AttributeError: 'DataFrame' object has no attribute 'name'

[islrnd]

AttributeError                            Traceback (most recent call last)
<ipython-input-32-8d38637b6cc6> in <module>
      6 
      7 oversampler=SMOTE(random_state=42)
----> 8 smote_train, smote_target = oversampler.fit_resample(X,y)
      9 
     10 print("Before OverSampling, counts of label '0', '1':", smote_target['label'].value_counts())

~\Anaconda3\lib\site-packages\imblearn\base.py in fit_resample(self, X, y)
     73         """
     74         check_classification_targets(y)
---> 75         X, y, binarize_y = self._check_X_y(X, y)
     76 
     77         self.sampling_strategy_ = check_sampling_strategy(

~\Anaconda3\lib\site-packages\imblearn\base.py in _check_X_y(self, X, y, accept_sparse)
    148         if hasattr(y, "loc"):
    149             # store information to build a series
--> 150             self._y_name = y.name
    151             self._y_dtype = y.dtype
    152         else:

~\Anaconda3\lib\site-packages\pandas\core\generic.py in __getattr__(self, name)
   5065             if self._info_axis._can_hold_identifiers_and_holds_name(name):
   5066                 return self[name]
-> 5067             return object.__getattribute__(self, name)
   5068 
   5069     def __setattr__(self, name, value):

AttributeError: 'DataFrame' object has no attribute 'name'

[chkoar]

@glemaitre I think that is related with the sanity check that we were discussing about, no? The user passes something that has .loc but it is not a Series as we expect. It is a DataFrame.

[glemaitre]

Yes but this is not pythonic to type check. We should probably look at attributes that are "couacing" better for Series.

[glemaitre]

@islrnd you should pass a numpy array or a pandas series, not a dataframe.

[00krishna]

This is a confusing issue. If I pass in a DataFrame then I get the error about no attribute name. If I pass in a Series I get a different error about ValueError: Found array with 0 feature(s) (shape=(7788867, 0)) while a minimum of 1 is required. So no matter how I am putting the data in, it generates a critical error.

[chkoar]

@00krishna post a sample code to reproduce you error.

[00krishna]

@chkoar Sorry, I realized I have all categorical features and that is causing the problem.

[parth-radonc]

@00krishna I am facing the similar issue.. Can you help me with it?

[chkoar]

@parth-mango You probably pass a data frame in y. You should pass a Series object. If you want provide a minimal reproducible code.

[flowersw]

I had the same issue, and was accidentally passing in a DataFrame for y instead of Series

[chkoar]

@flowersw #673 or a post PR hopefully will solve this problem.

[serjko]

Hi there,

I'm facing a similar issue that wasn't resolved after converting the initial DataFrame to Series. Please take a look at my question on SO https://datascience.stackexchange.com/questions/67141/passing-data-to-smote-after-applying-train-test-split

[chkoar]

@serjko please post a minimal reproducible example and your package versions.

[serjko]

Hi @chkoar, Thanks for jumping in so quickly. This is a false alarm. The root cause was my dataset and not SMOTE. I made it work after cleaning up the data and passing Series as 2nd variable for fit_sample. Apologies for bothering you and thanks again for the answer!