基于数据立方体的高效关联规则挖掘算法  

High efficiency algorithm based on data cube for mining association rules

作　　者：施化吉[1,2] 丁云磊[1] 

机构地区：[1]江苏大学计算机科学与通信工程学院,江苏镇江212013 [2]南京航空航天大学计算机应用研究所,江苏南京210016

出　　处：《计算机工程与设计》2009年第2期416-418,422,共4页Computer Engineering and Design

基　　金：国家火炬计划基金项目(2004EB33006[0]);江苏省高校自然科学指导性计划基金项目(05JKD520050)

摘　　要：关联规则挖掘过程中,为了得到侯选项集的支持度,需要将侯选项集与数据库中事务逐一进行比较,影响了算法的执行效率。针对该问题,提出一种基于数据立方体的关联规则挖掘算法ABDC。该算法结合了属性分组的思想,将生成的侯选项集进行编码后,利用数据库中事务的包含关系,在数据立方体中迅速查找到包含该侯选项集的所有事务在数据库中的出现频率,依次累加后根据最小支持量得到数据库中存在的频繁项集,不仅对事务数量的递增具有不敏感性,而且提高了算法的执行效率。实验结果表明,该算法对挖掘大量事务集中存在的关联规则是快速有效的。We need compare candidate items with every transaction for calculating the support of candidate items while mining association rules. This process influences the efficiency. To solve this problem, the algorithm based on data cube is proposed. This algorithm combines with the thinking of attributes dividing, coding the candidate items. Take advantage of inclusive relationship, the appearing frequency of all transaction including this candidate items is gotten through looking up the data cube. We get all frequent items through accumulating all the appearing frequency and comparing with the minimum support count. Not only non-sensitive to the transaction increasing, but also it improves the algorithm efficiency. The experiment proves that the algorithm is fast and effective for mining the association rules in the large-scale transaction collection.

关 键 词：关联规则 数据立方体 属性分组 包含关系 不敏感性 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]