finilize class LinearRegression plus tests (#11)

Co-authored-by: merzouk <merzoukoumedda@gmail.com>

src/main/scala/io/github/jsarni/CaraStage/CaraStageMapper.scala

@@ -17,6 +17,8 @@ trait CaraStageMapper {
     stageDescription.stageName match {
       case "LogisticRegression" =>
         LogisticRegression(stageDescription.params)
+      case "LinearRegression" =>
+        LinearRegression(stageDescription.params)
       case _ => throw
         new Exception(s"${stageDescription.stageName} is not a valid Cara Stage name. Please verify your Yaml File")
     }

src/main/scala/io/github/jsarni/CaraStage/ModelStage/LinearRegression.scala

@@ -0,0 +1,53 @@
+package io.github.jsarni.CaraStage.ModelStage
+import io.github.jsarni.CaraStage.Annotation.MapperConstructor
+import org.apache.spark.ml.PipelineStage
+import org.apache.spark.ml.regression.{LinearRegression => SparkLR}
+import scala.util.Try
+
+
+
+case class LinearRegression(MaxIter: Option[Int], RegParam: Option[Double], ElasticNetParam: Option[Double], LabelCol:Option[String], Loss: Option[String],
+                            FitIntercept: Option[Boolean], PredictionCol: Option[String], FeaturesCol: Option[String], Solver: Option[String],
+                            Standardization: Option[Boolean], Tol : Option[Double], WeightCol: Option[String])
+
+  extends CaraModel {
+
+  @MapperConstructor
+  def this(params: Map[String, String]) = {
+    this(
+      params.get("MaxIter").map(_.toInt),
+      params.get("RegParam").map(_.toDouble),
+      params.get("ElasticNetParam").map(_.toDouble),
+      params.get("LabelCol"),
+      params.get("Loss"),
+      params.get("FitIntercept").map(_.toBoolean),
+      params.get("PredictionCol"),
+      params.get("FeaturesCol"),
+      params.get("Solver"),
+      params.get("Standardization").map(_.toBoolean),
+      params.get("Tol").map(_.toDouble),
+      params.get("WeightCol")
+
+    )
+  }
+
+  override def build(): Try[PipelineStage] = Try {
+    val lr = new SparkLR()
+    val definedFields = this.getClass.getDeclaredFields.filter(f => f.get(this).asInstanceOf[Option[Any]].isDefined)
+    val names = definedFields.map(f => f.getName)
+    val values = definedFields.map(f => f.get(this))
+    val zipFields = names zip values
+    zipFields.map(f=>  getMethode(lr,f._2 match {case Some(s) => s },f._1).invoke(lr,(f._2 match {case Some(value) => value.asInstanceOf[f._2.type ] })))
+    lr
+
+  }
+}
+object LinearRegression
+{
+  def apply(params: Map[String, String]): LinearRegression = new LinearRegression(params)
+}
+
+
+
+
+

src/test/scala/io/github/jsarni/CaraStage/ModelStage/LinearRegressionTest.scala

@@ -0,0 +1,81 @@
+package io.github.jsarni.CaraStage.ModelStage
+import org.apache.spark.ml.regression.{LinearRegression => SparkLR}
+import io.github.jsarni.TestBase
+
+class LinearRegressionTest extends TestBase {
+  "build" should "Create an lr model and set all parameters with there args values or set default ones" in {
+    val params = Map(
+      "MaxIter" -> "10",
+      "RegParam" -> "0.3",
+      "ElasticNetParam" -> "0.1",
+      "FeaturesCol" -> "FeatureColname",
+      "FitIntercept" -> "True",
+      "PredictionCol" -> "Age",
+      "Standardization" -> "True",
+      "Tol" -> "0.13",
+      "WeightCol" -> "WeightColname",
+      "Loss" -> "huber",
+      "Solver" -> "normal",
+      "LabelCol" -> "LabelCol"
+    )
+    val lr = LinearRegression(params)
+    val lrWithTwoParams = new SparkLR()
+      .setRegParam(0.8)
+      .setStandardization(false)
+
+    val expectedResult = List(
+      new SparkLR()
+        .setMaxIter(10)
+        .setRegParam(0.3)
+        .setElasticNetParam(0.1)
+        .setFeaturesCol("FeatureColname")
+        .setFitIntercept(true)
+        .setPredictionCol("Age")
+        .setStandardization(true)
+        .setTol(0.13)
+        .setWeightCol("WeightColname")
+        .setLoss("huber")
+        .setSolver("normal")
+        .setLabelCol("LabelCol")
+    )
+    lr.build().isSuccess shouldBe true
+
+    val res = List(lr.build().get)
+    val resParameters = res.map(_.extractParamMap().toSeq.map(_.value))
+    val expectedParameters = expectedResult.map(_.extractParamMap().toSeq.map(_.value))
+
+    resParameters.head should contain theSameElementsAs expectedParameters.head
+
+    //    Test default values of unset params
+    lrWithTwoParams.getMaxIter shouldBe 100
+    lrWithTwoParams.getLoss shouldBe "squaredError"
+    lrWithTwoParams.getTol shouldBe 0.000001
+
+  }
+  "GetMethode" should "Return the appropriate methode by it's name" in {
+    val params = Map(
+      "MaxIter" -> "10",
+      "RegParam" -> "0.3",
+      "ElasticNetParam" -> "0.1",
+      "FeaturesCol" -> "FeatureColname",
+      "FitIntercept" -> "True",
+      "PredictionCol" -> "Age",
+      "Standardization" -> "True",
+      "Tol" -> "0.13",
+      "WeightCol" -> "WeightColname",
+      "Loss" -> "huber",
+      "Solver" -> "normal",
+      "LabelCol" -> "LabelCol"
+    )
+    val caraLr = LinearRegression(params)
+    val model =caraLr.build().get.asInstanceOf[SparkLR]
+
+    caraLr.getMethode(model,10,"MaxIter").getName shouldBe "setMaxIter"
+    caraLr.getMethode(model,0.0,"RegParam").getName shouldBe "setRegParam"
+    caraLr.getMethode(model, false ,"Standardization").getName shouldBe "setStandardization"
+
+  }
+
+
+
+}