thank for the knime 2.6 release. Smooth install and very promising.
I got a problem with new WEKA nodes (J48) as porting an old workflow to new version a metanode stopped working properly. Please find attached it. It sounded strange as it worked perfectly on knime 2.5.4
Thanks for any help.
the problem is that the WekaPredictor assumes the same number and order of attributes (i.e. table columns) as with which the model (e.g. the J48) was created. Hence, no column name mapping is done so far, which should definitely be changed in further versions. In previous versions, probably wrong result have been calculated instead of throwing an error.
But as a workaround you could just use the "Reference Column Filter" between the "Partitioning"-node and the "Weka Predictor" (see attachment).
I got the point. I will better off going back to check it all.
Anyway, the workaround seems to deliver same predictors selection and confusion matrix as with
Again, thanks a lot for your time and quick answer.