I am trying to predict classification problem. For that I have used Ranger, Xgboost and naive bayes. My Response class is imbalance . 92:8 ratio. My positive Response is only 8% of whole data.

Because of class imbalance I am getting more FP / FN.

I tried different sampling on training set and predicted with original set gives me more FP.

Can I do something like ?: - split data into train and test - apply sampling method individually on train and test - and use sampled test to predict.



1 Answer 1


Oversampling on your test set will only artifically improve your performance. What you may want to do instead is changing your objective function to give more importance to you imbalanced class. There are already a lot of question about class imbalance on this website, such as : Classification problem: custom minimization measure


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.