以下这些都是针对于电商网站的指标来分析业务的：以下这些指标都是使用SQL语句分析得出的结果，首先先创建一张表，然后将数据导入表中，字段根据分析的业务的需要而定，这里我的选择是我已经存在的表进行查询的，下面将给出所有查询的SQL语句
模块分析：
select * from web_log.weblog_etl where dateStr="20180709" limit 5;
1.统计每天每小时的PV数
select day,hour,count(request) as pvs from web_log.weblog_detail where dateStr="20180709" group by day,hour
2.统计每天的pv数
select dayStr,day,count(request) as pvs from web_log.weblog_detail where dateStr="20180709" group by dayStr,day
3.统计每月的pv数
select dayStr,month,count(request) as pvs from web_log.weblog_detail where dateStr="20180709" group by dayStr,month
4.统计不同终端维度的PV数
select dayStr,user_agent,count(request) as pvs from web_log.weblog_detail where dateStr="20180709" group by dayStr,user_agent
5.统计不同终端类型
select user_agent,count(user_agent) as iphone from web_log.weblog_detail where dateStr="20180709" group by user_agent
6.统计每个月每个终端的pv数
select dayStr,month,day,hour,user_agent,count(request) as pvs from web_log.weblog_detail where dateStr="20180709" group by dayStr,month,day,hour,user_agent
7.统计人均pv数
select dayStr,count(request)/count(distinct ip) from web_log.weblog_detail where dateStr="20180709" group by dayStr
8.统计不同来源的pv数
select
*
from
(
select t.ref_host,t.dayStr,t.hour,t.pv,row_number() over (partition by t.ref_host order by t.pv) as rnk
from
(
select
ref_host,dayStr,hour,count(request) as pv
from
web_log.weblog_detail
where
dateStr="20180709"
group by
ref_host,dayStr,hour
)t
)n
where n.rnk <4
---------------
select
*
from
(
select t.ref_host,t.pv,row_number() over (partition by t.ref_host order by t.pv) as rnk
from
(
select
ref_host,count(request) as pv
from
web_log.weblog_detail
where
dateStr="20180709"
group by
ref_host
)t
)n
where n.rnk <4
9.受访分析
9.1统计每日热门的受访页面的topN
9.1.1先统计每日每个网页访问次数
select dayStr,request,count(request) as counts from weblog_detail
where dateStr="20180709" group by dayStr,request
(2)row_number
select dayStr,request,count(request) as pv,row_number() over (partition by request order by request) from weblog_detail
where dateStr="20180709" group by dayStr,request
10.访客分析
10.1每个访客每小时的访问数(session个数)
select
ip,substring(s_time,12,2) as hour,count(distinct sessionId) as number
from web_log.weblog_pv
where
dateStr="20180709"
group by ip,substring(s_time,12,2)
10.2每个小时的ip访问数
select day,hour,count(ip) from web_log.weblog_detail where dateStr="20180709" group by day,hour
10.3每天的/每月的ip访问数
select month,day,count(ip) from web_log.weblog_detail where dateStr="20180709" group by month,day
10.4统计新增访客---需要维护 历史访客表
11.visit模型分析
11.1单次访客统计（sessionId只有一个）
select
substring(s_time,9,2),ip,count(distinct sessionId) as number
from
web_log.weblog_pv
where
dateStr="20180709"
group by
substring(s_time,9,2),ip
having number=1
11.2回头访客个数（sessionId >1)
select
substring(s_time,9,2),ip,count(distinct sessionId) as number
from
web_log.weblog_pv
where
dateStr="20180709"
group by
substring(s_time,9,2),ip
having number>1
11.3人均访问频率(一天访问多少次) 总session个数/总人数
select
substring(s_time,9,2) as day,count(distinct sessionId)/count(distinct ip) as number
from
web_log.weblog_pv
where
dateStr="20180709"
group by
substring(s_time,9,2)
11.4用户平均访问页数
select sum(pageNum)/count(distinct ip) from web_log.weblog_visit where dayStr="20180709"
12pageview模型分析
12.1每个用户的访问时长
select
ip,sum(stayTime) as alltime
from
web_log.weblog_pv
where dateStr="20180709"
group by ip
12.2 用户的平均访问时长
select sum(stayTime)/count(distinct ip) as avg_time
from
web_log.weblog_pv
where
dateStr="20180709"
12.3访问页面的漏斗
create table tb_num1 as
select "step1" as step, count(*) as number from web_log.weblog_pv where dateStr="20180709" and request like "%/hadoop-mahout%"
union all
select "step2" as step, count(*) as number from web_log.weblog_pv where dateStr="20180709" and request like "%/hadoop-hive%"
union all
select "step3" as step, count(*) as number from web_log.weblog_pv where dateStr="20180709" and request like "%/hadoop-zookeeper%"
union all
select "step4" as step,count(*) as number from web_log.weblog_pv where dateStr="20180709" and request like "%/black-ip-list%"
create table tb_num as
select "step1" as step, count(*) as number from web_log.weblog_pv where dayStr="20180709" and request like "%/hadoop-mahout%"
union all
select "step2" as step, count(*) as number from web_log.weblog_pv where dayStr="20180709" and request like "%/hadoop-hive%"
union all
select "step3" as step, count(*) as number from web_log.weblog_pv where dayStr="20180709" and request like "%/hadoop-zookeeper%"
union all
select "step4" as step,count(*) as number from web_log.weblog_pv where dayStr="20180709" and request like "%/black-ip-list%"
select
*
from
tb_num
-----全链接 笛卡尔积
select
*
from
(select tb_num.step as ak,tb_num.number as av from tb_num) a,
(select tb_num.step as bk,tb_num.number as bv from tb_num) b
-----漏斗
select
t.ak,t.av,t.bk,t.bv,t.av/t.bv as retain_rate
from
(
select
*
from
(select tb_num.step as ak,tb_num.number as av from tb_num) a,
(select tb_num.step as bk,tb_num.number as bv from tb_num) b
) t
where
t.ak='step1' and t.bk='step1' or
t.ak='step2' and t.bk='step1' or
t.ak='step3' and t.bk='step2' or
t.ak='step4' and t.bk='step3'
实例：不使用hive中的函数
公司有多个部门，统计出各个部门中 按照工资排序 前3的。（9条）
name deptNo sal
1-1 1 9000
1-2 1 6000
1-3 1 7000
2-1 2 8000
3-2 3 7600
2-3 2 8800
2-5 2 8800
3-3 3 5500
3-4 1 5000
SQL语句：
select deptno, name, sal
from emp e1
where
(
select count(1)
from emp e2
where e2.deptno=e1.deptno and e2.sal>=e1.sal
) <=3 /*这里的数值表示你想取前几名*/
order by deptno, sal desc;