Table 5 Classification performance metrics

Accuracy

How often are the model's predictions correct?

(TP + TN) / (TP + TN + FN + FP)

Precision

When the model predicts positive, how often is it correct?

TP / (TP + FP)

Sensitivity (Recall) or TPR

When it is actually positive, how often it correctly predicts?

TP / (TP + FN)

Specificity or TNR

When it is actually negative, how often it correctly predicts?

TN / (TN + FP)

Balanced accuracy

The mean of sensitivity and specificity

(TPR + TNR) / 2

F-Measure

The harmonic mean of precision and sensitivity

2 × Precision × Recall / (Precision + Recall)