如何用SQL实现用户行为漏斗分析

「目录」

• 需求一：用户活跃主题
• 需求二：用户新增主题
• 需求三：用户留存主题
• 需求四：沉默用户数
• 需求五：本周回流用户数
• 需求六：流失用户数
• 需求七：最近连续3周活跃用户数
• 需求八：最近七天内连续三天活跃用户数
• 需求九：GMV（Gross Merchandise Volume）一段时间内的成交总额
• 需求十：转化率=新增用户/日活用户
• 需求十一：用户行为漏斗分析
• 需求十二：品牌复购率
• 需求十三：ADS层品牌复购率报表分析
• 需求十四：求每个等级的用户对应的复购率前十的商品排行

需求一：用户活跃主题

DWS层--（用户行为宽表层） 目标：统计当日、当周、当月活动的每个设备明细

1 每日活跃设备明细 dwd_start_log--->dws_uv_detail_day

--把相同的字段collect_set到一个数组, 按mid_id分组(便于后边统计)

collect_set将某字段的值进行去重汇总，产生array类型字段。如: concat_ws('|', collect_set(user_id)) user_id,

建分区表dws_uv_detail_day：partitioned by ('dt' string)

代码语言：javascript复制

drop table if exists dws_uv_detail_day;
create table dws_uv_detail_day( 
    `mid_id` string COMMENT '设备唯一标识',
    `user_id` string COMMENT '用户标识', 
    `version_code` string COMMENT '程序版本号', 
    `version_name` string COMMENT '程序版本名', 
`lang` string COMMENT '系统语言', 
`source` string COMMENT '渠道号', 
`os` string COMMENT '安卓系统版本', 
`area` string COMMENT '区域', 
`model` string COMMENT '手机型号', 
`brand` string COMMENT '手机品牌', 
`sdk_version` string COMMENT 'sdkVersion', 
`gmail` string COMMENT 'gmail', 
`height_width` string COMMENT '屏幕宽高',
`app_time` string COMMENT '客户端日志产生时的时间',
`network` string COMMENT '网络模式',
`lng` string COMMENT '经度',
`lat` string COMMENT '纬度'
) COMMENT '活跃用户按天明细'
PARTITIONED BY ( `dt` string)
stored as  parquet
location '/warehouse/gmall/dws/dws_uv_detail_day/'
;

数据导入

按周分区；过滤出一周内的数据；按设备id分组；===>count（*）得到最终结果；

partition(dt='2019-02-10') from dwd_start_log where dt='2019-02-10' group by mid_id ( mid_id设备唯一标示 )

以用户单日访问为key进行聚合，如果某个用户在一天中使用了两种操作系统、两个系统版本、多个地区，登录不同账号，只取其中之一

代码语言：javascript复制

hive (gmall)>
set hive.exec.dynamic.partition.mode=nonstrict;

insert overwrite table dws_uv_detail_day  partition(dt='2019-02-10')
select  
    mid_id,
    concat_ws('|', collect_set(user_id)) user_id,
    concat_ws('|', collect_set(version_code)) version_code,
    concat_ws('|', collect_set(version_name)) version_name,
    concat_ws('|', collect_set(lang))lang,
    concat_ws('|', collect_set(source)) source,
    concat_ws('|', collect_set(os)) os,
    concat_ws('|', collect_set(area)) area, 
    concat_ws('|', collect_set(model)) model,
    concat_ws('|', collect_set(brand)) brand,
    concat_ws('|', collect_set(sdk_version)) sdk_version,
    concat_ws('|', collect_set(gmail)) gmail,
    concat_ws('|', collect_set(height_width)) height_width,
    concat_ws('|', collect_set(app_time)) app_time,
    concat_ws('|', collect_set(network)) network,
    concat_ws('|', collect_set(lng)) lng,
    concat_ws('|', collect_set(lat)) lat
from dwd_start_log
where dt='2019-02-10'  
group by mid_id;

查询导入结果；

代码语言：javascript复制

hive (gmall)> select * from dws_uv_detail_day limit 1;

###最后count(*)即是每日活跃设备的个数；
hive (gmall)> select count(*) from dws_uv_detail_day;

2 每周（dws_uv_detail_wk）活跃设备明细 partition(wk_dt)

周一到周日concat(date_add(next_day('2019-02-10', 'MO'), -7), '_', date_add(next_day('2019-02-10', 'MO'), -1))即 2019-02-04_2019-02-10

创建分区表：partitioned by('wk_dt' string)

代码语言：javascript复制

hive (gmall)>
drop table if exists dws_uv_detail_wk;

create table dws_uv_detail_wk( 
    `mid_id` string COMMENT '设备唯一标识',
    `user_id` string COMMENT '用户标识', 
    `version_code` string COMMENT '程序版本号', 
    `version_name` string COMMENT '程序版本名', 
`lang` string COMMENT '系统语言', 
`source` string COMMENT '渠道号', 
`os` string COMMENT '安卓系统版本', 
`area` string COMMENT '区域', 
`model` string COMMENT '手机型号', 
`brand` string COMMENT '手机品牌', 
`sdk_version` string COMMENT 'sdkVersion', 
`gmail` string COMMENT 'gmail', 
`height_width` string COMMENT '屏幕宽高',
`app_time` string COMMENT '客户端日志产生时的时间',
`network` string COMMENT '网络模式',
`lng` string COMMENT '经度',
`lat` string COMMENT '纬度',
    `monday_date` string COMMENT '周一日期',
    `sunday_date` string COMMENT  '周日日期' 
) COMMENT '活跃用户按周明细'
PARTITIONED BY (`wk_dt` string)
stored as  parquet
location '/warehouse/gmall/dws/dws_uv_detail_wk/'
;

导入数据：以周为分区；过滤出一个月内的数据，按设备id分组；

周一：date_add(next_day('2019-05-16','MO'),-7);

周日：date_add(next_day('2019-05-16','MO'),-1);

周一---周日：concat(date_add(next_day('2019-05-16', 'MO'), -7), "_", date_add(next_day('2019-05-16', 'MO'), -1));

代码语言：javascript复制

insert overwrite table dws_uv_detail_wk partition(wk_dt)
select mid_id,
concat_ws('|', collect_set(user_id)) user_id,
concat_ws('|', collect_set(version_code)) version_code,
concat_ws('|', collect_set(version_name)) version_name,
concat_ws('|', collect_set(lang)) lang,
concat_ws('|', collect_set(source)) source,
concat_ws('|', collect_set(os)) os,
concat_ws('|', collect_set(area)) area, 
concat_ws('|', collect_set(model)) model,
concat_ws('|', collect_set(brand)) brand,
concat_ws('|', collect_set(sdk_version)) sdk_version,
concat_ws('|', collect_set(gmail)) gmail,
concat_ws('|', collect_set(height_width)) height_width,
concat_ws('|', collect_set(app_time)) app_time,
concat_ws('|', collect_set(network)) network,
concat_ws('|', collect_set(lng)) lng,
concat_ws('|', collect_set(lat)) lat,
date_add(next_day('2019-02-10', 'MO'), -7),
date_add(next_day('2019-02-10', 'MO'), -1),
concat(date_add(next_day('2019-02-10', 'MO'), -7), '_', date_add(next_day('2019-02-10', 'MO'), -1))
from dws_uv_detail_day
where dt >= date_add(next_day('2019-02-10', 'MO'), -7) and dt <= date_add(next_day('2019-02-10', 'MO'), -1)
group by mid_id;

查询导入结果

代码语言：javascript复制

hive (gmall)> select * from dws_uv_detail_wk limit 1;
hive (gmall)> select count(*) from dws_uv_detail_wk;

