KNIME 2.11.1 and Weka 3.7 Bug

Hi all. First time forum poster. Been using KNIME for almost 2 years now and use the Weka extension quite frequently.

Ever since the 2.11.1 update, I have been getting the following error when using the Weka 3.7 Bagging meta classifier: "ERROR WekaSerializer Unexpected error while loading settings from class: weka.classifiers.meta.Bagging, using default values"

This error is occuring in workflows that previously used to work prior to the KNIME update.

Things actually seem to run and scores are produced DURING THE FIRST RUN, however, after changing the Bagging classifier settings for a 2nd/3rd... run KNIME throws the "The settings were not changed. The node will not be reset" message. Even after I reset the node and rerun with different settings (even different classifiers), I get the exact same scores!! strange.

Thus, not sure I can rely on Bagging at all.

Anybody else having this problem? Your help is much appreciated. Thanks all!

One more thing on Weka 3.7 and KNIME 2.11.1.

The Weka Predictor 3.7 node view shows nonsense results now, i.e.,


Total Number of Instances   0
Ignored Class Uknown Instances  n

where "n" is whatever number of obs you have.

I've also indicated that. WEKA 3.7 nodes ignore entered settings since introduction of driving them by flow variables (which doesn't work either).

same here!!!

Is it possible to put Weka 3.6 back up in extensions library until 3.7 is fixed? I built a new computer especially to run KNIME and started from a fresh install so I don't have 3.6 anymore unfortunately. Your help is much appreciated.

I think you can install it from the 2.10 site: 

I've not tested this though...

We just released KNIME 2.11.2 which fixes the settings problems you have discovered. Thanks again for reporting those issues and please keep us posted.


I've also found that the J48 node does not save the numFolds parameter.

Can such issues be fixed fast? Data mining cup is up and running!!!


Oh my fault, this parameter just'can not be not saved if the option for reduced-error pruning is not turned on.

Sorry folks.

Keep up the good work!