Iterative Mining for Rules with Constrained Antecedents

In this study we discuss the problem of mining association rules with constrained antecedents in a large database of sales transactions or clickstream records. For a user-defined set A of items, this problem asks for computing all association rules (satisfying suitable support and confidence thresholds) induced by A, where an association rule is said to be induced by A if its antecedent (i.e., LHS) is a subset of A while the consequent (i.e., RHS) contains no items in A. In particular, we are interested in a multi-step scenario where in each step A is incremented by one item and all association rules induced by the updated A are to be computed. We propose an efficient iterative algorithm that can exploit mining information gained in previous steps to efficiently answer subsequent queries.