[OSGeo-Discuss] Random Forest Classification

Ned Horning nedh at lightlink.com
Wed Oct 26 05:10:27 PDT 2011


Hi Mohammed,

I recently posted some random forests R scripts and guides on our web 
site: http://biodiversityinformatics.amnh.org/. On the site go to Open 
Source Resources => CBC Developed Resources. If you download and use 
these tools I would appreciate your feedback. I still need to upload 
this link to the OSGeo Educational Content Inventory.

I have had very good luck using random forests for classification. I 
prefer random forests classification much more than single decision 
trees if I'm using many predictor variables. Regression is a little more 
problematic but applying the correction to reduce the bias at the 
extreme values seems to help.

There are also classification and regression scripts to apply random 
forests to segmented images but those will be posted on another site in 
a few weeks along with some exercises.

All the best,

Ned

On 10/26/2011 02:40 AM, Mohammed Rashad wrote:
> Hi All,
> I wrant to do Random Forest classification. I installed R, 
> randomForest classifier package for R
> but dont know how to use it.
>
> Is there any Open Source Remote sensing application which do RF 
> classification on satellite images?
>
> Anyone r has random forest classification example?
>
> Any language or package  example no problem.
>
> Does anyone did it in R?
> if yes how?
>
> I google RF Classification but most of them are for medical disease 
> and research not for Remote Sensing
>
>
>
> -- 
> Regards,
>    Mohammed Rashad K M
>    M.S. (By Research) student
>    Lab for Spatial Informatics
>    Department of CSE
>    International Institute of Information Technology
>    Hyderabad, India
>
>
> _______________________________________________
> Discuss mailing list
> Discuss at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/discuss

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/discuss/attachments/20111026/6a399016/attachment-0002.html>


More information about the Discuss mailing list