Table 2.

Performance (CV-10 fold) of RF models both balanced and realistic datasets using AAC and DPC features as input at optimized parameters.

Dataset	Features	Thr	Sen	Spec	Acc	MCC	AUC	Parameters
Balanced dataset	AAC	0.5	89.44	94.37	91.90	0.84	0.96	mtry:8,ntree:500
	DPC	0.5	90.85	84.51	87.68	0.76	0.95	mtry:10,ntree:500
Realistic dataset	AAC	0.5	75.35	99.44	97.25	0.82	0.97	mtry:8,ntree:500
	DPC	0.5	58.45	99.79	96.04	0.73	0.95	mtry:10,ntree:500

It is apparent that AAC-based RF models could achieve maximum MCC value of 0.84 with an accuracy of 91.9% and displayed better performance than DPC-based RF models.