I am about to use geographically weighted random forest (GWRF) for regression prediction (I want to predict a continuous raster). I have one dependent and 4 independent variables. One of my independent variables is a categorical raster (several land use categories, named "1", "2" etc) while the rest of the variables are continuous rasters.

RF can handle both categorical and continuous data without problem.

GWR on the other hand can't handle categorical data. This means that I have to convert the categorical data to "dummy" and I should be okay (based on this post, and this).

So my question is this, should I convert the categorical variable to dummy before I use GWRF or I don't have to?

I am using the R package spatialML.



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.