Chapter 2. PMML model examples
PMML defines an XML schema that enables PMML models to be used between different PMML-compliant platforms. The PMML specification enables multiple software platforms to work with the same file for authoring, testing, and production execution, assuming producer and consumer conformance are met.
The following are examples of PMML Regression, Scorecard, Tree, and Mining models. These examples illustrate the supported types of models that you can integrate with your decision services in Red Hat Decision Manager.
For more PMML examples, see the DMG PMML Sample Files page.
Example PMML Regression model
<PMML version="4.2" xsi:schemaLocation="http://www.dmg.org/PMML-4_2 http://www.dmg.org/v4-2-1/pmml-4-2.xsd" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns="http://www.dmg.org/PMML-4_2"> <Header copyright="JBoss"/> <DataDictionary numberOfFields="5"> <DataField dataType="double" name="fld1" optype="continuous"/> <DataField dataType="double" name="fld2" optype="continuous"/> <DataField dataType="string" name="fld3" optype="categorical"> <Value value="x"/> <Value value="y"/> </DataField> <DataField dataType="double" name="fld4" optype="continuous"/> <DataField dataType="double" name="fld5" optype="continuous"/> </DataDictionary> <RegressionModel algorithmName="linearRegression" functionName="regression" modelName="LinReg" normalizationMethod="logit" targetFieldName="fld4"> <MiningSchema> <MiningField name="fld1"/> <MiningField name="fld2"/> <MiningField name="fld3"/> <MiningField name="fld4" usageType="predicted"/> <MiningField name="fld5" usageType="target"/> </MiningSchema> <RegressionTable intercept="0.5"> <NumericPredictor coefficient="5" exponent="2" name="fld1"/> <NumericPredictor coefficient="2" exponent="1" name="fld2"/> <CategoricalPredictor coefficient="-3" name="fld3" value="x"/> <CategoricalPredictor coefficient="3" name="fld3" value="y"/> <PredictorTerm coefficient="0.4"> <FieldRef field="fld1"/> <FieldRef field="fld2"/> </PredictorTerm> </RegressionTable> </RegressionModel> </PMML>
Example PMML Scorecard model
<PMML version="4.2" xsi:schemaLocation="http://www.dmg.org/PMML-4_2 http://www.dmg.org/v4-2-1/pmml-4-2.xsd" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns="http://www.dmg.org/PMML-4_2"> <Header copyright="JBoss"/> <DataDictionary numberOfFields="4"> <DataField name="param1" optype="continuous" dataType="double"/> <DataField name="param2" optype="continuous" dataType="double"/> <DataField name="overallScore" optype="continuous" dataType="double" /> <DataField name="finalscore" optype="continuous" dataType="double" /> </DataDictionary> <Scorecard modelName="ScorecardCompoundPredicate" useReasonCodes="true" isScorable="true" functionName="regression" baselineScore="15" initialScore="0.8" reasonCodeAlgorithm="pointsAbove"> <MiningSchema> <MiningField name="param1" usageType="active" invalidValueTreatment="asMissing"> </MiningField> <MiningField name="param2" usageType="active" invalidValueTreatment="asMissing"> </MiningField> <MiningField name="overallScore" usageType="target"/> <MiningField name="finalscore" usageType="predicted"/> </MiningSchema> <Characteristics> <Characteristic name="ch1" baselineScore="50" reasonCode="reasonCh1"> <Attribute partialScore="20"> <SimplePredicate field="param1" operator="lessThan" value="20"/> </Attribute> <Attribute partialScore="100"> <CompoundPredicate booleanOperator="and"> <SimplePredicate field="param1" operator="greaterOrEqual" value="20"/> <SimplePredicate field="param2" operator="lessOrEqual" value="25"/> </CompoundPredicate> </Attribute> <Attribute