从数据库中选择几个时段的平均值

Question

我需要在几个时间间隔内得到相同时期的平均计数。
所以现在我的查询看起来很难看，是否有可能让它变得更好？ 例如，如果我想获得7周的数据，我需要复制/粘贴相同的7行，唉！
另外，我有一个关于如何丢弃极值的问题（0,10,15,11 - 从样本中删除0）。

SELECT
  ROUND((w1 + w2 + w3 + w4) / 4)
FROM (
  SELECT COUNT(*) AS w1
  FROM a.b
  WHERE id = 1 AND DATE BETWEEN sysdate - 7 - (1 / 24 / 4) AND sysdate - 7),
  (SELECT COUNT(*) AS w2
   FROM a.b
   WHERE
     id = 1 AND DATE BETWEEN sysdate - 14 - (1 / 24 / 4) AND sysdate - 14),
  (SELECT COUNT(*) as w3
FROM a.b
WHERE id = 1 AND DATE BETWEEN sysdate - 21 - (1 / 24 / 4) AND sysdate - 21),
  (SELECT COUNT(*) as w4
FROM a.b
WHERE id = 1 AND DATE BETWEEN sysdate - 28 - (1 / 24 / 4) AND sysdate - 28);

如果这很重要，我正在使用Oracle DB。

Answer 1

也许尝试使用GROUP BY。

SELECT AVG(C) FROM (
SELECT SUM(1) AS C
,to_number(to_char(date,'WW')) as DT FROM a.b
GROUP BY to_number(to_char(date,'WW')))

Answer 2

SELECT count(*) / 4
FROM a.b
WHERE id = 1
AND (DATE BETWEEN sysdate - 7 - (1 / 24 / 4) AND sysdate - 7
 OR DATE BETWEEN sysdate - 14 - (1 / 24 / 4) AND sysdate - 14
 OR DATE BETWEEN sysdate - 21 - (1 / 24 / 4) AND sysdate - 21
 OR DATE BETWEEN sysdate - 28 - (1 / 24 / 4) AND sysdate - 28)

这是有效的，因为无论如何你总结了四个计数。

我担心我不理解你关于抛弃极端价值观的第二个问题。

Answer 3

设置测试数据：

drop table b;
create table b(event_id number, event_date date);

begin
for i in 1 .. 100000
loop
insert into b values(floor(dbms_random.value(1,5)),sysdate - abs(dbms_random.normal)*20);
end loop;
commit;
end;
/

示例查询：

with time_periods (start_time, end_time, period, tp_id)
     as 
     (select cast((sysdate - 7*(:start_week) - :time_interval) as date), 
             cast((sysdate - 7*(:start_week)) as date),
             cast(:start_week as number(3)),
             cast(1 as number(3))
        from dual
      union all
      select sysdate - 7*(period+1) - :time_interval,
             sysdate - 7*(period+1),
             period + 1,
             tp_id + 1
        from time_periods
        where tp_id < :num_weeks
        )
select avg(event_count) 
  from (
  select tp_id, 
         event_count, 
         stddev(event_count) over () std_dev,
         avg(event_count) over () average
    from (select tp.tp_id, 
                 count(*) event_count
            from b 
                 join time_periods tp
                   on b.event_date between tp.start_time and tp.end_time
           where b.event_id = 1
        group by tp.tp_id)
       )
 where event_count between average - std_dev and average + std_dev;