3 每月活跃设备明细 dws_uv_detail_mn partition(mn) - 把每日的数据插入进去

DWS层创建分区表 partitioned by(mn string)

代码语言：javascript复制

hive (gmall)>
drop table if exists dws_uv_detail_mn;

create  external table dws_uv_detail_mn( 
    `mid_id` string COMMENT '设备唯一标识',
    `user_id` string COMMENT '用户标识', 
    `version_code` string COMMENT '程序版本号', 
    `version_name` string COMMENT '程序版本名', 
`lang` string COMMENT '系统语言', 
`source` string COMMENT '渠道号', 
`os` string COMMENT '安卓系统版本', 
`area` string COMMENT '区域', 
`model` string COMMENT '手机型号', 
`brand` string COMMENT '手机品牌', 
`sdk_version` string COMMENT 'sdkVersion', 
`gmail` string COMMENT 'gmail', 
`height_width` string COMMENT '屏幕宽高',
`app_time` string COMMENT '客户端日志产生时的时间',
`network` string COMMENT '网络模式',
`lng` string COMMENT '经度',
`lat` string COMMENT '纬度'
) COMMENT '活跃用户按月明细'
PARTITIONED BY (`mn` string)
stored as  parquet
location '/warehouse/gmall/dws/dws_uv_detail_mn/'
;

数据导入 按月分区；过滤出一个月内的数据，按照设备id分组；

data_format('2019-03-10', 'yyyy-MM') ---> 2019-03

where date_format('dt', 'yyyy-MM') = date_format('2019-02-10', 'yyyy-MM') group by mid_id;

代码语言：javascript复制

hive (gmall)>
set hive.exec.dynamic.partition.mode=nonstrict;

insert  overwrite table dws_uv_detail_mn  partition(mn)
select  
    mid_id,
    concat_ws('|', collect_set(user_id)) user_id,
    concat_ws('|', collect_set(version_code)) version_code,
    concat_ws('|', collect_set(version_name)) version_name,
    concat_ws('|', collect_set(lang)) lang,
    concat_ws('|', collect_set(source)) source,
    concat_ws('|', collect_set(os)) os,
    concat_ws('|', collect_set(area)) area, 
    concat_ws('|', collect_set(model)) model,
    concat_ws('|', collect_set(brand)) brand,
    concat_ws('|', collect_set(sdk_version)) sdk_version,
    concat_ws('|', collect_set(gmail)) gmail,
    concat_ws('|', collect_set(height_width)) height_width,
    concat_ws('|', collect_set(app_time)) app_time,
    concat_ws('|', collect_set(network)) network,
    concat_ws('|', collect_set(lng)) lng,
    concat_ws('|', collect_set(lat)) lat,
    date_format('2019-02-10','yyyy-MM')
from dws_uv_detail_day
where date_format(dt,'yyyy-MM') = date_format('2019-02-10','yyyy-MM')   
group by mid_id;

查询导入结果

代码语言：javascript复制

hive (gmall)> select * from dws_uv_detail_mn limit 1;
hive (gmall)> select count(*) from dws_uv_detail_mn ;

DWS层加载数据脚本

在hadoop101的/home/kris/bin目录下创建脚本

[kris@hadoop101 bin]$ vim dws.sh

代码语言：javascript复制

#!/bin/bash

# 定义变量方便修改
APP=gmall
hive=/opt/module/hive/bin/hive

# 如果是输入的日期按照取输入日期；如果没输入日期取当前时间的前一天
if [ -n "$1" ] ;then
    do_date=$1
else 
    do_date=`date -d "-1 day"  %F`  
fi 


sql="
  set hive.exec.dynamic.partition.mode=nonstrict;

  insert overwrite table "$APP".dws_uv_detail_day partition(dt='$do_date')
  select  
    mid_id,
    concat_ws('|', collect_set(user_id)) user_id,
    concat_ws('|', collect_set(version_code)) version_code,
    concat_ws('|', collect_set(version_name)) version_name,
    concat_ws('|', collect_set(lang)) lang,
    concat_ws('|', collect_set(source)) source,
    concat_ws('|', collect_set(os)) os,
    concat_ws('|', collect_set(area)) area, 
    concat_ws('|', collect_set(model)) model,
    concat_ws('|', collect_set(brand)) brand,
    concat_ws('|', collect_set(sdk_version)) sdk_version,
    concat_ws('|', collect_set(gmail)) gmail,
    concat_ws('|', collect_set(height_width)) height_width,
    concat_ws('|', collect_set(app_time)) app_time,
    concat_ws('|', collect_set(network)) network,
    concat_ws('|', collect_set(lng)) lng,
    concat_ws('|', collect_set(lat)) lat
  from "$APP".dwd_start_log
  where dt='$do_date'  
  group by mid_id;


  insert  overwrite table "$APP".dws_uv_detail_wk partition(wk_dt)
  select  
    mid_id,
    concat_ws('|', collect_set(user_id)) user_id,
    concat_ws('|', collect_set(version_code)) version_code,
    concat_ws('|', collect_set(version_name)) version_name,
    concat_ws('|', collect_set(lang)) lang,
    concat_ws('|', collect_set(source)) source,
    concat_ws('|', collect_set(os)) os,
    concat_ws('|', collect_set(area)) area, 
    concat_ws('|', collect_set(model)) model,
    concat_ws('|', collect_set(brand)) brand,
    concat_ws('|', collect_set(sdk_version)) sdk_version,
    concat_ws('|', collect_set(gmail)) gmail,
    concat_ws('|', collect_set(height_width)) height_width,
    concat_ws('|', collect_set(app_time)) app_time,
    concat_ws('|', collect_set(network)) network,
    concat_ws('|', collect_set(lng)) lng,
    concat_ws('|', collect_set(lat)) lat,
    date_add(next_day('$do_date','MO'),-7),
    date_add(next_day('$do_date','SU'),-7),
    concat(date_add( next_day('$do_date','MO'),-7), '_' , date_add(next_day('$do_date','MO'),-1) 
  )
  from "$APP".dws_uv_detail_day 
  where dt>=date_add(next_day('$do_date','MO'),-7) and dt<=date_add(next_day('$do_date','MO'),-1) 
  group by mid_id; 


  insert overwrite table "$APP".dws_uv_detail_mn partition(mn)
  select  
    mid_id,
    concat_ws('|', collect_set(user_id)) user_id,
    concat_ws('|', collect_set(version_code)) version_code,
    concat_ws('|', collect_set(version_name)) version_name,
    concat_ws('|', collect_set(lang))lang,
    concat_ws('|', collect_set(source)) source,
    concat_ws('|', collect_set(os)) os,
    concat_ws('|', collect_set(area)) area, 
    concat_ws('|', collect_set(model)) model,
    concat_ws('|', collect_set(brand)) brand,
    concat_ws('|', collect_set(sdk_version)) sdk_version,
    concat_ws('|', collect_set(gmail)) gmail,
    concat_ws('|', collect_set(height_width)) height_width,
    concat_ws('|', collect_set(app_time)) app_time,
    concat_ws('|', collect_set(network)) network,
    concat_ws('|', collect_set(lng)) lng,
    concat_ws('|', collect_set(lat)) lat,
    date_format('$do_date','yyyy-MM')
  from "$APP".dws_uv_detail_day
  where date_format(dt,'yyyy-MM') = date_format('$do_date','yyyy-MM')   
  group by mid_id;
"

$hive -e "$sql"

增加脚本执行权限 chmod 777 dws.sh

脚本使用[kris@hadoop101 module]$ dws.sh 2019-02-11

查询结果

代码语言：javascript复制

hive (gmall)> select count(*) from dws_uv_detail_day;
hive (gmall)> select count(*) from dws_uv_detail_wk;
hive (gmall)> select count(*) from dws_uv_detail_mn ;

