L-MAX频繁项集及挖掘算法

提出了项集长度受限且生成项集对应事务信息的最大频繁项集挖掘问题,定义为L-MAX频繁项集挖掘,并重点研究了项集长度约束特征和事务集信息的存储与生成策略.首先研究了L-MAX频繁项集的性质,然后扩展FP-tree提出了ExFP-tree结构并给出ExFP-tree生成算法.ExFP-tree利用FP-tree共享前缀路径的性质通过共享子孙节点事务信息策略实现大量事务信息的压缩存储;最后基于FP-MAX算法,提出基于ExFP-tree的L-MAX频繁项集挖掘算法,核心思想是先根据L-MAX频繁项集长度约束性质进行前瞻剪枝再进行最大频繁项集挖掘,并通过回溯策略直接定位生成对应事务集.

L-MAX Frequent Itemsets and Algorithm of Mining L-MAX Frequent Itemsets

Presented a problem of mining maximal frequent itemsets with constrained itemsets-length and generating corresponding transaction information of itemsets ,which is defined as L-MAX frequent itemsets mining. The constraint characteristic of itemsets-length and the strategy of transaction sets storaged and generated is emphatically studied in this paper. Firstly the properties of L-MAX frequent itemsets is analysed. Then by extending FP-tree structure an ExFP-tree data structure is pre- sented and the algorithm of ExFP-tree generated is given. According to the character of prefixing paths shared in FP-tree, ExFP-tree implements compressingly storage of transaction information with sharing the transaction information of descendant nodes. Finally based on FP-MAX algorithm,an algorithm of mining L-MAX frequent itemsets based on ExFP-tree is presented. The main idea is firstly using look-ahead pruning method process ExFP-tree according to the properties of L-MAX frequent itemsets and then mining maximal frequent itemsets. To generate the corresponding transaction sets backdating and direct locating strategy is used.