Actually i think i found the problem, its something about the probability model
again as it seems, if you just take the normal predictions everythings good. Man
does that probability stuff absolutely not work properly. Any suggestions how to
do ROC curves without it?
Or am i just generally wrong with the asumption, that if i get those
probabilities, that 0.5 is the threshold to get "the default" result?
On 16.11.2012, at 16:32, Jessica Streicher wrote:
> Hi again!
>
> This might be more of a statistical question, but anyway:
> If i train several support vector machines with different degrees of
polynomials, and as result, get that higher degrees not only have a higher test
error, but also a higher in-sample error, why is that?
>
> I would assume i should get an in-sample error lower or at least the same
as the linear case.
>
> i'm worried i did something wrong there.
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.