用训练数据训练定标器

Question

我只是不知道如何训练标尺模特。

例如,

from sklearn.preprocessing import StandardScaler

object_ss = StandardScaler()

x_train_ss = obje_ss.fit_transform(x_train)
x_test_ss = obje_ss.transform(x_test)

这与下面的代码不同吗？

object_ss = StandardScaler()
object_ss.fit(x_train)

x_train_ss = obje_ss.transform(x_train)
x_test_ss = obje_ss.transform(x_test)

Answer 1

结果将是相同的，您可以验证，但尝试这两种方法和比较的结果。

from sklearn.preprocessing import StandardScaler

object_ss = StandardScaler()
x_train_ss = object_ss.fit_transform(x_train)
x_test_ss = object_ss.transform(x_test)

object_ss2 = StandardScaler()
object_ss2.fit(x_train)
x_train_ss2 = object_ss2.transform(x_train)
x_test_ss2 = object_ss2.transform(x_test)

print(np.abs(x_train_ss - x_train_ss2).mean())
print(np.abs(x_test_ss - x_test_ss2).mean())

打印的数字应该非常小(或者我甚至希望0被打印出来)。