just my2cent:
could you make a derived feature by including the desired predictor in
each of features you use to present to rpart?
(by the way, if I can understand your question right, I assume rpart
picks only one predictor each time at each node)
Best,
Weiwei
On 11/26/07, Wensui Liu <[EMAIL
If I understand correctly, rpart() will pick predictor at each node
automatically. I am wondering if there is a way to force rpart()
including a specific predictor. The reason I am asking is that I'd
like to use rpart() to detect interaction terms for some variables.
Thanks.
_
2 matches
Mail list logo
