Multiple-step rule discovery for associative classification

Associative classification has shown great promise over many other classification techniques. However, one of the major problems of using association rule mining for associative classification is the very large search space of possible rules which usually leads to a very complex rule discovery process. This paper proposes a multiple-step rule discovery approach for associative classification called Mstep-AC. The proposed Mstep-AC approach focuses on discovering effective rules for data samples that might cause misclassification in order to enhance classification accuracy. Although the rule discovery process in Mstep-AC is performed multiple times to mine effective rules, its complexity is comparable with conventional associative classification approach. In this paper, we present the proposed Mstep-AC approach and its performance evaluation