TABLE 4.

Objective functions for machine learning models.

Name	Prediction task	Formula
Accuracy	Classification	$Accuracy=\frac{TP+TN}{P+N}$
AUC (Area under the receiver operating characteristic curve)	Classification	$TPR=\frac{TP}{FN+TP}$
		$FPR=\frac{FP}{TN+FP}$
		$AUC=\int TPR d\left(FPR\right)$
LogAUC	Classification	$logAUC=\frac{{{\displaystyle \int }}_{0.001}^{0.1}TPR d\left(\text{log}\left(FPR\right)\right)}{{{\displaystyle \int }}_{0.001}^{0.1}d\left(\text{log}\left(FPR\right)\right)}$
MCC (Matthew’s correlation coefficient)	Classification	$MCC=\frac{TP\ast TN-FP\ast FN}{\sqrt{\left(TP+FP\right)\left(TP+FN\right)\left(TN+FP\right)\left(TN+FN\right)}}$
PPV (Positive predictive value)	Classification	$PPV=\frac{TP}{TP+FP}$
Enrichment factor	Classification	$EF\left(x\%\right)=\frac{PPV\left(x\%\right)}{PPV\left(100\%\right)}$
MAE (Mean absolute error)	Regression	$MAE=\frac{1}{N}{\displaystyle \sum }_i^N\left|f\left(x_i\right)-y_i\right|$
MAE_NMAD (MAE normalized by the mean absolute deviation)	Regression	$MA{E}_{NMAD}=\frac{MAE}{\frac{1}{N}{{\displaystyle \sum }}_i^N\left|y_i-\overline{y}\right|}$
RMSD (Root-mean-square deviation)	Regression	$RMSD=\sqrt{\frac{1}{N}{\displaystyle \sum }_i^N{\left(f\left(x_i\right)-y_i\right)}^2}$
NRMSD (RMSD normalized by the range)	Regression	$NRMSD=\frac{RMSD}{\max \left(y\right)-min\left(y\right)}$
RMSD_NSTD (RMSD normalized by the standard deviation)	Regression	$RMSD\_NSTD=\frac{RMSD}{\text{Stdev}\left(y\right)}$