In the workshop code, testing(splits) has no recipe.
https://github.com/business-science/modeltime_h2o_workshop/blob/master/forecasting_modeltime_h2o.R
Is this normal?
In tidymodels the parsnip::predict()-function implicitly adds recipes to new data.
Does modeltime_calibrate() or modeltime_forecast() the same?
Where can I find essential documentations, how to handle recipes correctly with modeltime?