Fig. 5

ROC curves of 1. the ensemble learning methods on 14 individual models, 2. the ensemble learning methods on the three “best-perform” models, and 3. the three “best-perform” individual models. The ensemble learning methods including the two-step stacking methods, the traditional stacking methods, and the weighted voting methods. The purple lines represent the individual models. The black lines represent the two-step stacking methods, the blue lines represent the traditional stacking methods, and the red lines represent the weighted voting methods, with the solid lines represents ensemble on 14 models, and the dashed lines represent ensemble on 3 models