I am having issues trying to understand what the size of a terminal node in a decision tree means? Could anyone give me an easy explanation?

I know a terminal node is a leaf node, one that has a label/class, however I can't get around my head why it has a size.


In this context, "size" refers to the number of training instances in the terminal node. That is, decision trees are built out until terminal nodes either have size less than or equal to the terminal node size, or are pure, whichever comes first.

  • $\begingroup$ Hi and thanks for the answer. I am starting to get an idea of what you mean. If it isnt too much to ask, could you give me an example? Thanks $\endgroup$ – user3276768 Jul 27 '15 at 19:36
  • $\begingroup$ lets say you have a decision tree, and that you start with 100 observatinons / cases. As you apply the decision rules, at each split a number of cases will move "right" and a number of cases will move "left. For example - the split is 80/20 (e.g. temp>80F) on the first criterium and 50/50 (e.g. colour=red) on the second one Then you will end up with 40 cases in the (temp>80F, colour=red) terminal node. $\endgroup$ – Wouter Jul 27 '15 at 21:17
  • $\begingroup$ @Wouter Perfect that was a great explanation. Thanks a lot! $\endgroup$ – user3276768 Jul 28 '15 at 10:23
  • $\begingroup$ @user777 in your answer above, you mean decision trees are built out[...], not Random Forests... $\endgroup$ – Antoine Sep 20 '15 at 19:14
  • $\begingroup$ @Antoine if one solely looks to the text, that appears to be the case. But it's not clear if the distinction is important, since cart and random-forest are both tags to the question, but decision tree is not. $\endgroup$ – Sycorax Sep 20 '15 at 19:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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