Beginning Anomaly Detection Using

true positive rate (TPR) = recall = sensitivity. The same as recall, the TPR  

tells us how many of the data points that are actually true were predicted as true by  

the model.

 

The false positive rate FPR

specificity

(

)

=

(

)

=

+

1–

FP

FP TN

 

The FPR tells us how many of the data points that are actually false were predicted to 

be positive by the model. The formula is similar to recall, but instead of the proportion 

of true positives to all of the true data points, it’s the proportion of false positives to all of 

the false data points.

 

Specificity

FPR

=

=

+

1–

TN

TN FP

 

Specificity is very similar to recall in that it tells us how many of the data points that 

are actually false were predicted as false by the model.

We can use the TPR and the FPR to form a graph known as a