基于激励的关联规则的挖掘

基于支持度的关联规则挖掘算法无法找到那些非频繁但效用很高的项集，基于效用的关联规则会漏掉那些效用不高但发生比较频繁、支持度和效用值的积（激励）很大的项集。提出了基于激励的关联规则挖掘问题及一种自下而上的挖掘算法HM-miner。激励综合了支持度与效用的优点，能同时度量项集的统计重要性和语义重要性。HM-miner利用激励的上界特性进行减枝，能有效挖掘高激励项集。

Motivation-based association rule mining

The existing algorithms for support-based Association Rule Mining (ARM) cannot find the itemsets that are not frequent but have high utility values, while Utility-Based Association Rule Mining (UBARM) cannot find the itemsets whose utility values are not high but the product of the support and utility of the same itemset (defined as motivation) is very large. This paper proposed motivation-based association rule and a down-top algorithm called HM-miner to find all high motivation itemsets efficiently. By integrating the advantages of support and utility, the new measure, i.e., motivation can measure both the statistical and semantic significance of an itemset. HM-miner adopted a new pruning strategy, which was based on the motivation upper bound property, to cut down the search space.