Required attribute Node@score is not defined error

[rahuls04]

I am trying to move from a logistic regression model to xgboost. However, when I run my test cases on my new xgboost PMML I get the following error - org.jpmml.evaluator.MissingAttributeException: Required attribute Node@score is not defined

Possible solutions?

[vruusmann]

Required attribute Node@score is not defined

Possible solutions?

Perhaps you should add Node@score attributes to all terminal (aka leaf) Node elements?

Where did you get this XGBoost PMML file? Was it generated using JPMML-XGBoost?

[rahuls04]

The xgboost pmml file was generated using the R pmml package

[rahuls04]

I do have Node score attributes at my leaf nodes like so -

[vruusmann]

The xgboost pmml file was generated using the R pmml package

It's probably incorrect/invalid then.

According to the PMML specification, the Node@score attribute is required.

Quoting http://dmg.org/pmml/v4-3/TreeModel.html: "it is the model provider's responsibility to make sure that it is not possible that the scoring process ends in a Node which does not have a score attribute. This includes scenarios with missing values where the final Node is not necessarily a leaf node."

[vruusmann]

The xgboost pmml file was generated using the R pmml package

Please re-export your xgboost model using the R2PMML package. The generated PMML file will be about 50% smaller, and correct.

[rahuls04]

Does the R2pmml package support complex feature transforms?

[vruusmann]

Does the R2pmml package support complex feature transforms?

Depends.

But what good are those complex feature transforms of the R pmml package, if the final model is incorrectly encoded?