Table 8.

The analysis of the cross-prediction performance of RF, ET, and SVC-based models for GP, GN, and GV ABPs, respectively, using AAB binary feature of NC-terminal on different validation datasets.

Validation Dataset	Performance Measures	Models
		Gram-Positive	Gram-Negative	Gram-Variable
Gram-positive ABPs	Sn	80.7	50.0	67.7
	Sp	90.3	89.3	87.6
	Acc	85.5	69.6	77.7
	AUC	0.93	0.85	0.88
	AUPRC	0.92	0.83	0.87
	MCC	0.71	0.43	0.57
Gram-negative ABPs	Sn	66.0	92.4	91.1
	Sp	89.4	94.9	87.3
	Acc	77.7	93.6	89.2
	AUC	0.89	0.98	0.93
	AUPRC	0.85	0.98	0.92
	MCC	0.57	0.87	0.78
Gram-variable ABPs	Sn	80.2	76.2	83.9
	Sp	87.4	92.0	90.1
	Acc	83.8	84.1	87.0
	AUC	0.91	0.93	0.94
	AUPRC	0.89	0.93	0.93
	MCC	0.68	0.69	0.74

Sn: Sensitivity, Sp: Specificity, Acc: Accuracy, MCC: Matthews correlation coefficient, AUC: Area under the receiver operating characteristic curve, AUPRC: Area under the precision-recall curve.