使用不同分类器的 sklearn precision_recall_curve 函数

Question

This may be an easy question, but I need help understanding how to use the precision_recall_curve function in sklearn .

I have a binary dataset and am using three classifiers ( SVM , RF , LR ) to classify it.

The example in sklearn's documentation shows to use the function like this:

y_score = classifier.decision_function(X_test)    
precision_recall_curve(y_test, y_score)

In the example, decision_function is a built-in function for SVM classifiers. However, I don't see a function like that for Random Forest classifiers or Linear Regression.

Can someone help me understand what the y_score and decision function really is, and how I can calculate it for any classifier?

Thanks!

Answer 1

Look at the second param description in documentation of precision_recall_curve :

probas_pred : array, shape = [n_samples]

Estimated probabilities or decision function.

When decision_function() is not present, you may use predict_proba() in its place.

Answer 2

对于所有其他没有内置decision_function ，您应该使用predict_proba函数，它基本上做同样的事情。

y_score = random_forest.predict_proba()