Incorporating Relevance Measures into Negative Association Rule Mining and its Acceleration

Abstract

The purpose of this study is to add new effective evaluation methods of negative rules with statistical measures. Also, we propose a branch and bound mining of top-k rules, and give a filtering method using weak relevance to negative rules mining. Besides, we propose a data structure for achieving high speed computation. Thus, we can realize efficient and effective negative rules mining. We also show some good results of experiments for evaluating our proposed method.