A Bayesian Network Model for Interesting Itemsets

Mining itemsets that are the most interesting under a statistical model of the underlying data is a frequently used and well-studied technique for exploratory data analysis. The most recent models of interestingness are predominantly based on maximum entropy distributions over items or tile entries with itemset constraints, and while computationally tractable are not easily interpretable. We therefore propose the first, to the best of our knowledge, generative model over itemsets, in the form of a Bayesian network, and an associated novel measure of interestingness. Our model is able to efficiently infer interesting itemsets directly from the transaction database using structural EM, in which the E-step employs the greedy approximation to weighted set cover. Our approach is theoretically simple, straightforward to implement, trivially parallelizable and exhibits competitive performance as we demonstrate on both synthetic and real-world examples.