我应该将目标类强制转换为float还是integer?

在SciKit学习中,将目标类强制转换为float或int类型(甚至字符串,请参见: Is numerical encoding necessary for the target variable in classification? )两者都是允许的。您只需注意分类目标将与输入保持相同的类型,因此如果您的输入是浮点类型,您将得到预测的浮点向量(请参见: https://scikit-learn.org/stable/tutorial/basic/tutorial.html#type-casting )

在本例中,您将直接验证 KNeighborsClassifier 将生成相同的类预测(但数据类型不同,具体取决于目标类输入类型):

from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier

data = load_iris()
(X_train, X_test, 
 y_train, y_test) = train_test_split(data.data,
                                     data.target,
                                     test_size=0.33,
                                     random_state=42)
neigh = KNeighborsClassifier(n_neighbors=3)

neigh.fit(X_train, y_train.astype(int))
int_preds = neigh.predict(X_test)

neigh.fit(X_train, y_train.astype(float))
float_preds = neigh.predict(X_test)

print(int_preds.dtype, float_preds.dtype)
print("Same classes:", (int_preds == float_preds).all())