如何使用 skLearn 构建基线 model 以预测具有多个值的 Y

Question

I have a sample data frame that looks like below. I would like to build a baseline model to predict y_combined using X=df.filter(regex='x_') :

df = pd.DataFrame({
    'x_1':[0.1,0.2,0.1,0],
    'x_2':[0.5,0.1,0.3,0.4],
    'x_3':[0.2,0.1,0.6,0.1],
    'x_4':[0,0.5,0.2,0.3],
    'y_1': [0, 1, 1, 0],
    'y_2': [0, 0, 1, 0],
    'y_3': [0, 1, 0, 1],
    'y_combined': [np.array([0, 0, 0]), np.array([1, 0, 1]),
                   np.array([1, 1, 0]), np.array([0, 0, 1])]
})

I am new to the baseline model building. To obtain y_predicted , how should I specify the DummyClassifer() model with strategy="constant" ? Or is there a different strategy I should be using?

For example, if the y_predicted = [1,1,1] , then I will see how well the prediction model performs by getting the average centroid between y_combined and y_predicted .

Answer 1

I will answer your question under the premise that you want to use the DummyClassifier using the strategy='constant' setting to build a baseline model for a multilabel classification problem, where the output equals y_combined of df . In this case, the following code will work:

from sklearn.dummy import DummyClassifier
import numpy as np
import pandas as pd


X = pd.DataFrame({
    'x_1': [0.1,0.2,0.1,0],
    'x_2': [0.5,0.1,0.3,0.4],
    'x_3': [0.2,0.1,0.6,0.1],
    'x_4': [0,0.5,0.2,0.3]
})
y = np.array([[0, 0, 0], [1, 0, 1], [1, 1, 0], [0, 0, 1]])

clf = DummyClassifier(strategy='constant', constant=np.array([1, 1, 1]))
clf.fit(X, y)

Notice that when you use strategy='constant' you also have to state the constant value that should be predicted via the constant=... parameter of the DummyClassifier . You will see that the baseline model will now always predict the specified constant value, no matter the input:

y_pred = clf.predict(X)
print(y_pred)

# output
[[1 1 1]
 [1 1 1]
 [1 1 1]
 [1 1 1]]

Since you also asked about other strategies, you can basically choose any of the strategies mentioned in the documentation of DummyClassifier . All have a common behaviour, as mentioned in the user guide:

Note that with all these strategies, the predict method completely ignores the input data!