How to make a tree plot in caret package?

25
votes

I'm using caret package to model the data using rpart package.

library('caret')
data(iris)
formula <- as.formula(Species ~.)
t <- train(formula,iris,method = "rpart",cp=0.002,maxdepth=8)
plot(t)

As a result I get object 't' and I'm trying to plot this object to get tree plot. But the result look like that: enter image description here

Are there any way to make a tree plot from caret train object?

3
rplotrpartr-caret
If your tree plot is simple another option could be using "tree map" visualizations. Not the same as a treeplot, but may be another interesting way to visualize the model. See treemapify in ggplot – cacti5

3 Answers

48
votes

nicer looking treeplot:

library(rattle)
fancyRpartPlot(t$finalModel)

enter image description here

31
votes

The object returned from caret::train() is a list. The element finalModel contains your model.

Try this:

plot(t$finalModel)
text(t$finalModel)

enter image description here

9
votes

Had the same problem, but the answers given here wouldn't solve it, since I used a random forest instead of a tree, the following is for all coming here having the same issue:

In short: A tree can only be displayed when the method is something like:

method = "rpart"

Using a random forest 

method = "rf"

will result in the following plot: enter image description here

Extended answer already here: Plot decision tree in R (Caret)