I fit random forest to my imbalanced dataset with minority class 1. I found that the AUC under the imbalanced data was better than that of re-sampled dataset (over/under sampling). Can someone help to put forward some theoretical explanation to justify this observation or please help to point out the underlying problem which caused this spurious output?

  • 1
    $\begingroup$ Why did you think balancing your data would imporve performance? $\endgroup$ – Matthew Drury Jan 6 '17 at 2:28
  • $\begingroup$ When I faced imbalanced data, I searched google and re-sampling was commonly suggested as a strategy for this situation. So presumably it should work somehow. $\endgroup$ – LUSAQX Jan 6 '17 at 2:38

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.