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.]]

这对我来说很管用:

pandas.factorize( ['B', 'C', 'D', 'B'] )[0]

输出:

[0, 1, 2, 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.]])

首先，最简单的热编码方法:使用Sklearn。

http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.OneHotEncoder.html

其次，我不认为使用熊猫进行一个热编码是那么简单(虽然未经证实)

在pandas中为python创建虚拟变量

最后，你需要一个热编码吗?一个热编码以指数方式增加了特征的数量，大大增加了任何分类器或任何你要运行的东西的运行时间。特别是当每个分类特征都有很多层次时。相反，你可以进行虚拟编码。

使用虚拟编码通常工作得很好，运行时间和复杂性要少得多。一位睿智的教授曾经告诉我，“少即是多”。

如果你愿意，这是我的自定义编码函数的代码。

from sklearn.preprocessing import LabelEncoder

#Auto encodes any dataframe column of type category or object.
def dummyEncode(df):
        columnsToEncode = list(df.select_dtypes(include=['category','object']))
        le = LabelEncoder()
        for feature in columnsToEncode:
            try:
                df[feature] = le.fit_transform(df[feature])
            except:
                print('Error encoding '+feature)
        return df

编辑:比较更清楚:

一热编码:将n层转换为n-1列。

Index  Animal         Index  cat  mouse
  1     dog             1     0     0
  2     cat       -->   2     1     0
  3    mouse            3     0     1

你可以看到，如果你的分类特征中有许多不同类型(或级别)，这会使你的记忆爆发式增长。记住，这只是一列。

伪代码:

Index  Animal         Index  Animal
  1     dog             1      0   
  2     cat       -->   2      1 
  3    mouse            3      2

转换为数字表示。极大地节省了特征空间，代价是准确性。

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]

为了补充其他问题，让我提供如何使用Numpy使用Python 2.0函数:

def one_hot(y_):
    # Function to encode output labels from number indexes 
    # e.g.: [[5], [0], [3]] --> [[0, 0, 0, 0, 0, 1], [1, 0, 0, 0, 0, 0], [0, 0, 0, 1, 0, 0]]

    y_ = y_.reshape(len(y_))
    n_values = np.max(y_) + 1
    return np.eye(n_values)[np.array(y_, dtype=np.int32)]  # Returns FLOATS

行n_values = np.max(y_) + 1可以硬编码，以便在使用小批量的情况下使用足够数量的神经元。

使用此函数的演示项目/教程: https://github.com/guillaume-chevalier/LSTM-Human-Activity-Recognition