查找首次将值增加1.85-SQL Server

jw133 发表于 Dev

jw133

我正在编写一些代码，以在7天的滚动窗口中寻找最小值。

首先，我要找到每天的最小值并将其插入到临时表中#minimumvalue这是要考虑到存在多个值的天数。

我正在通过使用找到rolling_min值min(Value) over(partition by ID order by Date_Value rows between 6 preceding and current row，这就是我得到的

ID | Date_Value |       Value |  rolling_min
1234  1-1-2018 12:15 pm  2.95   1.54    
1234 1-1-2018 3:30 pm    1.54   1.54    
1234 1-2-2018 10:45 am   1.57   1.54    
1234 1-3-2018 2:15 pm    2.02   1.54    
1234 1-4-2018 4:00 pm    2.80   1.54    
1234 1-5-2018            NULL   1.54    
1234 1-6-2018            NULL   1.54    
1234 1-7-2018            NULL   1.56    
1234 1-8-2018            NULL   1.57

前两行发生在同一天，因此rolling_min以1.54开始。实际上，它应该从2.95开始，因为这是当前的最小值。

从那里，我想找到最小增加了1.85的第一个实例。我尝试了一个案例陈述CASE WHEN Value >= 1.85 * MIN(Value) OVER(PARTITION BY ID ORDER BY Date_Value ROWS BETWEEN 6 PRECEDING and CURRENT ROW THEN 1 ELSE 0 END AS Increase YN

这是结果

ID | Date_Value |       Value |  rolling_min |  Increase YN
1234  1-1-2018 12:15 pm  2.95   1.54             1
1234 1-1-2018 3:30 pm    1.54   1.54             0
1234 1-2-2018 10:45 am   1.57   1.54             0
1234 1-3-2018 2:15 pm    2.02   1.54             0
1234 1-4-2018 4:00 pm    2.80   1.54             1
1234 1-5-2018            NULL   1.54             0
1234 1-6-2018            NULL   1.54             0
1234 1-7-2018            NULL   1.57             0
1234 1-8-2018            NULL   1.57             0

第一行显示1，因为它是1.85 * 1.54的最小值，但之后是1.54，所以这并不是我真正想要的。我怎样才能避免这种情况，而只在增加的地方返回1？

在此先感谢您的帮助：）

民町

我认为您正在尝试首次发现任何数据值比最近6天的最小值都高出85％。我希望那是对的。

这是我的代码：

create table data
(ID int not null,
Date_Value datetime not null,
Value decimal(5,2)
)
go
truncate table data
go
insert into data 
select 1234,  '2018-01-01 12:15:00',  2.95
union all select 1234,  '2018-01-01 12:15:00',  1.54
union all select 1234,  '2018-01-02 10:45:00',  1.57
union all select 1234,  '2018-01-03 14:15:00',  2.02
union all select 1234,  '2018-01-04 16:00:00',  2.80
union all select 1234,  '2018-01-05',  null
union all select 1234,  '2018-01-06',  null
union all select 1234,  '2018-01-07',  null
union all select 1234,  '2018-01-08',  null
union all select 1234,  '2018-01-09 09:00:00',  3.4
union all select 1234,  '2018-01-09 10:00:00',  5.4
union all select 1234,  '2018-01-09 11:00:00',  8.4
go
insert into data 
select 9999,  '2018-01-01 12:15:00',  2.95
union all select 9999,  '2018-01-01 12:15:00',  1.54
union all select 9999,  '2018-01-02 10:45:00',  1.57
union all select 9999,  '2018-01-03 14:15:00',  2.02
union all select 9999,  '2018-01-04 16:00:00',  2.80
union all select 9999,  '2018-01-05',  null
union all select 9999,  '2018-01-06',  null
union all select 9999,  '2018-01-07',  null
union all select 9999,  '2018-01-08',  null
union all select 9999,  '2018-01-09 09:00:00',  3.4
union all select 9999,  '2018-01-09 10:00:00',  5.4
union all select 9999,  '2018-01-09 11:00:00',  8.4
go

;with ByDayMin as 
(
    select ID, Date_Value = cast(Date_Value as date), Value=min(Value)
    from data
    group by ID, cast(Date_Value as date) 

),
IncBig as (
select ID, Date_Value,
        Value,
        sixDayMin =  MIN(Value) OVER(PARTITION BY ID ORDER BY Date_Value ROWS BETWEEN 6 PRECEDING and CURRENT ROW)
from ByDayMin

),Calcs as
(
    select d.*,
        i.SixDayMin,
        ValueIncreaseRatioLast6Days = d.Value / SixDayMin,
        Prior =   LAG(d.Date_Value, 1,null) OVER(PARTITION BY d.ID ORDER BY d.Date_Value)

from data d
join ByDayMin b
on b.ID = d.ID
    and b.Date_Value = cast(d.Date_Value as date) 
join IncBig i
on i.ID = d.ID
    and i.Date_Value = b.Date_Value

)

select c.ID, c.Date_Value, c.Value, c.SixDayMin, ValueIncreaseRatioLast6Days, 
        Increase_YN=case when c.ValueIncreaseRatioLast6Days >= 1.85
                            and not exists (select 1 from Calcs c2 
                                            where c2.ID = c.ID 
                                                and c2.ValueIncreaseRatioLast6Days >= 1.85
                                                and c2.Date_Value < c.Date_Value
                                                and Prior is not Null
                                            )
                            and Prior is not Null
                        then 1
                        else 0
                    end
    from Calcs c

order by c.ID, c.Date_Value