Machine learning-based mortality prediction models for smoker COVID-19 patients

Table 3 Performance results of top five “at admission” models

Rank	Algorithm	Feature set	Parameters	Calibration	Accuracy	AUC	Precision	Recall	F1 Score	Log Loss	Brier Score
1	XGBoost	7	Colsample_bytree = 0.3 Learning _rate = 0.01 n_estimators = 500 max_depth = 15	Uncalibrated	0.879	0.942	0.904	0.850	0.867	0.336	0.100
1	XGBoost	7		Calibrated	0.875	0.940	0.904	0.839	0.862	0.310	0.094
2	XGBoost	8	Colsample_bytree = 0.5 Learning _rate = 0.01 n_estimators = 300 max_depth = 15	Uncalibrated	0.867	0.929	0.860	0.872	0.864	0.366	0.109
2	XGBoost	8		Calibrated	0.859	0.927	0.870	0.837	0.849	0.329	0.100
3	XGBoost	5	Colsample_bytree = 0.3 Learning _rate = 0.01 n_estimators = 700 max_depth = 15	Uncalibrated	0.872	0.939	0.891	0.843	0.857	0.320	0.097
3	XGBoost	5		Calibrated	0.864	0.938	0.895	0.819	0.843	0.314	0.096
4	XGBoost	3	Colsample_bytree = 0.3 Learning _rate = 0.01 n_estimators = 900 max_depth = 15	Uncalibrated	0.873	0.936	0.894	0.841	0.855	0.306	0.094
4	XGBoost	3		Calibrated	0.872	0.934	0.906	0.824	0.849	0.316	0.095
5	Ensemble	5	XGBoost, MLP, Random Forest, Decision tree	Uncalibrated	0.860	0.916	0.870	0.842	0.850	0.361	0.113
5	Ensemble	5	XGBoost, MLP, Random Forest, Decision tree	Calibrated	0.853	0.923	0.874	0.815	0.832	0.356	0.108

ISSN: 1472-6947