Hello !
I would like to work on this issue.

I am wondering whether it would be better to add an "automatic_binning" parameter to the chi2 function or to clarify the documentation by explicitly stating that continuous data must be pre-binned before using chi2.

In the second case, it might also be helpful to include an example demonstrating how to bin continuous data before applying chi2. What do you think would be the preferred approach?

It is great to know that this issue is relevant.

As an immediate solution, some basic binning example is recommended.

In the long run, it is recommended that some binning technique has to be implemented. It is important to note, that the multivariate binning is non-trivial. For example, there are different binning strategies, e.g. R binning for Chi2 and handling infinite values.

Personally, I am not proficient enough in statistics to propose and theoretically justify any particular binning strategy and its influence on the feature selection performance.