Extensions of ant-miner algorithm to deal with class imbalance problem

Abstract

A database has class imbalance when there are more cases of one class then the others. Classification algorithms are sensitive of this imbalance and tend to valorize the majority classes and ignore the minority classes, which is a problem when the minority classes are the classes of interest. In this paper we propose two extensions of the Ant-Miner algorithm to find better rules to the minority classes. These extensions modify, mainly, how rules are constructed and evaluated. The results show that the proposed algorithms found better rules to the minority classes, considering predictive accuracy and simplicity of the discovered rule list. © 2012 Springer-Verlag.