Table 2.
Predicting clinical diagnosis: model performances for each physician in terms of AUC. Each column corresponds to no feature selection (“all”) or to one of three different feature selection methods. In parentheses the number of features used by each model.
| Physicians | All (24) | HITON_PC | HITON_MB | RFE |
|---|---|---|---|---|
| Exp1 | 0.94 | 0.92 (4) | 0.92 (5) | 0.95 (14) |
| Exp2 | 0.92 | 0.89 (7) | 0.90 (7) | 0.90 (12) |
| Exp3 | 0.98 | 0.95 (4) | 0.95 (4) | 0.97 (19) |
| NonExp1 | 0.92 | 0.89 (5) | 0.89 (6) | 0.90 (22) |
| NonExp2 | 1.00 | 0.99 (6) | 0.99 (6) | 0.98 (11) |
| NonExp3 | 0.89 | 0.89 (4) | 0.89 (6) | 0.87 (10) |