【Python学习】【scikit-learn】Pipeline报错fit_transform() takes 2 positional arguments but 3 were given

最近在读 Hands-On Machine Learning with Scikit-Learn & TensorFlow 这本书，在学到pipeline的时候，我模仿者写了这也的代码：

[python]  view plain  copy

1. num_attribs=list(housing_numerical)  
2. cat_attribs=["ocean_proximity"]  
3.   
4. num_pipeline=Pipeline([  
5.     ("selector",DataFrameSelector(num_attribs)),  
6.     ("imputer",Imputer(strategy="median")),  
7.     ("attribs_adder",CombinedAttributesAdder()),  
8.     ("std_scaler",StandardScaler()),  
9. ])  
10.   
11. cat_pipeline=Pipeline([  
12.     ("selector",DataFrameSelector(cat_attribs)),  
13.     <span style="color:#ff0000;">("label_binarizer",LabelBinarizer())</span>,  
14. ])  
15.   
16. full_pipeline=FeatureUnion(transformer_list=[  
17.     ("num_pipeline",num_pipeline),  
18.     ("cat_pipeline",cat_pipeline),  
19. ])  

但是会报错如下：

[plain]  view plain  copy

1. TypeError: fit_transform() takes 2 positional arguments but 3 were given  

我想，这应该是版本更新引起的问题，果然我在这里找到了答案。以下为引用：

The pipeline is assuming LabelBinarizer's fit_transform method is defined to take three positional arguments:

def fit_transform(self, x, y)
    ...rest of the code

while it is defined to take only two:

def fit_transform(self, x):
    ...rest of the code

所以，解决方法就是，自己写一个根据LabelBinarizer写一个MyLabelBinarizer,可以有三个参数self,X,y=None.

from sklearn.base import TransformerMixin #gives fit_transform method for free
class MyLabelBinarizer(TransformerMixin):
    def __init__(self, *args, **kwargs):
        self.encoder = LabelBinarizer(*args, **kwargs)
    def fit(self, x, y=0):
        self.encoder.fit(x)
        return self
    def transform(self, x, y=0):
        return self.encoder.transform(x)