spark数据清洗练习

通过编写Spark程序清洗酒店数据里的缺失数据、非法数据、重复数据

准备工作

1. 搭建 hadoop 伪分布或 hadoop 完全分布
2. 上传 hotal_data.csv 文件到 hadoop
3. idea 配置好 scala 环境

删除缺失值 >= 3 的数据

1. 读取 /hotel_data.csv
2. 删除缺失值 >= 3 的数据， 打印剔除的数量
3. 将清洗后的数据保存为/hotelsparktask1

import org.apache.spark.rdd.RDD
import org.apache.spark.{SparkConf, SparkContext}

object Demo01 {
  
  def main(args: Array[String]): Unit = {
    // System.setProperty("HADOOP_USER_NAME", "root")//解决保存文件权限不够的问题
    val config: SparkConf = new SparkConf().setMaster("local[1]").setAppName("1")
    val sc = new SparkContext(config)
    val hdfsUrl ="hdfs://192.168.226.129:9000"
    val filePath: String = hdfsUrl+"/file3_1/hotel_data.csv"
    val data: RDD[Array[String]] = sc.textFile(filePath).map(_.split(",")).cache()
    val total: Long = data.count()
    val dataDrop: RDD[Array[String]] = data.filter(_.count(_.equals("NULL")) <= 3)
    println("删除的数据条目有: " + (total - dataDrop.count()))
    dataDrop.map(_.mkString(",")).saveAsTextFile(hdfsUrl+ "/hotelsparktask1")
    sc.stop()
  }
}

删除星级、评论数、评分中任意字段为空的数据

1. 读取 /hotel_data.csv
2. 将字段{星级、评论数、评分}中任意字段为空的数据删除, 打印剔除的数量
3. 保存 /hotelsparktask2

import org.apache.spark.rdd.RDD
import org.apache.spark.{SparkConf, SparkContext}

object Demo02 {

  def main(args: Array[String]): Unit = {
    System.setProperty("HADOOP_USER_NAME", "root")
    val config: SparkConf = new SparkConf().setMaster("local[1]").setAppName("2")
    val sc = new SparkContext(config)
    val hdfsUrl ="hdfs://192.168.226.129:9000"
    val filePath: String = hdfsUrl+"/file3_1/hotel_data.csv"
    val data: RDD[Array[String]] = sc.textFile(filePath).map(_.split(",")).cache()
    val total: Long = data.count()
    val dataDrop: RDD[Array[String]] = data.filter {
      arr: Array[String] =>
        !(arr(6).equals("NULL") || arr(10).equals("NULL") || arr(11).equals("NULL"))
    }
    println("删除的数据条目有: " + (total - dataDrop.count()))
    dataDrop.map(_.mkString(",")).saveAsTextFile(hdfsUrl+ "/hotelsparktask2")
    sc.stop()
  }
}

删除非法数据

1. 读取第一题的 /hotelsparktask1
2. 剔除数据集中评分和星级字段的非法数据，合法数据是评分[0，5]的实数，星级是指星级字段内容中包含 NULL、二星、三星、四星、五星的数据
3. 剔除数据集中的重复数据
4. 分别打印 删除含有非法评分、星级以及重复的数据条目数
5. 保存 /hotelsparktask3

import org.apache.spark.rdd.RDD
import org.apache.spark.{SparkConf, SparkContext}

object Demo03 {

  def main(args: Array[String]): Unit = {
    System.setProperty("HADOOP_USER_NAME", "root")//解决权限问题
    val config: SparkConf = new SparkConf().setMaster(  "local[1]").setAppName("3")
    val sc = new SparkContext(config)
    val hdfsUrl ="hdfs://192.168.226.129:9000"
    val filePath: String = hdfsUrl+"/hotelsparktask1"
    val lines: RDD[String] = sc.textFile(filePath).cache()
    val data: RDD[Array[String]] = lines.map(_.split(","))

    val total: Long = data.count()
    val dataDrop: RDD[Array[String]] = data.filter {
      arr: Array[String] =>
        try {
          (arr(10).toDouble >= 0) && (arr(10).toDouble <= 5)
        } catch {
          case _: Exception => false
        }
    }

    val lab = Array("NULL", "一星", "二星", "三星", "四星", "五星")
    val dataDrop1: RDD[Array[String]] = data.filter { arr: Array[String] =>
      var flag = false
      for (elem <- lab) {
        if (arr(6).contains(elem)) {
          flag = true
        }
      }
      flag
    }

    val dataDrop2: RDD[String] = lines.distinct

    println("删除的非法评分数据条目有: " + (total - dataDrop.count()))
    println("删除的非法星级数据条目有: " + (total - dataDrop1.count()))
    println("删除重复数据条目有: " + (total - dataDrop2.count()))

    val wordsRdd: RDD[Array[String]] = lines.distinct.map(_.split(",")).filter {
      arr: Array[String] =>
        try {
          (arr(10).toDouble >= 0) && (arr(10).toDouble <= 5)
        } catch {
          case _: Exception => false
        }
    }.filter { arr: Array[String] =>
      var flag = false
      for (elem <- lab) {
        if (arr(6).contains(elem)) {
          flag = true
        }
      }
      flag
    }

    wordsRdd.map(_.mkString(","))
      .saveAsTextFile(hdfsUrl + "/hotelsparktask3")

    sc.stop()

  }
}

hotel_data.csv

下载数据：https://download.csdn.net/download/weixin_44018458/87437211