R Weka J48决策树无法处理数字类

Question

I found this document on the web: https://www.erpublication.org/admin/vol_issue1/upload%20Image/IJETR032129.pdf

There it uses on page 4 to build a decision tree with RWeka package and J48 function in R. In his example, he has both numerical and categorical values.

So, I made a test, with just on column trying to predict the other. Here is a sample:

VALUE CHURNED_F
    2         1
    2         0
    2         0
    2         0
    2         0
    1         0

This is my code:

m2 <- J48(`CHURNED_F` ~ ., data = head(train[, -c(1)]))

But I get this error:

Error in .jcall(o, "Ljava/lang/Class;", "getClass") : 
  weka.core.UnsupportedAttributeTypeException: weka.classifiers.trees.j48.C45PruneableClassifierTree: Cannot handle numeric class!

I don't understand the error, first of all it is a categorical class. Second, in the example in the document it perfectly uses both categorical and numerical columns.

How can I get this to work?

Answer 1

J48 requires the class be categorical, or in the case of R, a factor. I believe that your "Churned_F" variable is numeric. You can check what type your variables are by using the structure function:

str(train)  

The code below allows you to build a J48 tree. Here I ensure "CHURNED_F" is a factor.

library(RWeka)
train <- data.frame(VALUE = c(2,2,2,2,2,1), CHURNED_F = factor(c(1,0,0,0,0,0)))
m2 <- J48(CHURNED_F ~., data = train)

Answer 2

It means, that your answer column must be represented by a character instead of a numeric value. You can change it with this method:

Wine$X1=factor(Wine$X1,levels = c(1,2,3),labels = c("Uno","Dos","Tres"))

Wine is my data set. X1 is my answer column. 1, 2 and 3 are the answers. Uno, Dos and Tres are the wanted answer after the parsing of the numeric values.