如何解决YellowBrick SilhouetteVisualizer的‘Not Fitted’错误？

尝试用YellowBrick的SilhouetteVisualizer绘制KMeans聚类的轮廓图时持续报错，但同一数据集使用KElbowVisualizer时正常。以下是代码及错误信息：

KElbowVisualizer 代码 #

kmeans_kwargs = {"init": "random", "n_init": 10, "max_iter": 300, "random_state": 101}
kmeans = KMeans(n_clusters = k, **kmeans_kwargs)
visualizer = KElbowVisualizer(kmeans, k = (2, 31))
visualizer.fit(X)
visualizer.show()

SilhouetteVisualizer 代码 #

model = KMeans(n_clusters = 9)  # 根据肘部法则选择的最优k值
visualizer = SilhouetteVisualizer(model)
visualizer.fit(X)         
visualizer.show()

错误信息 #

---------------------------------------------------------------------------
NotFittedError                            Traceback (most recent call last)
File ~/miniforge3/envs/tensorflow/lib/python3.9/site-packages/yellowbrick/utils/helpers.py:50, in is_fitted(estimator)
     49 try:
---> 50     estimator.predict(np.zeros((7, 3)))
     51 except sklearn.exceptions.NotFittedError:

File ~/miniforge3/envs/tensorflow/lib/python3.9/site-packages/sklearn/cluster/_kmeans.py:1019, in _BaseKMeans.predict(self, X, sample_weight)
    999 """Predict the closest cluster each sample in X belongs to.
   1000 
   1001 In the vector quantization literature, `cluster_centers_` is called
   (...)
   1017     Index of the cluster each sample belongs to.
   1018 """
-> 1019 check_is_fitted(self)
   1021 X = self._check_test_data(X)

File ~/miniforge3/envs/tensorflow/lib/python3.9/site-packages/sklearn/utils/validation.py:1345, in check_is_fitted(estimator, attributes, msg, all_or_any)
   1344 if not fitted:
-> 1345     raise NotFittedError(msg % {"name": type(estimator).__name__})

NotFittedError: This KMeans instance is not fitted yet. Call 'fit' with appropriate arguments before using this estimator.

During handling of the above exception, another exception occurred:

AttributeError                            Traceback (most recent call last)
Input In [87], in <cell line: 5>()
      3 model = KMeans(n_clusters = 9)
      4 vis = SilhouetteVisualizer(model)
----> 5 visualizer.fit(rfm_scaled)
      6 visualizer.show()

File ~/miniforge3/envs/tensorflow/lib/python3.9/site-packages/yellowbrick/cluster/silhouette.py:136, in SilhouetteVisualizer.fit(self, X, y, **kwargs)
    129 """
    130 Fits the model and generates the silhouette visualization.
    131 """
    132 # TODO: decide to use this method or the score method to draw.
    133 # NOTE: Probably this would be better in score, but the standard score
    134 # is a little different and I'm not sure how it's used.
--> 136 if not check_fitted(self.estimator, is_fitted_by=self.is_fitted):
    137     # Fit the wrapped estimator
    138     self.estimator.fit(X, y, **kwargs)
    140 # Get the properties of the dataset

File ~/miniforge3/envs/tensorflow/lib/python3.9/site-packages/yellowbrick/utils/helpers.py:116, in check_fitted(estimator, is_fitted_by, **kwargs)
     85 """
     86 Determines whether or not to check if the model has been fitted, and will return
     87 ``True`` if so. The ``is_fitted_by`` argument is set to ``'auto'`` by default,
   (...)
    113     Whether or not the model is already fitted
    114 """
    115 if isinstance(is_fitted_by, str) and is_fitted_by.lower() == "auto":
--> 116     return is_fitted(estimator)
    117 return bool(is_fitted_by)

File ~/miniforge3/envs/tensorflow/lib/python3.9/site-packages/yellowbrick/utils/helpers.py:51, in is_fitted(estimator)
     49 try:
     50     estimator.predict(np.zeros((7, 3)))
---> 51 except sklearn.exceptions.NotFittedError:
     52     return False
     53 except AttributeError:
     54     # Some clustering models (LDA, PCA, Agglomerative) don't implement ``predict``

AttributeError: module 'sklearn' has no attribute 'exceptions'

核心错误是AttributeError: module 'sklearn' has no attribute 'exceptions'——你的YellowBrick版本过旧，内部代码仍在调用sklearn.exceptions路径，但新版本Scikit-learn已调整异常类的存放路径，导致兼容性冲突。而KElbowVisualizer未触发该检查逻辑，因此未报错。

方法1：升级YellowBrick到最新版本 #

执行以下命令更新YellowBrick，新版本已修复该兼容性问题：

pip install --upgrade yellowbrick

方法2：手动拟合KMeans后再传入可视化器 #

如果暂时无法升级库，可以先手动拟合KMeans模型，再传给SilhouetteVisualizer，跳过YellowBrick内部的拟合检查逻辑：

model = KMeans(n_clusters = 9)
model.fit(X)  # 先手动完成模型拟合
visualizer = SilhouetteVisualizer(model)
visualizer.fit(X)  # 可视化器将直接使用已拟合的模型
visualizer.show()

内容的提问来源于stack exchange，提问作者clpoh