移动平均与移动总计

PostgreSQL 汇总统计与窗口函数

Michel Semaan

Moving averages

概览

  • 移动平均(MA):最近 n 个周期的平均值
    • 例:销售中10日MA为最近10天售出数量的平均
    • 用于指示动量/趋势
    • 也可用于消除季节性
  • 移动总计:最近 n 个周期的总和
    • 例:最近3届奥运会奖牌数之和
    • 用于衡量表现;若总和下降,则整体表现下降
PostgreSQL 汇总统计与窗口函数

源表

查询

SELECT
  Year, COUNT(*) AS Medals
FROM Summer_Medals
WHERE
  Country = 'USA'
  AND Medal = 'Gold'
  AND Year >= 1980
GROUP BY Year
ORDER BY Year ASC;

结果

| Year | Medals |
|------|--------|
| 1984 | 168    |
| 1988 | 77     |
| 1992 | 89     |
| 1996 | 160    |
| 2000 | 130    |
| 2004 | 116    |
| 2008 | 125    |
| 2012 | 147    |
PostgreSQL 汇总统计与窗口函数

移动平均

查询

WITH US_Medals AS (...)

SELECT
  Year, Medals,
  AVG(Medals) OVER
    (ORDER BY Year ASC
     ROWS BETWEEN
     2 PRECEDING AND CURRENT ROW) AS Medals_MA
FROM US_Medals
ORDER BY Year ASC;

结果

| Year | Medals | Medals_MA |
|------|--------|-----------|
| 1984 | 168    | 168.00    |
| 1988 | 77     | 122.50    |
| 1992 | 89     | 111.33    |
| 1996 | 160    | 108.67    |
| 2000 | 130    | 126.33    |
| 2004 | 116    | 135.33    |
| 2008 | 125    | 123.67    |
| 2012 | 147    | 129.33    |
PostgreSQL 汇总统计与窗口函数

移动总计

查询

WITH US_Medals AS (...)

SELECT
  Year, Medals,
  SUM(Medals) OVER
    (ORDER BY Year ASC
     ROWS BETWEEN
     2 PRECEDING AND CURRENT ROW) AS Medals_MT
FROM US_Medals
ORDER BY Year ASC;

结果

| Year | Medals | Medals_MT |
|------|--------|-----------|
| 1984 | 168    | 168       |
| 1988 | 77     | 245       |
| 1992 | 89     | 334       |
| 1996 | 160    | 326       |
| 2000 | 130    | 379       |
| 2004 | 116    | 406       |
| 2008 | 125    | 371       |
| 2012 | 147    | 388       |
PostgreSQL 汇总统计与窗口函数

ROWS 与 RANGE

  • RANGE BETWEEN [START] AND [FINISH]
    • ROWS BETWEEN 功能几乎相同
    • RANGEOVERORDER BY 子句中的重复值视为一个实体

| Year | Medals | Rows_RT | Range_RT |
|------|--------|---------|----------|
| 1992 | 10     | 10      | 10       |
| 1996 | 50     | 60      | 110      |
| 2000 | 50     | 110     | 110      |
| 2004 | 60     | 170     | 230      |
| 2008 | 60     | 230     | 230      |
| 2012 | 70     | 300     | 300      |
  • 实务中几乎总是用 ROWS BETWEEN 而非 RANGE BETWEEN
PostgreSQL 汇总统计与窗口函数

Passons à la pratique !

PostgreSQL 汇总统计与窗口函数

Preparing Video For Download...