Enhancing Rule Importance Measure Using Concept Hierarchy
Jiye Li, Nick J. Cercone, Serene Wong, Lisa Jing, Yan · 2009
Abstract. A rule importance measure is used to evaluate how impor-tant are the rules which characterize a data set. This measure was de-signed based on association rules and it has been proven to be effective to enumerate the most important rules of all rules generated. However, since rule importance is an objective measure, its usage as a rule interesting-ness measure relies on the interpretation of domain experts. We propose to enhance the rule importance measure previously used by incorporat-ing a weight biased attribute concept hierarchy. The new measure better reflects the importance of a rule by integrating with the domain knowl-edge. A geriatric care data set is used as our experimental data set. We show that this enhanced rule importance measure provides a knowledge oriented distinction of rules classified as important.