Is there any keyword in scikit-learn for AdaBoostClassifier to return weights of the weak classifiers at each step? IDK if the features are chosen randomly (For the weak Decision Trees that this classifiers provide for further learning processes) if it is so, how can we observe the chosen features?

1 Answers

0
AI_Learning On Best Solutions

clf.estimator_weights_ would give the weights for each estimator in the ensemble.

Documentation

There is no sampling of random features in Adaboost classifier. There would be only sample weights based on the ensemble model error.

To know more about Adaboost technique, read here.

May be you can look at the feature importance of individual decision trees. You can start by tweaking this function.