Python3.7sklearn包中缺失Imputer函数

本人是win10下安装的anaconda3,相关版本如下：

然而在使用sklearn中的Imputer函数时，会出现报错：

>>> import numpy as np
>>> import sklearn
>>> from sklearn import preprocessing
>>> from sklearn.preprocessing import Imputer
Traceback (most recent call last):
  File "", line 1, in 
ImportError: cannot import name 'Imputer' from 'sklearn.preprocessing' (E:\python\lib\site-packages\sklearn\preprocessing\__init__.py)

利用dir()查看包内的函数，发现没有Imputer：

>>> dir(sklearn.preprocessing) 
['Binarizer', 'FunctionTransformer', 'KBinsDiscretizer', 'KernelCenterer', 'LabelBinarizer', 'LabelEncoder', 'MaxAbsScaler', 'MinMaxScaler', 'MultiLabelBinarizer', 'Normalizer', 'OneHotEncoder', 'OrdinalEncoder', 'PolynomialFeatures', 'PowerTransformer', 'QuantileTransformer', 'RobustScaler', 'StandardScaler', '__all__', '__builtins__', '__cached__', '__doc__', '__file__', '__loader__', '__name__', '__package__', '__path__', '__spec__', '_csr_polynomial_expansion', '_data', '_discretization', '_encoders', '_function_transformer', '_label', 'add_dummy_feature', 'binarize', 'label_binarize', 'maxabs_scale', 'minmax_scale', 'normalize', 'power_transform', 'quantile_transform', 'robust_scale', 'scale']

原因：0.22版本sklearn，imputer不在preprocessing里了，而是在sklearn.impute里，除了SimpleImputer外，还增加了KNNImputer。另外还有IterativeImputer用法，可以参考这个 https://scikit-learn.org/stable/modules/impute.html#impute 以及 https://scikit-learn.org/stable/modules/preprocessing.html#preprocessing 

from sklearn.impute import SimpleImputer

imp = SimpleImputer()
imp.fit([[1, 2], 
         [np.nan, 3], 
         [7, 6]])
imp.transform([[np.nan, 2], 
               [6, np.nan]])

array([[4.        , 2.        ],
       [6.        , 3.66666667]])