forecasting - get fitted values from evaluate function

[ArturDragunov]

Question: In issue #888, a similar question was raised regarding how to obtain fitted values from a model. I'm currently looking for a similar solution with the evaluate function. The documentation indicates that users can extract y_train, y_test, and y_pred values. Is there a way to also extract y_fitted using this function? The workaround could be to refit the model with the best parameters and obtain y_fitted separately from the evaluate function.
Also, is there a possibility to obtain fitted coefficient or other parameters of the fitted model from evaluate function?

[fkiraly]

Could you explain what you mean with y_fitted?

Regarding coeffs, there is currently no convenient way to get them from evaluate, that could be a feature request (appreciate an issue).

The advanced benchmarking framework should support that though, see https://github.com/sktime/sktime-workshop-pydata-prague-2023/blob/main/notebooks/05_evaluation_and_benchmarking.ipynb

FYI @hazrulakmal

[ArturDragunov]

Sure! "y_fitted" here refers to the in-sample predictions produced by the forecaster used during fitting inside the evaluate(). As extracting coefficients and y_fitted is a very possibly interrelated request, I created an issue #5452 with the description of both requests. I hope it's more clear now :) If not, please, let me know.