购物平台商品实时推荐系统（五）

一、使用hive load hdfs上的清洗的数据。 4861 + 9666
--------------------------------------------- ----------------------
   1.动态添加表分区
      $hive> alter table eshop.logs add partition(year=2018,month=11,day=25,hour=12,minute=51);

   2.load数据到表中。
      $hive> load data inpath '/data/eshop/cleaned/2018/11/25/12/58' into table eshop.logs  partition(year=2018,month=11,day=25,hour=12,minute=51);

   3.查询topN
      $hive> select * from logs ;
      //倒排序topN
      $hive> select request,count(*) as c from logs where year = 2018 and month = 11 and day = 25 group by request order by c desc ;

   4.创建统计结果表
      $hive> create table stats(request string, c int) row format DELIMITED FIELDS TERMINATED BY ',' LINES TERMINATED BY '\n' STORED AS TEXTFILE;
      $hive> insert into stats select request,count(*) as c from logs where year = 2018 and month = 11 and day = 25 group by request order by c desc ;

    5.Mysql中创建表
        mysql> create table stats (id int primary key auto_increment,request varchar(200), c int);

   5.使用sqoop将hive中的数据导出到mysql
      $>sqoop export --connect jdbc:mysql://192.168.43.1:3306/eshop --driver com.mysql.jdbc.Driver --username mysql --password mysql --table stats --columns request,c --export-dir hdfs://s100/user/hive/warehouse/eshop.db/stats

   6.将以上2-5部写成脚本,使用cron进行调度.
      a.描述
         每天的凌晨2点整,统计昨天的日志。

      b.创建bash脚本
         1.创建准备脚本 -- 动态创建hivesql脚本文件[stat.ql]。
            [/usr/local/bin/prestats.sh]
            #!/bin/bash
            y=`date +%Y`
            m=`date +%m`
            d=`date -d "-0 day" +%d`

            m=$(( m+0 ))
            d=$(( d+0 ))
            # 删除之前的hql文件
            rm -rf stat.ql

            #添加分区
            echo "alter table eshop.logs add if not exists partition(year=${y},month=${m},day=${d},hour=9,minute=28);" >> stat.ql

            #加载数据放到分区
            echo "load data inpath 'hdfs://s201/user/centos/eshop/cleaned/${y}/${m}/${d}/9/28' into table eshop.logs  partition(year=${y},month=${m},day=${d},hour=9,minute=28);" >> stat.ql

            #统计数据，并将结果插入到stats表
            echo "insert into eshop.stats select request,count(*) as c from eshop.logs where year = ${y} and month = ${m} and day = ${d} and hour=9 and minute = 28 group by request order by c desc ;" >> stat.ql

         2.创建执行脚本
            [/usr/local/bin/exestats.sh]
            #!/bin/bash
            # 创建hive脚本文件
            ./prestats.sh

            #执行hive的ql脚本
            hive -f stat.ql

            #执行sqoop导出到mysql
            sqoop export --connect jdbc:mysql://192.168.43.1:3306/eshop  --username mysql --password mysql --table stats --columns request,c --export-dir /user/hive/warehouse/eshop.db/stats
            #sqoop export --connect jdbc:mysql://192.168.43.1:3306/eshop  --username mysql --password mysql --table stats --export-dir /user/hive/warehouse/eshop.db/stats

         3.修改所有权限
            $>sudo chmod a+x /usr/local/bin/prestats.sh
            $>sudo chmod a+x /usr/local/bin/exestats.sh

    7.编写java客户端进行以上步骤
        a.在hive主机上启动hiveserver2
            $> hiveserver2 &

        b.编写java客户端通过jdbc访问hive数据
            1)新建HiveClient模块，添加maven支持
                
                
                    4.0.0

                    com.test
                    HiveClient
                    1.0-SNAPSHOT

                    

                        
                            org.apache.hive
                            hive-jdbc
                            2.1.0
                        

                        
                            mysql
                            mysql-connector-java
                            5.1.17
                        

                    


                
            2)编写类进行查询和插入StatDao.java
                package com.test.hiveclient;

                import org.apache.hadoop.hbase.client.Result;

                import java.sql.*;
                import java.util.HashMap;
                import java.util.Map;

                public class StatDao {

                    private static Map map = new HashMap();

                    public static void main(String [] args)
                    {
                        try {
                            Class.forName("org.apache.hive.jdbc.HiveDriver");
                            //建立连接
                            Connection conn = DriverManager.getConnection("jdbc:hive2://192.168.43.131:10000/eshop","","");
                            System.out.println(conn);
                            PreparedStatement ppst = conn.prepareStatement("select * from stats");
                            ResultSet set = ppst.executeQuery();
                            while (set.next()) {
                                map.put(set.getString(1), set.getInt(2));
                                System.out.print(set.getString(1) + " : ");
                                System.out.print(set.getInt(2));
                                System.out.println();
                            }


                            getMysqlConn();


