'SVC' object has no attribute '_probA'

[jayahm]

Hi,

I got the follwoing error:

~\Anaconda3\lib\site-packages\deslib\base.py in fit(self, X, y)
    241         self._setup_label_encoder(y)
    242         y_dsel = self.enc_.transform(y_dsel)
--> 243         self._set_dsel(X_dsel, y_dsel)
    244 
    245         # validate the value of k

~\Anaconda3\lib\site-packages\deslib\base.py in _set_dsel(self, X, y)
    336         self.n_features_ = X.shape[1]
    337         self.n_samples_ = self.DSEL_target_.size
--> 338         self.DSEL_processed_, self.BKS_DSEL_ = self._preprocess_dsel()
    339 
    340     def _set_region_of_competence_algorithm(self):

~\Anaconda3\lib\site-packages\deslib\base.py in _preprocess_dsel(self)
    671                    in DSEL.
    672         """
--> 673         BKS_dsel = self._predict_base(self.DSEL_data_)
    674         processed_dsel = BKS_dsel == self.DSEL_target_[:, np.newaxis]
    675 

~\Anaconda3\lib\site-packages\deslib\base.py in _predict_base(self, X)
    695 
    696         for index, clf in enumerate(self.pool_classifiers_):
--> 697             labels = clf.predict(X)
    698             predictions[:, index] = self._encode_base_labels(labels)
    699         return predictions

~\Anaconda3\lib\site-packages\sklearn\utils\metaestimators.py in <lambda>(*args, **kwargs)
    117 
    118         # lambda, but not partial, allows help() to work with update_wrapper
--> 119         out = lambda *args, **kwargs: self.fn(obj, *args, **kwargs)
    120         # update the docstring of the returned function
    121         update_wrapper(out, self.fn)

~\Anaconda3\lib\site-packages\sklearn\model_selection\_search.py in predict(self, X)
    485         """
    486         self._check_is_fitted('predict')
--> 487         return self.best_estimator_.predict(X)
    488 
    489     @if_delegate_has_method(delegate=('best_estimator_', 'estimator'))

~\Anaconda3\lib\site-packages\sklearn\svm\_base.py in predict(self, X)
    612             y = np.argmax(self.decision_function(X), axis=1)
    613         else:
--> 614             y = super().predict(X)
    615         return self.classes_.take(np.asarray(y, dtype=np.intp))
    616 

~\Anaconda3\lib\site-packages\sklearn\svm\_base.py in predict(self, X)
    333         X = self._validate_for_predict(X)
    334         predict = self._sparse_predict if self._sparse else self._dense_predict
--> 335         return predict(X)
    336 
    337     def _dense_predict(self, X):

~\Anaconda3\lib\site-packages\sklearn\svm\_base.py in _dense_predict(self, X)
    353             X, self.support_, self.support_vectors_, self._n_support,
    354             self._dual_coef_, self._intercept_,
--> 355             self._probA, self._probB, svm_type=svm_type, kernel=kernel,
    356             degree=self.degree, coef0=self.coef0, gamma=self._gamma,
    357             cache_size=self.cache_size)

AttributeError: 'SVC' object has no attribute '_probA'

scikit-learn version '0.24.1'

I tried to downgrade scikit-learn to older versions, but doing that may make the other libraries not work.

[Menelau]

Hello,

Without having an example and an explanation of what you were trying to do I cannot help you since the error stack alone is not informative enough. Are you loading a previously saved SVC? Methods saved with different sklearn versions may not be compatible with new ones.

[jayahm]

Very sorry for the lack of information. Because I thought it has something to with DESLIB compatibility with scikit learn.

To make it short, actually, I run the code with the latest version of scikit-learn.

But, that error came out.

If I downgrade to an older version, it is okay.

Can you confirm whether deslib is still compatible with the latest version of scikit-learn.

If yes, that means, the root of the problem is the saved SVM model from the older version of scikit-learn and not related to DESlib.

[Menelau]

Yes, the library is compatible with scikit-learn=0.24.1

I'm pretty sure your problem is the saved model. It is already known that SVM saved from previous sklearn versions are not compatible with the new ones. You can find some more information about that on stack overflow or even in the sklearn documentation/issues.

[jayahm]

I see. I'll check my scikit-learn. Thanks