Assume that I have two datasets $Train$ and $Test$. And there is the problem illustrated below: there are different feature distribution between two datasets

enter image description here

I want to find the train subset $A \subset Train$ of the given size (for example $n$) with distribution similar to this feature distribution on test.

I've already tried the following: using approach of adversarial validation I tried to assign new target variable 'is_train' which is 1 on train objects and 0 on test. Then I am sampling $n$ rows of train, concatenating them with test, splitting this new dataset into train and validation, fitting classifier on train trying to predict target 'is_tran' on validation, If ROC_AUC is not far from 0.5 wanted sample is found.

What can I try more? Is it a good idea just to reduce the number of objects in train that causes distribution differences?


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.