Mining Positive and Negative Association Rules from Large Databases

This paper is concerned with discovering positive and negative association rules, a problem which has been addressed by various authors from different angles, but for which no fully satisfactory solution has yet been proposed. We catalogue and critically examine the existing definitions and approaches, and we present an a priori-based algorithm that is able to find all valid positive and negative association rules in a support-confidence framework. Efficiency is guaranteed by exploiting an upward closure property that holds for the support of negative association rules under our definition of validity