Search code examples
rmachine-learningsvmwekaj48

Probabilities Classification - R

First of all, sorry about my english, I am brazilian and I am improving it yet.

I am using RWeka to create classification models (NaiveBayes, JRip, J48, SVM) and I need the probabilities of each instance.

For SVM (package "e1041") I simply do

model<-svm(classification~.,data=treino1, probability=TRUE)
pred <- predict(model, teste1[,-ncol(teste1)], probability = TRUE)
prob<-attr(pred, "probabilities")

And, then, the output is something like:

           0           1
1    9.126349e-01 0.087365132
2    9.085960e-01 0.091404044
3    9.414368e-01 0.058563209

What should I do to obtain the same type of output with J48 (package "RWeka")?

model<-J48(classification~.,data=treino1, probability=TRUE)

Error in J48(classification ~ ., data = treino1, probability = TRUE) : unused argument (probability = TRUE)

Thanks!

Solution

  • You should specify type = 'probability. Here's an example using the iris dataset:

    require(RWeka)
fit = J48(Species~., data=iris)
pred = predict(fit, iris, type = 'probability')
tail(p)

       setosa versicolor virginica
145      0 0.02173913 0.9782609
146      0 0.02173913 0.9782609
147      0 0.02173913 0.9782609
148      0 0.02173913 0.9782609
149      0 0.02173913 0.9782609
150      0 0.02173913 0.9782609