脚本执行时间；企业开发中一般在每日凌晨30分~1点

ADS层 目标：当日、当周、当月活跃设备数 使用 day_count表 join wk_count join mn_count , 把3张表连接一起

建表ads_uv_count表：

字段有day_count、wk_count、mn_count is_weekend if(date_add(next_day('2019-02-10', 'MO'), -1) = '2019-02-10', 'Y', 'N') is_monthend if(last_day('2019-02-10') = '2019-02-10', 'Y', 'N')

代码语言：javascript复制

drop table if exists ads_uv_count;
create external table ads_uv_count(
`dt` string comment '统计日期',
`day_count` bigint comment '当日用户量',
`wk_count` bigint comment '当周用户量',
`mn_count` bigint comment '当月用户量',
`is_weekend` string comment 'Y,N是否是周末,用于得到本周最终结果',
`is_monthend` string comment 'Y,N是否是月末,用于得到本月最终结果'
) comment '每日活跃用户数量'
stored as parquet
location '/warehouse/gmall/ads/ads_uv_count/';

导入数据：

代码语言：javascript复制

hive (gmall)>
insert  overwrite table ads_uv_count 
select  
  '2019-02-10' dt,
   daycount.ct,
   wkcount.ct,
   mncount.ct,
   if(date_add(next_day('2019-02-10','MO'),-1)='2019-02-10','Y','N') ,
   if(last_day('2019-02-10')='2019-02-10','Y','N') 
from 
(
   select  
      '2019-02-10' dt,
       count(*) ct
   from dws_uv_detail_day
   where dt='2019-02-10'  
)daycount   join 
( 
   select  
     '2019-02-10' dt,
     count (*) ct
   from dws_uv_detail_wk
   where wk_dt=concat(date_add(next_day('2019-02-10','MO'),-7),'_' ,date_add(next_day('2019-02-10','MO'),-1) )
)  wkcount  on daycount.dt=wkcount.dt
join 
( 
   select  
     '2019-02-10' dt,
     count (*) ct
   from dws_uv_detail_mn
   where mn=date_format('2019-02-10','yyyy-MM')  
)mncount on daycount.dt=mncount.dt
;

查询导入结果

hive (gmall)> select * from ads_uv_count ;

ADS层加载数据脚本

1）在hadoop101的/home/kris/bin目录下创建脚本

[kris@hadoop101 bin]$ vim ads.sh

代码语言：javascript复制

#!/bin/bash

# 定义变量方便修改
APP=gmall
hive=/opt/module/hive/bin/hive

# 如果是输入的日期按照取输入日期；如果没输入日期取当前时间的前一天
if [ -n "$1" ] ;then
    do_date=$1
else 
    do_date=`date -d "-1 day"  %F`  
fi 

sql="
  set hive.exec.dynamic.partition.mode=nonstrict;

insert into table "$APP".ads_uv_count 
select  
  '$do_date' dt,
   daycount.ct,
   wkcount.ct,
   mncount.ct,
   if(date_add(next_day('$do_date','MO'),-1)='$do_date','Y','N') ,
   if(last_day('$do_date')='$do_date','Y','N') 
from 
(
   select  
      '$do_date' dt,
       count(*) ct
   from "$APP".dws_uv_detail_day
   where dt='$do_date'  
)daycount   join 
( 
   select  
     '$do_date' dt,
     count (*) ct
   from "$APP".dws_uv_detail_wk
   where wk_dt=concat(date_add(next_day('$do_date','MO'),-7),'_' ,date_add(next_day('$do_date','MO'),-1) )
)  wkcount  on daycount.dt=wkcount.dt
join 
( 
   select  
     '$do_date' dt,
     count (*) ct
   from "$APP".dws_uv_detail_mn
   where mn=date_format('$do_date','yyyy-MM')  
)mncount on daycount.dt=mncount.dt;
"

$hive -e "$sql"

增加脚本执行权限 chmod 777 ads.sh

脚本使用 ads.sh 2019-02-11

查询导入结果 hive (gmall)> select * from ads_uv_count ;

需求二：用户新增主题

首次联网使用应用的用户。如果一个用户首次打开某APP，那这个用户定义为新增用户；卸载再安装的设备，不会被算作一次新增。新增用户包括日新增用户、周新增用户、月新增用户。

每日新增(老用户不算,之前没登陆过，今天是第一次登陆)设备--没有分区 -->以往的新增库里边没有他，但他今天活跃了即新增加的用户；

1 DWS层（每日新增设备明细表） 创建每日新增设备明细表：dws_new_mid_day

代码语言：javascript复制

hive (gmall)>
drop table if exists  dws_new_mid_day;
create  table  dws_new_mid_day
(
    `mid_id` string COMMENT '设备唯一标识',
    `user_id` string COMMENT '用户标识', 
    `version_code` string COMMENT '程序版本号', 
    `version_name` string COMMENT '程序版本名', 
`lang` string COMMENT '系统语言', 
`source` string COMMENT '渠道号', 
`os` string COMMENT '安卓系统版本', 
`area` string COMMENT '区域', 
`model` string COMMENT '手机型号', 
`brand` string COMMENT '手机品牌', 
`sdk_version` string COMMENT 'sdkVersion', 
`gmail` string COMMENT 'gmail', 
`height_width` string COMMENT '屏幕宽高',
`app_time` string COMMENT '客户端日志产生时的时间',
`network` string COMMENT '网络模式',
`lng` string COMMENT '经度',
`lat` string COMMENT '纬度',
    `create_date`  string  comment '创建时间' 
)  COMMENT '每日新增设备信息'
stored as  parquet
location '/warehouse/gmall/dws/dws_new_mid_day/';

dws_uv_detail_day(每日活跃设备明细) left join dws_new_mid_day nm(以往的新增用户表, 新建字段create_time2019-02-10) nm.mid_id is null;

导入数据

用每日活跃用户表 left join 每日新增设备表，关联的条件是mid_id相等。如果是每日新增的设备，则在每日新增设备表中为null。

from dws_uv_detail_day ud left join dws_new_mid_day nm on ud.mid_id=nm.mid_id

where ud.dt='2019-02-10' and nm.mid_id is null;

代码语言：javascript复制

hive (gmall)>
insert into table dws_new_mid_day  
select  
    ud.mid_id,
    ud.user_id , 
    ud.version_code , 
    ud.version_name , 
    ud.lang , 
    ud.source, 
    ud.os, 
    ud.area, 
    ud.model, 
    ud.brand, 
    ud.sdk_version, 
    ud.gmail, 
    ud.height_width,
    ud.app_time,
    ud.network,
    ud.lng,
    ud.lat,
    '2019-02-10'
from dws_uv_detail_day ud left join dws_new_mid_day nm on ud.mid_id=nm.mid_id
where ud.dt='2019-02-10' and nm.mid_id is null;

查询导入数据

hive (gmall)> select count(*) from dws_new_mid_day ;

2 ADS层（每日新增设备表） 创建每日新增设备表ads_new_mid_count

代码语言：javascript复制

hive (gmall)>
drop table if exists  `ads_new_mid_count`;
create  table  `ads_new_mid_count`
(
    `create_date`     string  comment '创建时间' ,
    `new_mid_count`   BIGINT comment '新增设备数量' 
)  COMMENT '每日新增设备信息数量'
row format delimited  fields terminated by 't' 
location '/warehouse/gmall/ads/ads_new_mid_count/';

导入数据 count（*） dws_new_mid_day表即可

加了create_date就必须group by create_time，否则报错：not in GROUP BY key 'create_date'

代码语言：javascript复制

hive (gmall)>
insert into table ads_new_mid_count 
select create_date , count(*)  from dws_new_mid_day
where create_date='2019-02-10'
group by create_date ;

