【Hive】（十四）Hive 项目实战之电子商务消费行为分析_hive项目实战(5)

网上学习资料一大堆，但如果学到的知识不成体系，遇到问题时只是浅尝辄止，不再深入研究，那么很难做到真正的技术提升。

需要这份系统化资料的朋友，可以戳这里获取

一个人可以走的很快，但一群人才能走的更远！不论你是正从事IT行业的老鸟或是对IT行业感兴趣的新人，都欢迎加入我们的的圈子（技术交流、学习资源、职场吐槽、大厂内推、面试辅导），让我们一起学习成长！

–add jar /opt/soft/data/md5.jar
–create temporary function md5 as ‘com.shopping.services.Encryption’

–select md5(‘abc’)
–drop temporary function encrymd5

**Clean and Mask customer\_details 创建明细表**

create table if not exists customer_details
as select customer_id,first_name,last_name,md5(email) email,gender,md5(address) address,country,job,credit_type,md5(credit_no)
from ext_customer_details

##### 对表内容进行检查，为数据清洗做准备


**Check ext\_transaction\_details data**  
 对`transaction`表的`transaction_id`进行检查，查看重复的、错误的、以及空值的数量。


这里从表中我们可以看到`transaction_id`存在100个重复的值。

with
t1 as (select ‘countrow’ as status,count(transaction_id) as val from ext_transaction_details),
t2 as (select ‘distinct’ as status,(count(transaction_id)-count(distinct transaction_id)) as val from ext_transaction_details),
t3 as (select ‘nullrow’ as status,count(transaction_id) as val from ext_transaction_details where transaction_id is null),
t4 as (select ‘errorexp’ as status,count(regexp_extract(transaction_id,‘^([0-9]{1,4})$’,0)) as val from ext_transaction_details)
select * from t1 union all select * from t2 union all select * from t3 union all select * from t4

![在这里插入图片描述](https://img-blog.csdnimg.cn/20200105134049227.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L0JlaWlzQmVp,size_1,color_FFFFFF,t_70)  
 **Clean transaction\_details into partition table**

create table if not exists transaction_details(
transaction_id string,
customer_id string,
store_id string,
price double,
product string,
buydate string,
buytime string
)
partitioned by (partday string)
row format delimited fields terminated by ‘,’
stored as rcfile

**开启动态分区**

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

##### 开启动态分区，通过窗口函数对数据进行清洗


**Clear data and import data into transaction\_details**

– partday 分区 transaction_id 重复
select if(t.ct=1,transaction_id,concat(t.transaction_id,‘_’,t.ct-1))
transaction_id,customer_id,store_id,price,product,buydate,buytime,date_format(buydate,‘yyyy-MM’)
as partday
from (select *,row_number() over(partition by transaction_id) as ct
from ext_transaction_details) t

insert into transaction_details partition(partday)
select if(t.ct=1,transaction_id,concat(t.transaction_id,‘_’,t.ct-1)) transaction_id,customer_id,store_id,price,product,buydate,buytime,date_format(regexp_replace(buydate,‘/’,‘-’),‘yyyy-MM’)
as partday from (select *,row_number() over(partition by transaction_id) as ct
from ext_transaction_details) t

* **row\_number() over(partition by transaction\_id)** 窗口函数 ：从1开始，按照顺序，生成分组内记录的序列，row\_number()的值不会存在重复，当排序的值相同时，按照表中记录的顺序进行排列 这里我们对分组的transaction\_id
* `if(t.ct=1,transaction_id,concat(t.transaction_id,'_',t.ct-1))` 如果满足`ct=1`,就是`transaction_id`,否则进行字符串拼接生成新的id


![在这里插入图片描述](https://img-blog.csdnimg.cn/20200105135856699.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L0JlaWlzQmVp,size_1,color_FFFFFF,t_70)  
 **Clean store\_review table**

create table store_review
as select transaction_id,store_id,nvl(review_score,ceil(rand()*5))
as review_score from ext_store_review

**NVL(E1, E2)的功能为：如果E1为NULL，则函数返回E2，否则返回E1本身。**  
 ![在这里插入图片描述](https://img-blog.csdnimg.cn/20200105140736190.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L0JlaWlzQmVp,size_1,color_FFFFFF,t_70)  
 我们可以看到表中的数据存在空值，通过NVL函数对数据进行填充。

show tables

![在这里插入图片描述](https://img-blog.csdnimg.cn/20200105140929270.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L0JlaWlzQmVp,size_1,color_FFFFFF,t_70)  
 通过清洗后的近源表和明细表如上。


##### 数据分析


##### Customer分析


* 找出顾客最常用的信用卡

select credit_type,count(credit_type) as peoplenum from customer_details
group by credit_type order by peoplenum desc limit 1

* 找出客户资料中排名前五的职位名称

select job,count(job) as jobnum from customer_details
group by job
order by jobnum desc
limit 5

