Extension of local association rules mining algorithm based on apriori algorithm

Abstract

The support is generally higher when the classical apriori algorithm is used as mining data based on association rules, if the support is small low then redundant frequent item set and redundant rules are produced large, so the local effective association rules has a larger confidence and a smaller support can not be mined out, which is the fatal defects of the classical apriori algorithm. According to the defects, the effectiveness of local rules is proved at first, meanwhile, two kinds of the correction algorithms are given: the one is apriori-con algorithm based on confidence and the other is apriori algorithm based on classification which is further divided into three kinds, apriori-class-int algorithm based on interest classification, apriori-class-pre algorithm based on forecast classification and apriori-class-clr algorithm based on clustering classification. The correctness of the theory is proved in the article and the effective of the correction algorithms is showed by cases.