Metric

Definition

Note

Accuracy

$\frac{\text{TP}+\text{TN}}{\text{TP}+\text{TN}+\text{FP}+\text{FN}}$

Overall correctness of the model. The total true results.

Precision

$\frac{\text{TP}}{\text{TP}+\text{FP}}$

Ratio of positive cases that we predicted appropriately.

The positive predicted value (PPV).

Recall/TP-Rate

$\frac{\text{TP}}{\text{TP}+\text{FN}}$

The ratio of correctly identified positive cases.

F1-Score

$\frac{2\text{TP}}{2\text{TP}+\text{FP}+\text{FN}}$

The weighted average of precision and recall.

False Negative Rate

(Miss Rate)

$\frac{\text{FP}}{\text{N}}=\frac{\text{FP}}{\text{FP}+\text{TN}}$

It is the probability that a true positive will be missed by the test

True Positive Rate (Sensitivity)

$\text{TPR}=\frac{\text{TP}}{\text{TP}+\text{FN}}$

It is the probability that an actual positive will test positive