Feature Request: give universal monotonic trend to features for automatic binning

[hermanashley]

Hi there,

It would be great for us if there was an option like this:

        return BinningProcess(
            categorical_variables=categorical_variables,
            variable_names=my_features,
            min_prebin_size=my_min_prebin_size,
            monotonic_trend='ascending'
        )

And ideally this would force an ascending trend for everything we want to bin.

If there is another way to quickly give all features the same monotonic trend please let me know as this would be very useful for our team! :)

[guillermo-navas-palencia]

Hi @hermanashley.

The BinningProcess class has the option binning_fit_params. In the meantime, you can follow this example: http://gnpalencia.org/optbinning/tutorials/tutorial_binning_process_FICO_xAI.html (cell 8).

I will add this feature in version 0.18.0 :)

[hermanashley]

Thank you very much!

If possible, would error handling be able to display which features are not being automatically binned? We sometimes have a feature sneak in that cannot be monotonically binned, it would be great to know which ones these are so we can handle them appropriately / adjust our approach as needed.

If not possible, adding the trend parameter to the BinningProcess is still a great start for us. Thanks again!