support for external regressors (if an additional regressor is in X, but not in the self.extra_regressor it will return ValueError) is it OK?
I had to violate the API a bit - changed static method _transform_data_to_tsmodel_input_format for an instance method. The reason is that unlike in prophet the helper functions for preprocessing the data frame likemake_future_dataframe are not just simple helpers but hides a huge complexity around data preprocessing and I couldn't get it working without them.
holidays windows features from HolidaTransformer are not passed yet
holidays are implemented as events with the method add_events, because the holidays' interface is not very flexible and there is no obvious way how to pass customer holidays, events should work the same.
The first implementation of the neural prophet:
make_future_dataframeare not just simple helpers but hides a huge complexity around data preprocessing and I couldn't get it working without them.eventswith the methodadd_events, because the holidays' interface is not very flexible and there is no obvious way how to pass customer holidays, events should work the same.