Question

我想为我真正愚蠢的问题道歉，但我的线性回归问题。我很挣钱。你能帮我吗？

这是我的主要代码。我目前正在使用一些外部库来绘制数据。

import com.fundtrml.config.ConfigSetUp
import org.apache.spark.ml.feature.LabeledPoint
import org.apache.spark.ml.linalg.Vectors
import org.apache.spark.ml.regression.LinearRegression
import org.apache.spark.sql.SparkSession

object SimpleLinearRegression {

  def main(args: Array[String]): Unit = {
    ConfigSetUp.HadoopBinariesConfig();

    val ss = SparkSession.builder().appName("DataSet Test")
      .master("local[*]").getOrCreate()

    import ss.implicits._

    var listOfData = List(40, 41, 45, 43, 42, 60, 61, 59, 50, 49, 47, 39, 41, 37, 36, 34, 33, 37)
    val data =  listOfData  //(1 to 21 by 1)                      // create a collection of Doubles
      .map(n => (n, n))                               // make it pairs
      .map { case (label, features) =>
      LabeledPoint(label, Vectors.dense(features)) } // create labeled points of dense vectors
      .toDF                                           // make it a DataFrame
    var splittedData = data.randomSplit(Array(0.6,0.4))
    var trainingData = splittedData(0)
    var testSetData = splittedData(1)

    trainingData.show()
    val lr = new LinearRegression()
        .setMaxIter(10)
        .setRegParam(0.3)
        .setElasticNetParam(0.8)

    //train
    val model = lr.fit(trainingData)
    println(s"model.intercept: ${model.intercept}")
    println(s"model.coefficients : ${model.coefficients}")
    // Summarize the model over the training set and print out some metrics
    val trainingSummary = model.summary
    println(s"numIterations: ${trainingSummary.totalIterations}")
    println(s"objectiveHistory: [${trainingSummary.objectiveHistory.mkString(",")}]")
    trainingSummary.residuals.show()
    println(s"RMSE: ${trainingSummary.rootMeanSquaredError}")
    println(s"r2: ${trainingSummary.r2}")

    val predictions = model.transform(testSetData)
    predictions.show()

    //Display the data
    import com.quantifind.charts.Highcharts._
    regression(listOfData) //using this external library with embeded functionality about regression

    var currentPredictions = predictions.select("prediction").rdd.map(r => r(0)).collect.toList
    println(currentPredictions)
//    regression(currentPredictions.map(_.toString.toDouble))
  }
}

我的训练集如下，标签列 - 值，应该预测，特征值，应该用来做预测：

+-----+--------+
|label|features|
+-----+--------+
| 43.0|  [43.0]|
| 45.0|  [45.0]|
| 42.0|  [42.0]|
| 60.0|  [60.0]|
| 50.0|  [50.0]|
| 59.0|  [59.0]|
| 61.0|  [61.0]|
| 47.0|  [47.0]|
| 49.0|  [49.0]|
| 41.0|  [41.0]|
| 34.0|  [34.0]|
+-----+--------+

评估回归模型，我得到以下数据：

model.intercept: 1.7363839862169372
model.coefficients : [0.9640297102666925]
numIterations: 3
objectiveHistory: [0.5,0.406233822167566,0.031956224821402285]
RMSE: 0.29784178261548705
r2: 0.9987061382565019 --> Extremely High Close to 1

最后，我得到了以下预测：

+-----+--------+------------------+
|label|features|        prediction|
+-----+--------+------------------+
| 40.0|  [40.0]| 40.29757239688463|
| 41.0|  [41.0]|41.261602107151326|
| 39.0|  [39.0]|39.333542686617946|
| 36.0|  [36.0]|36.441453555817866|
| 37.0|  [37.0]| 37.40548326608456|
| 33.0|  [33.0]| 33.54936442501779|
| 37.0|  [37.0]| 37.40548326608456|
+-----+--------+------------------+

很容易看出预测不在同一条线上。它不可能位于直线上。 This is whole data set, plotted using the Scala Library- WISP

Predicted data

Expected result, but done with the WISP

Answer 1

您绘制的内容似乎是Y轴上的标签和X轴列表中的索引，而不是X轴上的特征值。

当绘制特征与预测时，预测确实在同一条线上。这是我这样做的结果：link

线性回归Apache Spark与Scala甚至不是直线

1 个答案: