[R-sig-Geo] Important variable- SVM

Enoch Gyamfi Ampadu eg@mp@du @end|ng |rom gm@||@com
Mon Sep 2 12:45:41 CEST 2019

Dear List,
Please I need your help.

1. I used the SVM to undertake a forest cover classification. I wanted to
determine the important variable but when I checked the attributes of the
model it does not seem to have that function for that. I wanted to find if
there is a way I could determine it. Find below the attributes for the SVM

> attributes(modFit_svm)
 [1] "call"            "type"            "kernel"          "cost"
 [6] "gamma"           "coef0"           "nu"              "epsilon"
[11] "scaled"          "x.scale"         "y.scale"         "nclasses"
[16] "tot.nSV"         "nSV"             "labels"          "SV"
[21] "rho"             "compprob"        "probA"           "probB"
[26] "coefs"           "na.action"       "fitted"
 "decision.values" "terms"

I also applied the RF for the same image and I am able to determine the
important variable as it has that in the RF attributes. find below the RF

 [1] "call"            "type"            "predicted"       "err.rate"
 [6] "votes"           "oob.times"       "classes"         "importance"
[11] "local Importance" "proximity"       "ntree"           "mtry"
[16] "y"               "test"            "inbag"           "terms"

2. Please can you also assist me on how to determine the total area covered
by each of the classes. I am not finding my way out on how to determine it
as well.

I will be glad to have your advice.

Thank you.

Best regards,


*Enoch Gyamfi - Ampadu*

*Geography & Environmental Sciences*

*College of Agriculture, Engineering & Science*

*University of KwaZulu-Natal, Westville Campus*

*Private Bag X54001*
*Durban, South Africa **– 4000**.*
*Phone: +27 835 828255*

*email: egampadu using gmail.com <egampadu using gmail.com>*

*skype: enoch.ampadu*
*The highest evidence of nobility is self-control*.

*A simple act of kindness creates an endless ripple*.

	[[alternative HTML version deleted]]

More information about the R-sig-Geo mailing list