Using optimize parameters(Grid) on decision tree

Hi,

I used optimize parameters(Grid) to determine the optimal parameters used in decision tree. The result give me something like maximal_depth=14, minimal leaf size=92. But when i look in the model, i found that there are some leafs containing only 1 sample. Do anyone know if i did something wrong?

Find more posts tagged with

AI Studio

Accepted answers

MartinLiebig

Hi @Ben_Suen,

did you use a validation scheme like cross-validation? This sounds a bit over-trained.

Best,

Martin

All comments

MartinLiebig

Hi @Ben_Suen,

did you use a validation scheme like cross-validation? This sounds a bit over-trained.

Best,

Martin

Telcontar120

Also did you actually optimize minimal leaf size parameter or minimal size for split? They work a bit differently.

Ben_Suen

Hi @Telcontar120
I optimized minimal leaf size parameter. i think there should be at least 92 sample in each leaf after the optimize parameter operator give me these parameter. Do i misunderstand something here?

Telcontar120

@Ben_Suen can you post your process xml for us to review? I think @mschmitz whether or not the model is overtuned (although I agree with you) is a somewhat separate issue. If the tree is resulting in notes that are smaller than the minimal leaf size parameters it sounds like it could be a bug.