Reputation: 123
I found a way to plot a decision tree solution from rpart with the sunburstR-package. To plotting a sunburst it is necessary to have data.frame which represents a sequence. I modified the decision tree result to a sequence like below
Result of the decision tree
rpart(Species~.,data=iris)
n= 150
node), split, n, loss, yval, (yprob)
* denotes terminal node
1) root 150 100 setosa (0.33333333 0.33333333 0.33333333)
2) Petal.Length< 2.45 50 0 setosa (1.00000000 0.00000000 0.00000000) *
3) Petal.Length>=2.45 100 50 versicolor (0.00000000 0.50000000 0.50000000)
6) Petal.Width< 1.75 54 5 versicolor (0.00000000 0.90740741 0.09259259) *
7) Petal.Width>=1.75 46 1 virginica (0.00000000 0.02173913 0.97826087) *
Sequence for the sunburst:
sequences_1<-1
sequences_1<-data.frame(sequences_1)
sequences_1[1:3,]<-1
sequences_1$V1[1]<-"Petal.Length<_2.45-setosa"
sequences_1$V1[2]<-"Petal.Length>=2.45-Petal.Width<_1.75_54_5-versicolor"
sequences_1$V1[3]<-"Petal.Length>=2.45-Petal.Width>=1.75_46_1-virginica"
sequences_1$V2[1]<-50
sequences_1$V2[2]<-54
sequences_1$V2[3]<-46
sequences_1$sequences_1<-NULL
Plotting Sunburst:
library(sunburstR)
sunburst(sequences_1,count=TRUE)
The sequence for the sunburst plot, I did manually. Do someone know how to build the sequence automatically like above from the result of the rpart decision tree?
Upvotes: 3
Views: 259
Reputation: 6579
d3r provides a function d3_party
to convert rpart/partykit
into a d3 hierarchy. sunburst
can use the result of d3_party
with one minor modification to change "rule"
to "name"
. This is not ideal but in most cases will work flawlessly.
library(rpart)
library(d3r)
# d3_party requires partykit
# install.packages("partykit")
library(sunburstR)
rp <- rpart(Species~.,data=iris)
rp_d3 <- d3_party(rp)
# one trick/hack required since sunburst expects
# name but d3_party gives rule
# this is ugly but let's replace all "rule" with "name"
# with gsub
rp_d3 <- gsub(
x = rp_d3,
pattern = '"rule":',
replacement = '"name":'
)
sunburst(
rp_d3,
valueField = "n",
sumNodes = FALSE,
count = TRUE,
legend = FALSE
)
Upvotes: 2