Table 1 Model performance (AUROC, best sensitivity and specificity, PPV)
Model Type | Time split | AUROC (95%CI) | Specificity (balanced model) | Sensitivity (balanced model) | PPV (balanced model) | Sensitivity for 95% specificity | PPV at 95% specificity |
|---|---|---|---|---|---|---|---|
Logistic Regression with Lasso | 1, 2–5 | 0.736 (0.728–0.743) | 0.752 | 0.602 | 0.156 | 0.222 | 0.254 |
Naïve Bayes Classifier | 1, 2–5 | 0.682 (0.675–0.690) | 0.906 | 0.241 | 0.164 | 0.153 | 0.189 |
Support Vector Machine | 1, 2–5 | 0.737 (0.730–0.744) | 0.691 | 0.674 | 0.142 | 0.223 | 0.255 |
Random Forest | 1, 2–5 | 0.734 (0.726–0.740) | 0.653 | 0.700 | 0.134 | 0.210 | 0.239 |
Neural Network (3 × 139 nodes) | 1, 2–5 | 0.737 (0.730–0.743) | 0.781 | 0.619 | 0.178 | 0.298 | 0.312 |