The base model does not seem to be accepted by MapieQuantileRegressor

[masakljun]

Describe the bug I am trying to use my XGBoost model in combination with MapieQuantileRegressor. However, the error below comes up when I try to set estimator = my_xgboost_model

To Reproduce Steps to reproduce the behavior:

1. After the code I use (see screenshot below), this is the error that comes up.
2. ValueError: The base model does not seem to be accepted by MapieQuantileRegressor. Give a base model among: quantile_estimator_params.keys()Or, add your base model to quantile_estimator_params.

Expected behavior I see from the error that I should somehow add my model to the quantile_estimator_params, but I don't know how to do that.

Screenshots The relevant part of the code:

Desktop (please complete the following information):

• OS: Windows
• Browser: Chrome
• MAPIE Version: 0.4.1

Thanks for the help!

[LacombeLouis]

Hey @masakljun! It's important that for MapieQuantileRegressor you provide an estimator that can perform quantile regression. As such, I would suggest using LGBMRegressor(objective='quantile', alpha=0.5) or GradientBoostingRegressor(loss="quantile).

If you need to use XGBRegressor, you need to implement a quantile method as suggested in this article. Note that you would need to fit a wrapper on this method and add it to the MapieQuantileRegressor().quantile_estimator_params by defining the loss_name and alpha_name for that estimator.

[masakljun]

Ok, thanks for the clarification and help!