XGBoostRegressor not loading in java

[matigurten]

I created a XGBoostRegressor model in python and converted it using xgboost_to_pmml from nyoka, resulting in a PMML (version 4.4, then manually downgraded to 4.3 for java compatibility) file that ends in smth like this:

    <MiningModel modelName="XGBoostModel" functionName="regression">
        <MiningSchema>
            <MiningField name="x1" usageType="active" optype="continuous"/>
            <MiningField name="x2" usageType="active" optype="continuous"/>
            <MiningField name="x3" usageType="active" optype="continuous"/>
            <MiningField name="x4" usageType="active" optype="continuous"/>
            <MiningField name="prediction" usageType="target" optype="continuous"/>
        </MiningSchema>
        <Output>
            <OutputField name="predicted_prediction" optype="continuous" dataType="double" feature="predictedValue"/>
        </Output>
        <Segmentation multipleModelMethod="sum" missingThreshold="1"/>
    </MiningModel>
</PMML>

But when i try to create a MiningModelEvaluator in java, with this PMML I get the following error:

org.jpmml.evaluator.MissingElementException (at or around line 95 of the PMML document): Required element Segmentation/Segment is not defined

Line 95 is the one that reads <Segmentation multipleModelMethod="sum" missingThreshold="1"/>

Please help, thanks a lot!

[vruusmann]

I created a XGBoostRegressor model in python and converted it using xgboost_to_pmml from nyoka,

Please consider using the sklearn2pmml package instead.

org.jpmml.evaluator.MissingElementException (at or around line 95 of the PMML document): Required element Segmentation/Segment is not defined

The JPMML-Evaluator library is right when making this complaint, because you have an empty Segmentation element on line 95.

The PMML specification requires it to contain at least one Segment child element, see http://dmg.org/pmml/v4-3/MultipleModels.html#xsdElement_Segmentation