Weighted Multi-Measure Ranking For Associative Classifiers

This paper presents AC.Rank.AW, a rule-ranking method for associative classifiers based on weighted objective measures (OMs). It extends AC.Rank.A, an aggregation-based approach that combines multiple OMs assuming uniform importance. In contrast, AC.Rank.AW incorporates weighting methods, allowing differentiated contributions during aggregation. The method was evaluated on 43 datasets across 156 configurations, considering performance and interpretability. Statistical analyses using the Friedman and Nemenyi tests indicate that AC.Rank.AW improves interpretability without degrading performance in some configurations. The best results are obtained with the [GF][PM-SD][TS] combination, i.e., when using the [GF] OMs group, along with Topsis ([TS]) and Standard Deviation ([SD]) weighting method.