fitensemble adaboost m1 stump picking
4 views (last 30 days)
Show older comments
I am running fitensemble with adaboostm1 When I view the first tree both right and left decisions are class 1. I do have two classes in my data. I think this has no effect. Why would the algorithm pick such a stump or tree? Decision tree for classification 1 if x3<70.5 then node 2 elseif x3>=70.5 then node 3 else 1 2 class = 1 3 class = 1
0 Comments
Answers (1)
Ilya
on 13 Sep 2012
A decision tree in a boosting ensemble by default minimizes the Gini diversity index, not classification error. Two child nodes originating from the same parent can be dominated by the same class. This situation is not uncommon.
1 Comment
Reda
on 21 May 2014
Hello, I noticed the same problem as sedar sedar. When adaboosting a classification tree, the learners are all slumps. This is acounter-intuitive, specially that fitting a classification tree with the same parameters gives a much deeper tree. So why is the first learner not as deep ?
See Also
Categories
Find more on Classification in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!