Table 6.
The performance of ML models of COPD-EQ
| Model | AUC (95% CI) | Sensitivity (95% CI) | Specificity (95% CI) | Youden | Cut-off |
|---|---|---|---|---|---|
| SVM |
0.733 (0.718–0.748) |
0.709 (0.694–0.724) |
0.711 (0.696–0.726) |
0.421 |
0.166 |
| NB |
0.762 (0.747–0.777) |
0.621 (0.606–0.636) |
0.783 (0.768–0.798) |
0.405 |
0.011 |
| RF |
0.725 (0.710–0.740) |
0.679 (0.664–0.694) |
0.690 (0.675–0.705) |
0.370 |
0.229 |
| KNN |
0.715 (0.700–0.730) |
0.801 (0.786–0.816) |
0.556 (0.541–0.571) |
0.357 |
0.4 |
| LR |
0.775 (0.760–0.790) |
0.781 (0.765–0.797) |
0.670 (0.655–0.685) |
0.452 |
0.250 |
| AdaBoost |
0.783 (0.770–0.796) |
0.665(0.652–0.678) |
0.824 (0.811–0.837) |
0.490 |
0.492 |
| XgBoost |
0.785 (0.770–0.800 |
0.828 (0.813–0.843) |
0.671 (0.656–0.686) |
0.499 |
0.302 |
| LightGBM |
0.796 (0.781–0.811) |
0.812 (0.797–0.827) |
0.709 (0.694–0.724) |
0.521 |
0.281 |
| CatBoost | 0.801 (0.786–0.816) | 0.818 (0.803–0.833) | 0.711 (0.696–0.726) | 0.528 | 0.248 |
AdaBoost – adaptive boosting, AUC – area under curve, CaatBoost – categorical boosting, CI – confidence interval, KNN – K-nearest neighbours, LightGBM – light gradient boosting machine, LR – ogistic rRegression, NB – naïve Bayes, SVM – support vector machine, XGBoost – extreme gradient boosting