A human-centered approach for mining hybrid-dimensional association rules

This work presents a new method to mine hybrid-dimensional association rules in databases originated from multiple sources. We adopted an approach in which hybrid associations represent transactional rules that become either exceptionally weaker or exceptionally stronger in some subsets of an integrated database, which satisfy specific conditions over selected attributes. We propose new interest measures to evaluate hybrid-dimensional rules, as well as an algorithm to mine these patterns. This algorithm was applied to a real dataset that keeps information about purchases made by families residing in different Brazilian cities. The obtained results show that the proposed technique provides valuable information for decision making.