Hi everyone, I’m wondering if there’s a way to obtain in-sample predictions on the training set for a neural forecaster, similar to the forecast_fitted_values() function. Can anyone suggest a method to replicate this behavior? Thank you!

Hi @Asterios Tsiourvas, we are currently working on creating this function. In the meantime you can use the

predict_rolled

function. Another user had the same question and we discussed how to do it an a thread, I will tag you.