一种适应大型数据库的多支持度关联规则算法  

New algorithm for mining association rules with multiple supports in large databases

作　　者：薛永庆[1] 徐维祥[1] 

机构地区：[1]北京交通大学交通运输学院,北京100044

出　　处：《计算机工程与应用》2008年第2期182-185,共4页Computer Engineering and Applications

基　　金：北京市教委科技发展重点项目(No.KZ200710028014)。

摘　　要：关联规则挖掘一直是数据挖掘中的重要组成部分。提出一个新算法DPCFP-growth算法。DPCFP-growth算法是基于MSApirori算法,采用了CFP-growth分而治之的思想,并弥补了CFP-growth算法的不足。CFP-growth算法运行时要把整个数据库中的数据压缩到一个MIS-tree中然后进行频繁模式挖掘。在大型数据库中CFP-growth算法会建立一个深度很深宽度很宽的CFP-tree,以至于内存往往不能满足其要求,被迫使用大量的辅存,致使算法的运行效率急剧下降。DPCFP-growth算法根据CFP-tree的特征,有效地把大数据库分为若干个内存可以满足其要求的子数据库,然后在每个子数据库中进行局部频繁模式挖掘,最终汇总这些频繁模式生成全局频繁模式。实验表明该算法是正确的,并且在大型数据挖掘中,比CFP-growth算法有一定的优越性。Mining association rules from a large database have been described as an important problem of database mining.In this paper,a novel algorithm DPCFP-growth is proposes.The algorithm based upon the MSApirori takes advantage of the advantage and offsets the disadvantage of CFP-growth that CFP-growth always needs so much EMS memory in large database that the computer can＇t meet it.In DPCFP-growth algorithm,according to characters of MIS-tree,firstly partition a large database into some smaller databases which EMS memory of computer can meet them.Secondly the algorithm takes in the thinking of the CFPgrowth to mining local frequent patterns.Finally ,the algorithm parses the set of all the local frequent patterns to get final frequent patterns.The experiments show that the DPCFP-growth algorithm has more superiority to previous algorithm.

关 键 词：数据挖掘 数据库划分 多支持度 频繁模式 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]