An approach for mining association rules intersected with constraint itemsets

Abstract

When the number of association rules extracted from datasets is very large, using them becomes too complicated to the users. Thus, it is important to obtain a small set of association rules in direction to users. The paper investigates the problem of discovering the set of association rules intersected with constraint itemsets. Since the constraints usually change, we start the mining from the lattice of closed itemsets and their generators, mined only one time, instead of from the dataset. We first partition the rule set with constraint into disjoint classes of the rules having the same closures. Then, each class is mined independently. Using the set operators on the closed itemsets and their generators, we show the explicit representations of the rules intersected with constraints in two shapes: rules with confidence of equal to 1 and those with confidence of less than 1. Due to those representations, the algorithm IntARS-OurApp is proposed for mining quickly the rules without checking rules directly with constraints. The experiments proved its efficiency.