Bias corrections for Random Forest in regression using residual rotation

Abstract

This paper studies bias correction methods for Random Forest in regression. Random Forest is a special bagging trees that can be used in regression and classification. It is a popular method because of its high prediction accuracy. However, we find that Random Forest can have significant bias in regression at times. We propose a method to reduce the bias of Random Forest in regression using residual rotation. The real data applications show that our method can reduce the bias of Random Forest significantly. © 2015 The Korean Statistical Society.