i have an imbalance data set and I used smote to oversample the minority class and undersample the majority class. now, I want to check the test AUC using predict_proba of the model.

I have two questions: 1. Do I have to correct the probability if I am comparing AUCs? 2. How can I correct it (a combination of undersampling and oversampling!)

  1. No, any adjustment to the probabilities will presumably be monotonic, so the rank-ordering of the predictions will be the same, so the AUC will be the same.

  2. See, e.g., https://datascience.stackexchange.com/a/58899/55122

See also the more complex "probability calibration" techniques.

Also, if you see better results after smote+undersampling, and can share your data and work, I'd be very interested. I haven't yet seen an example where training on the original dataset doesn't do just as well (with appropriate thresholding).


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.