基于项权值变化的完全加权正负关联规则挖掘  

All-Weighted Positive and Negative Association Rules Mining Based on Dynamic Item Weight

作　　者：周秀梅[1] 黄名选[2] 

机构地区：[1]南宁地区教育学院数学与计算机科学系,广西崇左532200 [2]广西财经学院信息与统计学院,广西南宁530003

出　　处：《电子学报》2015年第8期1545-1554,共10页Acta Electronica Sinica

基　　金：国家自然科学基金(No.61262028;No.61363037);广西自然科学基金(No.2012GXNSFAA053235);广西财经学院数量经济学创新团队项目(No.2014CX01)

摘　　要：本文提出一种基于项权值变化的完全加权正负关联规则挖掘算法,解决了基于项权值变化的负模式挖掘问题.该算法考虑项权值依赖于事务记录的特点,采用新的项集剪枝方法和模式评价框架,通过项集的项内权值比和维数比的简单计算和比较,挖掘有效的完全加权正负关联规则.实验结果表明,与现有无加权正负关联规则挖掘算法比较,本文算法能避免无效的模式出现,其挖掘时间和候选项集数量明显减少,减幅最大分别可达94.09%和88.16%.This paper proposes a mining algorithm of aU-weighted positive and negative association rules based on dynamic item weight, which can solve the problems of negative patterns mining based on dynamic item weight. This algorithm took the dynamic item weight dependent on transaction records into consideration,and adopted the itemset pruning method and pattem evaluation framework so as to discover effective all-weighted positive ＆ negative association rules via simple calculation and comparison of weight ratio and dimension ratio from the itemset. The expe ＇rtmental results show that this algorithm can prevent ineffective patterns, which makes the maximal declines of the mining time and number of the candidate itemsets by up to 94.09% and 88.16% respectively compared with the existing unweighted positive and negative association rule mining algorithms.

关 键 词：数据挖掘 完全加权关联规则 负关联规则 频繁项集 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]