LinearRegressionWithSGD 训练模型

代码如下：

import org.apache.spark.mllib.classification.{SVMModel, SVMWithSGD}
import org.apache.spark.mllib.classification.{LogisticRegressionModel, LogisticRegressionWithLBFGS}
import org.apache.spark.mllib.evaluation.BinaryClassificationMetrics
import org.apache.spark.mllib.evaluation.MulticlassMetrics
import org.apache.spark.mllib.regression.LabeledPoint
import org.apache.spark.mllib.util.MLUtils


val updater = params.regType match {
  case NONE => new SimpleUpdater()
  case L1 => new L1Updater()
  case L2 => new SquaredL2Updater()
}

val algorithm = new LinearRegressionWithSGD()
algorithm.optimizer
  .setNumIterations(params.numIterations)
  .setStepSize(params.stepSize)
  .setUpdater(updater)
  .setRegParam(params.regParam)

val model = algorithm.run(training)

val prediction = model.predict(test.map(_.features))
val predictionAndLabel = prediction.html(test.map(_.label))

// Compute raw scores on the test set.
val scoreAndLabels = test.map { case LabeledPoint(label, features) =>
  val prediction = model.predict(features)
  (prediction, label)
}

// Get evaluation metrics.
val metrics = new BinaryClassificationMetrics(scoreAndLabels)
val auROC = metrics.areaUnderROC()
val auPR  = metrics.areaUnderPR()

metrics.precisionByThreshold().collect().foreach(println)
metrics.recallByThreshold().collect().foreach(println)
metrics.fMeasureByThreshold().collect().foreach(println)