partialScore="200"> <CompoundPredicate booleanOperator="and"> <SimplePredicate field="param1" operator="greaterOrEqual" value="20"/> <SimplePredicate field="param2" operator="greaterThan" value="25"/> </CompoundPredicate> </Attribute> </Characteristic> <Characteristic name="ch2" reasonCode="reasonCh2"> <Attribute partialScore="10"> <CompoundPredicate booleanOperator="or"> <SimplePredicate field="param2" operator="lessOrEqual" value="-5"/> <SimplePredicate field="param2" operator="greaterOrEqual" value="50"/> </CompoundPredicate> </Attribute> <Attribute partialScore="20"> <CompoundPredicate booleanOperator="and"> <SimplePredicate field="param2" operator="greaterThan" value="-5"/> <SimplePredicate field="param2" operator="lessThan" value="50"/> </CompoundPredicate> </Attribute> </Characteristic> </Characteristics> </Scorecard> </PMML>
Example PMML Tree model
<PMML version="4.2" xsi:schemaLocation="http://www.dmg.org/PMML-4_2 http://www.dmg.org/v4-2-1/pmml-4-2.xsd" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns="http://www.dmg.org/PMML-4_2"> <Header copyright="JBOSS"/> <DataDictionary numberOfFields="5"> <DataField dataType="double" name="fld1" optype="continuous"/> <DataField dataType="double" name="fld2" optype="continuous"/> <DataField dataType="string" name="fld3" optype="categorical"> <Value value="true"/> <Value value="false"/> </DataField> <DataField dataType="string" name="fld4" optype="categorical"> <Value value="optA"/> <Value value="optB"/> <Value value="optC"/> </DataField> <DataField dataType="string" name="fld5" optype="categorical"> <Value value="tgtX"/> <Value value="tgtY"/> <Value value="tgtZ"/> </DataField> </DataDictionary> <TreeModel functionName="classification" modelName="TreeTest"> <MiningSchema> <MiningField name="fld1"/> <MiningField name="fld2"/> <MiningField name="fld3"/> <MiningField name="fld4"/> <MiningField name="fld5" usageType="predicted"/> </MiningSchema> <Node score="tgtX"> <True/> <Node score="tgtX"> <SimplePredicate field="fld4" operator="equal" value="optA"/> <Node score="tgtX"> <CompoundPredicate booleanOperator="surrogate"> <SimplePredicate field="fld1" operator="lessThan" value="30.0"/> <SimplePredicate field="fld2" operator="greaterThan" value="20.0"/> </CompoundPredicate> <Node score="tgtX"> <SimplePredicate field="fld2" operator="lessThan" value="40.0"/> </Node> <Node score="tgtZ"> <SimplePredicate field="fld2" operator="greaterOrEqual" value="10.0"/> </Node> </Node> <Node score="tgtZ"> <CompoundPredicate booleanOperator="or"> <SimplePredicate field="fld1" operator="greaterOrEqual" value="60.0"/> <SimplePredicate field="fld1" operator="lessOrEqual" value="70.0"/> </CompoundPredicate> <Node score="tgtZ"> <SimpleSetPredicate booleanOperator="isNotIn" field="fld4"> <Array type="string">optA optB</Array> </SimpleSetPredicate> </Node> </Node> </Node> <Node score="tgtY"> <CompoundPredicate booleanOperator="or"> <SimplePredicate field="fld4" operator="equal" value="optA"/> <SimplePredicate field="fld4" operator="equal" value="optC"/> </CompoundPredicate> <Node score="tgtY"> <CompoundPredicate booleanOperator="and"> <SimplePredicate field="fld1" operator="greaterThan" value="10.0"/> <SimplePredicate field="fld1" operator="lessThan" value="50.0"/> <SimplePredicate field="fld4" operator="equal" value="optA"/> <SimplePredicate field="fld2" operator="lessThan" value="100.0"/> <SimplePredicate field="fld3" operator="equal" value="false"/> </CompoundPredicate> </Node> <Node score="tgtZ"> <CompoundPredicate booleanOperator="and"> <SimplePredicate field="fld4" operator="equal" value="optC"/> <SimplePredicate field="fld2" operator="lessThan" value="30.0"/> </CompoundPredicate> </Node> </Node> </Node> </TreeModel> </PMML>
