给定数字的频率查询中位数

Numbers 表保存数字的值及其频率。

+----------+-------------+
|  Number  |  Frequency  |
+----------+-------------|
|  0       |  7          |
|  1       |  1          |
|  2       |  3          |
|  3       |  1          |
+----------+-------------+

在此表中,数字为 0, 0, 0, 0, 0, 0, 0, 1, 2, 2, 2, 3,所以中位数是 (0 + 0) / 2 = 0。

+--------+
| median |
+--------|
| 0.0000 |
+--------+

请编写一个查询来查找所有数字的中位数并将结果命名为 median 。

审题:查询所有数字的中位数,数字的个数不同。可能奇数也可能偶数。

思考:对所有数字排序,求Frequency 列个数的和,然后计算中位数位置。然后用中位数地址位置减去Frequency里的数字,等于零的时候在那个数字位置,就是中位数。

解题:

解法一

  • 确定每个数字,在展开后的升序列中,起始和结束下标;
  • 在展开后的升序列中, 中位数的起始和结束下标;
  • 由1和2,筛选出是中位数的行

注意,起始下标从0开始,结束下标为开区间。

应用排名算法:表left join自连接,group by对数字分组,按数字升序。

再筛选出比每个数字小的所有其它数字的行。将这些行的中频率值相加。其值即为起始下标。

结果命名为A。

(
	SELECT N1.NUMBER, SUM(IF(N2.frequency IS NULL,0,N2.frequency)) AS `beg`
	FROM numbers AS N1
	LEFT JOIN numbers AS N2 ON (N1.NUMBER > N2.NUMBER)
	GROUP BY N1.number       ##按照number分组
	ORDER BY N1.NUMBER       ##按照升序排序
) AS A

再算每个数字,在展开后的升序列中,结束下标。

再用排名算法: 表left join自连接,group by对数字分组,按数字升序。

再筛选出比每个数字小于等于的所有其它数字的行。将这些行的中频率值相加。其值即为结束下标,且是开区间。

结果命名为B。

(
	SELECT N1.NUMBER, SUM(IF(N2.frequency IS NULL,0,N2.frequency)) AS `end`
	FROM numbers AS N1
	LEFT JOIN numbers AS N2 ON (N1.NUMBER >= N2.NUMBER)
	GROUP BY N1.number
	ORDER BY N1.number
) AS B

那么,连接表A和表B,关联于相同的数字。每个数字,在展开后升序列中的位置区间为[beg,end)。

SELECT *
FROM 
(
	SELECT N1.NUMBER, SUM(IF(N2.frequency IS NULL,0,N2.frequency)) AS `beg`
	FROM numbers AS N1
	LEFT JOIN numbers AS N2 ON (N1.NUMBER > N2.NUMBER)
	GROUP BY N1.number
	ORDER BY N1.NUMBER
) AS A
JOIN 
(
	SELECT N1.NUMBER, SUM(IF(N2.frequency IS NULL,0,N2.frequency)) AS `end`
	FROM numbers AS N1
	LEFT JOIN numbers AS N2 ON (N1.NUMBER >= N2.NUMBER)
	GROUP BY N1.number
	ORDER BY N1.number
) AS B
	ON (A.NUMBER = B.NUMBER)

再算在展开后的升序列中, 中位数的起始和结束下标

数字总数N是偶数时,下标(N-1)/2和N/2位置处为中位数。N是奇数时,下标(N-1)/2为中位数。

数字的频率之和为N。确定中位数区间[beg,beg+cnt+1),beg从0开始。

beg = (N-1)/2

cnt = 0或1,N为偶数时为1,N为奇数时为0。

结果命名为表C。

(
	SELECT FLOOR((SUM(N.frequency)-1)/2) AS `beg`,
        if(SUM(N.frequency)%2=1,0,1) AS `cnt`
	FROM numbers AS N
) AS C