查询导入数据

hive (gmall)> select * from ads_new_mid_count;

扩展每月新增：

代码语言：javascript复制

--每月新增
drop table if exists dws_new_mid_mn;
create table dws_new_mid_mn(
    `mid_id` string COMMENT '设备唯一标识',
    `user_id` string COMMENT '用户标识', 
    `version_code` string COMMENT '程序版本号', 
    `version_name` string COMMENT '程序版本名', 
    `lang` string COMMENT '系统语言', 
    `source` string COMMENT '渠道号', 
    `os` string COMMENT '安卓系统版本', 
    `area` string COMMENT '区域', 
    `model` string COMMENT '手机型号', 
    `brand` string COMMENT '手机品牌', 
    `sdk_version` string COMMENT 'sdkVersion', 
    `gmail` string COMMENT 'gmail', 
    `height_width` string COMMENT '屏幕宽高',
    `app_time` string COMMENT '客户端日志产生时的时间',
    `network` string COMMENT '网络模式',
    `lng` string COMMENT '经度',
    `lat` string COMMENT '纬度'
)comment "每月新增明细"
partitioned by(mn string)
stored as parquet
location "/warehouse/gmall/dws/dws_new_mid_mn";

insert overwrite table dws_new_mid_mn partition(mn)
select
    um.mid_id,
    um.user_id , 
    um.version_code , 
    um.version_name , 
    um.lang , 
    um.source, 
    um.os, 
    um.area, 
    um.model, 
    um.brand, 
    um.sdk_version, 
    um.gmail, 
    um.height_width,
    um.app_time,
    um.network,
    um.lng,
    um.lat,
    date_format('2019-02-10', 'yyyy-MM')
from dws_uv_detail_mn um left join dws_new_mid_mn nm on um.mid_id = nm.mid_id
where um.mn =date_format('2019-02-10', 'yyyy-MM') and nm.mid_id = null; ----为什么加上它就是空的？？查不到数据了呢
--##注意这里不能写出date_format(um.mn, 'yyyy-MM') =date_format('2019-02-10', 'yyyy-MM') 
    |

需求三：用户留存主题

如果不考虑2019-02-11和2019-02-12的新增用户：2019-02-10新增100人，一天后它的留存率是30%，2天12号它的留存率是25%，3天后留存率32%；

站在2019-02-12号看02-11的留存率：新增200人，12号的留存率是20%；

站在2019-02-13号看02-12的留存率：新增100人，13号即一天后留存率是25%；

用户留存率的分析：昨日的新增且今天是活跃的 / 昨日的新增用户量

如今天11日，要统计10日的 用户留存率---->10日的新设备且是11日活跃的 / 10日新增设备 　　分母：10日的新增设备(每日活跃 left join 以往新增设备表(nm) nm.mid_id is null ) 　　分子：每日活跃表(ud) join 每日新增表(nm) where ud.dt='今天' and nm.create_date = '昨天'

① DWS层（每日留存用户明细表dws_user_retention_day） 用户1天留存的分析：===>>

留存用户=前一天新增 join 今天活跃

代码语言：javascript复制

   用户留存率=留存用户/前一天新增

创建表：dws_user_retention_day

代码语言：javascript复制

hive (gmall)>
drop table if exists  `dws_user_retention_day`;
create  table  `dws_user_retention_day` 
(
    `mid_id` string COMMENT '设备唯一标识',
    `user_id` string COMMENT '用户标识', 
    `version_code` string COMMENT '程序版本号', 
    `version_name` string COMMENT '程序版本名', 
`lang` string COMMENT '系统语言', 
`source` string COMMENT '渠道号', 
`os` string COMMENT '安卓系统版本', 
`area` string COMMENT '区域', 
`model` string COMMENT '手机型号', 
`brand` string COMMENT '手机品牌', 
`sdk_version` string COMMENT 'sdkVersion', 
`gmail` string COMMENT 'gmail', 
`height_width` string COMMENT '屏幕宽高',
`app_time` string COMMENT '客户端日志产生时的时间',
`network` string COMMENT '网络模式',
`lng` string COMMENT '经度',
`lat` string COMMENT '纬度',
   `create_date`       string  comment '设备新增时间',
   `retention_day`     int comment '截止当前日期留存天数'
)  COMMENT '每日用户留存情况'
PARTITIONED BY ( `dt` string)
stored as  parquet
location '/warehouse/gmall/dws/dws_user_retention_day/'
;

导入数据（每天计算前1天的新用户访问留存明细）

from dws_uv_detail_day每日活跃设备 ud join dws_new_mid_day每日新增设备 nm on ud.mid_id =nm.mid_id where ud.dt='2019-02-11' and nm.create_date=date_add('2019-02-11',-1);

代码语言：javascript复制

hive (gmall)>
insert  overwrite table dws_user_retention_day  partition(dt="2019-02-11")
select  
    nm.mid_id,
    nm.user_id , 
    nm.version_code , 
    nm.version_name , 
    nm.lang , 
    nm.source, 
    nm.os, 
    nm.area, 
    nm.model, 
    nm.brand, 
    nm.sdk_version, 
    nm.gmail, 
    nm.height_width,
    nm.app_time,
    nm.network,
    nm.lng,
    nm.lat,
nm.create_date,
1 retention_day 
from  dws_uv_detail_day ud join dws_new_mid_day nm   on ud.mid_id =nm.mid_id 
where ud.dt='2019-02-11' and nm.create_date=date_add('2019-02-11',-1);

查询导入数据（每天计算前1天的新用户访问留存明细）

hive (gmall)> select count(*) from dws_user_retention_day;

② DWS层（1,2,3,n天留存用户明细表）直接插入数据：dws_user_retention_day 用union all连接起来，汇总到一个表中；1）直接导入数据（每天计算前1,2,3，n天的新用户访问留存明细） 直接改变这个即可以，date_add('2019-02-11',-3); -1是一天的留存率；-2是两天的留存率、-3是三天的留存率

代码语言：javascript复制

hive (gmall)>
insert  overwrite table dws_user_retention_day  partition(dt="2019-02-11")
select  
    nm.mid_id,
    nm.user_id , 
    nm.version_code , 
    nm.version_name , 
    nm.lang , 
    nm.source, 
    nm.os, 
    nm.area, 
    nm.model, 
    nm.brand, 
    nm.sdk_version, 
    nm.gmail, 
    nm.height_width,
    nm.app_time,
    nm.network,
    nm.lng,
    nm.lat,
    nm.create_date,
    1 retention_day 
from dws_uv_detail_day ud join dws_new_mid_day nm  on ud.mid_id =nm.mid_id 
where ud.dt='2019-02-11' and nm.create_date=date_add('2019-02-11',-1)

union all
select  
    nm.mid_id,
    nm.user_id , 
    nm.version_code , 
    nm.version_name , 
    nm.lang , 
    nm.source, 
    nm.os, 
    nm.area, 
    nm.model, 
    nm.brand, 
    nm.sdk_version, 
    nm.gmail, 
    nm.height_width,
    nm.app_time,
    nm.network,
    nm.lng,
    nm.lat,
    nm.create_date,
    2 retention_day 
from  dws_uv_detail_day ud join dws_new_mid_day nm   on ud.mid_id =nm.mid_id 
where ud.dt='2019-02-11' and nm.create_date=date_add('2019-02-11',-2)

union all
select  
    nm.mid_id,
    nm.user_id , 
    nm.version_code , 
    nm.version_name , 
    nm.lang , 
    nm.source, 
    nm.os, 
    nm.area, 
    nm.model, 
    nm.brand, 
    nm.sdk_version, 
    nm.gmail, 
    nm.height_width,
    nm.app_time,
    nm.network,
    nm.lng,
    nm.lat,
    nm.create_date,
    3 retention_day 
from  dws_uv_detail_day ud join dws_new_mid_day nm   on ud.mid_id =nm.mid_id 
where ud.dt='2019-02-11' and nm.create_date=date_add('2019-02-11',-3);

