Question

我已经为多类分类执行了标签二进制化，它运行正常：

y_test
1
3
4
2
0

from sklearn.preprocessing import label_binarize

y_test_binarize = label_binarize(y_test, classes=[0, 1, 2, 3, 4])
y_test_binarize

0   1   0   0   0
0   0   0   1   0
0   0   0   0   1
0   0   1   0   0
1   0   0   0   0

接下来，我想做一个反向过程，从y_test变量中获取y_test_binarize。

有预定义的方法吗？

Answer 1

您还可以使用LabelBinarizer，它将label_binarize函数包装在一个类中，并提供转换为二进制数据的方法，并将它们反转换为原始类。

y_test = [1, 3, 4, 2, 0]

from sklearn import preprocessing
lb = preprocessing.LabelBinarizer()

y_test_binarize = lb.fit_transform(y_test)
#Output: y_test_binarize
array([[0   1   0   0   0],
       [0   0   0   1   0],
       [0   0   0   0   1],
       [0   0   1   0   0],
       [1   0   0   0   0]])

y_test_original = lb.inverse_transform(y_test_binarize)
#Output: y_test_original
array([1, 3, 4, 2, 0])

希望这会有所帮助。随意询问是否有任何问题。

Answer 2

一种简单的方法是计算二值化数据和类的矩阵乘积：

@IBAction func signUp(_ sender: AnyObject) {

//I first check to see if the users left any of the fields blank

if firstName.text == "" || lastName.text == "" || email.text == "" || userName.text == "" || password.text == "" {

    createAlert(title: "Error in form", message: "Please fill in all text fields")


//If everything is filled in
}else{


    let user = PFUser()
    user.username = userName.text
    user["firstname"] = firstName.text
    user["lastname"] = lastName.text
    user.email = email.text
    user.password = password.text


    user.signUpInBackground(block: { (success, error) in                
        if error != nil {

            var displayErrorMessage = "Please try again later."

            if let errorMessage = (error! as NSError).userInfo["error"] as? String {

                displayErrorMessage = errorMessage
            }

            self.createAlert(title: "Signup error", message: displayErrorMessage)

        }else{
            print("User signed up")
        }
    })
}

}

Answer 3

使用numpy的argmax函数：

np.argmax(y_test_binarized, axis=1)

Python中标签二值化的逆过程

3 个答案: