Table 2. Residue-based evaluation of different machine learning models and training sets on DB312 (RB264).

Training set	Classifier a	Recall	Precision	F1	ACC	MCC	AUC
Balanced	NB	0.625 (0.575)	0.243 (0.291)	0.349 (0.386)	0.728 (0.755)	0.257 (0.277)	0.723 (0.708)
	NN	0.482 (0.474)	0.384 (0.456)	0.426 (0.462)	0.849 (0.853)	0.344 (0.379)	0.813 (0.801)
	RF	0.487 (0.535)	0.405 (0.465)	0.442 (0.497)	0.857 (0.855)	0.363 (0.415)	0.820 (0.822)
	SVM	0.541 (0.533)	0.448 (0.481)	0.489 (0.506)	0.869 (0.862)	0.418 (0.426)	0.847 (0.832)
Unbalanced	NB	0.586 (0.590)	0.252 (0.285)	0.351 (0.383)	0.748 (0.746)	0.257 (0.274)	0.717 (0.701)
	NN	0.443 (0.489)	0.441 (0.479)	0.440 (0.482)	0.869 (0.860)	0.367 (0.403)	0.812 (0.799)
	RF	0.468 (0.519)	0.429 (0.488)	0.447 (0.502)	0.865 (0.863)	0.372 (0.424)	0.806 (0.817)
	SVM	0.544 (0.549)	0.505 (0.513)	0.523 (0.530)	0.885 (0.871)	0.458 (0.456)	0.862 (0.845)

^aNB: naive Bayes, NN: neural networks, RF: random forest, and SVM: support vector machines.