Table 7.
Survival performance results with training set (cross-validation approach) and test set (new unseen data) with feature selection.
Classifier | c-Index | #F | c-Index | c-Index IPWC | AUCD_ROC. | ICI |
---|---|---|---|---|---|---|
Cox proportional hazard | 0.700 (0.048) | 5(N), 4(C) | 0.658 | 0.670 | 0.669 | 0.175 |
Cox proportional hazard IPCW | 0.704 (0.051) | 2(N), 5(C) | 0.647 | 0.646 | 0.654 | 0.293 |
Random survival forest | 0.706 (0.035) | 4(N), 4(C) | 0.675 | 0.723 | 0.684 | 0.235 |
Extra survival trees | 0.696 (0.054) | 4(N), 1(C) | 0.657 | 0.704 | 0.666 | 0.406 |
Survival support vector machine | 0.698 (0.057) | 3(N), 5(C) | 0.638 | 0.631 | 0.649 | 0.233 |
Gradient boosting models | 0.714 (0.013) | 2(N), 5(C) | 0.724 | 0.762 | 0.748 | 0.298 |
Gradient boosting models* | 0.714 (0.018) | 2(N), 2(C) | 0.711 | 0.754 | 0.733 | 0.476 |
C-index, concordance index; C-index IPCW, concordance index inverse probability of censoring weights; AUCD_ROC, area under the cumulative/dynamic ROC; ICI, interpretability concordance index.
*Model extracted by manual inspection.