【大数据处理技术】「#4」利用Spark预测回头客行为

预处理test.csv和train.csv数据集

预测回头客

启动hadoop

启动MySQL服务

启动spark-shell

• 下载安装MySQL的JDBC驱动

cd ~/下载/
unzip mysql-connector-java-5.1.40.zip -d /usr/local/spark/jars

• 正式启动spark-shell

cd /usr/local/spark
./bin/spark-shell --jars /usr/local/spark/jars/mysql-connector-java-5.1.40/mysql-connector-java-5.1.40-bin.jar --driver-class-path /usr/local/spark/jars/mysql-connector-java-5.1.40/mysql-connector-java-5.1.40-bin.jar

支持向量机SVM分类器预测回头客

• 导入需要的包

import org.apache.spark.SparkConf
import org.apache.spark.SparkContext
import org.apache.spark.mllib.regression.LabeledPoint
import org.apache.spark.mllib.linalg.{Vectors,Vector}
import org.apache.spark.mllib.classification.{SVMModel, SVMWithSGD}
import org.apache.spark.mllib.evaluation.BinaryClassificationMetrics
import java.util.Properties
import org.apache.spark.sql.types._
import org.apache.spark.sql.Row

• 读取训练数据

val train_data = sc.textFile("/dbtaobao/dataset/train_after.csv")
val test_data = sc.textFile("/dbtaobao/dataset/test_after.csv")

• 构建模型

# 构建模型
val train= train_data.map{line =>
  val parts = line.split(',')
  LabeledPoint(parts(4).toDouble,Vectors.dense(parts(1).toDouble,parts
(2).toDouble,parts(3).toDouble))
}
val test = test_data.map{line =>
  val parts = line.split(',')
  LabeledPoint(parts(4).toDouble,Vectors.dense(parts(1).toDouble,parts(2).toDouble,parts(3).toDouble))
}

# 训练模型
val numIterations = 1000
val model = SVMWithSGD.train(train, numIterations)

• 模型评估

model.clearThreshold()
val scoreAndLabels = test.map{point =>
  val score = model.predict(point.features)
  score+" "+point.label
}
scoreAndLabels.foreach(println)

# 设定阀值
model.setThreshold(0.0)
scoreAndLabels.foreach(println)

• 结果添加到MySQL中

model.clearThreshold()
val scoreAndLabels = test.map{point =>
  val score = model.predict(point.features)
  score+" "+point.label
}
//设置回头客数据
val rebuyRDD = scoreAndLabels.map(_.split(" "))
//下面要设置模式信息
val schema = StructType(List(StructField("score", StringType, true),StructField("label", StringType, true)))
//下面创建Row对象，每个Row对象都是rowRDD中的一行
val rowRDD = rebuyRDD.map(p => Row(p(0).trim, p(1).trim))
//建立起Row对象和模式之间的对应关系，也就是把数据和模式对应起来
val rebuyDF = spark.createDataFrame(rowRDD, schema)
//下面创建一个prop变量用来保存JDBC连接参数
val prop = new Properties()
prop.put("user", "root") //表示用户名是root
prop.put("password", "root") //表示密码是hadoop
prop.put("driver","com.mysql.jdbc.Driver") //表示驱动程序是com.mysql.jdbc.Driver
//下面就可以连接数据库，采用append模式，表示追加记录到数据库dbtaobao的rebuy表中
rebuyDF.write.mode("append").jdbc("jdbc:mysql://localhost:3306/dbtaobao", "dbtaobao.rebuy", prop)

model.clearThreshold()
val scoreAndLabels = test.map{point =>
  val score = model.predict(point.features)
  score+" "+point.label
}
val rebuyRDD = scoreAndLabels.map(_.split(" "))
val schema = StructType(List(StructField("score", StringType, true),StructField("label", StringType, true)))
val rowRDD = rebuyRDD.map(p => Row(p(0).trim, p(1).trim))
val rebuyDF = spark.createDataFrame(rowRDD, schema)
val prop = new Properties()
prop.put("user", "root")
prop.put("password", "123")
prop.put("driver","com.mysql.jdbc.Driver")
rebuyDF.write.mode("append").jdbc("jdbc:mysql://localhost:3306/dbtaobao", "dbtaobao.rebuy", prop)

• 因为是Arm架构，这里数据传输打MySQL上出了点问题