Docstring not enough informative for Pipeline class

[IronFarm]

Description

The docstrings for multiple methods in the Pipeline class state that the methods "Apply transformers/samplers...". However, most of the methods ignore the samplers e.g. the predict method:

imbalanced-learn/imblearn/pipeline.py

Lines 349 to 350 in eafae67

	"""Apply transformers/samplers to the data, and predict with the final
	estimator

vs.

imbalanced-learn/imblearn/pipeline.py

Lines 367 to 368 in eafae67

	if hasattr(transform, "fit_sample"):
	pass

The behaviour appears to be correct (sampling should only be applied when fitting) but the docstrings are misleading.

Affected Methods

• predict
• predict_proba
• decision_function
• predict_log_proba
• score

[glemaitre]

This is the expected behaviour: only sample at fit time and do nothing during predict and scoring.

[glemaitre]

I closed a bit fast here. The docstring could be improved mentioning that sampling only occur at fit time.

Thanks for reporting