CODEWITHSUNDEEP
rtreerpart
Paweł Orliński
Paweł Orliński

Reputation: 239

R classification tree with Rpart

I have some data which I would like to segment. My first thought was classification tree in R from Rpart package. My training data consists of many explanatory variables and one 0-1 response variable named "sold". The response value "1" appears in about 80% of rows. When I try to build a tree with rpart(sold~., training_data, method = "class"), R is unable to create a tree. I suppose that the reason is that it can't find any segments which differ very much from one another. After quick inspection of the data, I expect that my tree should look like that left node will have 85% of sold and right node will have 75% of sold.

Is there any way to create a classification tree on such data set?

Upvotes: 2

Views: 1858

Answers (1)

hyunwoo jeong
hyunwoo jeong

Reputation: 1624

I had same problem. It seems to be a problem about 'cp'. Refer to my code:

tr1<-rpart(bad~group+amount, data=ra, 
control=rpart.control(minsplit=5, cp=0.001),method='class')

When I run this, it worked. when I increase cp level (e.g. 0.005), it didn't work.

Upvotes: 1

Related Questions