Fig. 3

In training and validation sets, multiple ML classification models are integrated for analysis. A ROC curves evaluated the classification accuracy of the 9 models in the train set: XGBoost (AUROC = 0.902), LightGBM (AUROC = 0.631), RF (AUROC = 1.000), AdaBoost (AUROC = 0.803), GNB (AUROC = 0.723), CNB (AUROC = 0.587), MLP (AUROC = 0.510), SVM (AUROC = 0.507), and KNN (AUROC = 1.000). B ROC curves evaluated the classification accuracy of the 9 models in the validation set: XGBoost (AUROC = 0.717), LightGBM (AUROC = 0.576), RF (AUROC = 0.693), AdaBoost (AUROC = 0.707), GNB (AUROC = 0.709), CNB (AUROC = 0.565), MLP (AUROC = 0.500), SVM (AUROC = 0.548), and KNN (AUROC = 0.581). C The calibration curve for different models in the validation set, the abscissa represents the average prediction probability, the ordinate represents the actual probability of the event, the dashed diagonal is the reference line, and the other smooth solid lines represent the different model fitting lines. Brier scores evaluated the calibration of the 9 models: XGBoost (Brier score = 0.193), LightGBM (Brier score = 0.211), RF (Brier score = 0.194), AdaBoost (Brier score = 0.234), GNB (Brier score = 0.198), CNB (Brier score = 0.265), MLP (Brier score = 0.222), SVM (Brier score = 0.217), and KNN (Brier score = 0.252). D The decision curve for different models in the validation set. The solid lines represent different models. E The AUPR curve for different models in the training set, the y‐axis is precision and the x‐axis is recall. AUPR evaluated the overall performance of the 9 models in the train set: XGBoost (AUPR = 0.953), LightGBM (AUPR = 0.753), RF (AUPR = 1.000), AdaBoost (AUPR = 0.901), GNB (AUPR = 0.858), CNB (AUPR = 0.752), MLP (AUPR = 0.706), SVM (AUPR = 0.712), and KNN (AUPR = 1.000). F The AUPR curve for different models in the validation set, the y‐axis is precision and the x‐axis is recall. AUPR evaluated the overall performance of the 9 models in the validation set: XGBoost (AUPR = 0.842), LightGBM (AUPR = 0.734), RF (AUPR = 0.831), AdaBoost (AUPR = 0.835), GNB (AUPR = 0.836), CNB (AUPR = 0.750), MLP (AUPR = 0.699), SVM (AUPR = 0.732), and KNN (AUPR = 0.721). Abbreviations: XGBoost, Extreme Gradient Boosting; LightGBM, Light Gradient Boosting Machine; RF, Random Forest; GNB, Gaussian Naive Bayes; CNB, Complement Naive Bayes; MLP, Multilayer Perceptron; SVM, Support Vector Machine; KNN, K-Nearest Neighbour; AdaBoost, Adaptive Boost; AUROC, area under the receiver-operating characteristic curve; ROC, receiver operating characteristic; AUPR, area under the precision-recall curve; PR, precision-recall curve; AUC, area under curve