Geek asks and answers

Random review of the forest code

I am engaged in a random forest research project. I found many implementations of the algorithm, but the bulk of the code is often written in Fortran, while I am completely naive in it.

I need to edit the code, change the main parameters (for example, the depth of the tree, the number of functional variables, ...) and track the performance of the algorithm during each run.

I am currently using Windows-Precompiled-RF_MexStandalone-v0.02- . The train and prediction functions are matlab mex files and cannot be opened or edited. Can anyone give me advice on what to do, or there is a valid and fully Matlab-based version of random forests.

I carefully read randomforest-matlab. Unfortunately, the main part of training is a dll file. Thanks to reading more, most of my miracles are now permitted. Basically my question was how to manage multiple trees at the same time.

+3
source share
4 answers

Have you looked at these libraries?

+5
source

If you are doing a research project on this, it is probably best to implement individual tree formation yourself in C, and then write Mex wrappers. I would start with the ID3 tree (before trying to use C4.5, for example). Then write the random forest code itself, which, as soon as you write the tree code, is not so hard.

You will:

  • .
  • ,

, , . , - .

? . , ? Undoubtably.

+2

R , randomForest. Breiman Fortran, C.

http://cran.r-project.org/web/packages/randomForest/index.html

, ( , ,...), .

+1

, , - Weka. java based lucid.Performance , R. http://www.cs.waikato.ac.nz/ml/weka/

0

More articles:


All Articles