0
$\begingroup$

I have gone through few of the paper for AUC calculation, following paper has mentioned AUC can be calculated as

AUC = (Sensitivity +Specificity)/2.

A systematic analysis of performance measures for classification tasks

Please let me know that whether it is correct or not.

$\endgroup$
1
$\begingroup$

The article you mentioned is referring to the AUC of Receiver Operating Characteristic Curve (ROC-Curve), that is different from the AUC of the Precision-Recall Curve (PR-Curve). See this post to know about the differences between the PR-Curve and the ROC-Curve. If what you want is that same value for the PR-Curve you should look at the F1-Score, see this wikipedia article for an explanation of the relation between the AUC of the ROC-Curve, the Youden Index and the F1-Score. Also a good indicator of the AUC of the PR-Curve is the Average Precision (AP) as explained here.

$\endgroup$
  • $\begingroup$ Thanks... I agree that it is AUC of ROC curve. However, I didn't find in any of the mentioned paper that they have mentioned that we can take an average of Sensitivity and Specificity as AUC. Can you please specify any paper which holds this formulation true? $\endgroup$ – Chandan Gautam Dec 1 '17 at 3:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.