关联规则挖掘的PredictiveApriori算法的研究及改进  

Research andimprovement on PredictiveApriori's algorithm mining with association rules

作　　者：吴芝明[1] 钱程[1] 伍少梅[1] 

机构地区：[1]四川大学计算机学院,成都610064

出　　处：《四川大学学报（自然科学版）》2012年第1期97-101,共5页Journal of Sichuan University(Natural Science Edition)

摘　　要：在关联规则挖掘算法中,PredictiveApriori算法只需设定输出最好的N个规则,就可挖掘出N个预测精度最大的规则,从而极大地减轻用户参数设置的负担.但该算法需要不断对事务数据库进行扫描,如果事务数据库很大将会大大增加I/O时间,降低挖掘效率.本文融合事务压缩技术,提出一种称为Fast PredictiveApriori的改进版算法.实验结果表明,随着事务数据量的增大,改进后的算法只需扫描压缩事务集.因此,在事务数据库很大时,该算法能明显提高执行效率.PredictiveApriori algorithm can find out the most accuracy N rules when setting the parameter N, so the algorithm greatly reduces the burden of the user＇s parameters setting. However, this algorithm increases the I/0 time and hinders the mining efficiency because of scanning the transaction database continuously. This paper provides an improved method by combining affairs compression technology and PredictiveApriori. The method simply scans the compressed transaction sets and the experiment results show that the improved method can improve the efficiency significantly with transaction database increases.

关 键 词：关联规则 PredictiveApriori 事务压缩 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]