JPMML Classifier node error when appending probability columns

KNIME Extensions
1

Hello everyone,

I am getting an error when using the JPMML Classifier node. I am modeling binary classes (0/1) and I am using a PMML export from IBM SPSS Modeler. The data have 82 fields. I read the PMML with the "PMML Reader" node and score my data with the "JPMML Classifier" node. Using the "JPMML Classifier" node as-is is working as expected: I get a prediction field appended to my data with binary labels.

But, when I check the "Append columns with normalized class distribution" under the "Options" tab in the node dialog, the node outputs the following error upon execution:

ERROR JPMML Classifier     0:12       Execute failed: Cell count in row "Row0" is not equal to length of column names array: 84 vs. 82

Looks like the node cannot write the probability columns to the data, or change the data model.

Any ideia how to fix this issue?

Thanks in advance!

2

Hi,

is it possible for you to share the workflow? It would be best if I could have a look at it myself. What kind of model does your PMML contain?

Kind regards,

Alexander

3

Hi ,

i tried to solve this by using below flow.

Tree Ensemble Model Extract -> Table to PMML -> PMML to Cell -> Cell to PMML -> JPMML Classifier.

 

Hope this will work.

 

Regards

Deb