Reduced error pruning

I am using the Decision Tree Learner node and trying different pruning options. The MDL option works fine for pruning. However, “reduced error pruning” seems to have no effect. This is similar to a previous post that also found no solution (Reduced error pruning). I would appreciate any help to understand why reduced error pruning has no impact on the prediction performance. Thanks.

Hello @saddas ,
To see if REP works, I tested it on this workflow, and I could see different values when I used MDL and REP.

Could you please let us know more about the problem you are trying to solve using Decision Tree?

Thanks,
Sanket

1 Like

Hi @sanket_2012 ,
Thanks for your reply and for sharing the workflow. I ran your workflow with 4 options:

  1. MDL = no, REP = no
  2. MDL = no, REP = yes
  3. MDL = yes , REP = no
  4. MDL = yes , REP = yes

There were indeed minor differences in the resulting confusion matrix among the different options, but options 3 and 4 returned the same result (REP didn’t have an effect when MDL is performed) - see enclosed screenshot.