Support for random forest fit by ranger

[mikajoh]

First of all, your package is awesome! I merely have a friendly feature request.

The ranger package, which you are probably familiar with, have huge speed improvements compared to randomForest and randomForestSRC. I often find the latter two to be too slow for large (or even medium-sized) data sets.

Do you have any plans to implement support for random forest fit by ranger, or do you know if anybody else is working on that?

[zmjones]

I didn't really have plans to add support for it here, but I just rewrote the package in such a way that doing this probably wouldn't be very hard. So maybe I will. Also basically all of this functionality has been implemented in mlr which supports ranger and a ton of other methods.

[mikajoh]

I understand! Sure, I tend to use mlr for this, but I often don't need the extended functionality, and the more standard approach with one rf function call + your pkg seems like a faster workflow for quick eda on new data. Your code looks pretty manageable, so I might just take a stab at it myself, if OK :)

[zmjones]

No worries! I am almost done with it now. Will probably push the change tonight On Fri, Nov 25, 2016 at 4:02 PM Mikael Poul Johannesson < notifications@github.com> wrote:

I understand! Sure, I tend to use mlr https://github.com/mlr-org/mlr for this, but I often don't need the extended functionality, and the more standard approach with one rf function call + your pkg seems like a faster workflow for quick eda on new data. Your code looks pretty manageable, so I might just take a stab at myself, if OK :)

— You are receiving this because you commented.

Reply to this email directly, view it on GitHub https://github.com/zmjones/edarf/issues/54#issuecomment-263022134, or mute the thread https://github.com/notifications/unsubscribe-auth/ABEBhSTtsTmgX0_4mb8OqicEMiAmK-iPks5rB0zdgaJpZM4K8Ux1 .

[mikajoh]

Awesome!

[zmjones]

just pushed the addition. let me know if there are problems.