Example PMML Mining model (modelChain)
<PMML version="4.2" xsi:schemaLocation="http://www.dmg.org/PMML-4_2 http://www.dmg.org/v4-2-1/pmml-4-2.xsd" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns="http://www.dmg.org/PMML-4_2"> <Header> <Application name="Drools-PMML" version="7.0.0-SNAPSHOT" /> </Header> <DataDictionary numberOfFields="7"> <DataField name="age" optype="continuous" dataType="double" /> <DataField name="occupation" optype="categorical" dataType="string"> <Value value="SKYDIVER" /> <Value value="ASTRONAUT" /> <Value value="PROGRAMMER" /> <Value value="TEACHER" /> <Value value="INSTRUCTOR" /> </DataField> <DataField name="residenceState" optype="categorical" dataType="string"> <Value value="AP" /> <Value value="KN" /> <Value value="TN" /> </DataField> <DataField name="validLicense" optype="categorical" dataType="boolean" /> <DataField name="overallScore" optype="continuous" dataType="double" /> <DataField name="grade" optype="categorical" dataType="string"> <Value value="A" /> <Value value="B" /> <Value value="C" /> <Value value="D" /> <Value value="F" /> </DataField> <DataField name="qualificationLevel" optype="categorical" dataType="string"> <Value value="Unqualified" /> <Value value="Barely" /> <Value value="Well" /> <Value value="Over" /> </DataField> </DataDictionary> <MiningModel modelName="SampleModelChainMine" functionName="classification"> <MiningSchema> <MiningField name="age" /> <MiningField name="occupation" /> <MiningField name="residenceState" /> <MiningField name="validLicense" /> <MiningField name="overallScore" /> <MiningField name="qualificationLevel" usageType="target"/> </MiningSchema> <Segmentation multipleModelMethod="modelChain"> <Segment id="1"> <True /> <Scorecard modelName="Sample Score 1" useReasonCodes="true" isScorable="true" functionName="regression" baselineScore="0.0" initialScore="0.345"> <MiningSchema> <MiningField name="age" usageType="active" invalidValueTreatment="asMissing" /> <MiningField name="occupation" usageType="active" invalidValueTreatment="asMissing" /> <MiningField name="residenceState" usageType="active" invalidValueTreatment="asMissing" /> <MiningField name="validLicense" usageType="active" invalidValueTreatment="asMissing" /> <MiningField name="overallScore" usageType="predicted" /> </MiningSchema> <Output> <OutputField name="calculatedScore" displayName="Final Score" dataType="double" feature="predictedValue" targetField="overallScore" /> </Output> <Characteristics> <Characteristic name="AgeScore" baselineScore="0.0" reasonCode="ABZ"> <Extension name="cellRef" value="$B$8" /> <Attribute partialScore="10.0"> <Extension name="cellRef" value="$C$10" /> <SimplePredicate field="age" operator="lessOrEqual" value="5" /> </Attribute> <Attribute partialScore="30.0" reasonCode="CX1"> <Extension name="cellRef" value="$C$11" /> <CompoundPredicate booleanOperator="and"> <SimplePredicate field="age" operator="greaterOrEqual" value="5" /> <SimplePredicate field="age" operator="lessThan" value="12" /> </CompoundPredicate> </Attribute> <Attribute partialScore="40.0" reasonCode="CX2"> <Extension name="cellRef" value="$C$12" /> <CompoundPredicate booleanOperator="and"> <SimplePredicate field="age" operator="greaterOrEqual" value="13" /> <SimplePredicate field="age" operator="lessThan" value="44" /> </CompoundPredicate> </Attribute> <Attribute partialScore="25.0"> <Extension name="cellRef" value="$C$13" /> <SimplePredicate field="age" operator="greaterOrEqual" value="45" /> </Attribute> </Characteristic> <Characteristic