[R] MetaCost (and RandomForest)

Annelen Brunner annelen_brunner at gmx.de
Fri Jul 27 03:05:56 CEST 2012


Hello!

Is there an implementation of the MetaCost algorithm in R?
I found a project here: https://r-forge.r-project.org/scm/viewvc.php/src/mlrCost/metacost.r?root=mlr&view=log
But I'm not sure about its status - it seems unfinished to me.

My goal is to use MetaCost for cost-sensitive learning of an unbalanced data set with RandomForest. I did this in Weka with good results and now want to implement the same procedure in R. 
If MetaCost is not available, I would also welcome suggestions how to do cost-sensitive learning with RandomForest. I am aware that you can play around with the sample size (sampsize) parameter, but this is quite a different strategy and produces quite different (and worse) results on my data.

Best regards,

Annelen


More information about the R-help mailing list