[QUESTION]What is the difference between predict_proba, decision_function and predict function and how to choose

[danny95928]

in chapter 3, you use the predict_proba to get list of probabilities for each instance of each class and then to threshold the predicted decision score you have used the decision_function() rather than predict() i'm suck at programming would you please tell me what is the differences between these and how to choose the right one

[messierandromeda]

predict_proba gives the probabilities of the image belonging to each class, in comparison to predict, which predicts a single digit. The decision_function relates to the Support Vector Classification (SVC), and it shows us the side of the hyperplane (generated by the classifier) that a particular point is on, and also the distance to it.
An explanation of decision_function vs. predict: https://stackoverflow.com/questions/20113206/scikit-learn-svc-decision-function-and-predict