Figure 2.

Comparison of the performance of the XGBoost and baseline models. Performance of the XGBoost classifier by ROC curves (left) and PR curves (right) on the unimputed data set (red) for mortality (top) and critical event (bottom) prediction versus the three baseline models: XGBoost classifier on the imputed data set (purple), LASSO (green), and LR (orange). LASSO: least absolute shrinkage and selection operator; PRC: precision-recall curve; ROC: receiver operating characteristic; XGB: Extreme Gradient Boosting.