Skip to content

extract decision tree from rpart

1 message · Tim

Tim
# 
Hi,
I am using rpart for my decision stump. I am trying to extract the learned
stump from the output of rpart.
For example:
    cntrl <- rpart.control(maxdepth = 1, minsplit = learn-1, 
                           maxsurrogate = 0, usesurrogate=0, maxcompete = 1,

                           cp = 0, xval = 0)
    fit  <- rpart(y~bx, weights = w/mean(w), control = cntrl)

After some while of searching rpart document, playing the code and asking
questions, I came to this tentative conclusion that:
if my stump is for classifiction, fit[[1]]$var[[1]] is the selected feature
ID, and fit$split[1,4] is the splitting value
if my stump is for regression, fit[[1]]$var[[1]] is the selected feature ID,
fit$split[1,4] is the splitting value, fit[[1]]$yval[[2]] and 
fit[[1]]$yval[[3]] are the two constant values for the two leaves.

However I get stuck again. For example, for my classification stump, here is
a value of fit, where feature No. 59 is selected and split at 0.0065
n= 236 
node), split, n, deviance, yval
      * denotes terminal node
1) root 236 236.0000 -2.831422e-17  
  2) bx.59< 0.0065 156   0.0000 -1.000000e+00 *
  3) bx.59>=0.0065 80 116.7475  5.053073e-01 *
[1] bx.59
60 Levels: <leaf> bx.1 bx.2 bx.3 bx.4 bx.5 bx.6 bx.7 bx.8 bx.9 bx.10 ...
bx.59

if I give fit[[1]]$var[[1]] to a variable rr, then it is wrong
[1] 60

Can someone tell me how to correctly get the decision tree info from the
output of rpart?
Thanks and regards!