Research of Mining Effective and Weighted Association Rules Based on Dual Confidence

Abstract

Association rule is an important model in data mining. However, traditional association rules are mostly based on the support and confidence metrics, and most algorithms and researches assumed that each attribute in the database is equal. In fact, because the user preference to the item is different, the mining rules using the existing algorithms are not always appropriate to users. By introducing the concept of weighted dual confidence, a new algorithm which can mine effective weighted rules is proposed in this paper, which is on the basis of the dual confidence association rules used in algorithm. The case studies show that the algorithm can reduce the large number of meaningless association rules and mine interesting negative association rules in real life.