Empirical Study on the Predictive Power of Rotation Forest
Bagus Sartono, Mulianto Raharjo, Cici Suhaeni · IOP Conference Series Earth and Environmental Science · 2018
Abstract This paper presents the comparison of the predictive power of rotation forest and other classification techniques in the case of classification problem. Rotation forest is an ensemble of many decision trees that utilizing principal component analysis to do some rotation to the original predictor variables before constructing the trees. An empirical study using fourteen different datasets was held to observe how good the prediction resulted by the rotation forest. The authors found that in most of the cases rotation forest perform better compared to logistic regression, tree, and discriminant analysis. We also revealed that the rotation forest fails to have excellent prediction when the predictors are dominated by categorical variables. In general, rotation forest could be a competitive approach to handle classification tasks.