本文主要是介绍特征选择错误：The classifier does not expose coef_ or feature_importances_ attributes，希望对大家解决编程问题提供一定的参考价值，需要的开发者们随着小编来一起学习吧！
在利用RFE进行特征筛选的时候出现问题，源代码如下：
from sklearn.svm import SVR
model_SVR = SVR(C=1.0, cache_size=200, coef0=0.0, degree=3, epsilon=0.1, gamma='auto',kernel='rbf', max_iter=-1, shrinking=True, tol=0.001, verbose=False)
selector = RFE(estimator = model_SVR, step=1)
selector = selector.fit(df_train_X, df_train_y.values.ravel())
selector.support_
selector.ranking_
错误如下：
---------------------------------------------------------------------------
RuntimeError Traceback (most recent call last)
<ipython-input-49-ca6dfcbf43ee> in <module>38 kernel='rbf', max_iter=-1, shrinking=True, tol=0.001, verbose=False)39 selector = RFE(estimator = model_SVR, step=1)
---> 40 selector = selector.fit(df_train_X, df_train_y.values.ravel())41 selector.support_42 selector.ranking_~/anaconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/feature_selection/rfe.py in fit(self, X, y)142 The target values.143 """
--> 144 return self._fit(X, y)145 146 def _fit(self, X, y, step_score=None):~/anaconda3/envs/tensorflow/lib/python3.7/site-packages/sklearn/feature_selection/rfe.py in _fit(self, X, y, step_score)189 coefs = getattr(estimator, 'feature_importances_', None)190 if coefs is None:
--> 191 raise RuntimeError('The classifier does not expose '192 '"coef_" or "feature_importances_" '193 'attributes')RuntimeError: The classifier does not expose "coef_" or "feature_importances_" attributes
经过百度发现问题，原文在stackflow，链接如下：
KNN with RFECV returns: “The classifier does not expose ”coef_“ or ”feature_importances_“ attributes”
“knn不提供进行特征选择的逻辑。除非您为KNN定义自己的特性重要性度量，否则您不能使用它（sklearn的实现）来实现这样的目标。据我所知-没有这样的通用对象，所以-scikit learn没有实现它。另一方面，支持向量机，像每一个线性模型-提供这样的信息。”
也就是用rbf核进行筛选是不可以的，SVR中不提供rbf特征选择的逻辑，对于支持向量回归可以用linear进行筛选
这篇关于特征选择错误：The classifier does not expose coef_ or feature_importances_ attributes的文章就介绍到这儿，希望我们推荐的文章对编程师们有所帮助！
