Interpretable machine learning in predicting drug-induced liver injury among tuberculosis patients: model development and validation study

Table 2 Comparison of performance ability of the three models in the validation set

Performance ability	Indicators	Logistic	Random Forest	XGBoost
Discrimination	Optimal cutoff (Youden)	0.144	0.182	0.167
	True negative	1332	1424	1461
	False positive	444	352	315
	False negative	79	75	83
	True positive	266	270	262
	Specificity	0.750	0.802	0.823
	Sensitivity	0.771	0.782	0.760
	Precision	0.375	0.434	0.454
	Recall	0.771	0.782	0.760
	F1 score	0.505	0.558	0.568
	Accuracy	0.753	0.799	0.812
	AUROC	0.848	0.877	0.887
	AUPR	0.670	0.727	0.750
Calibration	Brier score	0.085	0.093	0.072

Abbreviations AUROC Area under receiver operative curve, AUPR Area under the precision-recall curve

ISSN: 1471-2288