Scikit-learn 扩展的自定义 one-hot 编码矩阵 - 不是从数据集构建的

Question

我正在尝试构建一个 one-hot 编码矩阵，该矩阵表示我的样本中未找到的其他类别。

如果使用以下代码：

s = np.array(['man', 'man', 'woman', 'woman', 'son', 'son', 'son', 'son', 'son'])
label_encoder = LabelEncoder()
integer_encoded = label_encoder.fit_transform(s)
onehot_encoder = OneHotEncoder(sparse=False)
integer_encoded = integer_encoded.reshape(len(integer_encoded), 1)
Y = onehot_encoder.fit_transform(integer_encoded)
print(Y)

结果是这样的：

[[1. 0. 0.]
 [1. 0. 0.]
 [0. 0. 1.]
 [0. 0. 1.]
 [0. 1. 0.]
 [0. 1. 0.]
 [0. 1. 0.]
 [0. 1. 0.]
 [0. 1. 0.]]

但实际上我有以下类别，其中一些在我的数据集中不存在，但我需要考虑它们：

categories = np.array(['man', 'woman', 'son', 'daughter', 'boy', 'girl', 'king', 'queen', 'baby', 'child'])

所以我需要的是这样的：

[[1. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
 [1. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
 [0. 0. 1. 0. 0. 0. 0. 0. 0. 0.]
 [0. 0. 1. 0. 0. 0. 0. 0. 0. 0.]
 [0. 1. 0. 0. 0. 0. 0. 0. 0. 0.]
 [0. 1. 0. 0. 0. 0. 0. 0. 0. 0.]
 [0. 1. 0. 0. 0. 0. 0. 0. 0. 0.]
 [0. 1. 0. 0. 0. 0. 0. 0. 0. 0.]
 [0. 1. 0. 0. 0. 0. 0. 0. 0. 0.]]

因此我想弄清楚如何在这段代码中实现 OneHotEncoder(sparse=False, categories=categories)：

categories = np.array(['man', 'woman', 'son', 'daughter', 'boy', 'girl', 'king', 'queen', 'baby', 'child'])
s = np.array(['man', 'man', 'woman', 'woman', 'son', 'son', 'son', 'son', 'son'])
label_encoder = LabelEncoder()
integer_encoded = label_encoder.fit_transform(s)
onehot_encoder = OneHotEncoder(sparse=False, categories=categories)
integer_encoded = integer_encoded.reshape(len(integer_encoded), 1)
Y = onehot_encoder.fit_transform(integer_encoded)
print(Y)

但它给出了以下错误：

ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

如果我改变：

integer_encoded = integer_encoded.reshape(len(integer_encoded), 1)

to

integer_encoded = integer_encoded.reshape(len(integer_encoded), 1).all()

我收到以下错误：

Reshape your data either using array.reshape(-1, 1) if your data has a single feature or array.reshape(1, -1) if it contains a single sample.

谁能帮我解决这个问题？

Answer 1

这里的问题是OneHotEncoder 的categories 参数
您的categories 变量是ndarray，它正在提高ValueError。
尝试使用常规排序 list。
而且，您不需要在您的情况下使用LabelEncoder。

from sklearn.preprocessing import LabelEncoder, OneHotEncoder
import numpy as np

categories = [sorted(['man', 'woman', 'son',
                      'daughter', 'boy', 'girl',
                      'king', 'queen', 'baby', 'child'])]
print(f'sorted categories: {categories}')

s = np.array(['man', 'man', 'woman', 'woman',
              'son', 'son', 'son', 'son', 'son']).reshape(-1, 1)
onehot_encoder = OneHotEncoder(sparse=False, categories=categories)
Y = onehot_encoder.fit_transform(s)
print(Y)

sorted categories: [['baby', 'boy', 'child', 'daughter', 'girl', 'king', 'man', 'queen', 'son', 'woman']]
[[0. 0. 0. 0. 0. 0. 1. 0. 0. 0.]
 [0. 0. 0. 0. 0. 0. 1. 0. 0. 0.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 1.]
 [0. 0. 0. 0. 0. 0. 0. 0. 0. 1.]
 [0. 0. 0. 0. 0. 0. 0. 0. 1. 0.]
 [0. 0. 0. 0. 0. 0. 0. 0. 1. 0.]
 [0. 0. 0. 0. 0. 0. 0. 0. 1. 0.]
 [0. 0. 0. 0. 0. 0. 0. 0. 1. 0.]
 [0. 0. 0. 0. 0. 0. 0. 0. 1. 0.]]