RE: class prediction with weka KnowledgeFlow

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view

RE: class prediction with weka KnowledgeFlow

Mark Hall-11
Hi Christian,

This works for me. I set up the same flow as you did. After running
the cross validation I disconnected the train and test connections
from the CrossValidationFoldMaker, added a new ArffLoader,
ClassAssigner and TestSetMaker and then connected these up to the
classifier. I loaded a test set and passed this through the
Classifier, ClassifierPerformanceEvaluator and got the results in the
TextViewer. There is a slight bug in that some of the connection links
are shown in grey rather than red (as there is only is test set
connected and no training set).

I'm not sure if this is exactly what you want anyway. The model that
gets applied to the new test set in this example is the model that has
been built on the last fold of the cross validation.


> Hi,
> I want to make a class prediction on a built model (crossvalidation,
> Meta.CostSensitiveClassifier) an label the test data with the predicted
> class. I am able to build the model and vizualize it in textual form
> (components are:arff loader, class assigner, crossvalidationfoldmaker,
> costsensitiveClassifier, classifierPerformanceEvaluator, Textviewer). Now I
> want to make a class prediction with this model. I loaded another arff file
> and tried to supply it as a test set to the classifier, but this doesn't
> works. Can anybody please help?!
> Is it possible to save the result from an classification (after
> PerformanceEvaluator) to a csv file?
> Thanks a lot,
> Christian

Wekalist mailing list
[hidden email]