2）查询导入数据（每天计算前1,2,3天的新用户访问留存明细）

hive (gmall)> select retention_day , count(*) from dws_user_retention_day group by retention_day;

③ ADS层 留存用户数 ads_user_retention_day_count 直接count( * )即可 1）创建 ads_user_retention_day_count表：

代码语言：javascript复制

hive (gmall)>
drop table if exists  `ads_user_retention_day_count`;
create  table  `ads_user_retention_day_count` 
(
   `create_date`       string  comment '设备新增日期',
   `retention_day`     int comment '截止当前日期留存天数',
   `retention_count`    bigint comment  '留存数量'
)  COMMENT '每日用户留存情况'
stored as  parquet
location '/warehouse/gmall/ads/ads_user_retention_day_count/';

导入数据 按创建日期create_date 和 留存天数retention_day进行分组group by；

代码语言：javascript复制

hive (gmall)>
insert into table ads_user_retention_day_count 
select   
    create_date, 
    retention_day, 
    count(*) retention_count  
from dws_user_retention_day
where dt='2019-02-11' 
group by create_date,retention_day;

查询导入数据

hive (gmall)> select * from ads_user_retention_day_count;

---> 2019-02-10 1 112

④ 留存用户比率 retention_count / new_mid_count 即留存个数 / 新增个数 创建表 ads_user_retention_day_rate

代码语言：javascript复制

hive (gmall)>
drop table if exists  `ads_user_retention_day_rate`;
create  table  `ads_user_retention_day_rate` 
(
     `stat_date`          string comment '统计日期',
     `create_date`       string  comment '设备新增日期',
     `retention_day`     int comment '截止当前日期留存天数',
     `retention_count`    bigint comment  '留存数量',
     `new_mid_count`     string  comment '当日设备新增数量',
     `retention_ratio`   decimal(10,2) comment '留存率'
)  COMMENT '每日用户留存情况'
stored as  parquet
location '/warehouse/gmall/ads/ads_user_retention_day_rate/';

导入数据

代码语言：javascript复制

join ads_new_mid_countt --->每日新增设备表

代码语言：javascript复制

hive (gmall)>
insert into table ads_user_retention_day_rate
select 
    '2019-02-11' , 
    ur.create_date,
    ur.retention_day, 
    ur.retention_count , 
    nc.new_mid_count,
    ur.retention_count/nc.new_mid_count*100
from 
(
    select   
        create_date, 
        retention_day, 
        count(*) retention_count  
    from `dws_user_retention_day` 
    where dt='2019-02-11' 
    group by create_date,retention_day
)  ur join ads_new_mid_count nc on nc.create_date=ur.create_date;

查询导入数据

代码语言：javascript复制

hive (gmall)>select * from ads_user_retention_day_rate;

2019-02-11 2019-02-10 1 112 442 25.34

需求四：沉默用户数

沉默用户：指的是只在安装当天启动过，且启动时间是在一周前

使用日活明细表dws_uv_detail_day作为DWS层数据

建表语句

代码语言：javascript复制

hive (gmall)>
drop table if exists ads_slient_count;
create external table ads_slient_count( 
    `dt` string COMMENT '统计日期',
    `slient_count` bigint COMMENT '沉默设备数'
) 
row format delimited fields terminated by 't'
location '/warehouse/gmall/ads/ads_slient_count';

导入数据

代码语言：javascript复制

hive (gmall)>
insert into table ads_slient_count
select 
    '2019-02-20' dt,
    count(*) slient_count
from 
(
    select mid_id
    from dws_uv_detail_day
    where dt<='2019-02-20'
    group by mid_id
    having count(*)=1 and min(dt)<date_add('2019-02-20',-7)
) t1;

需求五：本周回流用户数

本周回流=本周活跃-本周新增-上周活跃

使用日活明细表dws_uv_detail_day作为DWS层数据

本周回流(上周以前活跃过，上周没活跃，本周活跃了)=本周活跃-本周新增-上周活跃 本周回流=本周活跃left join 本周新增 left join 上周活跃，且本周新增id为null，上周活跃id为null；

建表：

代码语言：javascript复制

hive (gmall)>
drop table if exists ads_back_count;
create external table ads_back_count( 
    `dt` string COMMENT '统计日期',
    `wk_dt` string COMMENT '统计日期所在周',
    `wastage_count` bigint COMMENT '回流设备数'
) 
row format delimited fields terminated by 't'
location '/warehouse/gmall/ads/ads_back_count';

导入数据

代码语言：javascript复制

hive (gmall)> 
insert into table ads_back_count
select 
   '2019-02-20' dt,
   concat(date_add(next_day('2019-02-20','MO'),-7),'_',date_add(next_day('2019-02-20','MO'),-1)) wk_dt,
   count(*)
from 
(
    select t1.mid_id
    from 
    (
        select    mid_id
        from dws_uv_detail_wk
        where wk_dt=concat(date_add(next_day('2019-02-20','MO'),-7),'_',date_add(next_day('2019-02-20','MO'),-1))
    )t1
    left join
    (
        select mid_id
        from dws_new_mid_day
        where create_date<=date_add(next_day('2019-02-20','MO'),-1) and create_date>=date_add(next_day('2019-02-20','MO'),-7)
    )t2
    on t1.mid_id=t2.mid_id
    left join
    (
        select mid_id
        from dws_uv_detail_wk
        where wk_dt=concat(date_add(next_day('2019-02-20','MO'),-7*2),'_',date_add(next_day('2019-02-20','MO'),-7-1))
    )t3
    on t1.mid_id=t3.mid_id
    where t2.mid_id is null and t3.mid_id is null
)t4;

需求六：流失用户数

流失用户：最近7天未登录我们称之为流失用户

使用日活明细表dws_uv_detail_day作为DWS层数据

建表语句

代码语言：javascript复制

hive (gmall)>
drop table if exists ads_wastage_count;
create external table ads_wastage_count( 
    `dt` string COMMENT '统计日期',
    `wastage_count` bigint COMMENT '流失设备数'
) 
row format delimited fields terminated by 't'
location '/warehouse/gmall/ads/ads_wastage_count';

导入数据

代码语言：javascript复制

hive (gmall)>
insert into table ads_wastage_count
select
     '2019-02-20',
     count(*)
from 
(
    select mid_id
from dws_uv_detail_day
    group by mid_id
    having max(dt)<=date_add('2019-02-20',-7)
)t1;

需求七：最近连续3周活跃用户数

最近3周连续活跃的用户：通常是周一对前3周的数据做统计，该数据一周计算一次。

使用周活明细表dws_uv_detail_wk作为DWS层数据

建表语句

代码语言：javascript复制

hive (gmall)>
drop table if exists ads_continuity_wk_count;
create external table ads_continuity_wk_count( 
    `dt` string COMMENT '统计日期,一般用结束周周日日期,如果每天计算一次,可用当天日期',
    `wk_dt` string COMMENT '持续时间',
    `continuity_count` bigint
) 
row format delimited fields terminated by 't'
location '/warehouse/gmall/ads/ads_continuity_wk_count';

