移動平均

PostgreSQLで学ぶ時系列分析

Jasmin Ludolf

Content Developer, DataCamp

移動平均とは？

スライディングウィンドウ

移動平均

平均する行範囲を指定
ROWS BETWEEN 句を使用
UNBOUNDED PRECENDING AND CURRENT ROW: 現在行までの全過去行

SELECT 
id,
views,
AVG(views) OVER(
    PARTITION BY id
    ORDER BY ts
    ROWS BETWEEN
    UNBOUNDED PRECEDING AND CURRENT ROW)
    AS running_avg_views
FROM dc_news_fact;

移動平均を計算

直近5行の移動平均を計算
- ROWS BETWEEN 4 PRECEDING AND CURRENT ROW

(1) 1行前

(2) 2行前

(3) 3行前

(4) 4行前

(5) 現在行

移動平均を計算

SELECT
    id,
    ts,
    views,

    AVG(views) OVER (

      ORDER BY ts

      ROWS BETWEEN 4 PRECEDING AND CURRENT ROW) AS moving_average_five

FROM dc_news_fact;

移動平均を計算：結果

|   id|                 ts|views|moving_average_five|
|-----|-------------------|-----|-------------------|
|12121|2015-12-27 17:14:16| null|               null|
|12121|2015-12-27 17:34:16|   23|  23.00000000000000|
|12121|2015-12-27 17:54:16|   10|  16.50000000000000|
|12121|2015-12-27 18:14:16|    8|  13.66666666666666|
|12121|2015-12-27 18:34:16|    4|  11.25000000000000|
|12121|2015-12-27 18:54:16|    7|  10.40000000000000|
|12121|2015-12-27 19:14:16|    0| 5.8000000000000000|
...

現在行を含めない移動平均の計算

ROWS BETWEEN 5 PRECEDING AND 1 PRECEDING

SELECT
    stock_symbol,
    day,
    daily_avg,
    AVG(daily_avg) OVER(
        PARTITION BY stock_symbol
        ORDER BY day
        ROWS BETWEEN 5 PRECEDING AND 1 PRECEDING
        ) AS avg_prev_five
FROM daily_stock_averages
ORDER BY stock_symbol, day;

Let's practice!

PostgreSQLで学ぶ時系列分析