An Empirical Study of Learning from Imbalanced Data Using Random Forest

TLDR

Random forest classifiers are a relatively recent learning method, and prior studies on their use with imbalanced data are limited. This study empirically evaluates random forest learners on imbalanced data, determining optimal tree and attribute counts and comparing performance to other common learners. The authors conduct a comprehensive suite of experiments using the Weka implementation of random forest to assess its performance.

Abstract

This paper discusses a comprehensive suite of experiments that analyze the performance of the random forest (RF) learner implemented in Weka. RF is a relatively new learner, and to the best of our knowledge, only preliminary experimentation on the construction of random forest classifiers in the context of imbalanced data has been reported in previous work. Therefore, the contribution of this study is to provide an extensive empirical evaluation of RF learners built from imbalanced data. What should be the recommended default number of trees in the ensemble? What should the recommended value be for the number of attributes? How does the RF learner perform on imbalanced data when compared with other commonly-used learners? We address these and other related issues in this work.