导入数据

代码语言：javascript复制

hive (gmall)>
insert into table ads_continuity_wk_count
select 
     '2019-02-20',
     concat(date_add(next_day('2019-02-20','MO'),-7*3),'_',date_add(next_day('2019-02-20','MO'),-1)),
     count(*)
from 
(
    select mid_id
    from dws_uv_detail_wk
    where wk_dt>=concat(date_add(next_day('2019-02-20','MO'),-7*3),'_',date_add(next_day('2019-02-20','MO'),-7*2-1)) 
    and wk_dt<=concat(date_add(next_day('2019-02-20','MO'),-7),'_',date_add(next_day('2019-02-20','MO'),-1))
    group by mid_id
    having count(*)=3
)t1;

需求八：最近七天内连续三天活跃用户数

说明：最近7天内连续3天活跃用户数

使用日活明细表dws_uv_detail_day作为DWS层数据

建表

代码语言：javascript复制

hive (gmall)>
drop table if exists ads_continuity_uv_count;
create external table ads_continuity_uv_count( 
    `dt` string COMMENT '统计日期',
    `wk_dt` string COMMENT '最近7天日期',
    `continuity_count` bigint
) COMMENT '连续活跃设备数'
row format delimited fields terminated by 't'
location '/warehouse/gmall/ads/ads_continuity_uv_count';

导入数据

代码语言：javascript复制

hive (gmall)>
insert into table ads_continuity_uv_count
select
    '2019-02-12',
    concat(date_add('2019-02-12',-6),'_','2019-02-12'),
    count(*)
from
(
    select mid_id
    from
    (
        select mid_id      
        from
        (
            select 
                mid_id,
                date_sub(dt,rank) date_dif
            from
            (
                select 
                    mid_id,
                    dt,
                    rank() over(partition by mid_id order by dt) rank
                from dws_uv_detail_day
                where dt>=date_add('2019-02-12',-6) and dt<='2019-02-12'
            )t1
        )t2 
        group by mid_id,date_dif
        having count(*)>=3
    )t3 
    group by mid_id
)t4;

ODS层跟原始字段要一模一样；

DWD层 　　dwd_order_info订单表 　　dwd_order_detail订单详情(订单和商品) 　　dwd_user_info用户表 　　dwd_payment_info支付流水 　　dwd_sku_info商品表(增加分类)

每日用户行为宽表 dws_user_action

字段：user_id、order_count、order_amount、payment_count、payment_amount 、comment_count

代码语言：javascript复制

drop table if exists dws_user_action;
create external table dws_user_action(
user_id string comment '用户id',
order_count bigint comment '用户下单数',
order_amount decimal(16, 2) comment '下单金额',
payment_count bigint comment '支付次数',
payment_amount decimal(16, 2) comment '支付金额',
comment_count bigint comment '评论次数'
)comment '每日用户行为宽表'
partitioned by(`dt` string)
stored as parquet
location '/warehouse/gmall/dws/dws_user_action/'
tblproperties("parquet.compression"="snappy");

导入数据

0占位符，第一个字段要有别名

代码语言：javascript复制

with tmp_order as(
select user_id, count(*) order_count, sum(oi.total_amount) order_amount from dwd_order_info oi
where date_format(oi.create_time, 'yyyy-MM-dd')='2019-02-10' group by user_id
),
tmp_payment as(
select user_id, count(*) payment_count, sum(pi.total_amount) payment_amount from dwd_payment_info pi
where date_format(pi.payment_time, 'yyyy-MM-dd')='2019-02-10' group by user_id
),
tmp_comment as(
select user_id, count(*) comment_count from dwd_comment_log c
where date_format(c.dt, 'yyyy-MM-dd')='2019-02-10' group by user_id
)
insert overwrite table dws_user_action partition(dt='2019-02-10')
select user_actions.user_id, sum(user_actions.order_count), sum(user_actions.order_amount), 
sum(user_actions.payment_count),
sum(user_actions.payment_amount),
sum(user_actions.comment_count) from(
select user_id, order_count, order_amount, 0 payment_count, 0 payment_amount, 0 comment_count from tmp_order
union all select user_id, 0, 0, payment_count, payment_amount, 0 from tmp_payment
union all select user_id, 0, 0, 0, 0, comment_count from tmp_comment
) user_actions group by user_id;

GMV拍下订单金额；包括付款和未付款；

建表ads_gmv_sum_day语句：

代码语言：javascript复制

drop table if exists ads_gmv_sum_day;
create table ads_gmv_sum_day(
`dt` string comment '统计日期',
`gmv_count` bigint comment '当日GMV订单个数',
`gmv_amount` decimal(16, 2) comment '当日GMV订单总额',
`gmv_payment` decimal(16, 2) comment '当日支付金额'
) comment 'GMV'
row format delimited fields terminated by 't'
location '/warehouse/gmall/ads/ads_gmv_sum_day';

导入数据：from用户行为宽表dws_user_action

sum(order_count) gmv_count 、 sum(order_amount) gmv_amount 、sum(payment_amount) payment_amount 过滤日期，以dt分组；

代码语言：javascript复制

insert into table ads_gmv_sum_day 
select '2019-02-10' dt, sum(order_count) gmv_count, sum(order_amount) gmv_amount, sum(payment_amount) gmv_payment
from dws_user_action where dt='2019-02-10' group by dt;

编写脚本：

代码语言：javascript复制

#/bin/bash
APP=gmall
hive=/opt/module/hive/bin/hive
if [ -n "$1" ]; then
    do_date=$1
else
    do_date=`date -d "-1 day"  %F`
fi    
sql="
insert into table "$APP".ads_gmv_sum_day 
select '$do_date' dt, sum(order_count) gmv_count, sum(order_amount) gmv_amount, sum(payment_amount) gmv_payment
from "$APP".dws_user_action where dt='$do_date' group by dt;
"
$hive -e "$sql";

需求十：转化率=新增用户/日活用户

https://img2018.cnblogs.com/blog/1247221/201905/1247221-20190516210029788-275690435.png

代码语言：javascript复制

ads_user_convert_day
  dt
  uv_m_count   当日活跃设备
  new_m_count  当日新增设备
  new_m_ratio  新增占日活比率

ads_uv_count      用户活跃数（在行为数仓中；） day_count dt
ads_new_mid_count 用户新增表（行为数仓中） new_mid_count create_date

建表ads_user_convert_day

代码语言：javascript复制

drop table if exists ads_user_convert_day;
create table ads_user_convert_day(
`dt` string comment '统计日期',
`uv_m_count` bigint comment '当日活跃设备',
`new_m_count` bigint comment '当日新增设备',
`new_m_radio` decimal(10, 2) comment '当日新增占日活比率'
)comment '转化率'
row format delimited fields terminated by 't'
location '/warehouse/gmall/ads/ads_user_convert_day/';

数据导入 cast(sum( uc.nmc)/sum( uc.dc)*100 as decimal(10,2)) new_m_ratio ；使用union all

代码语言：javascript复制

insert into table ads_user_convert_day select '2019-02-10', sum(uc.dc) sum_dc, sum(uc.nmc) sum_nmc, 
cast(sum(uc.nmc)/sum(uc.dc) * 100 as decimal(10, 2)) new_m_radio
from(select day_count dc, 0 nmc from ads_uv_count where dt='2019-02-10'
union all select 0 dc, new_mid_count from ads_new_mid_count where create_date='2019-02-10'
)uc;

访问到下单转化率| 下单到支付转化率

代码语言：javascript复制

ads_user_action_convert_day
dt
total_visitor_m_count                 总访问人数
order_u_count                        下单人数
visitor2order_convert_ratio         访问到下单转化率
payment_u_count                     支付人数
order2payment_convert_ratio            下单到支付转化率

dws_user_action （宽表中）
    user_id
    order_count
    order_amount
    payment_count
    payment_amount 
    comment_count
ads_uv_count 用户活跃数（行为数仓中）
    dt
    day_count 
    wk_count
    mn_count
    is_weekend
    is_monthend

