I have two questions:

  1. just read this answer and I don't think I totally understand this term ... does sensitivity and specificity and other measures derived from these two such as the geometric mean of these two values considered as "point-wise" metrics ? is there any paper you can refer me to which describe such term and the difference between them and the AUC-ROC/PR when evaluating a classifier or choosing the best model?

  2. If I want to compare the a binary classifier with different classes ratio (1:2, 1:3, .... , 1:7) which AUC is better .. is it AUC-ROC or AUC-PR ? from this answer it seems that in such situation the AUC-ROC is better ! but why some academic papers such as this suggest using AUC-PR in imbalanced dataset cases ? below is a quote from the mentioned paper:

    A large number change in the number of false positives can lead to a small change in the false positive rate used in ROC analysis. Precision, on the other hand, by comparing false positives to true positives rather than true negatives, captures the effect of the large number of negative examples on the algorithm's performance.


Your Answer

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

Browse other questions tagged or ask your own question.