r - How can you reduce the default ntree=500 parameter passed to RF from caret? -
i believe "rf" (randomforest) method in caret sets default number of trees @ 500. unfortunately, causes time complexity grow out of control larger datasets. there quick way reduce number of trees without creating custom method? know tuneable parameter rf mtry.
just clarify: i'm not looking tune on number of trees. want fix lower value can run rf in reasonable amount of time.
you can specify ntree parameter when call train so:
rf <- train(x, y, method="rf", preprocess=c("center","scale"), ntree=100, trcontrol=fitcontrol)
Comments
Post a Comment