假设在10个变量中，在数据帧中有3个分类变量，分别为cname1、cname2和cname3。 然后下面的代码将自动在新的数据框架中创建一个热编码变量。

import category_encoders as ce
encoder_var=ce.OneHotEncoder(cols=['cname1','cname2','cname3'],handle_unknown='return_nan',return_df=True,use_cat_names=True)
new_df = encoder_var.fit_transform(old_df)

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

简短的回答

这里有一个函数，可以在不使用numpy、pandas或其他包的情况下进行一次性编码。它接受一个整数、布尔值或字符串(也可能是其他类型)的列表。

import typing


def one_hot_encode(items: list) -> typing.List[list]:
    results = []
    # find the unique items (we want to unique items b/c duplicate items will have the same encoding)
    unique_items = list(set(items))
    # sort the unique items
    sorted_items = sorted(unique_items)
    # find how long the list of each item should be
    max_index = len(unique_items)

    for item in items:
        # create a list of zeros the appropriate length
        one_hot_encoded_result = [0 for i in range(0, max_index)]
        # find the index of the item
        one_hot_index = sorted_items.index(item)
        # change the zero at the index from the previous line to a one
        one_hot_encoded_result[one_hot_index] = 1
        # add the result
        results.append(one_hot_encoded_result)

    return results

例子:

one_hot_encode([2, 1, 1, 2, 5, 3])

# [[0, 1, 0, 0],
#  [1, 0, 0, 0],
#  [1, 0, 0, 0],
#  [0, 1, 0, 0],
#  [0, 0, 0, 1],
#  [0, 0, 1, 0]]

one_hot_encode([True, False, True])

# [[0, 1], [1, 0], [0, 1]]

one_hot_encode(['a', 'b', 'c', 'a', 'e'])

# [[1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 1, 0], [1, 0, 0, 0], [0, 0, 0, 1]]

长(er)回答

I know there are already a lot of answers to this question, but I noticed two things. First, most of the answers use packages like numpy and/or pandas. And this is a good thing. If you are writing production code, you should probably be using robust, fast algorithms like those provided in the numpy/pandas packages. But, for the sake of education, I think someone should provide an answer which has a transparent algorithm and not just an implementation of someone else's algorithm. Second, I noticed that many of the answers do not provide a robust implementation of one-hot encoding because they do not meet one of the requirements below. Below are some of the requirements (as I see them) for a useful, accurate, and robust one-hot encoding function:

单热编码函数必须:

处理各种类型的列表(例如，整数，字符串，浮点数等)作为输入 处理带有重复项的输入列表 返回与输入相对应(与输入顺序相同)的列表的列表 返回一个列表的列表，其中每个列表都尽可能短

我测试了这个问题的许多答案，其中大多数都未能满足上述要求之一。

熊猫的热编码非常简单:

def one_hot(df, cols):
    """
    @param df pandas DataFrame
    @param cols a list of columns to encode 
    @return a DataFrame with one-hot encoding
    """
    for each in cols:
        dummies = pd.get_dummies(df[each], prefix=each, drop_first=False)
        df = pd.concat([df, dummies], axis=1)
    return df

编辑:

使用sklearn的LabelBinarizer实现one_hot的另一种方法:

from sklearn.preprocessing import LabelBinarizer 
label_binarizer = LabelBinarizer()
label_binarizer.fit(all_your_labels_list) # need to be global or remembered to use it later

def one_hot_encode(x):
    """
    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 label_binarizer.transform(x)

我在我的声学模型中使用了这个: 也许这对你的模型有帮助。

def one_hot_encoding(x, n_out):
    x = x.astype(int)  
    shape = x.shape
    x = x.flatten()
    N = len(x)
    x_categ = np.zeros((N,n_out))
    x_categ[np.arange(N), x] = 1
    return x_categ.reshape((shape)+(n_out,))

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