I am working with Random Forests in Weka. I thought the ID3 algorithm is used to find the best split attribute at each level. But after reading a bit I noticed that ID3 can not handle numerical attributes (one needs for instance C4.5 for that). Random Forests in Weka do have no problems with numerical features as input... How are they handeled?

(Small additional question: Is Weka still something you can work with today or do I need to switch the tool for current research? Weka seems a bit "old-fashioned" sometimes...)


1 Answer 1


Weka does use its own implementation of C4.5, named J48.

That said, its RandomForest calls on RandomTree, which appears to be implemented independently, not ever touching the J48 code. (That's from what I can tell, and I don't normally work in Java so take it with a grain of salt.) I can't tell from a glance whether there are substantial differences aside from the obvious (that RandomTree subsets features available to split on at each node).

Class hierarchy of module tree
tree folder of source
(click through to see J48, RandomForest, RandomTree, etc.)

  • $\begingroup$ As to the relevance of Weka, I can't say with any authority. FWIW, a quick look for job postings does find hits including it, but always in a list as "experience with toolkits such as ...". $\endgroup$
    – Ben Reiniger
    Commented Apr 28, 2020 at 12:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.