Performing Random Forest using EnMap Box?

Question

I am new to the software EnMap Box that incorporates Random Forest (RF) as one of its capability. To learn the Enmap, I followed the RF tutorial provided online.

My goal is to use RF as a tool to classify Landsat images using more than 100 explanatory variables.

Link to RF: https://bitbucket.org/enmapbox/enmap-box/wiki/imageRF%20-%20Manual%20for%20Application#!variable-importance-of-rfcrfr-models

In the EnMap tutorial (link provided above), under the "Parameterization of RFC/RFR Models", it only says "Select the Image to be classified...." But what image in particular? Am I going to use the image that stacked all 100 variables that I want in the analysis? In other words, how would I input the various explanatory variables into EnMap, so that I could then apply the tool "Variable Importance" to select important layers for classification?

Is there anyone familiar with EnMap and could guide me through this?

Answer 1

Although this is an old question and you have probably found out a long time ago, maybe someone else can use my answer.

With the imageRF, you can use as many variables or layers as you want, you just need to have them stacked up together in one (ENVI) file. Then, you parameterize your model using that layer stack. Afterwards, you can calculate variable importances for that model. It is advisable to have a look at those first and classify the image/layer stack when you are sure you have found a good combination of variables to save time.