建表

代码语言：javascript复制

drop table if exists ads_user_action_convert_day;
create table ads_user_action_convert_day(
`dt` string comment '统计日期',
`total_visitor_m_count` bigint comment '总访问人数',
`order_u_count` bigint comment '下单人数',
`visitor2order_convert_radio` decimal(10, 2) comment '访问到下单转化率',
`payment_u_count` bigint comment '支付人数',
`order2payment_convert_radio` decimal(10, 2) comment '下单到支付的转化率'
)COMMENT '用户行为漏斗分析'
row format delimited  fields terminated by 't' 
location '/warehouse/gmall/ads/ads_user_convert_day/'
;

插入数据

代码语言：javascript复制

insert into table ads_user_action_convert_day
select '2019-02-10', uv.day_count, ua.order_count, 
cast(ua.order_count/uv.day_count * 100 as decimal(10, 2)) visitor2order_convert_radio,
ua.payment_count,
cast(ua.payment_count/ua.order_count * 100 as decimal(10, 2)) order2payment_convert_radio
from(
select sum(if(order_count>0, 1, 0)) order_count,
sum(if(payment_count>0, 1, 0)) payment_count
from dws_user_action where dt='2019-02-10'
)ua, ads_uv_count  uv where uv.dt='2019-02-10';

需求十一：品牌复购率

需求：以月为单位统计，购买2次以上商品的用户，用户购买商品明细表 dws_sale_detail_daycount：（宽表）建表dws_sale_detail_daycount

代码语言：javascript复制

drop table if exists dws_sale_detail_daycount;

create external table dws_sale_detail_daycount(
user_id   string  comment '用户 id',
sku_id    string comment '商品 Id',
user_gender  string comment '用户性别',
user_age string  comment '用户年龄',
user_level string comment '用户等级',
order_price decimal(10,2) comment '商品价格',
sku_name string   comment '商品名称',
sku_tm_id string   comment '品牌id',
sku_category3_id string comment '商品三级品类id',
sku_category2_id string comment '商品二级品类id',
sku_category1_id string comment '商品一级品类id',
sku_category3_name string comment '商品三级品类名称',
sku_category2_name string comment '商品二级品类名称',
sku_category1_name string comment '商品一级品类名称',
spu_id  string comment '商品 spu',
sku_num  int comment '购买个数',
order_count string comment '当日下单单数',
order_amount string comment '当日下单金额'
) comment  '用户购买商品明细表'
partitioned by(`dt` string)
stored as parquet
location '/warehouse/gmall/dws/dws_sale_detail_daycount'
tblproperties("parquet.compression"="snappy");

数据导入

ods_order_detail订单详情表、dwd_user_info用户表、dwd_sku_info商品表

代码语言：javascript复制

with tmp_detail as(
select user_id, sku_id, sum(sku_num) sku_num, count(*) order_count, sum(od.order_price*sku_num) order_amount
from ods_order_detail od where od.dt='2019-02-10' and user_id is not null group by user_id, sku_id
)
insert overwrite table dws_sale_detail_daycount partition(dt='2019-02-10')
select
tmp_detail.user_id,
tmp_detail.sku_id,
u.gender,
months_between('2019-02-10', u.birthday)/12 age,
u.user_level,
price,
sku_name,
tm_id,
category3_id ,  
category2_id ,  
category1_id ,  
category3_name ,  
category2_name ,  
category1_name ,  
spu_id,
tmp_detail.sku_num,
tmp_detail.order_count,
tmp_detail.order_amount 
from tmp_detail 
left join dwd_user_info u on u.id=tmp_detail.user_id and u.dt='2019-02-10'
left join dwd_sku_info s on s.id=tmp_detail.sku_id and s.dt='2019-02-10';

ADS层 品牌复购率报表分析 建表ads_sale_tm_category1_stat_mn

buycount 购买人数、buy_twice_last两次以上购买人数、

buy_twice_last_ratio '单次复购率'、

buy_3times_last '三次以上购买人数',

buy_3times_last_ratio 多次复购率'

代码语言：javascript复制

drop table ads_sale_tm_category1_stat_mn;
create  table ads_sale_tm_category1_stat_mn
(   
    tm_id string comment '品牌id ' ,
    category1_id string comment '1级品类id ',
    category1_name string comment '1级品类名称 ',
    buycount   bigint comment  '购买人数',
    buy_twice_last bigint  comment '两次以上购买人数',
    buy_twice_last_ratio decimal(10,2)  comment  '单次复购率', 
    buy_3times_last   bigint comment   '三次以上购买人数',
    buy_3times_last_ratio decimal(10,2)  comment  '多次复购率' ,
    stat_mn string comment '统计月份',
    stat_date string comment '统计日期' 
)   COMMENT '复购率统计'
row format delimited  fields terminated by 't' 
location '/warehouse/gmall/ads/ads_sale_tm_category1_stat_mn/'
;

插入数据

代码语言：javascript复制

insert into table ads_sale_tm_category1_stat_mn
select mn.sku_tm_id,
mn.sku_category1_id,
mn.sku_category1_name,
sum(if(mn.order_count >= 1, 1, 0)) buycount,
sum(if(mn.order_count >= 2, 1, 0)) buyTwiceLast,
sum(if(mn.order_count >= 2, 1, 0)) / sum(if(mn.order_count >= 1, 1, 0)) buyTwiceLastRatio,
sum(if(mn.order_count >= 3, 1, 0)) buy3timeLast,
sum(if(mn.order_count >= 3, 1, 0)) / sum(if(mn.order_count >= 1, 1, 0)) buy3timeLastRadio,
date_format ('2019-02-10' ,'yyyy-MM') stat_mn,
'2019-02-10' stat_date
from (
select sd.sku_tm_id, sd.sku_category1_id, sd.sku_category1_name, user_id, sum(order_count) order_count
from dws_sale_detail_daycount sd where date_format(dt, 'yyyy-MM') <= date_format('2019-02-10', 'yyyy-MM')
group by sd.sku_tm_id, sd.sku_category1_id, user_id, sd.sku_category1_name
) mn
group by mn.sku_tm_id, mn.sku_category1_id, mn.sku_category1_name
;

数据导入脚本

1）在/home/kris/bin目录下创建脚本ads_sale.sh

[kris@hadoop101 bin]$ vim ads_sale.sh

代码语言：javascript复制

#!/bin/bash

# 定义变量方便修改
APP=gmall
hive=/opt/module/hive/bin/hive

# 如果是输入的日期按照取输入日期；如果没输入日期取当前时间的前一天
if [ -n "$1" ] ;then
    do_date=$1
else 
    do_date=`date  -d "-1 day"   %F`  
fi 

sql="

set hive.exec.dynamic.partition.mode=nonstrict;

insert into table "$APP".ads_sale_tm_category1_stat_mn
select   
    mn.sku_tm_id,
    mn.sku_category1_id,
    mn.sku_category1_name,
    sum(if(mn.order_count>=1,1,0)) buycount,
    sum(if(mn.order_count>=2,1,0)) buyTwiceLast,
    sum(if(mn.order_count>=2,1,0))/sum( if(mn.order_count>=1,1,0)) buyTwiceLastRatio,
    sum(if(mn.order_count>=3,1,0))  buy3timeLast  ,
    sum(if(mn.order_count>=3,1,0))/sum( if(mn.order_count>=1,1,0)) buy3timeLastRatio ,
    date_format('$do_date' ,'yyyy-MM') stat_mn,
    '$do_date' stat_date
from 
(     
    select od.sku_tm_id, 
        od.sku_category1_id,
        od.sku_category1_name,  
        user_id , 
        sum(order_count) order_count
    from  "$APP".dws_sale_detail_daycount  od 
    where date_format(dt,'yyyy-MM')<=date_format('$do_date' ,'yyyy-MM')
    group by od.sku_tm_id, od.sku_category1_id, user_id, od.sku_category1_name
) mn
group by mn.sku_tm_id, mn.sku_category1_id, mn.sku_category1_name;

