Mining interesting sets and rules in relational databases

Abstract

In this paper we propose a new and elegant approach toward the generalization of frequent itemset mining to the multi-relational case. We define relational itemsets that contain items from several relations, and a support measure that can easily be interpreted based on the key dependencies as defined in the relational scheme. We present an efficient depth-first algorithm, which mines relational itemsets directly from arbitrary relational databases. Several experiments show the practicality and usefulness of the proposed approach.