I'm trying to create a Random Forest Classifier for selecting ~ 700 features.

I have a highly imbalanced dataset to select features from. There are significantly fewer positive cases (1%) compared with negative cases.

I'm trying to select an appropriate evaluation metric between the following options; F1, PR-AUC and balanced accuracy. Can anyone give some insight as to which is the most suitable?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.