                        } catch (Exception e) {
                            e.printStackTrace();
                        }
                    }


                    public static void getMysqlConn()
                    {
                        try {
                            //加载类(加载驱动程序)
                            Class.forName("com.mysql.jdbc.Driver");
                            //数据库连接url
                            String url = "jdbc:mysql://192.168.43.1:3306/eshop" ;
                            //username
                            String user = "mysql";
                            //password
                            String pass = "mysql" ;

                            //得到连接
                            Connection conn = DriverManager.getConnection(url, user, pass);
                            //创建语句对象
                            Statement st = conn.createStatement();
                            for(String key : map.keySet() )
                            {
                                PreparedStatement pt =  conn.prepareStatement("insert into stats (request,count) values(?,?)");
                                pt.setString(1, key);
                                pt.setInt(2, map.get(key));
                                pt.executeUpdate();
                            }

                        } catch (Exception e) {
                            e.printStackTrace();
                        }
                    }
                }


二、JFreeChart生成统计图表
----------------------------------------------------------
    1.pom.xml
      
          jfree
          jfreechart
          1.0.13
      

    2.使用JFreechart生成图片
        package com.test.eshop.test;

        import org.jfree.chart.ChartFactory;
        import org.jfree.chart.ChartUtilities;
        import org.jfree.chart.JFreeChart;
        import org.jfree.chart.plot.PiePlot;
        import org.jfree.chart.plot.PiePlot3D;
        import org.jfree.data.general.DefaultPieDataset;
        import org.jfree.data.general.PieDataset;
        import org.junit.Test;

        import java.awt.*;
        import java.io.File;
        import java.io.IOException;

        /**
         * 测试饼图
         */
        public class TestJfreechart {

            @Test
            public void pie() throws Exception {
                File f = new File("d:/pie.png");

                //数据集
                DefaultPieDataset ds = new DefaultPieDataset();
                ds.setValue("HuaWei",3000);
                ds.setValue("Apple",5000);
                ds.setValue("Mi",1890);

                JFreeChart chart = ChartFactory.createPieChart("饼图演示", ds, false, false, false);

                Font font = new Font("宋体",Font.BOLD,15);
                chart.getTitle().setFont(font);
                //背景透明

                ((PiePlot)chart.getPlot()).setForegroundAlpha(0.2f);
                ((PiePlot)chart.getPlot()).setExplodePercent("Apple",0.1f);
                ((PiePlot)chart.getPlot()).setExplodePercent("HuaWei",0.2f);
                ((PiePlot)chart.getPlot()).setExplodePercent("Mi",0.3f);


                //创建3D饼图
                ChartUtilities.saveChartAsJPEG(f, chart,400,300);
            }
        }


三、引入Spark推荐系统
--------------------------------------------------------------------
    1.设计用户商品表 -- Mysql
        create table useritems(id int primary key auto_increment ,userid int, itemid int, score int , time timestamp);

    2.添加映射文件UserItem.hbm.xml

    3.Dao + Service

    4.controller

    5.spark部分
      a)通过sqoop到处mysql数据到hdfs
         $> sqoop import --connect jdbc:mysql://192.168.43.1:3306/eshop --driver com.mysql.jdbc.Driver --username mysql --password mysql --table useritems --columns userid,itemid,score -m 2 --target-dir /data/eshop/recommends --check-column id --incremental append --last-value 0

      b)启动spark集群

      c)启动spark-shell
         $> spark-shell --master spark://s100:7077

         #内置SparkSession--spark
         $scala>
            import org.apache.spark.ml.evaluation.RegressionEvaluator
            import org.apache.spark.ml.recommendation.ALS
            import spark.implicits._

            case class UserItem(userId: Int, itemId: Int, score : Int);

            def parseRating(str: String): UserItem = {
            val fields = str.split(",")
            UserItem(fields(0).toInt, fields(1).toInt, fields(2).toInt)
            }

            val useritems = spark.read.textFile("hdfs://s100/data/eshop/recommends").map(parseRating).toDF()

            //val test = spark.read.textFile("hdfs://s100/data/eshop/testdata.txt").map(parseRating).toDF()
            val Array(training, test) = useritems.randomSplit(Array(0.8, 0.2))

            val als = new ALS()
            .setMaxIter(5)
            .setRegParam(0.01)
            .setUserCol("userId")
            .setItemCol("itemId")
            .setRatingCol("score")
            val model = als.fit(training)

            val predictions = model.transform(test)

            val evaluator = new RegressionEvaluator().setMetricName("rmse").setLabelCol("score").setPredictionCol("prediction")
            val rmse = evaluator.evaluate(predictions)
            println(s"Root-mean-square error = $rmse")


            //保存ALS模型
            model.save("hdfs://s100/data/eshop/rec/model");
            spark.stop()


            //加载模型
            import org.apache.spark.ml.recommendation.ALSModel;
            val model = ALSModel.load("hdfs://s201/user/centos/eshop/rec/model");
            val predictions = model.transform(test)