pmml learning column missing error

KNIME Analytics Platform
1

Hi, I have just built a workflow with a pmml writer node using a cross-validation decision tree learner based on the classification-cuisine example (pmml learner workflow picture as attached). The pmml reader workflow predicts the system classifier for airplane maintenance data.

When I  run new data to predict a classifier on the pmml reader workflow I get the error as attached which I can get avoid if I include the original training data in the new data.

Can someone please advise if I am missing a step or steps in the pmml writer creation. 

2

Hi,

the problem seems to be that you train your decision tree on data with more columns than those available in the new data. Your decision tree has split criteria based on the attribute "onoff" and when this is not in the table coming into the predictor, it fails. So either you include the column to your new data (if that is possible), or you exclude it from the learning data. I hope this helps.

Regards,

Alexander