使用MultiLabelBinarizer实现SMOTE

Question

这是我的第一个NLP项目。我试着用SMOTE作为一个有14个类的分类器。在使用SMOTE之前，我需要将类转换为数组。我试过使用MultiLinearBinarizer，但它似乎不起作用。从堆栈跟踪来看，似乎所有的东西都被转换了。我需要将其他东西转换成数组吗？我该怎么做？

from sklearn.naive_bayes import MultinomialNB
from sklearn.pipeline import Pipeline
from sklearn.feature_extraction.text import TfidfTransformer
from sklearn.preprocessing import MultiLabelBinarizer

nb = Pipeline([('vect', CountVectorizer()),
               ('tfidf', TfidfTransformer()),
               ('clf', MultinomialNB()),
              ])
nb.fit(X_train, y_train)


mlb = MultiLabelBinarizer()
print(mlb.fit_transform(df["Technique"].str.split(",")))
print(mlb.classes_)

import imblearn 
from imblearn.over_sampling import SMOTE 

smote = SMOTE('minority')

x_sm, y_sm = smote.fit_sample(X_train, y_train)
#print(x_sm.shape, y_sm.shape)
pd.DataFrame(x_sm.todense(), columns=tv.get_feature_names())

我得到了错误ValueError:无法将字符串转换为浮动：“左中间”

这是堆栈跟踪

[[1 0 0 ... 0 0 0]
 [1 0 0 ... 0 0 0]
 [0 0 0 ... 0 0 0]
 ...
 [0 0 0 ... 0 0 0]
 [0 0 0 ... 0 0 0]
 [0 0 0 ... 0 0 0]]
['Appeal_to_Authority' 'Appeal_to_fear-prejudice' 'Bandwagon'
 'Black-and-White_Fallacy' 'Causal_Oversimplification' 'Doubt'
 'Exaggeration' 'Flag-Waving' 'Labeling' 'Loaded_Language' 'Minimisation'
 'Name_Calling' 'Red_Herring' 'Reductio_ad_hitlerum' 'Repetition'
 'Slogans' 'Straw_Men' 'Thought-terminating_Cliches' 'Whataboutism']
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-45-68681190410f> in <module>()
     10 smote = SMOTE('minority')
     11 
---> 12 x_sm, y_sm = smote.fit_sample(X_train, y_train)
     13 #print(x_sm.shape, y_sm.shape)
     14 pd.DataFrame(x_sm.todense(), columns=tv.get_feature_names())

8 frames
/usr/local/lib/python3.6/dist-packages/numpy/core/_asarray.py in asarray(a, dtype, order)
     83 
     84     """
---> 85     return array(a, dtype, copy=False, order=order)
     86 
     87 

ValueError: could not convert string to float: 'left left center'
```

Answer 1

如果没有编码，就不能将X_train安装到y_train中。在功能上尝试这样的方法：

from sklearn.preprocessing import OneHotEncoder
enc = OneHotEncoder(handle_unknown='ignore')
enc.fit(X)

标签的编码。