Table 3.

Comparison of RF and SVM classifiers under the two types of tenfold cross-validation.

Cross-validation		Classification algorithm	Model	Accuracy	Precision	Recall	F1-measure	MCCa
Entire tenfold cross-validation		Random forest	Addition + subtraction	89.54%	92.70%	87.20%	89.86%	79.25%
		Support vector machine		81.92%	85.92%	79.57%	82.62%	64.06%
Composition tenfold cross-validation	ODITb test dataset	Random forest	Addition + Hadamard	79.86%	75.20%	82.99%	78.89%	60.06%
		Support vector machine		64.65%	61.20%	65.53%	63.06%	29.51%
	NDITc test dataset	Random forest	Addition + Hadamard	64.49%	41.56%	76.82%	53.73%	32.64%
		Support vector machine		57.45%	48.26%	58.41%	52.46%	15.07%

^aMCC: Mathews correlation coefficient. ^bODIT: One Drug In Train set. ^cNDIT: No Drug In Train set.