Removing unimportant variables before building Random Forest

Question

I am using the 'fgl' data-set from the MASS package and building the Random Forest model. 

model1<-randomForest(type ~ .,data=fgl)

Now, I'd want to know can i remove the unimportant variables beforehand so that i can build a better Random Forest algorithm?

Answer 1

Build the randomForest model on top of the 'fgl' data-set as usual and then use the importance() function to find the relative importance measure of these variables.

rf<-randomForest(type ~ .,data=fgl)
importance(rf)

This gives you the following result:

 MeanDecreaseGini
RI         23.00797
Na         16.73084
Mg         25.26359
Al         24.97366
Si         13.05597
K          13.86039
Ca         20.30383
Ba         13.24304
Fe          6.80955

Now, you can go ahead and start building the randomForest model by removing the least important variables