Warn if label leakage

[noahho]

Describe the workflow you want to enable

Print a warning if the features contain the label or are perfectly correlated to it.

Describe your proposed solution

Check correlation of the columns in X to y in training set and print warning

Describe alternatives you've considered, if relevant

No response

Additional context

No response

Impact

None

[RobertoCorti]

Hi! I'd like to work on this issue. My plan is to add a validation check in ensure_compatible_fit_inputs() that runs after sklearn validation (so X and y are already numpy arrays). The check would:

• Detect if any feature column is identical to the target (exact match after handling NaNs)
• Detect if any numeric feature is perfectly correlated with the target

It would issue a UserWarning with the offending column indices (and feature names if available). This covers both the classifier and regressor paths since they both go through ensure_compatible_fit_inputs.

Happy to hear if there are some preferences