"
$hive -e "$sql"

增加脚本执行权限
[kris@hadoop101 bin]$ chmod 777 ads_sale.sh
执行脚本导入数据
[kris@hadoop101 bin]$ ads_sale.sh 2019-02-11
查看导入数据
hive (gmall)>select * from ads_sale_tm_category1_stat_mn limit 2;

品牌复购率结果输出到MySQL

1）在MySQL中创建ads_sale_tm_category1_stat_mn表

代码语言：javascript复制

create  table ads_sale_tm_category1_stat_mn
(   
    tm_id varchar(200) comment '品牌id ' ,
    category1_id varchar(200) comment '1级品类id ',
    category1_name varchar(200) comment '1级品类名称 ',
    buycount   varchar(200) comment  '购买人数',
    buy_twice_last varchar(200) comment '两次以上购买人数',
    buy_twice_last_ratio varchar(200) comment  '单次复购率', 
    buy_3times_last   varchar(200) comment   '三次以上购买人数',
    buy_3times_last_ratio varchar(200)  comment  '多次复购率' ,
    stat_mn varchar(200) comment '统计月份',
    stat_date varchar(200) comment '统计日期' 
)

2）编写Sqoop导出脚本

在/home/kris/bin目录下创建脚本sqoop_export.sh

[kris@hadoop101 bin]$ vim sqoop_export.sh

代码语言：javascript复制

#!/bin/bash

db_name=gmall

export_data() {
/opt/module/sqoop/bin/sqoop export 
--connect "jdbc:mysql://hadoop101:3306/${db_name}?useUnicode=true&characterEncoding=utf-8"  
--username root 
--password 123456 
--table $1 
--num-mappers 1 
--export-dir /warehouse/$db_name/ads/$1 
--input-fields-terminated-by "t"  
--update-key "tm_id,category1_id,stat_mn,stat_date" 
--update-mode allowinsert 
--input-null-string '\N'    
--input-null-non-string '\N'  
}

case $1 in
  "ads_sale_tm_category1_stat_mn")
     export_data "ads_sale_tm_category1_stat_mn"
;;
   "all")
     export_data "ads_sale_tm_category1_stat_mn"
;;
esac

3）执行Sqoop导出脚本

[kris@hadoop101 bin]$ chmod 777 sqoop_export.sh

[kris@hadoop101 bin]$ sqoop_export.sh all

4）在MySQL中查看结果

SELECT * FROM ads_sale_tm_category1_stat_mn;

需求十二：求每个等级的用户对应的复购率前十的商品排行

1）每个等级，每种商品，买一次的用户数，买两次的用户数=》得出复购率

2）利用开窗函数，取每个等级的前十

3）形成脚本

用户购买明细宽表 dws_sale_detail_daycount

① t1--按user_leval, sku_id, user_id统计下单次数

代码语言：javascript复制

select 
    user_level, 
    sku_id, 
    user_id, 
    sum(order_count) order_count_sum
from dws_sale_detail_daycount
where date_format(dt, 'yyyy-MM') = date_format('2019-02-13', 'yyyy-MM')
group by user_level, sku_id, user_id limit 10;

② t2 --求出每个等级，每种商品，买一次的用户数，买两次的用户数 得出复购率

代码语言：javascript复制

select 
    t1.user_level,
    t1.sku_id,
    sum(if(t1.order_count_sum > 0, 1, 0)) buyOneCount,
    sum(if(t1.order_count_sum > 1, 1, 0)) buyTwiceCount,
    sum(if(t1.order_count_sum > 1, 1, 0)) / sum(if(t1.order_count_sum > 0, 1, 0)) * 100 buyTwiceCountRatio,
    '2019-02-13' stat_date
from(
select 
    user_level, 
    sku_id, 
    user_id, 
    sum(order_count) order_count_sum
from dws_sale_detail_daycount
where date_format(dt, 'yyyy-MM') = date_format('2019-02-13', 'yyyy-MM')
group by user_level, sku_id, user_id
) t1
group by t1.user_level, t1.sku_id;

③ t3 --按用户等级分区，复购率排序

代码语言：javascript复制

select
    t2.user_level,
    t2.sku_id,
    t2.buyOneCount,
    t2.buyTwiceCount,
    t2.buyTwiceCountRatio,
    t2.stat_date
from(
select 
    t1.user_level,
    t1.sku_id,
    sum(if(t1.order_count_sum > 0, 1, 0)) buyOneCount,
    sum(if(t1.order_count_sum > 1, 1, 0)) buyTwiceCount,
    sum(if(t1.order_count_sum > 1, 1, 0)) / sum(if(t1.order_count_sum > 0, 1, 0)) * 100 buyTwiceCountRatio,
    '2019-02-13' stat_date
from(
select 
    user_level, 
    sku_id, 
    user_id, 
    sum(order_count) order_count_sum
from dws_sale_detail_daycount
where date_format(dt, 'yyyy-MM') = date_format('2019-02-13', 'yyyy-MM')
group by user_level, sku_id, user_id
) t1
group by t1.user_level, t1.sku_id
)t2

④ -分区排序 rank()

代码语言：javascript复制

select
    t2.user_level,
    t2.sku_id,
    t2.buyOneCount,
    t2.buyTwiceCount,
    t2.buyTwiceCountRatio,
rank() over(partition by t2.sku_id order by t2.buyTwiceCount) rankNo
from(
select 
    t1.user_level,
    t1.sku_id,
    sum(if(t1.order_count_sum > 0, 1, 0)) buyOneCount,
    sum(if(t1.order_count_sum > 1, 1, 0)) buyTwiceCount,
    sum(if(t1.order_count_sum > 1, 1, 0)) / sum(if(t1.order_count_sum > 0, 1, 0)) * 100 buyTwiceCountRatio,
    '2019-02-13' stat_date
from(
select 
    user_level, 
    sku_id, 
    user_id, 
    sum(order_count) order_count_sum
from dws_sale_detail_daycount
where date_format(dt, 'yyyy-MM') = date_format('2019-02-13', 'yyyy-MM')
group by user_level, sku_id, user_id
) t1
group by t1.user_level, t1.sku_id
)t2

⑤ 作为子查询取前10

代码语言：javascript复制

select t3.user_level, t3.sku_id, t3.buyOneCount, t3.buyTwiceCount, t3.buyTwiceCountRatio, t3.rankNo
from(
select
    t2.user_level,
    t2.sku_id,
    t2.buyOneCount,
    t2.buyTwiceCount,
    t2.buyTwiceCountRatio,
rank() over(partition by t2.sku_id order by t2.buyTwiceCount) rankNo
from(
select 
    t1.user_level,
    t1.sku_id,
    sum(if(t1.order_count_sum > 0, 1, 0)) buyOneCount,
    sum(if(t1.order_count_sum > 1, 1, 0)) buyTwiceCount,
    sum(if(t1.order_count_sum > 1, 1, 0)) / sum(if(t1.order_count_sum > 0, 1, 0)) * 100 buyTwiceCountRatio,
    '2019-02-13' stat_date
from(
select 
    user_level, 
    sku_id, 
    user_id, 
    sum(order_count) order_count_sum
from dws_sale_detail_daycount
where date_format(dt, 'yyyy-MM') = date_format('2019-02-13', 'yyyy-MM')
group by user_level, sku_id, user_id
) t1
group by t1.user_level, t1.sku_id
)t2
) t3 where rankNo <= ;

bash bash指令 hadoop hive

0 人点赞