Hive 案例分析（B站用户行为大数据分析）

一、案例需求

b站现在积累有用户数据和视频列表数据，为了配合市场部门做好用户运营工作，需要对b站的用户行为进行分析，其具体需求如下所示：

• 统计b站视频不同评分等级(行转列)的视频数。
• 统计上传b站视频最多的用户Top 10，以及这些用户上传的视频观看次数在前10的视频。
• 统计b站每个类别视频观看数top 10。
• 统计b站视频分类热度top 10。
• 统计b站视频观看数top 10。

二、设计数据表结构

2.1 user 表结构

2.2 video 表结构

三、创建数据表

3.1 创建 video 数据库

create database if not exists video;

3.2 创建外表

用于保存原始数据

3.1.2 创建 external_user

create external table if not exists external_user
(uid int,
name string,
regtime string,
visitnum int,
lastvisit string,
gender int,
birthday string,
country string,
province string,
city string,
uploadvideos int)
row format delimited fields terminated by ","
stored as textfile;

3.1.3 创建 external_video

create external table if not exists  external_video
(vid string,
uid int,
vday int,
vtype string,
vlength int,
visit int,
score int,
comments int,
collection int,
fabulous int,
forward int)
row format delimited fields terminated by ","
stored as textfile;

3.2 创建内表

用于计算获取结果

3.2.1 创建 orc_user

create table if not exists orc_user
(uid int,
name string,
regtime string,
visitnum int,
lastvisit string,
gender int,
birthday string,
country string,
province string,
city string,
uploadvideos int)
row format delimited fields terminated by ","
stored as orc
tblproperties("orc.compress"="SNAPPY");

3.2.2 创建 orc_video

create table if not exists orc_video
(vid string,
uid int,
vday int,
vtype string,
vlength int,
visit int,
score int,
comments int,
collection int,
fabulous int,
forward int)
row format delimited fields terminated by ","
stored as orc
tblproperties("orc.compress"="SNAPPY");

四、准备工作

4.1 准备user和video 数据

4.2 数据导入外表

4.2.1 导入 external_user 表

load data local inpath '/usr/local/data/tmp/user.txt' overwrite into table external_user;

4.2.1 导入 external_video 表

load data local inpath '/usr/local/data/tmp/video.txt' overwrite into table external_video;

4.3 数据导入 内表

4.3.1 导入 ocl_user

insert into table orc_user select * from external_user;

4.3.1 导入 ocl_video

insert into table orc_video select * from external_video;

五、统计分析

5.1 统计B站视频观看数前 top 10

select vid,visit from orc_video order by visit desc limit 10;

5.2 统计B站视频分类热度 top 10

select vtype,count(vid) hot from orc_video group by vtype order by hot desc limit 10;

5.3 统计每个类别视频观看人数前 top 3

select v.vtype,v.vid,v.visit from (select vtype,vid,visit,rank() over(partition by vtype order by visit desc) rk ) v where rk <= 3;

5.4 统计上传B站视频最多的用户前 top 10，以及这些用户上传的视频观看次数在前10的视频

select v.vid,v.visit,v.uid from (select uid,uploadvideos from orc_user order by uploadvideos desc limit 10) u join orc_video v on u.uid=v.uid order by v.visit desc limit 10;

5.5 统计B站视频不同评分等级（等级作为字段显示）的视频数量

select
max(case v.score when 1 then v.num else 0 end) 1star,
max(case v.score when 2 then v.num else 0 end) 2star,
max(case v.score when 3 then v.num else 0 end) 3star,
max(case v.score when 4 then v.num else 0 end) 4star,
max(case v.score when 5 then v.num else 0 end) 5star
from (select score,count(*) as num from orc_video group by score) v;