Dilated Chi-square: a novel interestingness measure to build accurate and compact decision list

Abstract

Associative classification has aroused significant attention in recent years. This paper proposed a novel interestingness measure, named dilated chi-square, to statistically reveal the interdependence between the antecedents and the consequent of classification rules. Using dilated chi-square, instead of confidence, as the primary ranking criterion for rules under the framework of popular CBA algorithm, the adapted algorithm presented in this paper can empirically generate more accurate and much more compact decision lists.