Table 3.
Performance metrics of the different classifiers for predicting treatment response or remission in the NeuroPharm dataset. AUC: Area under the ROC Curve. Values are given as mean (SD), aside from the p-values.
| Response | Remission | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Classifier | AUC | Balanced Accuracy | Sensitivity | Specificity | AUC p-value | AUC | Balanced Accuracy | Sensitivity | Specificity |
AUC p-value |
| Elastic Net | 0.61 (0.04) | 59.4 (2.5) | 60.0 (7.0) | 58.8 (6.5) | 0.067 | 0.50 (0.05) | 48.4 (4.7) | 54.5 (7.2) | 42.2 (8.0) | 0.440 |
| Random Forest | 0.62 (0.03) | 58.0 (3.2) | 37.1 (5.4) | 79.0 (2.7) | 0.048 | 0.59 (0.03) | 55.2 (3.0) | 75.3 (4.6) | 35.1 (4.4) | 0.115 |
| SVM | 0.46 (0.04) | 48.0 (3.5) | 41.0 (13.3) | 55.0 (16.8) | 1.000 | 0.50 (0.03) | 49.5 (4.8) | 84.0 (9.9) | 15.1 (9.6) | 0.470 |
| Boosted Trees | 0.55 (0.04) | 53.8 (3.7) | 40.4 (4.1) | 67.2 (5.5) | 0.220 | 0.56 (0.06) | 54.2 (6.0) | 66.0 (6.6) | 42.4 (9.1) | 0.160 |