Table 3.
Summary of the performance of the five different machine learning models on the training and test datasets.
| ML Models | AUROCtraining[95% CI] | AUPRCtraining[95% CI] | AUROCtest[95% CI] | AUPRCtest[95% CI] |
|---|---|---|---|---|
| Elastic Net | 0.715 [0.575-0.844] |
0.361 [0.243-0.524] |
0.721 [0.493-0.938] |
0.380 [0.171-0.662] |
| Ridge | 0.737 [0.612-0.859] |
0.406 [0.257-0.584] |
0.751 [0.575-0.927] |
0.326 [0.164-0.558] |
| Lasso | 0.754 [0.630-0.874] |
0.402 [0.256-0.576] |
0.748 [0.554-0.932] |
0.346 [0.168-0.620] |
| PLS | 0.732 [0.605-0.853] |
0.406 [0.256-0.579] |
0.744 [0.567-0.924] |
0.312 [0.156-0.566] |
| svmLin |
0.744
[0.600-0.881] |
0.431
[0.278-0.610] |
0.764
[0.536-0.960] |
0.431
[0.214-0.720] |
Area Under the Receiver Operating Characteristics Curves and Area Under the Precision Recall Curves with their respective 95% confidence intervals calculated for 5 different machine learning models on the training and test datasets to predict 28-day survival.
Performances of the best model are highlighted in bold.