基于OWSFP-Tree的最大频繁项目集挖掘算法  

Maximal frequent itemsets mining algorithm based on OWSFP-tree

作　　者：赵志刚[1] 王芳[1] 万军[1] 

机构地区：[1]广西大学计算机与电子信息学院,广西南宁530004

出　　处：《计算机工程与设计》2013年第5期1687-1690,1807,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(60973074);广西教育厅科研基金项目(201106LX035)

摘　　要：为了解决FP-Max算法挖掘最大频繁项目集时递归生成条件模式树和超集检验的问题,提出了基于单向有序FP-Tree的NCFP-Max算法。该算法在挖掘过程中采用预剪枝策略减少挖掘结点,利用单向有序FP-Tree避免每次存储当前挖掘出的频繁项目集之前都需要超集检验,利用项目表格避免递归生成条件模式树减少时空资源。实验结果表明,在事务条数多、项的数量大的情况下,NCFP-Max算法的挖掘时间比FP-Max算法缩短了50%左右。To avoid generating recursively conditional pattern tree and superset checking in mining maximal frequent itemsets （MFS） for FP-Max, A new algorithm based on FP-Tree and one-way sequential FP-Tree （OWSFP-Tree） is put forward to improve the efficiency of mining MFI by several means. Firstly node number is reduced by applying the pruning strategy. Second OWSFP-Tree is used to avoid superset checking before storing current frequent itemsets every time. Third the item form is employed to avoid generating recursively conditional pattern tree. Experimental results show that the new algorithm when the transactional database have a large number of transactions and items can improve mining efficiency and shorten computational time about 50 % than FP-Max.

关 键 词：数据挖掘 关联规则 最大频繁项目集 FP-TREE 条件模式树 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]