扩展@Martin Thoma的答案

def one_hot_encode(y):
    """Convert an iterable of indices to one-hot encoded labels."""
    y = y.flatten() # Sometimes not flattened vector is passed e.g (118,1) in these cases
    # the function ends up creating a tensor e.g. (118, 2, 1). flatten removes this issue
    nb_classes = len(np.unique(y)) # get the number of unique classes
    standardised_labels = dict(zip(np.unique(y), np.arange(nb_classes))) # get the class labels as a dictionary
    # which then is standardised. E.g imagine class labels are (4,7,9) if a vector of y containing 4,7 and 9 is
    # directly passed then np.eye(nb_classes)[4] or 7,9 throws an out of index error.
    # standardised labels fixes this issue by returning a dictionary;
    # standardised_labels = {4:0, 7:1, 9:2}. The values of the dictionary are mapped to keys in y array.
    # standardised_labels also removes the error that is raised if the labels are floats. E.g. 1.0; element
    # cannot be called by an integer index e.g y[1.0] - throws an index error.
    targets = np.vectorize(standardised_labels.get)(y) # map the dictionary values to array.
    return np.eye(nb_classes)[targets]

使用Pandas进行基本的单热编码要容易得多。如果您正在寻找更多的选项，您可以使用scikit-learn。

对于Pandas的基本单热编码，您可以将数据帧传递给get_dummies函数。

例如，如果我有一个名为imdb_movies的数据帧:

.．.和我想要一个热编码的评级列，我这样做:

pd.get_dummies(imdb_movies.Rated)

这将返回一个新的数据框架，其中包含一个列，表示存在的每个评级“级别”，以及一个1或0，指定给定观察值的评级。

通常，我们希望它是原始数据框架的一部分。在本例中，我们使用“列绑定”将新的虚拟编码框架附加到原始框架上。

我们可以使用Pandas concat函数进行列绑定:

rated_dummies = pd.get_dummies(imdb_movies.Rated)
pd.concat([imdb_movies, rated_dummies], axis=1)

现在我们可以对完整的数据框架进行分析。

简单效用函数

我建议你自己做一个效用函数来快速做到这一点:

def encode_and_bind(original_dataframe, feature_to_encode):
    dummies = pd.get_dummies(original_dataframe[[feature_to_encode]])
    res = pd.concat([original_dataframe, dummies], axis=1)
    return(res)

用法:

encode_and_bind(imdb_movies, 'Rated')

结果:

另外，根据@pmalbu的评论，如果你想让函数删除原来的feature_to_encode，那么使用这个版本:

def encode_and_bind(original_dataframe, feature_to_encode):
    dummies = pd.get_dummies(original_dataframe[[feature_to_encode]])
    res = pd.concat([original_dataframe, dummies], axis=1)
    res = res.drop([feature_to_encode], axis=1)
    return(res) 

你可以在同一时间编码多个特征，如下所示:

features_to_encode = ['feature_1', 'feature_2', 'feature_3',
                      'feature_4']
for feature in features_to_encode:
    res = encode_and_bind(train_set, feature)

你也可以做以下事情。注意，对于下面的内容，您不必使用pd.concat。

import pandas as pd 
# intialise data of lists. 
data = {'Color':['Red', 'Yellow', 'Red', 'Yellow'], 'Length':[20.1, 21.1, 19.1, 18.1],
       'Group':[1,2,1,2]} 

# Create DataFrame 
df = pd.DataFrame(data) 

for _c in df.select_dtypes(include=['object']).columns:
    print(_c)
    df[_c]  = pd.Categorical(df[_c])
df_transformed = pd.get_dummies(df)
df_transformed

还可以将显式列更改为分类列。例如，这里我正在更改颜色和组

import pandas as pd 
# intialise data of lists. 
data = {'Color':['Red', 'Yellow', 'Red', 'Yellow'], 'Length':[20.1, 21.1, 19.1, 18.1],
       'Group':[1,2,1,2]} 

# Create DataFrame 
df = pd.DataFrame(data) 
columns_to_change = list(df.select_dtypes(include=['object']).columns)
columns_to_change.append('Group')
for _c in columns_to_change:
    print(_c)
    df[_c]  = pd.Categorical(df[_c])
df_transformed = pd.get_dummies(df)
df_transformed

您可以使用numpy。眼睛的功能。

import numpy as np

def one_hot_encode(x, n_classes):
    """
    One hot encode a list of sample labels. Return a one-hot encoded vector for each label.
    : x: List of sample Labels
    : return: Numpy array of one-hot encoded labels
     """
    return np.eye(n_classes)[x]

def main():
    list = [0,1,2,3,4,3,2,1,0]
    n_classes = 5
    one_hot_list = one_hot_encode(list, n_classes)
    print(one_hot_list)

if __name__ == "__main__":
    main()

结果

D:\Desktop>python test.py
[[ 1.  0.  0.  0.  0.]
 [ 0.  1.  0.  0.  0.]
 [ 0.  0.  1.  0.  0.]
 [ 0.  0.  0.  1.  0.]
 [ 0.  0.  0.  0.  1.]
 [ 0.  0.  0.  1.  0.]
 [ 0.  0.  1.  0.  0.]
 [ 0.  1.  0.  0.  0.]
 [ 1.  0.  0.  0.  0.]]

下面是使用DictVectorizer和Pandas datafframe .to_dict('records')方法的解决方案。

>>> import pandas as pd
>>> X = pd.DataFrame({'income': [100000,110000,90000,30000,14000,50000],
                      'country':['US', 'CAN', 'US', 'CAN', 'MEX', 'US'],
                      'race':['White', 'Black', 'Latino', 'White', 'White', 'Black']
                     })

>>> from sklearn.feature_extraction import DictVectorizer
>>> v = DictVectorizer()
>>> qualitative_features = ['country','race']
>>> X_qual = v.fit_transform(X[qualitative_features].to_dict('records'))
>>> v.vocabulary_
{'country=CAN': 0,
 'country=MEX': 1,
 'country=US': 2,
 'race=Black': 3,
 'race=Latino': 4,
 'race=White': 5}

>>> X_qual.toarray()
array([[ 0.,  0.,  1.,  0.,  0.,  1.],
       [ 1.,  0.,  0.,  1.,  0.,  0.],
       [ 0.,  0.,  1.,  0.,  1.,  0.],
       [ 1.,  0.,  0.,  0.,  0.,  1.],
       [ 0.,  1.,  0.,  0.,  0.,  1.],
       [ 0.,  0.,  1.,  1.,  0.,  0.]])

在这里我尝试了这个方法:

import numpy as np
#converting to one_hot





def one_hot_encoder(value, datal):

    datal[value] = 1

    return datal


def _one_hot_values(labels_data):
    encoded = [0] * len(labels_data)

    for j, i in enumerate(labels_data):
        max_value = [0] * (np.max(labels_data) + 1)

        encoded[j] = one_hot_encoder(i, max_value)

    return np.array(encoded)