* 在美国女性最常用的信用卡

select credit_type,count(credit_type) as femalenum from customer_details
where gender=‘Female’
group by credit_type
order by femalenum desc
limit 1

* 按性别和国家进行客户统计

select count(*) as customernum,country,gender from customer_details
group by country,gender

##### Transaction分析


* 计算每月总收入

select partday,sum(price) as countMoney from transaction_details group by partday

* 计算每个季度的总收入  
 **Create Quarter Macro 定义季度宏**，将时间按季度进行划分

create temporary macro
calQuarter(dt string)
concat(year(regexp_replace(dt,‘/’,‘-’)),‘年第’,ceil(month(regexp_replace(dt,‘/’,‘-’))/3),‘季度’)

select calQuarter(buydate) as quarter,sum(price) as sale
from transaction_details group by calQuarter(buydate)

![在这里插入图片描述](https://img-blog.csdnimg.cn/20200105143513847.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L0JlaWlzQmVp,size_16,color_FFFFFF,t_70)


* 按年计算总收入

create temporary macro calYear(dt string) year(regexp_replace(dt,‘/’,‘-’))

select calYear(buydate) as year,sum(price) as sale from transaction_details group by calYear(buydate)

* 按工作日计算总收入

create temporary macro calWeek(dt string) concat(‘星期’,dayofweek(regexp_replace(dt,‘/’,‘-’))-1)

select concat(‘星期’,dayofweek(regexp_replace(buydate,‘/’,‘-’))-1) as week,sum(price) as sale
from transaction_details group by dayofweek(regexp_replace(buydate,‘/’,‘-’))

![在这里插入图片描述](https://img-blog.csdnimg.cn/20200105143839896.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L0JlaWlzQmVp,size_1,color_FFFFFF,t_70)


* 按时间段计算总收入(需要清理数据)

select concat(regexp_extract(buytime,‘[0-9]{1,2}’,0),‘时’) as time,sum(price) as sale from transaction_details group by regexp_extract(buytime,‘[0-9]{1,2}’,0)

![在这里插入图片描述](https://img-blog.csdnimg.cn/20200105144015410.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L0JlaWlzQmVp,size_1,color_FFFFFF,t_70)


* 按时间段计算平均消费  
 **Time macro**

create temporary macro calTime(time string) if(split(time,’ ‘)[1]=‘PM’,regexp_extract(time,’[0-9]{1,2}‘,0)+12,
if(split(time,’ ‘)[1]=‘AM’,regexp_extract(time,’[0-9]{1,2}‘,0),split(time,’😂[0]))

select calTime(buytime) as time,sum(price) as sale from transaction_details group by calTime(buytime)

![在这里插入图片描述](https://img-blog.csdnimg.cn/20200105144212459.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L0JlaWlzQmVp,size_1,color_FFFFFF,t_70)

–define time bucket
–early morning: (5:00, 8:00]
–morning: (8:00, 11:00]
–noon: (11:00, 13:00]
–afternoon: (13:00, 18:00]
–evening: (18:00, 22:00]
–night: (22:00, 5:00] --make it as else, since it is not liner increasing
–We also format the time. 1st format time to 19:23 like, then compare, then convert minites to hours
with
t1 as
(select calTime(buytime) as time,sum(price) as sale from transaction_details group by calTime(buytime) order by time),
t2 as
(select if(time>5 and time<=8,‘early morning’,if(time >8 and time<=11,‘moring’,if(time>11 and time <13,‘noon’,
if(time>13 and time <=18,‘afternoon’,if(time >18 and time <=22,‘evening’,‘night’))))) as sumtime,sale
from t1)
select sumtime,sum(sale) from t2
group by sumtime

![在这里插入图片描述](https://img-blog.csdnimg.cn/20200105144356779.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L0JlaWlzQmVp,size_1,color_FFFFFF,t_70)


* 按工作日计算平均消费

select concat(‘星期’,dayofweek(regexp_replace(buydate,‘/’,‘-’))-1)
as week,avg(price) as sale from transaction_details
where dayofweek(regexp_replace(buydate,‘/’,‘-’))-1 !=0 and dayofweek(regexp_replace(buydate,‘/’,‘-’))-1 !=6
group by dayofweek(regexp_replace(buydate,‘/’,‘-’))

![在这里插入图片描述](https://img-blog.csdnimg.cn/2020010514445327.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L0JlaWlzQmVp,size_1,color_FFFFFF,t_70)


* 计算年、月、日的交易总数

select buydate as month,count(*) as salenum from transaction_details group by buydate

* 找出交易量最大的10个客户

select c.customer_id,c.first_name,c.last_name,count(c.customer_id) as custnum from customer_details c
inner join transaction_details t
on c.customer_id=t.customer_id
group by c.customer_id,c.first_name,c.last_name
order by custnum desc
limit 10

