0

iam certainly working on decision algorithm trees and in my research this article came up http://www.slideshare.net/aorriols/lecture6-c45.

If you go to slide 12 top-right corner there is an algorithm which i understand completely. The thing i dont understand is (and it doesnt say it anywhere) when you prune a subtree which class takes the position.

In slide 12 classification bad takes the place of the subtree. Now, "bad" is the most common class of the parent or not????

I hope you enlight me .Thanks

Nick
  • 2,818
  • 5
  • 42
  • 60
  • I would say it's the most likely one given that it already passed the path up to the parent (based on some probability distribution). – Nico Schertler Dec 08 '15 at 15:29
  • meaning what exactly? – Nick Dec 08 '15 at 15:58
  • You have an incoming probability distribution on the node you want to prune. Based on this distribution, calculate the probabilities for each class and take the class with the highest probability. A very simple approximation for that is counting occurrences of the classes. – Nico Schertler Dec 08 '15 at 16:00
  • So the subtree is replaced by the most common class of that subtree. Am i right? – Nick Dec 08 '15 at 16:04
  • At least, that's what I would do. Not sure if it's what your lecturer had in mind. – Nico Schertler Dec 08 '15 at 16:20

0 Answers0