rpart not creating Decision Tree in R, SVM works

Question

I'm trying to create a decision tree for classification but it doesn't get created. The same data performs with 0.85 accuracy using a SVM (train == test data), "play" is the target...

Any idea what I'm doing wrong? Here is the data and code: https://gist.github.com/romeokienzler/c471819cbf156a69f73daf49f8c700c6

outlook,temp,humidity,windy,play

sunny,hot,high,false,no

sunny,hot,high,true,no

overcast,hot,high,false,yes

rainy,mild,high,false,yes

rainy,cool,normal,false,yes

rainy,cool,normal,true,no

overcast,cool,normal,true,yes

sunny,mild,high,false,no

sunny,cool,normal,false,yes

rainy,mild,normal,false,yes

sunny,mild,normal,true,yes

overcast,mild,high,true,yes

overcast,hot,normal,false,yes

rainy,mild,high,true,no

For using the SVM I've encoded the data: https://gist.github.com/romeokienzler/9bfce4182eda3d7662315621462c9cc6

outlook,temp,humidity,windy,play

1,1,2,FALSE,FALSE

1,1,2,TRUE,FALSE

2,1,2,FALSE,TRUE

3,2,2,FALSE,TRUE

3,3,1,FALSE,TRUE

3,3,1,TRUE,FALSE

2,3,1,TRUE,TRUE

1,2,2,FALSE,FALSE

1,3,1,FALSE,TRUE

3,2,1,FALSE,TRUE

1,2,1,TRUE,TRUE

2,2,2,TRUE,TRUE

2,1,1,FALSE,TRUE

3,2,2,TRUE,FALSE

This is the SVM case:

library(e1071)
df= read.csv("5.tennis_encoded.csv")
attach(df)
x <- subset(df, select=-play)
y <- play
detach(df)
model = svm(x,y,type = "C")
pred = predict(model,x)
truthVector = pred == y
good = length(truthVector[truthVector==TRUE])
bad = length(truthVector[truthVector==FALSE])
good/(good+bad)

[1] 0.8571429

And this one for the Decision Tree

df= read.csv("5.tennis_encoded.csv")
library(rpart)
model = rpart(play ~ .,method = "class", data=df)
print(model)

1) root 14 5 TRUE (0.3571429 0.6428571) *

So I basically get a tree with only a root and 0.64% probability for play == yes

Any ideas what I'm doing wrong?

comendeiro comendeiro · Accepted Answer · 2017-04-12T11:25:39

Most likely you are passing to the algorithm too few data for splitting.

Check the rpart.control function for more details

rpart.control(minsplit = 20, minbucket = round(minsplit/3), cp = 0.01, 
              maxcompete = 4, maxsurrogate = 5, usesurrogate = 2, xval = 10,
              surrogatestyle = 0, maxdepth = 30, ...)

As you see, the minimum split size is 20.

If you

model = rpart(play ~ .,method = "class", data=df, control= rpart.control(minsplit=2))

You should get more splits

rpart not creating Decision Tree in R, SVM works

1 Answers