Fix the defect of modelChain with Output

autodeployai · Dec 9, 2024 · 5e4a672 · 5e4a672
1 parent 67c9bb7
commit 5e4a672
Show file tree

Hide file tree

Showing 2 changed files with 86 additions and 1 deletion.
diff --git a/src/main/scala/org/pmml4s/model/MiningModel.scala b/src/main/scala/org/pmml4s/model/MiningModel.scala
@@ -143,6 +143,7 @@ class MiningModel(
                 case ResultFeature.probability    => x._2.value.foreach(y => {
                   probabilities += (y -> x._1.toDouble)
                 })
+                case _ =>
               }
             })
             outputs.probabilities = probabilities.toMap
@@ -151,7 +152,7 @@ class MiningModel(
             if (outputs.probabilities.nonEmpty && outputs.predictedValue == null) {
               outputs.evalPredictedValueByProbabilities()
             }
-            result(series, outputs)
+            result(last, outputs)
           } else last
         }
         case method        => {

diff --git a/src/test/scala/org/pmml4s/model/MiningModelTest.scala b/src/test/scala/org/pmml4s/model/MiningModelTest.scala
@@ -89,4 +89,88 @@ class MiningModelTest extends BaseModelTest {
     assert(r(2) === 0.5555556666666667)
     assert(r(3) === 3)
   }
+
+  test("a modelChain model with Output") {
+    val model = Model.fromString(
+      """
+        |<PMML xmlns="http://www.dmg.org/PMML-4_4" version="4.4">
+        |    <Header description="Test Rule Model"/>
+        |    <DataDictionary numberOfFields="4">
+        |        <DataField name="X" optype="categorical" dataType="string">
+        |            <Value value="A"/>
+        |            <Value value="B"/>
+        |        </DataField>
+        |        <DataField name="_predictedValue1" optype="continuous" dataType="double"/>
+        |        <DataField name="_predictedValue2" optype="continuous" dataType="double"/>
+        |        <DataField name="_predictedValueSum" optype="continuous" dataType="double"/>
+        |        <DataField name="_predictedValueFinal" optype="continuous" dataType="double"/>
+        |        <DataField name="predictedValueFinal" optype="continuous" dataType="double"/>
+        |    </DataDictionary>
+        |    <MiningModel functionName="regression">
+        |        <MiningSchema>
+        |            <MiningField name="X" usageType="active"/>
+        |            <MiningField name="predictedValueFinal" usageType="target"/>
+        |        </MiningSchema>
+        |        <Segmentation multipleModelMethod="modelChain">
+        |            <Segment id="0">
+        |                <True/>
+        |                <RuleSetModel functionName="regression" algorithmName="RuleSet">
+        |                    <MiningSchema>
+        |                        <MiningField name="X" usageType="active"/>
+        |                        <MiningField name="_predictedValue1" usageType="target"/>
+        |                    </MiningSchema>
+        |                    <Output>
+        |                        <OutputField name="_predictedValue1" optype="continuous" dataType="double" feature="predictedValue"/>
+        |                    </Output>
+        |                    <RuleSet>
+        |                        <RuleSelectionMethod criterion="firstHit"/>
+        |                        <SimpleRule score="100" confidence="1" weight="1">
+        |                            <SimplePredicate field="X" operator="equal" value="A" />
+        |                        </SimpleRule>
+        |                        <SimpleRule score="-100" confidence="1" weight="1">
+        |                            <True/>
+        |                        </SimpleRule>
+        |                    </RuleSet>
+        |                </RuleSetModel>
+        |            </Segment>
+        |            <Segment id="1">
+        |                <True/>
+        |                <RuleSetModel functionName="regression" algorithmName="RuleSet">
+        |                    <MiningSchema>
+        |                        <MiningField name="X" usageType="active"/>
+        |                        <MiningField name="_predictedValue1" usageType="active"/>
+        |                        <MiningField name="_predictedValue2" usageType="active"/>
+        |                    </MiningSchema>
+        |                    <Output>
+        |                        <OutputField name="_predictedValue2" optype="continuous" dataType="double" feature="predictedValue"/>
+        |                        <OutputField name="_predictedValueSum" optype="continuous" dataType="double" feature="transformedValue">
+        |                            <Apply function="+">
+        |                                <FieldRef field="_predictedValue2"/>
+        |                                <FieldRef field="_predictedValue1"/>
+        |                            </Apply>
+        |                        </OutputField>
+        |                    </Output>
+        |                    <RuleSet>
+        |                        <RuleSelectionMethod criterion="firstHit"/>
+        |                        <SimpleRule score="10" confidence="1" weight="1">
+        |                            <SimplePredicate field="X" operator="equal" value="A" />
+        |                        </SimpleRule>
+        |                        <SimpleRule score="-10" confidence="1" weight="1">
+        |                            <True/>
+        |                        </SimpleRule>
+        |                    </RuleSet>
+        |                </RuleSetModel>
+        |            </Segment>
+        |        </Segmentation>
+        |        <Output>
+        |            <OutputField name="predictedValueFinal" optype="continuous" dataType="double" feature="transformedValue">
+        |                <FieldRef field="_predictedValueSum"/>
+        |            </OutputField>
+        |        </Output>
+        |    </MiningModel>
+        |</PMML>
+        |""".stripMargin)
+    assert(model.predict(Map("X" -> "A"))("predictedValueFinal") === 110.0)
+    assert(model.predict(Map("X" -> "B"))("predictedValueFinal") === -110.0)
+  }
 }