R help - Oct 2011 - handling constant factors in prediction using svm

Hi users!

I am fitting a model with several factor variables as independents using
svm. since there are lots of categorical variables,the training and test
data sets have been created using dummy.data.frame option from dummies
package. I have a factor A in the training data set with 2 levels (0,1).In
the test set, this factor A has only 1 level (1) and hence when applying
dummy.data.frame, the variable gets dropped(and that's how i want it too).
The problem comes when I am trying to predict the test data as an error is
thrown saying A0 object is not found. Is there anyway  to solve this
problem?

Thanks
Divya

--
View this message in context:
http://r.789695.n4.nabble.com/handling-constant-factors-in-prediction-using-svm-tp3870093p3870093.html
Sent from the R help mailing list archive at Nabble.com.

On 04.10.2011 08:53, Divyam wrote:
> Hi users!
>
> I am fitting a model with several factor variables as independents using
> svm. since there are lots of categorical variables,the training and test
> data sets have been created using dummy.data.frame option from dummies
> package. I have a factor A in the training data set with 2 levels (0,1).In
> the test set, this factor A has only 1 level (1) and hence when applying
> dummy.data.frame, the variable gets dropped(and that's how i want it
too).
> The problem comes when I am trying to predict the test data as an error is
> thrown saying A0 object is not found. Is there anyway  to solve this
> problem?
Errr, if you learned a model that predicts based on several variables, 
including A0, what do you expect what happens if A0 is not given? Well, 
you cannot predict. So if A0 is constant in your test cases, just supply it!

To simplify, consider a linear model y=bX+e. Now one column of X is 
missing for prediction. y will be undefined, obviously.

Uwe Ligges




> Thanks
> Divya
>
> --
> View this message in context:
http://r.789695.n4.nabble.com/handling-constant-factors-in-prediction-using-svm-tp3870093p3870093.html
> Sent from the R help mailing list archive at Nabble.com.
>
> ______________________________________________
> 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.

Ok. I should have made the question clearer. By asking is there anyway to
solve this problem, what I meant was, how to give exceptions to such
scenarios. My mistake and apologies for that. I can understand why
prediction cannot be done when a trained variable is missing in the test set
:). thanks for responding to a not very clear question Uwe.

Divya

--
View this message in context:
http://r.789695.n4.nabble.com/handling-constant-factors-in-prediction-using-svm-tp3870093p3881504.html
Sent from the R help mailing list archive at Nabble.com.