Table 2.

Classification performances of the random forest models.

	Training set					Test set
	AUROCa (95% CI)	PRAUCb	Accuracy	F1-score	AUROC (95% CI)		PRAUC	Accuracy	F1-score
Cephalosporin	0.777 (0.775-0.779)	0.725	0.715	0.676	0.638 (0.635-0.642)		0.547	0.603	0.556
TZPc	0.864 (0.862-0.867)	0.688	0.808	0.652	0.630 (0.626-0.634)		0.332	0.641	0.313
Carbapenem	0.877 (0.874-0.880)	0.539	0.822	0.493	0.665 (0.659-0.671)		0.222	0.725	0.220
TMP-SMXd	0.881 (0.879-0.882)	0.829	0.822	0.781	0.670 (0.666-0.673)		0.568	0.638	0.560
Fluoroquinolone	0.884 (0.884-0.885)	0.938	0.802	0.832	0.721 (0.718-0.724)		0.813	0.657	0.706

^aAUROC: area under the receiver operating characteristic curve.

^bPRAUC: precision-recall area under the curve.

^cTZP: piperacillin-tazobactam.

^dTMP-SMX: trimethoprim-sulfamethoxazole.