负关联规则挖掘中的频繁项集爆炸问题  

Exploding number of frequent itemsets in the mining of negative association rules

作　　者：马占欣[1] 陆玉昌[2] 

机构地区：[1]周口师范学院计算机科学系,周口466001 [2]清华大学计算机科学与技术系智能技术与系统国家重点实验室,北京100084

出　　处：《清华大学学报（自然科学版）》2007年第7期1212-1215,共4页Journal of Tsinghua University(Science and Technology)

基　　金：国家自然科学基金资助项目(60473115)

摘　　要：为了解决负关联规则挖掘中包含负项目的频繁项集数量爆炸问题,引入过频繁项集的概念,证明过频繁项集及其所有超集均不可能产生有趣的关联规则。在频繁项集生成过程中,必须对项集中包含负项目的个数进行限制。使用最小支持度、最大支持度、项集中最多包含负项目的个数2个参数来约束频繁项集生成过程的算法。实验结果表明:所增加的2个参数是必要的,特别是当数据集中所包含的项目个数比较多时,只有恰当地设置2个额外参数,才能够保证挖掘过程的正常进行。This paper uses the over-frequent itemset concept to solve the problem of an exploding number of frequent itemset including negative items when mining negative association rules. The analysis shows that with the over-frequent itemset and all its interesting association rules cannot be generated and an upper bound to the negative items must be included in the itemset during the generation of frequent itemset. The paper brings an algorithm which restricts the generation of frequent itemset by using 3 parameters： min support, max support, and the maximum number of negative item in the itemset. Test results show that the latter two parameters are necessary. More specifically, when the data pool has a large number of items, the normal mining process can only be ensured by appropriately setting these two additional parameters.

关 键 词：信息处理 数据挖掘 负关联规则 过频繁项集 最大支持度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]