Hiya,

I'd like to improve the performance of my SVM regression model.

I believe I've tuned the parameters of the RBF and cost successfully however expect that I can improve performance using a random forest approach to feature selection.

Also perhaps adding some transforms of my input space prior to the kernel application.

I'm working in R.

I'm really looking for some help on understanding first then building the random forest algorithm for the SVM.

I'm restricted in resource in that I'm working on a personal laptop with 4Gb ram.

There are a few papers on this topic but are a bit too heavy for an introductory understanding of the random forest algorithm.

Regards,

Clancy.