[BUG] 'BalancedBaggingClassifier' and 'EasyEnsembleClassifier' objects have no attribute 'n_features_in_'

[breisfeld]

Describe the bug

Error when using BalancedBaggingClassifier or EasyEnsembleClassifier methods shown in the documentation: https://imbalanced-learn.org/stable/ensemble.html

Steps/Code to Reproduce

Following examples from the documentation...

model = BalancedBaggingClassifier(base_estimator=DecisionTreeClassifier(),
                                sampling_strategy='auto',
                                replacement=False,
                                random_state=0)
model.fit(X_train, y_train)

or

model = EasyEnsembleClassifier(random_state=0)
model.fit(X_train, y_train)

Actual Results

AttributeError                            Traceback (most recent call last)
~\AppData\Local\Temp/ipykernel_247760/4241213219.py in <module>
      8                                 random_state=0)
      9 #model = LGBMClassifier(is_unbalance=True, learning_rate=0.6, n_estimators=500,  random_state=random_state)
---> 10 model.fit(X_train, y_train)

~\anaconda3\envs\iarc_chemicals\lib\site-packages\imblearn\ensemble\_bagging.py in fit(self, X, y)
    319         # RandomUnderSampler is not supporting sample_weight. We need to pass
    320         # None.
--> 321         return self._fit(X, y, self.max_samples, sample_weight=None)
    322 
    323     def _more_tags(self):

~\anaconda3\envs\iarc_chemicals\lib\site-packages\sklearn\ensemble\_bagging.py in _fit(self, X, y, max_samples, max_depth, sample_weight)
    335             max_features = self.max_features
    336         elif isinstance(self.max_features, float):
--> 337             max_features = self.max_features * self.n_features_in_
    338         else:
    339             raise ValueError("max_features must be int or float")

AttributeError: 'BalancedBaggingClassifier' object has no attribute 'n_features_in_'

Similar error for EasyEnsembleClassifier

Versions

Windows-10-10.0.19043-SP0
Python 3.9.7 | packaged by conda-forge | (default, Sep 29 2021, 19:15:42) [MSC v.1916 64 bit (AMD64)]
NumPy 1.20.3
SciPy 1.7.1
Scikit-Learn 1.0.1
Imbalanced-Learn 0.8.1

[leaphan]

i have the same question ,Have you found a solution

[YasCoMa]

You have to downgrade the scikit-learn package using: pip3 install scikit-learn==1.0 -U
The attribute n_features_in_ is deprecated and its support was lost in sklearn version 1.2

[YasCoMa]

i have the same question ,Have you found a solution

I just posted the solution for me if you still need it!

[leaphan]

You have to downgrade the scikit-learn package using: pip3 install scikit-learn==1.0 -U The attribute n_features_in_ is deprecated and its support was lost in sklearn version 1.2

okay,i will try it,thank u very much!

[glemaitre]

We solve the problem in master but we would need to release with a more constraint limitation regarding the version of scikit-learn

[glemaitre]

I solved the issue in master. I will release it shortly. Basically, we broke some expectations in scikit-learn by moving some validation outside from _fit.

[lopezjuana]

I solved the issue in master. I will release it shortly. Basically, we broke some expectations in scikit-learn by moving some validation outside from _fit.

Cómo se resuelve el problema?

[jaymon0703]

I solved the issue in master. I will release it shortly. Basically, we broke some expectations in scikit-learn by moving some validation outside from _fit.

Hi @glemaitre which commit and release was the fix deployed with? Thank you!

[glemaitre]

certainly 0.9.0. We realeased 0.9.1 since