I am using sklearn.ensemble.RandomForestClassifier to do classification. I have 14 classes (14 labels) in total. Now my code is like
clf = RandomForestClassifier(n_estimators = 50)
clf.fit(train_data, labels)
importances = clf.feature_importances_
indices = np.argsort(importances)[-20:]
by doing this I can get the indices for top 20 important features. However, since I have 14 classes (0,1,2,3,4,5,6,7,8,9,10,11,12,13) and now I want to know for each class, what its most important features are, then what should I do?