2
$\begingroup$

Please I want to know if there is any SVM R package that can handle more than one response variable (y) at a time. that is to train one model for predicting more than one response variable. it could be regression or multi class classification problem.

Thanks for your help

$\endgroup$
2
  • 2
    $\begingroup$ Just a little thought here. Why don't you just train one SVM for each response variable? $\endgroup$ – Pablo Suau Feb 27 '15 at 9:15
  • $\begingroup$ I second @PabloSuau's thought. Many packages have the capability to train multi-class classifiers, but often what's happening under the hood is that n binary classifiers are trained, and these are used individually to make predictions for each class. The SO question here also seems relevant: stackoverflow.com/questions/22009871/… $\endgroup$ – kylerthecreator Jun 29 '15 at 19:32
1
$\begingroup$

e1071 is a highly popular and therefore safe solution for this purpose.

But if data starts to get bigger I would go with classyfire. It's a quite fresh package built on e1071. It is highly optimized and have some other nice feature too. In my experience it produced similar results to e1071 but with a much faster speed as it was able to utilize all 4 core of my processor.

$\endgroup$
0
$\begingroup$

Check out the e1071 package, here is the manual http://cran.r-project.org/web/packages/e1071/e1071.pdf

$\endgroup$
0
$\begingroup$

Clarification: are we talking about multiclass

https://en.wikipedia.org/wiki/Multiclass_classification

or multilabel?

https://en.wikipedia.org/wiki/Multi-label_classification

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.