一种改进的基于关联图的关联规则挖掘算法

A revised graph-based algorithm for mining association rules

关联规则是数据挖掘研究的一个重要课题 ,而最大频繁项集的生成是影响关联规则挖掘的关键问题 .在已有的频繁集发现算法中 ,DLG算法通过减少事务数据库的扫描次数 ,进而有效减少挖掘过程的I/O代价 .在阐述DLG算法的实现原理与执行过程的基础上 ,为进一步减少候选项集的数量 ,提出一种改进算法DLG .其主要思想是在关联图构造阶段 ,统计每一个频繁项目的入度 ,以此作为剪枝的依据 .

Mining association rules is an important part of data mining field. An d generating the frequent itemsets is a key problem of mining association rules. Among the proposed algorithms of finding frequent itemsets, DLG is a efficient algorithm to controls I/O cost by reducing the number of database passes. The pr inciple and implemental process are discussed, and then a revised algorithm is p resented based on DLG in order to cut down the number of candidates further. The p rinciple of DLG is to count the in-degree of each frequent itemset on which a p runing is based in the phase of constructing graphs. Finally, the performance an alysis and comparison experiments are done and the result shows the algorithm is excellent.