Extreme Gradient Boosting with Squared Logistic Loss Function

Tree boosting has empirically proven to be a highly effective and versatile approach for predictive modeling. The core argument is that tree boosting can adaptively determine the local neighborhoods of the model thereby taking the bias-variance trade-off into consideration during model fitting. Recently, a tree boosting method known as XGBoost has gained popularity by providing higher accuracy. XGBoost further introduces some improvements which allow it to deal with the bias-variance trade-off even more carefully. In this manuscript, performance accuracy of XGBoost is further enhanced by applying a loss function named squared logistics loss (SqLL). Accuracy of the proposed algorithm, i.e., XGBoost with SqLL, is evaluated using test/train method, K-fold cross-validation, and stratified cross-validation method.