Figure 1

Flowchart for developing predictive models. Numbers above the arrows indicate the order of the processes: (1) data stratified random splitting, (2) data preprocessing, (3) model construction, (4) model calibration, and (5) performance evaluation. MICE multivariate imputation by chained equation, SVM support vector machine, XGBoost extreme gradient boosting, LightGBM light gradient boosting machine, MLP multilayer perceptron.