Re: what is the significance of this Warning?

On 27/06/12 7:55 PM, Bill wrote:
> Hello All,
>
> I get this warning
> "[InputMappedClassifier] Warning: incoming nominal attribute Color does
> not have the same number of values as model attribute Color"
>
> The attribute Color is the class attribute in my case (although the
> warning message does not seem specify that.) I have read that
> "InputMappedClassifier that can
>
> handle discrepancies between train and test headers by building a
> mapping between train and test attributes and the values of nominal
> attributes. However, it can't handle the case where there are different
> class values between the two data sets. "
>
>
> Question:
> Since this is a warning and not an error I am not sure if this will ruin
> the prediction/classification that is being attempted. Any comments?

It isn't a problem as long as the class attribute values that appear in 
the test set have all been seen/declared in the training data. If there 
are class labels in the test data that have not been seen during 
training then these test instances have their class mapped to missing 
value since evaluation can only be performed with respect to what was 
known at training time.

Cheers,
Mark.

>
> Also, the way I have attempted to fix this is to put more records into
> the test set even though they are being artificially generated. I
> suppose that the fact that the records are artificial will not effect
> the predictions on the non-artificial records since both types of
> records are in just the test set and not effecting model generation.
> Does this sound correct?

_______________________________________________
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