一种新的关联规则抽样算法  

A new sampling algorithm for association rule

作　　者：秦如新[1] 陈静[1] 冯一宁[1] 

机构地区：[1]中国农业大学理学院,北京100083

出　　处：《中国农业大学学报》2007年第3期85-88,共4页Journal of China Agricultural University

基　　金：国家自然科学基金资助项目(1037113160573158)

摘　　要：针对目前经典的关联规则挖掘Apriori算法需对数据库多次扫描费时多计算量大,而抽样扫描会造成挖掘精确度下降等问题,采用控制样本频繁项目集的方法,利用频繁1项集进行抽样处理,对关联规则挖掘的抽样操作和精度控制进行研究,提出了基于抽样操作的关联规则挖掘算法——HAC算法。理论分析及性能试验结果表明:HAC算法能够有效缩减数据库规模,至少少扫描数据库1次,提高了关联规则挖掘的效率,同时其计算精度不受影响。In order to reduce the long time spent for scanning the database by using Apriori algorithm, which may descend the mining accuracy, the research on the sample operation and precision control with the help of frequent itemset, especially, the frequent 1-item-set is presented in this paper. The HAC algorithm based on sampling was de- signed. The results in theory and capability experiment indicated that HAC algorithm could decrease the scanning times by at least once, promote the efficiency of mining and improve the computation precision.

关 键 词：关联规则 抽样 准则系数 APRIORI算法 HAC算法 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]