Sensitivity and Specificity

We hope that you are already aware about below 4 terms:

 

• True Positive
• False Positive
• True Negative
• False Negative

 

In case you are not, then please read this short article on confusion matrix first and then proceed with this one. 

 

There are few ratio related variables where we find out Positive or Negative Rate. Four metrics are designed for this. Let us check them out.

 

• True Positive Rate (TPR)
• False Negative Rate (FNR)
• True Negative Rate (TNR)
• False Positive Rate (FPR)

 

1) True Positive Rate (TPR):

TPR = TP / P = 1 - FNR

Here P stands for actual Positive class available in the data.

It simply means how many correct predictions happened for the Positive class divided by the count of actual Positive class. Or we can call it as percentage of actual Positives correctly predicted by the model.

  

2) False Negative Rate (FNR):

FNR = FN / P = 1 - TPR

It means that how many wrong predictions happened for the Positive class divided by the count of actual Positive class. Or we can call it as percentage of actual Positives incorrectly predicted by the model.

 

TPR + FNR = 100% of Actual Positives

 

3) True Negative Rate (TNR):

TNR = TN / N = 1 - FPR

Here N stands for actual Negative class available in the data.

It means how many correct predictions happened for the Negative class divided by the count of actual Negative class. Or we can call it as percentage of actual Negatives correctly predicted by the model.

 

4) False Positive Rate (FPR):

FPR = FP / N = 1 - TNR

It means that how many wrong predictions happened for the Negative class divided by the count of actual Negative class. Or we can call it as percentage of actual Negatives incorrectly predicted by the model.

 

TNR + FPR = 100% of Actual Negatives

 

Now you are almost done with this article, but wait!! We have not covered the topic for which this article is dedicated.

No worries. 

You have already learnt the actual concept, now you just need to know which metric belongs to Sensitivity and which metric to Specificity. That’s it.

 

Sensitivity or Recall:

If you look at the definitions above, you may realize that the definition of TPR looks like the definition of Recall which we have studied already. The True Positive Rate is also known as Sensitivity.

 

Specificity or TNR:

The True Negative Rate is also known as Specificity.