基于等价类的关联规则挖掘矩阵算法  

An Association Mining Matrix Algorithm Based on Equivalence Class

作　　者：王鸿铭[1] 沈夏炯[1] 李国雁[1] 臧国轻[1] 

机构地区：[1]河南大学计算机与信息工程学院,河南开封475004

出　　处：《计算机技术与发展》2008年第4期55-58,共4页Computer Technology and Development

基　　金：河南省自然科学基金项目(0311011700)

摘　　要：关联规则挖掘算法中的Apriori算法利用查找频繁项集来发现数据集中的关联规则,算法思路简单容易实现;但在由k-1次频繁项集生成k次频繁项集时需反复查找数据库,效率较低,在寻找高次频繁项集时其低效性更加明显;矩阵算法是通过直接查找高次频繁项集,避免了反复查找数据库,但要存储大量的非频繁项集,且查找低次频繁项集速度较慢。文中提出的矩阵等价类算法,利用等价关系进一步降低矩阵算法的时间空间复杂度,然后通过项目相似度直接求取所有最大频繁项集。实验结果证明了算法的可行性、高效性。Apriori algorithm,one of associated rules of mining algorithms, utilizes searching results on frequent itemsets to find associated rules in datasets. Though the algorithm is easy to implement, it spends enormous time and becomes much more unefficient on searching in the databases, especially when k times frequent itemsets are created by k-1 ones and searching for the high level frequent itemsets. The matrix algorithm avoids these situation but needs store substantive unfrequent itemsets and is low velocity when searching for low level frequent itemsets. The matrix-equivalence genus algorithm mentioned in this paper, is proved that can reduce the complexity on time and room effectively on finding all most frequent itemsets.

关 键 词：等价类 项目相似度 矩阵算法 APRIORI算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]