R help - Aug 2010 - decision tree finetune

My decision tree grows only with one split and based on what I see in
E-Miner it should split on more variables. How can I adjust splitting
criteria in R?

Also is there way to indicate that some variables are binary, like variable
Info_G is binary so in the results would be nice to see "2) Info_G=0"
instead of "2) Info_G<0.5".

Thank you in advance!
And thanks for Eric who helped with my previous question about starting
"rpart".

Olga
> fit <- rpart(Retention ~
Info_G+AOPD+Mail+Xref_Umbr+Ins_Age+Discount+Xref_A   +     Con6      +
+ Con5   +    Con4     +
+ Con3  +  Con2  +
+ Con1 , data=Home,control=rpart.control(minsplit=5))
>
> fit
n= 48407
node), split, n, deviance, yval
      * denotes terminal node
1) root 48407 4730.642 0.8902225
  2) Info_G< 0.5 14280 1999.293 0.8316527 *
  3) Info_G>=0.5 34127 2661.865 0.9147303 *
>
	[[alternative HTML version deleted]]

I figured it out myself, here it is: control=rpart.control(cp=.001))

Thank you!


On Fri, Aug 13, 2010 at 12:58 PM, Olga Shaganova
<olga.shagan@gmail.com>wrote:
> My decision tree grows only with one split and based on what I see in
> E-Miner it should split on more variables. How can I adjust splitting
> criteria in R?
>
> Also is there way to indicate that some variables are binary, like variable
> Info_G is binary so in the results would be nice to see "2)
Info_G=0"
> instead of "2) Info_G<0.5".
>
> Thank you in advance!
> And thanks for Eric who helped with my previous question about starting
> "rpart".
>
> Olga
>
> > fit <- rpart(Retention ~
> Info_G+AOPD+Mail+Xref_Umbr+Ins_Age+Discount+Xref_A   +     Con6      +
> + Con5   +    Con4     +
> + Con3  +  Con2  +
> + Con1 , data=Home,control=rpart.control(minsplit=5))
> >
> > fit
> n= 48407
> node), split, n, deviance, yval
>       * denotes terminal node
> 1) root 48407 4730.642 0.8902225
>   2) Info_G< 0.5 14280 1999.293 0.8316527 *
>   3) Info_G>=0.5 34127 2661.865 0.9147303 *
> >
>
	[[alternative HTML version deleted]]