Table 2.
Performance metrics of the algorithms.
| Algorithm | Precision | Recall | F1 | Accuracy | MCC | AUROC (95% CI) | AUPRC | |
|---|---|---|---|---|---|---|---|---|
| mRS | XGBoost | 0.75 | 0.923 | 0.828 | 0.865 | 0.729 | 0.923 (0.825–1) | 0.917 |
| LightGBM | 0.812 | 0.929 | 0.867 | 0.892 | 0.781 | 0.958 (0.886–1) | 0.958 | |
| CatBoost | 0.562 | 0.9 | 0.692 | 0.784 | 0.574 | 0.872 (0.748–0.996) | 0.871 | |
| Random Forest | 0.625 | 0.909 | 0.741 | 0.811 | 0.626 | 0.929 (0.835–1) | 0.917 | |
| Mean | 0.687 | 0.915 | 0.782 | 0.838 | 0.678 | 0.921 | 0.916 | |
| NIHSS Shift | XGBoost | 0.737 | 0.7 | 0.718 | 0.711 | 0.422 | 0.767 (0.615–0.92) | 0.813 |
| LightGBM | 0.895 | 0.63 | 0.739 | 0.684 | 0.406 | 0.787 (0.64–0.934) | 0.761 | |
| CatBoost | 0.737 | 0.737 | 0.737 | 0.737 | 0.474 | 0.820 (0.683–0.956) | 0.833 | |
| Random Forest | 0.684 | 0.765 | 0.722 | 0.737 | 0.476 | 0.834 (0.702–0.965) | 0.870 | |
| Mean | 0.763 | 0.708 | 0.729 | 0.717 | 0.445 | 0.802 | 0.819 |
Abbreviations: MCC: Matthew’s correlation coefficient; AUROC: Area under the receiver operating characteristic curve; CI: Confidence interval; AUPRC: The area under the precision-recall curve; mRS: Modified Rankin Score; NIHSS: National Institutes of Health Stroke Scale.