使用子查询和分组依据每天计算每个国家/地区的 DAU 平均值

Question

我正在尝试计算每个国家/地区在 1 个月内的 DAU 平均值。查询的工作是：

1. 识别唯一用户
2. 查找上次登录的所有用户 月
3. 将它们分成不同的日子
4. 将它们细分为 各国
5. 计算每个国家/地区的平均值。

到目前为止，我已经完成了第 1、2、3 和 4 步，但最后一步被证明很棘手。

该查询应该首先计算子查询，在该子查询中计算有多少活跃用户在上个月打开了应用程序，然后将它们分组到日期和国家/地区。 在此之后，它应该使用它在子查询中计算的所有 30 天数据来计算每个国家的平均 DAU。 结果将是一个国家列表及其平均 DAU。

到目前为止的查询如下所示：

SELECT Country, AVG(User_ID)
FROM usersession
WHERE User_ID IN
    (SELECT count(distinct us.User_ID)
     FROM usersession us
     WHERE Opened > current_timestamp - interval 1 month
     GROUP BY DAY(Opened), Country)
GROUP BY Country ORDER BY Country;

子查询执行第 1、2、3、4 步，但子查询之外的辅助查询并没有按预期工作。

表格如下（只是相关信息的一个简短示例）：

ID    |  UserID  | Opened              | Country
-----------------------------------------------
233231          1   2017-11-20 08:00:00      NA
223214          2   2017-11-20 08:53:00      DK

预期结果（总共约 230 个国家/地区）：

Country |  Average  
------------------
     NA    150354
     DK     60345
     FI     50242

实际结果：

+---------+--------------+
| Country | AVG(User_ID) |
+---------+--------------+
| NULL    |  804397.7297 |
|         |  746046.7500 |
| BR      |  893252.0000 |
| GB      |  935599.0000 |
| RU      |  993311.0000 |
| US      |  735568.0000 |
+---------+--------------+

Answer 1

我想这就是你想要的：

select
    country,
    sum(number_of_users) / count(distinct day_of_month) as daily_average_users
from
    (
        select 
           country,
           day(opened)             as day_of_month,
           count(distinct user_id) as number_of_users
        from
           user_session
        where
           opened > current_timestamp - interval 1 month
        group by
           country,
           day_of_month
    ) x
group by 
    country
order by 
    country;

我在 MySQL 5.7 上对此进行了测试：

create table user_session
(
    id       int,
    user_id  int,
    opened   timestamp,
    country  varchar(2)
);

insert into user_session (id, user_id, opened, country) values ( 1, 100, '2017-12-20 08:00:00', 'NA');
insert into user_session (id, user_id, opened, country) values ( 2, 100, '2017-12-20 08:00:00', 'NA');
insert into user_session (id, user_id, opened, country) values ( 3, 100, '2017-12-20 08:00:00', 'NA');
insert into user_session (id, user_id, opened, country) values ( 4, 100, '2017-12-21 08:00:00', 'NA');
insert into user_session (id, user_id, opened, country) values ( 5, 100, '2017-12-22 08:00:00', 'NA');
insert into user_session (id, user_id, opened, country) values ( 6, 200, '2017-12-20 08:00:00', 'NA');
insert into user_session (id, user_id, opened, country) values ( 7, 300, '2017-12-21 08:00:00', 'NA');
insert into user_session (id, user_id, opened, country) values ( 8, 400, '2017-12-20 08:00:00', 'NA');
insert into user_session (id, user_id, opened, country) values ( 9, 500, '2017-12-20 08:00:00', 'NA');
insert into user_session (id, user_id, opened, country) values (10, 600, '2017-12-20 08:00:00', 'DK');
insert into user_session (id, user_id, opened, country) values (11, 600, '2017-12-21 08:00:00', 'DK');
insert into user_session (id, user_id, opened, country) values (12, 700, '2017-12-20 08:00:00', 'DK');
insert into user_session (id, user_id, opened, country) values (13, 800, '2017-12-20 08:00:00', 'DK');
insert into user_session (id, user_id, opened, country) values (14, 800, '2017-12-21 08:00:00', 'DK');
insert into user_session (id, user_id, opened, country) values (15, 800, '2017-12-21 08:00:00', 'DK');
insert into user_session (id, user_id, opened, country) values (16, 900, '2017-12-20 08:00:00', 'DK');
insert into user_session (id, user_id, opened, country) values (17, 900, '2017-12-20 08:00:00', 'DK');
insert into user_session (id, user_id, opened, country) values (18, 900, '2017-12-22 08:00:00', 'DK');
insert into user_session (id, user_id, opened, country) values (19, 900, '2017-12-22 08:00:00', 'DK');
insert into user_session (id, user_id, opened, country) values (19, 1000, '2017-12-22 08:00:00', 'DK');

结果：

+---------+---------------------+
| country | daily_average_users |
+---------+---------------------+
| DK      |              2.6667 |
| NA      |              2.3333 |
+---------+---------------------+
2 rows in set (0.00 sec)

要获得正确的每日平均值，您需要在数据中表示该月的每一天（否则平均值超过了表示的天数）。如果不是这种情况，那么我们需要计算所考虑期间的天数。