Figure 2

Comparisons of the performance of the seven ML models. The seven ML models were compared against each other based on their ROC curves, and their performances were evaluated by several important metrics such as AUC, accuracy, precision, recall, and f1 score, whose values of Mean and 95% CI were shown in Table S1. (A) ROC curves of the XGBoost model, AUC [mean]: 0.809 (0.806–0.811). (B) ROC curves of the RF model, AUC [Mean]: 0.797 (0.795–0.800). (C) ROC curves of the SVM model, AUC [mean]: 0.552 (0.547–0.557). (D) ROC curves of the DNN model, AUC [mean]: 0.666 (0.607–0.724). (E) ROC curves of the KNN model, AUC [mean]: 0.645 (0.640–0.650). (F) ROC curves of the Decision Tree model, AUC [mean]: 0.753 (0.748–0.758). (G) ROC curves of the LA model, AUC [mean]: 0.687 (0.684–0.690).