J48 Weka in Knime error Cannot handle string class

Hello everyone,

This is my first time using KNIME. I'm trying to classify some tweets into 3 sentiment class: positive, negative and neutral. The data includes 2 columns: the tweet and the manually classified class. In an attempt to increase the accuracy, I used J48 (3.7) node from the KNIME node repository, but it failed with the following error:

ERROR J48 (3.7)            0:293      Execute failed: Weka classifier cannot work with given data. Reason: weka.classifiers.trees.J48: Cannot handle string class!

How can I resolve this problem with this node? The actual workflow I am working on is in the attachment and KNIME 3.1.1 is currently being used.

Thank you!

Hi,

would be better if you export your workflow with some datas.

Fabien
 

Hi,

my bad. Certainly it is better to go with data. Here you are. The workflow with Weka J48 Learner / Predictor. I understand that the "Document Class" is now of String type. How can I change it / is there anyway that I can get J48 or other Weka learner to work with this kind of data type e.g. weka random forest?

Thank you very much.

Duc

Hi see the workflow in attachment.

Just filter the document column.

Best regards,

Fabien
 

Hi, Thank you very much :). It's working