Reference Issues/PRs

Fixes #26711

What does this implement/fix? Explain your changes.

This PR fixes the LabelEncoder.set_output() method availability issue where the method was listed in documentation but threw an AttributeError when called.

Problem:

• LabelEncoder inherited set_output from TransformerMixin but it was conditionally unavailable due to @available_if(_auto_wrap_is_configured) decorator
• The condition required get_feature_names_out method and proper auto_wrap_output_keys configuration, both of which were missing
• Users saw set_output in documentation but got runtime errors when trying to use it

Solution:

1. Removed auto_wrap_output_keys=None from the LabelEncoder class definition, which was explicitly blocking the auto-wrapping functionality
2. Added get_feature_names_out() method that returns appropriate feature names for the single output that LabelEncoder produces
3. The method handles both cases where input feature names are provided and when they need to be generated

Implementation Details:

• get_feature_names_out() returns a single-element array with either the first input feature name (if provided) or a default name 'labelencoder_output'
• This satisfies the _auto_wrap_is_configured() condition in sklearn/utils/_set_output.py
• Maintains backward compatibility - all existing functionality remains unchanged
• Follows the same pattern as other sklearn transformers

Testing:
The fix enables the following workflow that previously failed:
from sklearn.preprocessing import LabelEncoder
le = LabelEncoder()
le.set_output(transform='pandas') # This now works without AttributeError
le.fit(['cat', 'dog', 'bird'])
result = le.transform(['cat', 'dog']) # Can now return pandas Series when configured

text

Any other comments?

This is a minimal, targeted fix that addresses the API consistency issue without breaking changes. The implementation follows sklearn's established patterns for transformers and maintains the principle that LabelEncoder is designed for target variable transformation (1D output).

The fix enables users to use set_output with LabelEncoder as they would expect from reading the documentation, resolving the confusion between documented and actual behavior