Table 2.
DeLong test-based comparison of AUC between model 1 and model 2 across different algorithms.
| Algorithm | Model 1 (AUC) | Model 2 (AUC) | DeLong Test (z) | p-value |
|---|---|---|---|---|
| LR | 0.72 | 0.79 | − 4.3391 | < 0.001 |
| Decision Tree | 0.60 | 0.65 | − 4.2912 | < 0.001 |
| Extra Trees | 0.73 | 0.82 | − 5.8787 | < 0.001 |
| Gradient Boosting | 0.76 | 0.83 | − 4.6764 | < 0.001 |
| KNN | 0.68 | 0.76 | − 4.9441 | < 0.001 |
| Naive Bayes | 0.74 | 0.74 | − 0.3877 | 0.698 |
| Random Forest | 0.74 | 0.82 | − 5.9244 | < 0.001 |
| SVM | 0.75 | 0.82 | − 4.3505 | < 0.001 |
LR Logistic Regression, KNN K-Nearest Neighbors, SVM Support Vector Machine, AUC area under the receiver operating characteristic curve.