Table 2.

Performance of top ten binary classification models for the training and external test sets ¹.

Data Set	Model	CA	AUC	SE	SP	TP	TN	FP	FN
Training set	Ext-RF	0.865	0.926	0.88	0.85	44	46	8	6
	Ext-LR	0.885	0.922	0.90	0.87	45	47	7	5
	Ext-ANN	0.846	0.913	0.88	0.81	44	44	10	6
	Ext-SVM	0.865	0.912	0.88	0.85	44	46	8	6
	Graph-RF	0.817	0.897	0.84	0.80	42	43	11	8
	PubChem-LR	0.798	0.887	0.74	0.85	37	46	8	13
	Ext-Tree	0.837	0.879	0.78	0.89	39	48	6	11
	PubChem-RF	0.750	0.871	0.68	0.81	34	44	10	16
	Graph-LR	0.779	0.870	0.76	0.80	38	43	11	12
	PubChem-Tree	0.827	0.867	0.82	0.83	41	45	9	9
External test set	Ext-RF	0.840	0.930	0.75	0.92	9	12	1	3
	Ext-LR	0.840	0.974	0.75	0.92	9	12	1	3
	Ext-ANN	0.800	0.962	0.67	0.92	8	12	1	4
	Ext-SVM	0.880	0.904	0.83	0.92	10	12	1	2
	Graph-RF	0.880	0.920	0.92	0.85	11	11	2	1
	PubChem-LR	0.840	0.936	0.92	0.77	11	10	3	1
	Ext-Tree	0.880	0.901	0.83	0.92	10	12	1	2
	PubChem-RF	0.800	0.917	0.75	0.85	9	11	2	3
	Graph-LR	0.840	0.936	0.75	0.92	9	12	1	3
	PubChem-Tree	0.640	0.667	0.67	0.62	8	8	5	4

¹ CA, classification accuracy; AUC, the area under the ROC curve; SE, sensitivity; SP, specificity; TP, the number of true positive compounds; TN, the number of true negative compounds; FP, the number of false positive compounds; FN, the number of true negative compounds.