I have this X_train and test distribution for the 4 features 'X', 'Y', 'TX' and 'TY'. I realize the range of the distribution is widely varying .. Can you suggest a good way to clean/ transform that data so that the model generalizes pretty well. I have tried (data-mean)/std .. didn't help much with performance Training Data Features:- enter image description here Test Data Features:- enter image description here

  • $\begingroup$ Those plots aren't very helpful, with the wild vertical jumps being connected by line segments. I'd suggest histograms instead, except that training's TX and TY seem to have some dependence on the sample number; is there a time dependency (and the samples are in temporal order)? The distribution in training and testing sets are rather different too, which is generally not going to go well. But especially the test set's TX and TY seem scaled while the training's are not?? We need some more context here. $\endgroup$ – Ben Reiniger Oct 3 at 2:21
  • $\begingroup$ @BenReiniger: Thanks for your response, I will edit the question with more context and the histograms $\endgroup$ – Zakir Oct 3 at 2:56

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.