pandas dtypes "boolean" not supported in classification target

Describe the bug

Pandas has extended NumPy’s dtypes and these dtypes extensions are not all supported as targets in a sklearn classifier. In particular, if the target y is a Pandas “boolean” dtype, a classifier such as LogisticRegression fails whereas if the target is a numpy “bool” dtype, the classifier will not fail.

Steps/Code to Reproduce

import pandas as pd
from sklearn.linear_model import LogisticRegression

df = pd.DataFrame({"boolean_col": pd.Series([False, True, False, True], dtype="boolean"),
                   "bool_col": pd.Series([False, True, False, True], dtype="bool"),
                   "num_col": pd.Series([1, 2, 3, 4])})

clf = LogisticRegression()

Expected Results

clf.fit(df[["num_col"]], df.bool_col)
--------------------------------------------------------------------------------------------------------------
LogisticRegression()

Actual Results

clf.fit(df[["num_col"]], df.boolean_col)
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-4-9a0c5abf260e> in <module>
----> 1 clf.fit(df[["num_col"]], df.boolean_col)

~\AppData\Local\Continuum\anaconda3\envs\lognode\lib\site-packages\sklearn\linear_model\_logistic.py in fit(self, X, y, sample_weight)
   1343                                    order="C",
   1344                                    accept_large_sparse=solver != 'liblinear')
-> 1345         check_classification_targets(y)
   1346         self.classes_ = np.unique(y)
   1347 

~\AppData\Local\Continuum\anaconda3\envs\lognode\lib\site-packages\sklearn\utils\multiclass.py in check_classification_targets(y)
    170     if y_type not in ['binary', 'multiclass', 'multiclass-multioutput',
    171                       'multilabel-indicator', 'multilabel-sequences']:
--> 172         raise ValueError("Unknown label type: %r" % y_type)
    173 
    174 

ValueError: Unknown label type: 'unknown'

Versions

System: python: 3.8.2 (default, Apr 14 2020, 19:01:40) [MSC v.1916 64 bit (AMD64)] executable: ~\AppData\Local\Continuum\anaconda3\envs\lognode\python.exe machine: Windows-10-10.0.18362-SP0

Python dependencies: pip: 20.0.2 setuptools: 46.1.3.post20200330 sklearn: 0.23.1 numpy: 1.18.1 scipy: 1.4.1 Cython: None pandas: 1.0.3 matplotlib: 3.2.1 joblib: 0.14.1 threadpoolctl: 2.1.0

Built with OpenMP: True

[foxale]

I can confirm it’s still present in 0.24.2

[se-jaeger]

Still failing with version ‘1.0.1’ when using precision_recall_curve with boolean data.

I debugged and found that in utils.multiclass.type_of_target something goes wrong. (See comment earlier.) There is still a y = np.asarray(y), which is the cause?

It’s because in utils.multiclass.type_of_target, the casting y = np.asarray(y) converts to object dtype. But since dtype boolean allows the pd.NA and is experimental, I’m not sure how sklearn should deal with it.