An Interestingness Measure and Computation Method of Association Rules Based on Frequent Itemsets Relatedness

To address inadequacy of association rules interestingness measure method currently, we present a novel method to measure interestingness with relatedness among items in frequent itemsets. It firstly computed relatedness between frequent k-itemsets and each subset of frequent 2-itemsets, which is a linear combination of Complementarity Intensity (CI), Substitutability Intensity (SI) and Mutual Interaction (MI). The mean of relatedness of all frequent 2-itemsets subsets was regarded as relatedness of frequent k-itemsets. Finally weighted computation method of association rule interestingness was given according to principle of objective interestingness of association rule is inversely proportional to relatedness of frequent itemsets. The method can not only sort rules, but also analyze actual relationship among all items in frequent 2-itemsets, which is conductive to selection of users on rules.