[Question] Viewable Preprocessors and Regressors (internal mechanisms)?

[BradKML]

Short Question Description

• How did this question come about? I would like to see which regression model would discord unneeded data columns (e.g. Lasso and ElasticNet), and which regressors (e.g. Ridge) relies on data preprocessors and feature preprocessors to function as the filter.
• Would a small code snippet help? See below, currently trying to see if one can directly check the objects and have them display their internal parameters.
• What have you already looked at? Other tools such as LazyPredict, but their preprocessors did not show any array parameters. Test Run with Example Predictor shankarpandala/lazypredict#387

[eddiebergman]

Hi @BrandonKMLee,

The first step would be to limit your search to just the regressors of interest.

from autosklearn.classification import AutoSklearnRegresssor

regr = AutoSklearnRegressor(..., include={"regressor": [<names here>]})
regr.fit(X, y)

From here, you could use regr.show_models() to see what pipelines are in the final ensemble and what components it consists off. However I would imagine this would just end up focusing on one regressor. You could do that process above for each regressor of interest and see what final pipelines you get out and if it prefers any particular feature preprocessors or not.

Best,
Eddie