This message was deleted.

Hi @Siavash Sakhavi! Thanks for using our libraries. Use the

cross_validation

instead for precisely this process. Merge your train and test data in one dataframe, and specify the length of the test set with

test_size

or

n_windows

+

step_size

. The

futr_df

is used only by the

predict

method to specify only the future exogenous covariates, so it will not take the values of the target variable.