【机器学习算法】线性回归

环境
　　spark-1.6
　　python3.5

一、线性回归

二、spark MLLIB案例

package com.wjy.df

import org.apache.spark.SparkConf
import org.apache.spark.SparkContext
import org.apache.spark.ml.regression.LinearRegressionModel
import org.apache.spark.mllib.linalg.Vectors
import org.apache.spark.mllib.regression.LabeledPoint
import org.apache.spark.mllib.regression.LinearRegressionWithSGD

/**
 * @author Administrator
 * 线性回归案例
 */
object LinearRegression {
  def main(args: Array[String]): Unit = {
    val conf = new SparkConf().setMaster("local").setAppName("LinearRegressionWithSGD")
    val sc = new SparkContext(conf)
    sc.setLogLevel("WARN")
    //读取样本数据 官方样例文件
    val data = sc.textFile("./data/lpsa.data")
    val examples = data.map{ line => 
      val parts = line.split(",")
      val y = parts(0)
      val xs = parts(1)
      LabeledPoint(parts(0).toDouble,Vectors.dense(parts(1).split(" ").map(_.toDouble)))
    }.cache()
    val train2TestData = examples.randomSplit(Array(0.8, 0.2), 1)
    
    val lsr = new LinearRegressionWithSGD()
    //让训练出来的模型有w0参数，就是有截距
    lsr.setIntercept(true)
    
    //在每次迭代的过程中 梯度下降算法的下降步长大小    0.1 0.2 0.3 0.4
    val stepSize = 1
    //设置步长
    lsr.optimizer.setStepSize(stepSize)
    
    /*
     *  迭代次数
     *  训练一个多元线性回归模型收敛（停止迭代）条件：
     *      1、error值小于用户指定的error值
     *      2、达到一定的迭代次数
     */
    val numIterations = 100
    //设置迭代次数
    lsr.optimizer.setNumIterations(numIterations)
    
    //每一次下山后，是否计算所有样本的误差值,1代表所有样本,默认就是1.0
    val miniBatchFraction = 1
    lsr.optimizer.setMiniBatchFraction(miniBatchFraction)
   
    //使用80%数据训练
    val model = lsr.run(train2TestData(0))
    println(model.weights)
    println(model.intercept)
    
    //使用20%数据对样本进行测试
    val prediction = model.predict(train2TestData(1).map(_.features))
    val predictionAndLabel = prediction.zip(train2TestData(1).map(_.label))
   
    //打印前20条数据
    val print_predict = predictionAndLabel.take(20)
    println("prediction" + "	" + "label")
    for(i <- 0 to print_predict.length-1){
      println(print_predict(i)._1+"	"+print_predict(i)._2)
    }
    
    //计算测试集平均误差
    val loss = predictionAndLabel.map{
      case(p,v) =>
        val err = p-v
        Math.abs(err)
    }.reduce(_+_)
    val error = loss / train2TestData(1).count
    println("Test RMSE = " + error)
    
    // 模型保存
    val ModelPath = "model"
    model.save(sc, ModelPath)
    //val sameModel = LinearRegressionModel.load(sc,ModelPath)
    
    sc.stop()
  }
}

结果：

[0.7296067051590363,0.23094665849041549,-0.1359562285885802,0.19004800201024025,0.2745413011485292,-0.31515879010131637,-0.04672248486523373,0.30883491480399367]
2.4764583366071977
prediction    label
1.749456972317874    0.3715636
1.8633537772490665    1.3480731
2.6325111666721064    1.7137979
2.3720657017536393    1.8484548
1.011168768081166    2.0476928
2.6730070097763634    2.5533438
3.011702574063707    2.7180005
2.2693119088733686    2.8063861
2.4416666667211793    2.8419982
3.1092859129401047    2.9626924
3.3123201208597277    3.2752562
2.6098535244026935    3.5876769
Test RMSE = 0.5736895056295152