Measures of Class Membership in Association Rule Based Classification

摘要

In this work, we focus on the measures of class membership defined for classifiers based on associative classification rules. We revisit the chi 2 test for defining an effective measure. For comparison, we adapt the weight of evidence (Wang and Wong TKDE 2003) for a system based on the notions of support and confidence. Some variants of those measures are also defined. The effect of the defined measures is verified through the experimentation on categorical UCI datasets.