0

Possible Duplicate:
Suggestions for speeding up Random Forests

I want to build random forest on my data 129600 X 900. Moreover, I want to have not less than 1000 trees for regression. I executed the following code on my data, where all_reg is the 129600 X 900 data matrix and train_resp_reg is 129600 label matrix

train_reg <- randomForest(x=as.data.frame(all_reg[,2:dim(all_reg [2]]),
                      y=as.numeric(train_resp_reg),ntree=100)

the code have been running on a single processor for more than 3 days and it's still running, so I wonder if there is any solution to run it much faster, I can have access to a server with 64 cores, so how to make this code run as fast as possible?

any suggestions are appreciated

Community
  • 1
  • 1
DOSMarter
  • 1,485
  • 5
  • 21
  • 29
  • I saw that post, but still I need to parallelize not only building the trees, but also processing the data, especially because my data matrix is so huge, so I was wondering if this is possible? – DOSMarter Dec 10 '12 at 18:02
  • 2
    You've been shown how to parallelize RFs. If there is _other_ stuff that you need to speed up, you should ask a separate question about that code. – joran Dec 10 '12 at 18:46

0 Answers0