Table 2.
Performance metrics of all models with their optimal hyperparameters based on k-folds cross-validation
| Classification Model | AUC (95% CI) |
AUC (95% CI) (SMOT) |
Specificity (95% CI) |
Specificity (95% CI) (SMOT) |
Sensitivity (95% CI) |
Sensitivity (95% CI) (SMOT) |
|---|---|---|---|---|---|---|
| Neural Network | 0.586 (0.557, 0.615) | 0.628 (0.583, 0.673) | 0.969 (0.943, 0.994) | 0.880 (0.835, 0.925) | 0.204 (0.137, 0.271) | 0.375 (0.282, 0.467) |
| XGBoost | 0.663 (0.624, 0.702) | 0.685 (0.652, 0.718) | 0.964 (0.948, 0.979) | 0.917 (0.905, 0.929) | 0.363 (0.287, 0.439) | 0.452 (0.387, 0.517) |
| Random Forest Classifier | 0.625 (0.584, 0.666) | 0.637 (0.600, 0.676) | 0.969 (0.965, 0.973) | 0.952 (0.932, 0.972) | 0.279 (0.197, 0.361) | 0.209 (0.156, 0.262) |
| Logistic Regression | 0.589 (0.560, 0.618) | 0.667 (0.628, 0.706) | 0.971 (0.961, 0.980) | 0.744 (0.704, 0.783) | 0.209 (0.156, 0.262) | 0.591 (0.526, 0.656) |
| Balanced Bagging Classifier | 0.672 (0.627, 0.717) | 0.657 (0.624, 0.690) | 0.814 (0.784, 0.843) | 0.858 (0.842, 0.874) | 0.529 (0.456, 0.602) | 0.457 (0.396, 0.518) |
| Balanced Random Forest Classifier | 0.684 (0.653, 0.715) | 0.681 (0.642, 0.720) | 0.727 (0.709, 0.744) | 0.819 (0.792, 0.846) | 0.642 (0.577, 0.707) | 0.542 (0.466, 0.618) |
AUC Area under the receiver operating characteristics curve, CI Confidence Interval, SMOTE Synthetic Minority Oversampling Technique, bolded numbers indicate best performance for that metric