基于矩阵相乘的Apriori改进算法  

An Improved Apriori Algorithm Based On Matrix Multiplication

作　　者：王蒙 方睿[1] 邹书蓉[1] WANG Meng;FANG Rui;ZOU Shurong(College of Computer Sciences,Chengdu University of Information Technology,Chengdu 610225)

机构地区：[1]成都信息工程大学计算机学院,成都610225

出　　处：《计算机与数字工程》2018年第10期1974-1979,共6页Computer & Digital Engineering

基　　金：科技厅重点研发项目(编号:2017GZ0331)资助

摘　　要：Apriori算法是一种经典的关联规则挖掘算法,算法能够很好地挖掘出关联规则,通过对频繁项集的连接步和剪枝步得到候选集,但是还要对大量候选集进行多次重复扫描数据库,产生庞大的候选集,严重影响了算法执行效率。论文提出一种基于矩阵的改进算法,通过事务矩阵和项集矩阵相乘来改进反复回扫数据库的问题,建立事务数组统计删除在算法执行过程中不能生成下一频繁集的事务,优化Apriori算法对频繁项集的连接步和剪枝步过程。通过实验验证改进算法不仅能准确地挖掘出频繁项集而且大大地缩短挖掘时间。Apriori algorithm is a kind of classical association rule mining algorithm. Although the algorithm has a lot of pruning of candidate sets, it is necessary to scan the database repeatedly, which seriously affects the efficiency of algorithm execution. In this paper, an improved algorithm based on matrix is proposed to improve the problem of repeating the scan database by multiplying the transaction matrix and the itemsets matrix. The establishment of transaction array to delete the transaction which can not generate the next frequent set, and optimize the Apriori algorithm for judging and connecting frequent itemsets. Experiments show that the efficiency of the improved algorithm is significantly higher than that of the Apriori algorithm.

关 键 词：关联规则 APRIORI算法 矩阵相乘 频繁项集 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]