[Python] sklearn's model.predict_proba() does not seem to make the same prediction as...

From my research (see e.g. here), model.predict_proba(X) should return class probabilities under the same order as the classes listed with model.classes. However, when I test this by comparing the outputs of model.predict(X) and model.predict_proba(X), I become confused.

Here is an example:

from sklearn.tree import DecisionTreeClassifier
model = sklearn.tree.DecisionTreeClassifier(criterion='gini', max_depth=10)
model.fit(X_train, y_train)
print('Predicted class:', model.predict(X_test)) # X_test is single vector
print('Predicted probabilities:', model.predict_proba(X_test))
print('Corresponding classes:', model.classes_)


The output is:

Predicted class: ['Class B']
Predicted probabilities: [0.33514444 0.28186712 0.38298844]
Corresponding classes: ['Class B' 'Class C' 'Class A']


Clearly, the predicted class should be Class A, since Class A was assigned the highest probability. However, the predicted class is Class B. Am I missing something? If I am wrong that the ordering of predicted probabilities should be the same as the ordering of model.classes_, how can I find the classes that correspond to each probabilities output by predict_proba?

Continue reading...