Knowledge Compilation for Itemset Mining

Mining frequently occurring patterns or itemsets is a fundamental task in datamining. Many ad-hoc itemset mining algorithms have been proposed for enumerating frequent, maximal and closed itemsets. The datamining community has been particularly interested in finding itemsets that satisfy additional constraints, which is a challenging task for existing techniques. In this paper we present a novel approach to itemset mining whereby the set of all itemsets are compiled into a compact form, closely related to binary decision diagrams. While there were previous attempts to utilize decision diagrams for storing the set of frequent itemsets, to the best of our knowledge, this is the first approach that does not rely on backtrack search to generate such a set. Instead, we use a pure knowledge compilation inference mechanism based on pairwise conjunctions of decision diagrams. Our empirical evaluation demonstrates that our approach is complementary to current state-of-the-art approaches. While we do not dominate these approaches, we show that our pure compilation approach can handle instances that remain out of reach for current state-of-the-art itemset miners.