A Weighted Approach for Class Association Rules

Abstract

Class association rule mining is one of the most important studies supporting classification and prediction. Multiple researches recently focus on mining class association rules using support and confidence user-defined thresholds. However, in the real datasets, each attribute is associated with an indicator value. Based on the actual needs, in this paper, we propose a new approach which combines support, confidence and an interestingness measure (weight) to quickly improve the accuracy of class association rules.