Table 3.
Performance of the best model using all features vs. a subset of features found with the RFE procedure.
| dataset | features | F1 score | ||
|---|---|---|---|---|
| median | 95% CI | |||
| CHECK | 379 | (all) | 0.584 | (0.583, 0.586) |
| 10–15 | (subset) | 0.573 | (0.570, 0.575) | |
| OAI | 1299 | (all) | 0.69 | (0.689, 0.690) |
| 209–364 | (subset) | 0.676 | (0.675, 0.676) | |
We report the median model score for a median CV-repeat and the 95% confidence interval around it (from binomial distribution). For the size of the selected subset of features, we report a range across all CV-repeats.