Importing SMOTE raise AttributeError: module 'sklearn.metrics._dist_metrics' has no attribute 'DistanceMetric32'

Question

Running from imblearn.over_sampling import SMOTE will raise following error.

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
d:\A\OneDrive - UBC\ENGR\518 Machine Learning\Project\codes\model_training_laptop - Copy.ipynb Cell 2 in <cell line: 1>()
----> 1 from imblearn.over_sampling import SMOTE

File e:\Anaconda\lib\site-packages\imblearn\__init__.py:52, in <module>
     48     sys.stderr.write("Partial import of imblearn during the build process.\n")
     49     # We are not importing the rest of scikit-learn during the build
     50     # process, as it may not be compiled yet
     51 else:
---> 52     from . import combine
     53     from . import ensemble
     54     from . import exceptions

File e:\Anaconda\lib\site-packages\imblearn\combine\__init__.py:5, in <module>
      1 """The :mod:`imblearn.combine` provides methods which combine
      2 over-sampling and under-sampling.
      3 """
----> 5 from ._smote_enn import SMOTEENN
      6 from ._smote_tomek import SMOTETomek
      8 __all__ = ["SMOTEENN", "SMOTETomek"]

File e:\Anaconda\lib\site-packages\imblearn\combine\_smote_enn.py:10, in <module>
      7 from sklearn.base import clone
      8 from sklearn.utils import check_X_y
---> 10 from ..base import BaseSampler
     11 from ..over_sampling import SMOTE
     12 from ..over_sampling.base import BaseOverSampler

File e:\Anaconda\lib\site-packages\imblearn\base.py:15, in <module>
     12 from sklearn.preprocessing import label_binarize
     13 from sklearn.utils.multiclass import check_classification_targets
---> 15 from .utils import check_sampling_strategy, check_target_type
     16 from .utils._validation import ArraysTransformer
     17 from .utils._validation import _deprecate_positional_args

File e:\Anaconda\lib\site-packages\imblearn\utils\__init__.py:7, in <module>
      1 """
      2 The :mod:`imblearn.utils` module includes various utilities.
      3 """
      5 from ._docstring import Substitution
----> 7 from ._validation import check_neighbors_object
      8 from ._validation import check_target_type
      9 from ._validation import check_sampling_strategy

File e:\Anaconda\lib\site-packages\imblearn\utils\_validation.py:15, in <module>
     12 import numpy as np
     14 from sklearn.base import clone
---> 15 from sklearn.neighbors._base import KNeighborsMixin
     16 from sklearn.neighbors import NearestNeighbors
     17 from sklearn.utils import column_or_1d

File e:\Anaconda\lib\site-packages\sklearn\neighbors\__init__.py:6, in <module>
      1 """
      2 The :mod:`sklearn.neighbors` module implements the k-nearest neighbors
      3 algorithm.
      4 """
----> 6 from ._ball_tree import BallTree
      7 from ._kd_tree import KDTree
      8 from ._distance_metric import DistanceMetric

File sklearn\neighbors\_ball_tree.pyx:1, in init sklearn.neighbors._ball_tree()

AttributeError: module 'sklearn.metrics._dist_metrics' has no attribute 'DistanceMetric32'

What is the result of `import imblearn; imblearn.show_versions()`? — Alexander L. Hayes, Dec 09 '22 at 23:09
This is probably a case where upgrading `scikit-learn` and `imbalanced-learn` will resolve the problem: `pip install --upgrade scikit-learn` and `pip install --upgrade imbalanced-learn` — Alexander L. Hayes, Dec 09 '22 at 23:10
@AlexanderL.Hayes Thanks, now the import problem has been solved using your answer. — Leo, Dec 15 '22 at 20:09
Good to know! I've turned the comment into an answer. Feel free to accept it, or at least our notes will be here for future people who end up in this thread. — Alexander L. Hayes, Dec 16 '22 at 01:03

score 2 · Accepted Answer · answered Dec 16 '22 at 01:02

This is probably a case where upgrading scikit-learn and imbalanced-learn will resolve the problem.

pip install --upgrade scikit-learn
pip install --upgrade imbalanced-learn

Not all versions of scikit-learn and imbalanced-learn are compatible with one another. Version 0.10.0 should be compatible with scikit-learn>=1.0.0 (e.g. discussion here).

Importing SMOTE raise AttributeError: module 'sklearn.metrics._dist_metrics' has no attribute 'DistanceMetric32'

1 Answers1

Linked