Re: svm() output interpretation

[David Henderson <DNADave@Insightful.Com>]

Hi Kim:

A colleague of mine, Lixin Gong, was also looking into this.  Here is a
summary of what we both found:

The "accuracies" attribute of an output object when using a positive
non-zero value for the cross argument in svm is the prediction accuracy
in percent.  Each element is indeed the result of multiplying the
proportion of correct classifications by 100.  This will not be evident
when looking at the S code as the multiplication is performed in the C++
code itself.

You indicated that you were having some problems tuning your svm.  Could
you elaborate more on this?

Thanks!!

Dave H
-- 
David A. Henderson, Ph.D. 
Research Scientist 
Insightful Corporation 
1700 Westlake Avenue North, Suite 500 
Seattle, WA 98109-3044 
Tel: 206-802-2307 
Fax: 206-283-8691 
DNADave@Insightful.Com 
http://www.insightful.com