Table 2. Statistics for machine learning models with only BCRAT inputs and for the BCRAT.
| AUC | Sensitivity | Specificity | Precision | |
|---|---|---|---|---|
| LR | 0.561 (0.525-0.596 95% CI) | 0.599 (0.540-0.657 95% CI) | 0.509 (0.501-0.518 95% CI) | 0.0258 (0.0234-0.0284 95% CI) |
| NB | 0.559 (0.524-0.595 95% CI) | 0.602 (0.544-0.661 95% CI) | 0.504 (0.496-0.513 95% CI) | 0.0257 (0.0233-0.0282 95% CI) |
| DT | 0.510 (0.474-0.547 95% CI) | 0.387 (0.328-0.445 95% CI) | 0.616 (0.607-0.624 95% CI) | 0.0213 (0.0184-0.0248 95% CI) |
| LDA | 0.561 (0.525-0.596 95% CI) | 0.587 (0.529-0.646 95% CI) | 0.512 (0.503-0.521 95% CI) | 0.0254 (0.0230-0.0281 95% CI) |
| SVM | 0.447 (0.411-0.483 95% CI) | 0.993 (0.982-1.00 95% CI) | 0.00718 (0.00571-0.00865 95% CI) | 0.0212 (0.0210-0.0214 95% CI) |
| NN | 0.567 (0.532-0.603 95% CI) | 0.621 (0.563-0.679 95% CI) | 0.474 (0.465-0.482 95% CI) | 0.0249 (0.0227-0.0273 95% CI) |
| BCRAT | 0.563 (0.528-0.597 95% CI) | 0.647 (0.590-0.704 95% CI)* | 0.461 (0.452-0.470 95% CI)* | 0.0254 (0.0232-0.0277 95% CI)* |
Abbreviations: CI = confidence interval, LR = logistic regression, NB = naive Bayes, DT = decision tree, LDA = linear discriminant analysis, SVM = support vector machine, NN = neural network
* Calculated using sensitivity / specificity values based on the threshold that maximized the sum of testing data set sensitivities and specificities rather than the sum of training data set sensitivities and specificities.