Table 2.
Performance of the classifier for different time gaps. The numbers in parenthesis indicate 95% confidence intervals which have been calculated by the bootstrap method. The best hyperparameter settings for each model are described in Table S2 (Supplementary Material S4).
| T (time gap in units of 6 months) | Validation accuracy | Validation AUC | Test accuracy | Test AUC | ||
|---|---|---|---|---|---|---|
| 4 (2 years) | 0.752 (0.7238, 0.779) | 0.8832 (0.8629, 0.9026) | 0.7715 (0.7455, 0.7975) | 0.896 (0.875, 0.9146) | ||
| 4, Confusion matrix |
|
|
||||
| 6 (3 years) | 0.7356 (0.7062, 0.765) | 0.9059 (0.888, 0.9213) | 0.7465 (0.7143, 0.7773) | 0.8988 (0.8795, 0.9176) | ||
| 6, Confusion matrix |
|
|
||||
| 8 (4 years) | 0.7724 (0.7416, 0.8032) | 0.9199 (0.905, 0.9337) | 0.7688 (0.7382, 0.8008) | 0.9209 (0.9050, 0.9353) | ||
| 8, Confusion matrix |
|
|
||||
