A Hellinger-Based Importance Measure of Association Rules for Classification Learning

Abstract

Classification learning with association rules has been an active research area during recent years. Thus, it is important to establish some numerical importance measure for association rules. In this paper, we propose a new rule importance measure, called a HD measure, using information theory. A num ber of properties of the new measure are analyzed, and its classification performances are compared with that of other rule measures. (C) 2014 Wiley Periodicals, Inc.