第三步,筛出落在中位数区间中的数字

已经有每个数字的位置区间S=[A.beg,B.end)。中位数位置区间T=[beg,beg+cnt+1) 。

易知,区间S和区间T相交位置的数字是中位数。

区间S与区间T的长度大小关系有两种。

第一种,T区间长度 >= S区间长度。

那么,判断区间S与区间T是否相交,逻辑为:

if(S的起点落在区间T中 或 S的尾部落在区间T中)
{
    满足此条件的数据行为中位数行
}
(
	(C.beg <= A.beg AND A.beg < (C.beg +C.cnt+1)) 
	OR 
	(C.beg < B.END AND B.END <= (C.beg +C.cnt+1))
)

第二种,T区间长度 < S区间长度。

那么,判断区间S与区间T是否相交,逻辑为:

if(T的起点落在区间S中 或 T的尾部落在区间S中)
{
    满足此条件的数据行为中位数行
}
(
	(A.beg <= C.beg AND C.beg < B.end) 
	OR 
	(A.beg < (C.beg+C.cnt+1) AND (C.beg+C.cnt+1) <= B.END)
)

合起来,判断中位数的逻辑是:

(
	(
		(C.beg <= A.beg AND A.beg < (C.beg +C.cnt+1)) 
		OR 
		(C.beg < B.END AND B.END <= (C.beg +C.cnt+1))
	)
	OR
	(
		(A.beg <= C.beg AND C.beg < B.end) 
		OR 
		(A.beg < (C.beg+C.cnt+1) AND (C.beg+C.cnt+1) <= B.END)
	)
)

连接表A,表B和表C,得出中位数的数字:

SELECT *
FROM 
(
	SELECT N1.NUMBER, SUM(IF(N2.frequency IS NULL,0,N2.frequency)) AS `beg`
	FROM numbers AS N1
	LEFT JOIN numbers AS N2 ON (N1.NUMBER > N2.NUMBER)
	GROUP BY N1.number
	ORDER BY N1.NUMBER
) AS A
JOIN 
(
	SELECT N1.NUMBER, SUM(IF(N2.frequency IS NULL,0,N2.frequency)) AS `end`
	FROM numbers AS N1
	LEFT JOIN numbers AS N2 ON (N1.NUMBER >= N2.NUMBER)
	GROUP BY N1.number
	ORDER BY N1.number
) AS B
	ON (A.NUMBER = B.NUMBER)
JOIN 
(
	SELECT FLOOR((SUM(N.frequency)-1)/2) AS `beg`, if(SUM(N.frequency)%2=1,0,1) AS `cnt`
	FROM numbers AS N
) AS C
	ON (
			(
				(C.beg <= A.beg AND A.beg < (C.beg +C.cnt+1)) 
				OR 
				(C.beg < B.END AND B.END <= (C.beg +C.cnt+1))
			)
			OR
			(
				(A.beg <= C.beg AND C.beg < B.end) 
				OR 
				(A.beg < (C.beg+C.cnt+1) AND (C.beg+C.cnt+1) <= B.END)
			)
		)

在结果集中,

当数字总数N为偶数是,最多有两行数据。

当数字总数N为奇数是,只有一行数据。

因此可知,中位数 = 数据行的数字总和 / 数据总行数 = 数字的平均数

SELECT AVG(A.NUMBER) AS `median`
FROM
...

最终结果:

SELECT AVG(A.NUMBER) AS `median`
FROM 
(
	SELECT N1.NUMBER, SUM(IF(N2.frequency IS NULL,0,N2.frequency)) AS `beg`
	FROM numbers AS N1
	LEFT JOIN numbers AS N2 ON (N1.NUMBER > N2.NUMBER)
	GROUP BY N1.number
	ORDER BY N1.NUMBER
) AS A
JOIN 
(
	SELECT N1.NUMBER, SUM(IF(N2.frequency IS NULL,0,N2.frequency)) AS `end`
	FROM numbers AS N1
	LEFT JOIN numbers AS N2 ON (N1.NUMBER >= N2.NUMBER)
	GROUP BY N1.number
	ORDER BY N1.number
) AS B
	ON (A.NUMBER = B.NUMBER)
JOIN 
(
	SELECT FLOOR((SUM(N.frequency)-1)/2) AS `beg`, if(SUM(N.frequency)%2=1,0,1) AS `cnt`
	FROM numbers AS N
) AS C
	ON (
			(
				(C.beg <= A.beg AND A.beg < (C.beg +C.cnt+1)) 
				OR 
				(C.beg < B.END AND B.END <= (C.beg +C.cnt+1))
			)
			OR
			(
				(A.beg <= C.beg AND C.beg < B.end) 
				OR 
				(A.beg < (C.beg+C.cnt+1) AND (C.beg+C.cnt+1) <= B.END)
			)
		)
GROUP BY C.cnt

