. 2021 Sep 24;22(19):10291. doi: 10.3390/ijms221910291

Table 5.

Description of performance metrics. Here, a “positive” label can, for example, correspond to “having CVD”, while a “negative” may correspond to “healthy”.

Metric	Description	Math. Definition
True positive	A positive sample correctly	$T P$
(TP)	predicted by the model.
True negative	A negative sample correctly	$T N$
(TN)	predicted by the model.
False positive	A sample wrongly classified as	$F P$
(FP)	positive by the model.
False negative	A sample wrongly classified as	$F N$
(FN)	negative by the model.
Precision	Fraction of true positives among	$T P$
	the predicted positives.	$T P + F P$
Recall	Fraction of positives that are	$T P$
(Sensitivity)	correctly predicted.	$T P + F N$
Specificity	Fraction of negatives that are	$T N$
	correctly predicted.	$T N + F P$
Accuracy	Fraction of correctly predicted	$T N + T P$
	positives and negatives.	$T P + T N + F P + F N$
ROC curve	A curve indicating performance
(Receiver Operating	of a classifier. The Y-axis shows	$R (s)$
Characteristic)	recall and the X-axis shows
	s = (1-specificity)
AUC-ROC	Quantitative performance
(Area Under the Curve	measure based on ROC curve.
- ROC)	Ranges from 0 to 1, where 1	$\int R (s) d s$
	corresponds to perfect, and 0.5
	to random, classification.
C-statistic	Equivalent to AUC-ROC. Can
	be used for censored data
	(missing patient outcomes).
	$r_{i}$ —predicted risk of patient i	$\sum_{i j} ε_{i} 1 (r_{i} > r_{j}) 1 (t_{i} < t_{j})$
	$t_{i}$ —time to event, patient i	$\sum_{i j} ε_{i} 1 (t_{i} < t_{j})$
	$ε_{i} \in {0, 1}$ - whether (event)
	information exists.
	$1 (x) = {1 if x, 0 otherwise}$
PR curve	Similar to ROC curve. Y-axis
(Precision Recall)	shows precision and X-axis	$P (r)$
	recall (r).
AUC-PR	Quantitative performance
(AUC—Precision-Recall)	measure based on PR curve.	$\int P (r) d r$
	Alternative to AUC-ROC.