yhd日志分析(二)

yhd日志分析(二)

继续yhd日志分析，统计数据

日期	uv	pv	登录人数	游客人数	平均访问时长	二跳率	独立ip数

1 分析

登录人数

count(distinct endUserId)

游客人数

count(distinct guid) - count(distinct endUserId)

平均访问时长

先把tracktime转换为unix timestamp, 相同sessionId的tracktime中，max(tracktime)-min(tracktime)，得到用户停留时间。所有用户的停留时间相加，得到总停留时间。总停留时间和总访问次数的比例，就是平均访问时长

select sum(stay_time) as total_stay_time
from (select max(to_unix_timestamp(trackTime)) - min(to_unix_timestamp(trackTime)) as stay_time from yhd_log group by sessionId) stay

用户访问总数

count(distinct sessionId)

二跳率

sessionViewNo=2的用户，即为二跳用户。统计出二跳用户和uv的比例

select count(distinct guid) from yhd_log where sessionViewNo=2

独立ip数

count(distinct ip)

实现

1. 借助中间表，分别存放停留时间和二次跳用户总数

    // 存放总停留时间
    
    create table if not exists yhd_log_total_stay_time(
    	date string,
    	total_stay_time bigint
    )
    row format delimited fields terminated by '	'
    stored as textfile;
    
    // 存放二次跳用户总数
    
    create table if not exists yhd_log_total_second_jump(
    	date string,
    	total_second_jump bigint
    )
    row format delimited fields terminated by '	'
    stored as textfile;
   

2. 计算总停留时间，存放在yhd_log_total_stay_time, 按日期分组

    insert overwrite table yhd_log_total_stay_time
    select date, sum(stay_time) as total_stay_time
    from (select max(to_unix_timestamp(trackTime)) - min(to_unix_timestamp(trackTime)) as stay_time, date from yhd_log group by date, sessionId) stay
    group by date
   

3. 计算二次跳用户总数，存放在yhd_log_total_second_jump， 按日期分组

    insert overwrite table yhd_log_total_second_jump
    select date, count(distinct guid)
    from yhd_log
    where sessionViewNo=2 
    group by date
   

4. 统计

    把yhd_log_total_stay_time，yhd_log_total_second_jump和yhd_log按照 date连接查询
   
    select date, pv, uv, user_count, guest_count, 
    	total_stay_time/total_visit as average_stay_time,
    	total_second_jump/ uv as second_jump_rate, indepent_ip
    from (
    select log.date, 
    	count(url) as pv, 
    	count(distinct guid) as uv,
    	count(distinct endUserId) as user_count,
    	count(distinct guid) - count(distinct endUserId) as guest_count,
    	count(distinct sessionId) as total_visit,
    	min(stay.total_stay_time) as total_stay_time,
    	min(second.total_second_jump) as total_second_jump,
    	count(distinct ip) as indepent_ip
    from yhd_log log inner join yhd_log_total_stay_time stay on stay.date=log.date inner join yhd_log_total_second_jump second on second.date=log.date
    group by log.date
    ) stat
   

结果

date	pv	uv	user_count	guest_count	average_stay_time	second_jump_rate	indepent_ip
20150828	126134	39007	17687	21320	745.9797393244751	0.13118158279283207	30462