最大频集的挖掘方法  

Methods for Maximal Frequent Pattern Mining

作　　者：刘君强[1] 孙晓莹[1] 王勋[1] 

机构地区：[1]杭州商学院计算机信息工程学院,杭州310035

出　　处：《计算机工程》2003年第11期25-26,173,共3页Computer Engineering

基　　金：浙江省自然科学基金项目(602140);国家"863"计划 (2002AA121064);浙江省教育厅科技计划项目(20020635)

摘　　要：在对树生成的策略、模式支持集表示与投影、闭合性剪裁方法研究的基础上,提出了挖掘最大频繁集的两个新算法cap-max和op-max.cap-max是基于闭合频集的二阶段挖掘算法,先挖掘闭合频集,再从中求得最大频集;op-max是直接剪裁完全频集的算法,在频集生成过程中,通过前窥和包含关系检查及时剪裁搜索空间.实验表明,cap-max的时间效率比maxminer高1.5～4倍,op-cap的时间效率比maxminer高2～10倍.Two new algorithms for maximal frequent pattern mining, cap-max and op-max, are proposed in this paper. They employ the hybrid tree search strategy, heterogeneous representations and projections of transaction subsets, and tree pruning methods based on closure inclusion that were developed in the previous s tudy. cap-max discovers the closed frequent pattern set first, which is then use d to derive the maximal frequent pattern set. op-max generates the frequent patt ern set by opportunistic projection, the same time the frequent pattern set is p runed by looking ahead and checking general inclusion directly. Comparative expe riments on real world and artificial datasets show that cap-max outperforms maxm iner by a factor of one and half to four, and op-max outperforms maxminer by a f actor of two to ten.

关 键 词：知识发现 数据挖掘 最大频繁模式 关联规则 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论] TP391[自动化与计算机技术—计算机科学与技术]