Improving the performance of association classifiers by rule prioritization

Abstract

Numerous associative classification algorithms have been proposed but none considers the rule dependence problem, which directly influences the classification accuracy. Since finding the optimal execution order of class association rules (CARs) is a combinatorial problem, this study proposes a polynomial-time algorithm that re-ranks the execution order of CARs by rule priority to reduce the influence of rule dependence. The classification accuracy and recall rate of the associative classification algorithm are thus improved. The experimental results show that the proposed association classifier yields better classification results than those of an association classifier that does not consider rule dependence.