在 SQL 中分析并形成分位数并计算落在各个分位数中的值的百分比

Question

您好，我在 MySQL 数据库中有以下数据

txnid amount type
    1    500 A
    2    600 B
    3    700 A
    4    800 C
    5    900 D
    6   1000 A
    7   1100 B
    8   1200 B

我想在第一个 bin 中构建四个包含 0-25 分位数的 bin，在第二个 bin 中构建 25-50 个分位数，依此类推。接下来我要计算落在各个 bin 中的值的百分比。

Answer 1

如果我正确地遵循了这一点，您可以使用ntile() 和算术来构建存储桶，然后使用条件聚合来透视：

select min_amount, max_amount, 
    sum(type = 'A') / total_cnt a,
    sum(type = 'B') / total_cnt b,
    sum(type = 'C') / total_cnt c,
    sum(type = 'D') / total_cnt d,
    count(*) / total_cnt all_types
from (
    select t.*, 
        min(amount) over() + (ntile(4) over(order by amount) - 1) / 4 * (max(amount) over() - min(amount) over()) min_amount,
        min(amount) over() +  ntile(4) over(order by amount)      / 4 * (max(amount) over() - min(amount) over()) max_amount,
        count(*) over() total_cnt
    from mytable t
) t
group by min_amount, max_amount

我将存储桶边界放在两个不同的列中，因为这对我来说更有意义 - 如果您愿意，可以将它们连接起来。比率表示为0 和1 之间的值，如果需要百分比，可以乘以100。

Demo on DB Fiddle：

最小金额 |最大金额 |一个 |乙 | c | d |所有类型 ---------: | ---------: | -----: | -----: | -----: | -----: | --------: 500.0000 | 675.0000 | 0.1250 | 0.1250 | 0.0000 | 0.0000 | 0.2500 675.0000 | 850.0000 | 0.1250 | 0.0000 | 0.1250 | 0.0000 | 0.2500 850.0000 | 1025.0000 | 0.1250 | 0.0000 | 0.0000 | 0.1250 | 0.2500 1025.0000 | 1200.0000 | 0.0000 | 0.2500 | 0.0000 | 0.0000 | 0.2500

Answer 2

需要首先确定结果中每一行的间隔。由于表中没有这样的值。可能会形成一个包含UNION ALL 运算符的子查询。然后应该构造一个动态查询，考虑到type 列的不同值可能会更改。因此，您可以使用以下语句：

SET SESSION group_concat_max_len = 4096;
SET @sql = NULL;

SELECT GROUP_CONCAT(
       CONCAT('CONCAT(ROUND( 100 * SUM( CASE WHEN amount BETWEEN THEN type = ''',type,''' END)/COUNT(*), 2),"%") AS ',type )
       )
  INTO @sql
  FROM ( SELECT DISTINCT type FROM t) tt;

SELECT GROUP_CONCAT(
       CONCAT(
               'SELECT "',REPLACE(intrval,' AND ','-'),'" AS "Amount Bins", ',
                          REPLACE( @sql, 'BETWEEN THEN', 
                                  CONCAT('BETWEEN ',intrval,' THEN' )),
                      ', CONCAT(ROUND( 100 * SUM( CASE WHEN amount BETWEEN ',intrval,' THEN 1 END)/COUNT(*), 2),"%") AS "All Types"',         
               ' FROM t '                    
              )
       SEPARATOR ' UNION ALL '       
       )
  INTO @sql
  FROM (SELECT @i := @i + 1 AS i, 
               CONCAT(@i * ((max_amount - min_amount) / 4) + min_amount,' AND ',
                     (@i + 1) * ((max_amount - min_amount) / 4) + min_amount ) AS intrval
          FROM information_schema.tables i
          JOIN (SELECT @i := -1) AS ii
          JOIN (SELECT MIN(amount) AS min_amount, MAX(amount) AS max_amount
                  FROM t ) AS t_agg  
         WHERE @i < 3) i;

PREPARE stmt FROM @sql;
EXECUTE stmt;
DEALLOCATE PREPARE stmt;

结果是

Amount Bins A       B       C       D       All Types
----------- ------  ------  ------  ------  ---------
500-675     12.50%  12.50%  0.00%   0.00%   25.00%
675-850     12.50%  0.00%   12.50%  0.00%   25.00%
850-1025    12.50%  0.00%   0.00%   12.50%  25.00%
1025-1200   0.00%   25.00%  0.00%   0.00%   25.00%

Demo