Let say I have a Time series, I'm using sliding/expanding window method to split to train and test data: train would be all the data I have until day x and test is day x+1.

To avoid Data leakage I'm re-fitting the scaler for each day with the data I have until that day.

My question is: Why not fit the scaler with the data using all the training points and the Features from the next day ( Test Data)?

It wont be data leakage as I already have the test features available to me at the day of the forecast..



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.