【机器学习笔记三】回归分析 - 岭回归

package com.fredric.spark.lr

import org.apache.spark.mllib.linalg.Vectors
import org.apache.spark.mllib.regression.{RidgeRegressionWithSGD, LinearRegressionWithSGD, LabeledPoint}
import org.apache.spark.{SparkContext, SparkConf}

import scala.util.Random

/*-
 * 回归分析 岭回归
 * Fredric 2017
 */
object lrl2 {
  def main(args:Array[String]): Unit ={
    val conf = new SparkConf().setMaster("local").setAppName("Rr01")
    val sc   = new SparkContext(conf)

    //以二元线性回归y = 7*x1 + 5*x2 + 3为例，初始化数据
    val Array = new Array[LabeledPoint](500)

    for(t <- 1 to 500){

      val random = new Random();

      val x1 = random.nextInt(4) + random.nextDouble()
      val x2 = random.nextInt(3) + random.nextDouble()
      val y = 7 * x1 + 5 * x2 + (new Random()).nextDouble() + 3

      Array(t-1) = new LabeledPoint(y,  Vectors.dense(x1, x2, 1))
    }

    val data = sc.makeRDD(Array)
    
    //采用岭回归模型
    val model = new RidgeRegressionWithSGD()
    model.optimizer.setNumIterations(100);//迭代100次

    val res = model.run(data)

    //输出结果为 [7.016836776795048,5.045698277235657,3.3398441790354525]
    println(res.weights)
  }
}