02 util: AttributeError: 'SelectFromModel' object has no attribute 'estimator_'

[kzfm]

feature_name[abs(selector_en.estimator.coef_)>selector_en.threshold_]

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-18-58d17a399d71> in <module>()
----> 1 feature_name[abs(selector_en.estimator.coef_)>selector_en.threshold_]

~/anaconda3/lib/python3.6/site-packages/sklearn/feature_selection/from_model.py in threshold_(self)
    172     @property
    173     def threshold_(self):
--> 174         scores = _get_feature_importances(self.estimator_, self.norm_order)
    175         return _calculate_threshold(self.estimator, scores, self.threshold)
    176 

AttributeError: 'SelectFromModel' object has no attribute 'estimator_'

[MNaumanBootter]

Sometimes, this kind of issue occurs because of scikit-learn's version mismatch. You may have to use the same version as of author's scikit-learn version.

I was having error: "AttributeError: 'SelectFromModel' object has no attribute 'importance_getter'".

I downgraded the scikit-learn from version 1.0.2 to 0.23.2 which was used to train the model and it fixed the issue.

Hope it helps.