Hi Mark,

 

Thank you so much for the helpful answer.

It would be very nice if you have some comments whether there is much difference when using logistic models or not using for AUC calculation.

 

Thank you in advance,

Kanthida

>>> Mark Hall <mhall <at> pentaho.com> 06/30/10 11:51 PM >>>
On 1/07/10 12:53 AM, Kanthida Kusonmano wrote:
> Hi,
> I am wondering how the AUC is calculated in WEKA for Support Vector
> Machine (SVM) that has no probabilistic model.

Weka computes the AUC using the Mann-Whitney U statistic (actually the p
statistic from the U statistic). There are many ties in the ranking produced by
SVMs when there is no probabilistic model fit to the output. It is best, if
using SMO, to turn on the -M option, to fit logistic models to SMO's output when
evaluating with respect to AUC.

Cheers,
Mark.

_______________________________________________
Wekalist mailing list
Send posts to: Wekalist <at> list.scms.waikato.ac.nz
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html