I am having a dataset which has responses to various questions recorded.Some of the records have no response/missing values recorded as 0.For example a sample ques has the distribution.
table(vodka$Q13) 0 1 2 3 4 96 750 232 69 17
0 is quite high here(96 out of 1164) records.So I was thinking of using decision trees to predict the value for records which have this value as 0.This is my code:
loantime.train <- vodka[which(vodka$Q13 != 0),] loantime.test <- vodka[which(vodka$Q13 == 0),] library(rpart) loantime.rpart <- rpart(Q13 ~ .,data = loantime.train,method = "class") printcp(loantime.rpart)
Can I say that the accuracy is 1-0.29775 or is there some other parameter which can be used for that.
How are such cases really dealt with ?
Can someone please help me with this??