how to define model name in a for loop?

Question

I want to train n random forest on different samples. sample 1 gives rf1, sample 2 gives rf2, etc.

BUT this kind of code doesn't work (error object of type 'closure' is not subsettable)

for (i in 1:3) {
       rf$i <- train(Y~.,data=trainingData,method="rf",
              ntree = 100,
              tuneGrid=data.frame(mtry = mtry),
              trControl = controle,
              metric='ROC')
}

How can I create the n random forest models? Yours sincerely Loïc

Answer 1

It does not work because rf does not exist yet and you can't subset it.

1. Use a list as a container

The following should work.

# define the length of your random forest trials
N = 3
rf = vector( "list", N)
for (i in seq_len( N ) {
   rf[[ i ]] <- train( Y ~. , data = trainingData, method = "rf",
          ntree = 100,
          tuneGrid=data.frame(mtry = mtry),
          trControl = controle,
          metric='ROC')
}

The above code stores a list rf which contains three element according to N . You can access to each object with rf[[ 1 ]] , rf[[ 2 ]] , rf[[ 3 ]] .

2. Store objects independently

If you want to physically store independent rf objects in your Global Enviroment, the you have to use assign() as follows:

# define the length of your random forest trials
N = 3
for (i in seq_len( N ) {
       assign( paste0( "rf", i) ,
               train( Y ~. , 
                      data = trainingData, method = "rf",
                      ntree = 100,
                      tuneGrid=data.frame(mtry = mtry),
                      trControl = controle,
                      metric='ROC')
    }

This stores three objects rf1 , rf2 , and rf3 in your environment and you can work on them independently.