摘要： 针对在某些限定项目数与交易长度数据的关联规则挖掘中FP-growth算法执行效率很低的问题，提出一种最大频繁模式挖掘算法，该算法引入与FP-tree结构类似的All-subset tree存储所有的最大频繁项目集，无需在扫描数据库前指定最小支持度，可以动态给定最小支持度而不用重新扫描数据库。实验结果表明，该算法在这些特定数据的挖掘中，与FP-growth相比明显提高了挖掘效率。

关键词: 数据挖掘, 关联规则, 频繁模式树, 最大频繁项目集

Abstract: Aiming at the low mining efficiency problem existing in the mining association rules of FP-growth algorithm on data limited with item counts and transaction length, an improved algorithm(All-subset tree) for mining maximal frequent patterns is proposed, a novel data structure, All-subset tree, which is similar to FP-tree is introduced to store all maximal frequent item sets. The algorithm is in no need of appointed minimum support before scanning the database, and need not rescan the database when assigned minimum support dynamically. Experimental results show that all-subset tree algorithm greatly improves the mining efficiency compared with FP-growth algorithm when mining specific database.

Key words: data mining, association rule, Frequent Pattern tree(FP-tree), maximal frequent itemsets

中图分类号: 

• TP311.12