3 ms·
>> CART showed slightly better results for this task but its main advantage is that it gives you a decision tree that is easy to understand, explain and impleme
by YeGoblynQueenne 10y ago
>> CART showed slightly better results for this task but its main advantage is that it gives you a decision tree that is easy to understand, explain and implement vs SVM or Random Forests that work like a black box and require using heavy ML libraries in production.
I don't get that. Random Forests are ensembles of decision trees and can be used for regression. I've not used the R CART library, so I might be missing something, but the interpretability of Random Forests should be the same as that of, well, any other model built by a decision-tree learner (i.e. just as unreadable as any other classifier's model, once their parameters become sufficiently many).