name="OccupationScore" baselineScore="0.0"> <Extension name="cellRef" value="$B$16" /> <Attribute partialScore="-10.0" reasonCode="CX2"> <Extension name="description" value="skydiving is a risky occupation" /> <Extension name="cellRef" value="$C$18" /> <SimpleSetPredicate field="occupation" booleanOperator="isIn"> <Array n="2" type="string">SKYDIVER ASTRONAUT</Array> </SimpleSetPredicate> </Attribute> <Attribute partialScore="10.0"> <Extension name="cellRef" value="$C$19" /> <SimpleSetPredicate field="occupation" booleanOperator="isIn"> <Array n="2" type="string">TEACHER INSTRUCTOR</Array> </SimpleSetPredicate> </Attribute> <Attribute partialScore="5.0"> <Extension name="cellRef" value="$C$20" /> <SimplePredicate field="occupation" operator="equal" value="PROGRAMMER" /> </Attribute> </Characteristic> <Characteristic name="ResidenceStateScore" baselineScore="0.0" reasonCode="RES"> <Extension name="cellRef" value="$B$22" /> <Attribute partialScore="-10.0"> <Extension name="cellRef" value="$C$24" /> <SimplePredicate field="residenceState" operator="equal" value="AP" /> </Attribute> <Attribute partialScore="10.0"> <Extension name="cellRef" value="$C$25" /> <SimplePredicate field="residenceState" operator="equal" value="KN" /> </Attribute> <Attribute partialScore="5.0"> <Extension name="cellRef" value="$C$26" /> <SimplePredicate field="residenceState" operator="equal" value="TN" /> </Attribute> </Characteristic> <Characteristic name="ValidLicenseScore" baselineScore="0.0"> <Extension name="cellRef" value="$B$28" /> <Attribute partialScore="1.0" reasonCode="LX00"> <Extension name="cellRef" value="$C$30" /> <SimplePredicate field="validLicense" operator="equal" value="true" /> </Attribute> <Attribute partialScore="-1.0" reasonCode="LX00"> <Extension name="cellRef" value="$C$31" /> <SimplePredicate field="validLicense" operator="equal" value="false" /> </Attribute> </Characteristic> </Characteristics> </Scorecard> </Segment> <Segment id="2"> <True /> <TreeModel modelName="SampleTree" functionName="classification" missingValueStrategy="lastPrediction" noTrueChildStrategy="returnLastPrediction"> <MiningSchema> <MiningField name="age" usageType="active" /> <MiningField name="validLicense" usageType="active" /> <MiningField name="calculatedScore" usageType="active" /> <MiningField name="qualificationLevel" usageType="predicted" /> </MiningSchema> <Output> <OutputField name="qualification" displayName="Qualification Level" dataType="string" feature="predictedValue" targetField="qualificationLevel" /> </Output> <Node score="Well" id="1"> <True/> <Node score="Barely" id="2"> <CompoundPredicate booleanOperator="and"> <SimplePredicate field="age" operator="greaterOrEqual" value="16" /> <SimplePredicate field="validLicense" operator="equal" value="true" /> </CompoundPredicate> <Node score="Barely" id="3"> <SimplePredicate field="calculatedScore" operator="lessOrEqual" value="50.0" /> </Node> <Node score="Well" id="4"> <CompoundPredicate booleanOperator="and"> <SimplePredicate field="calculatedScore" operator="greaterThan" value="50.0" /> <SimplePredicate field="calculatedScore" operator="lessOrEqual" value="60.0" /> </CompoundPredicate> </Node> <Node score="Over" id="5"> <SimplePredicate field="calculatedScore" operator="greaterThan" value="60.0" /> </Node> </Node> <Node score="Unqualified" id="6"> <CompoundPredicate booleanOperator="surrogate"> <SimplePredicate field="age" operator="lessThan" value="16" /> <SimplePredicate field="calculatedScore" operator="lessOrEqual" value="40.0" /> <True /> </CompoundPredicate> </Node> </Node> </TreeModel> </Segment> </Segmentation> </MiningModel> </PMML>