* 找出消费最多的前10位顾客

select c.customer_id,c.first_name,c.last_name,sum(price) as sumprice from customer_details c
inner join transaction_details t
on c.customer_id=t.customer_id
group by c.customer_id,c.first_name,c.last_name
order by sumprice desc
limit 10

* 统计该期间交易数量最少的用户

select c.customer_id,c.first_name,c.last_name,count(*) as custnum from customer_details c
inner join transaction_details t
on c.customer_id=t.customer_id
group by c.customer_id,c.first_name,c.last_name
order by custnum asc
limit 1

* 计算每个季度的独立客户总数

select calQuarter(buydate) as quarter,count(distinct customer_id) as uninum
from transaction_details
group by calQuarter(buydate)

* 计算每周的独立客户总数

select calWeek(buydate) as quarter,count(distinct customer_id) as uninum
from transaction_details
group by calWeek(buydate)

* 计算整个活动客户平均花费的最大值

select sum(price)/count(*) as sale
from transaction_details
group by customer_id
order by sale desc
limit 1

* 统计每月花费最多的客户

with
t1 as
(select customer_id,partday,count(distinct buydate) as visit from transaction_details group by partday,customer_id),
t2 as
(select customer_id,partday,visit,row_number() over(partition by partday order by visit desc) as visitnum from t1)
select * from t2 where visitnum=1

* 统计每月访问次数最多的客户

with
t1 as
(select customer_id,partday,sum(price) as pay from transaction_details group by partday,customer_id),
t2 as
(select customer_id,partday,pay,row_number() over(partition by partday order by pay desc) as paynum from t1)
select * from t2 where paynum=1

* 按总价找出最受欢迎的5种产品

select product,sum(price) as sale from transaction_details
group by product
order by sale desc
limit 5

* 根据购买频率找出最畅销的5种产品

select product,count(*) as num from transaction_details
group by product
order by num desc
limit 5

* 根据客户数量找出最受欢迎的5种产品

select product,count(distinct customer_id) as num from transaction_details
group by product
order by num desc
limit 5

* 验证前5个details

select * from transaction_details where product in (‘Goat - Whole Cut’)

##### Store分析


* 按客流量找出最受欢迎的商店

with
t1 as (select store_id,count(*) as visit from transaction_details
group by
store_id order by visit desc limit 1)
select s.store_name,t.visit
from t1 t
inner join
ext_store_details s
on t.store_id=s.store_id

* 根据顾客消费价格找出最受欢迎的商店

with
t1 as (select store_id,sum(price) as sale from transaction_details
group by
store_id order by sale desc limit 1)
select s.store_name,t.sale
from t1 t
inner join
ext_store_details s
on t.store_id=s.store_id

* 根据顾客交易情况找出最受欢迎的商店

with
t1 as
(select store_id,store_name from ext_store_details)
select t.store_id,store_name,count(distinct t.customer_id) as num
from transaction_details t
inner join t1 s
on s.store_id=t.store_id
group by t.store_id,store_name
order by num desc
limit 1

* 根据商店和唯一的顾客id获取最受欢迎的产品

with
t1 as (select store_id,product,count(distinct customer_id) as num from transaction_details
group by store_id,product order by num desc limit 1)
select s.store_name,t.num,t.product
from t1 t
inner join
ext_store_details s
on t.store_id=s.store_id

![img](https://i-blog.csdnimg.cn/blog_migrate/4fc4ff0113f6cc3261399bad20183793.png)
![img](https://img-blog.csdnimg.cn/img_convert/386a409c77d5ac5c62a3d9fff2f4cf79.png)

**网上学习资料一大堆，但如果学到的知识不成体系，遇到问题时只是浅尝辄止，不再深入研究，那么很难做到真正的技术提升。**

**[需要这份系统化资料的朋友，可以戳这里获取](https://bbs.csdn.net/forums/4f45ff00ff254613a03fab5e56a57acb)**


**一个人可以走的很快，但一群人才能走的更远！不论你是正从事IT行业的老鸟或是对IT行业感兴趣的新人，都欢迎加入我们的的圈子（技术交流、学习资源、职场吐槽、大厂内推、面试辅导），让我们一起学习成长！**

t customer_id) as num from transaction_details
group by store_id,product order by num desc limit 1)
select s.store_name,t.num,t.product 
from t1 t 
inner join 
ext_store_details s 
on t.store_id=s.store_id

[外链图片转存中…(img-5uci3QeN-1715350344445)]
[外链图片转存中…(img-Dg4LsOJ0-1715350344445)]

网上学习资料一大堆，但如果学到的知识不成体系，遇到问题时只是浅尝辄止，不再深入研究，那么很难做到真正的技术提升。

需要这份系统化资料的朋友，可以戳这里获取

一个人可以走的很快，但一群人才能走的更远！不论你是正从事IT行业的老鸟或是对IT行业感兴趣的新人，都欢迎加入我们的的圈子（技术交流、学习资源、职场吐槽、大厂内推、面试辅导），让我们一起学习成长！