I just read this discussion here and here.

I have a dataset of 977 records where class proportion is 77:23.

My balanced accuracy is 75.5, AUC is 81% but my average_precision_score is only 56%.

Does it make sense to use PR score for my dataset? Is it heavily imbalanced?

Actually, in our domain, missing a positive case is costly. So, in order to identify all positives correctly, our model does make some mistakes (false positives).

but am not sure what should I decide based on PR score? Is my model useless?

but my AUC, Lift and gain charts show some degree of seperation between classes etc



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.