesn't contribute to the target pattern,instead,it only preserves the related information,so the size of the TFP-tree is greatly reduced.According to the TFP-tree's properties,we put forward a maximum target frequent pattern mining algorithm,which have contributions in following two things:(1)directly mining target patterns which can satisfy the user's require;(2)mining the maximum patterns.The experiments results show that TFP-tree is very effective and it also can greatly improve the FP-tree's performance.