An Efficient Approach to Mine Rare Association Rules Using Maximum Items' Support Constraints

摘要

Rare association rule is an association rule consisting of rare items. It is difficult to mine rare association rules with a single minimum support (minsup) constraint because low minsup can result in generating too many rules (or frequent patterns) in which some of them are uninteresting. In the literature, "maximum constraint model," which uses multiple minsup constraints has been proposed and extended to Apriori approach for mining frequent patterns. Even though this model is efficient, the Apriori-like approach raises performance problems. With this motivation, we propose an FP-growth-like approach for this model. This FP-growth-like approach utilizes the prior knowledge provided by the user at the time of input and discovers frequent patterns with a single scan on the transactional dataset. Experimental results on both synthetic and real-world datasets show that the proposed approach is efficient.