解法二

先算在展开后的升序列中, 中位数的起始和结束下标

借鉴解法一,逻辑为:

(
	SELECT FLOOR((SUM(N.frequency)-1)/2) AS `beg`,
        if(SUM(N.frequency)%2=1,0,1) AS `cnt`
	FROM numbers AS N
) AS B

定义用户变量:@fre_sum——数字升序列中频率前缀和,从0开始。

在这里插入代码片
(SELECT @fre_sum:=0) AS C

连接数字表A,表B和表C,并按照数字升序。

(
        SELECT *
	FROM 
	numbers AS A,
	(
		SELECT 
			FLOOR((SUM(N.frequency)-1)/2) AS `beg`, 
			IF(SUM(N.frequency)%2=1,0,1) AS `cnt`
		FROM numbers AS N
	) AS B,
	(SELECT @fre_sum:=0) AS C
	ORDER BY A.number
) AS D

再选出中位数的数字。

每个数字的@fre_sum的值,确定了一个区间S=[@fre_sum,@fre_sum+A.frequency)。

只要中位数区间T=[B.beg,B.beg+B.cnt+1)与区间T相交。相交的数字即是中位数。

判断相交的逻辑:

if(T的起点落在区间S中 或 T的终点落在区间S中){
    此数字是中位数
}
	if(
		@fre_sum<=B.beg AND B.beg < (@fre_sum + A.Frequency),
		1,
		if(
			@fre_sum < (B.beg+B.cnt+1) AND (B.beg+B.cnt+1) <= (@fre_sum + A.Frequency),
			1,
			0
		)
	) AS wanted
@fre_sum:=@fre_sum+A.Frequency AS fre
(
	SELECT A.*,B.*,
	if(
		@fre_sum<=B.beg AND B.beg < (@fre_sum + A.Frequency),
		1,
		if(
			@fre_sum < (B.beg+B.cnt+1) AND (B.beg+B.cnt+1) <= (@fre_sum + A.Frequency),
			1,
			0
		)
	) AS wanted,
	@fre_sum:=@fre_sum+A.Frequency AS fre
	FROM 
	numbers AS A,
	(
		SELECT 
			FLOOR((SUM(N.frequency)-1)/2) AS `beg`, 
			IF(SUM(N.frequency)%2=1,0,1) AS `cnt`
		FROM numbers AS N
	) AS B,
	(SELECT @fre_sum:=0) AS C
	ORDER BY A.number
) AS D

从表D中选出wanted=1的数字,并求平均值,为最终结果。

SELECT AVG(D.NUMBER) AS `median`
FROM 
(
	SELECT A.*,B.*,
	if(
		@fre_sum<=B.beg AND B.beg < (@fre_sum + A.Frequency),
		1,
		if(
			@fre_sum < (B.beg+B.cnt+1) AND (B.beg+B.cnt+1) <= (@fre_sum + A.Frequency),
			1,
			0
		)
	) AS wanted,
	@fre_sum:=@fre_sum+A.Frequency AS fre
	FROM 
	numbers AS A,
	(
		SELECT 
			FLOOR((SUM(N.frequency)-1)/2) AS `beg`, 
			IF(SUM(N.frequency)%2=1,0,1) AS `cnt`
		FROM numbers AS N
	) AS B,
	(SELECT @fre_sum:=0) AS C
	ORDER BY A.number
) AS D
WHERE wanted = 1

知识点:

1.round()函数遵循四舍五入原则,用于把数值字段舍入为指定的小数位数

2.floor(value)函数返回小于或等于指定值(value)的最小整数

3.ceiling(value)函数返回大于或等于指定值(value)的最小整数

  • 0
    点赞
  • 1
    收藏
    觉得还不错? 一键收藏
  • 0
    评论

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值