[scikit-learn] About the Boston housing prices dataset

Juan Nunez-Iglesias jni at fastmail.com
Tue Oct 13 06:37:46 EDT 2020

I very much like your paragraph, Olivier. I might recommend additionally raising it as a warning when calling the data creation function.

For reference, in scikit-image when we removed Lena we raised a warning and returned an alternative (the now-famous `data.astronaut()`) for two versions, before removing the image altogether.

I think that was a good approach for us, but I like your preference of using the dataset as an educational opportunity in this case. I lean in that direction also, but with the caveat that I think the message should be included in a warning, not just in the docstring.


> On 13 Oct 2020, at 8:59 pm, Olivier Grisel <olivier.grisel at ensta.org> wrote:
> Hi all,
> Thanks to the sustained effort of several contributors (thanks Maria
> and Lucy in particular), the Boston housing price dataset is no longer
> used in the examples of scikit-learn (nor in the test suite) in the
> master branch.
> To give some context on why this dataset is problematic, please have a
> look at this discussion and  the blog post linked in it:
> https://github.com/scikit-learn/scikit-learn/issues/16155
> Now that we no longer use sklearn.datasets.load_boston internally, we
> have to make a decision about what to do with the loader function
> itself: deprecate it? just silently hide it from our documentation
> from our documentation (probably a bad idea)? keep it but educate our
> users about its ethical problem?
> Personally, I would be slightly in favor of the latter option and I
> drafted a short paragraph here:
> https://github.com/scikit-learn/scikit-learn/pull/18594#issuecomment-707601448
> Please feel free to share your thoughts so that we can hopefully make
> a consensual decision before the 0.24 release.
> Regards,
> -- 
> Olivier
> _______________________________________________
> scikit-learn mailing list
> scikit-learn at python.org
> https://mail.python.org/mailman/listinfo/scikit-learn

More information about the scikit-learn mailing list