Question

I want to calculate a 12 month moving average from a MySQL column. The data represents time-series power measurements, it is a largish dataset (every 10 minutes for several years). A high performance query would be nice but speed is something I can work out later.

DTE                  Active
2012-1-3 00:10       500
2012-1-3 00:20       520
... etc

The following query gives me the total Active for each month :

SELECT YEAR(DTE) AS year, MONTH(DTE) AS month, SUM(Active)/6 as total FROM saturne s GROUP BY YEAR(DTE), MONTH(DTE)

The following query gives me the moving average Active for a given month and year – say october 2011

SELECT SUM(Active)/6 AS average FROM saturne 
WHERE (YEAR(DTE) = 2011 AND MONTH(DTE) <= 10) OR (YEAR(DTE) = 2010 AND MONTH(DTE) > 10) 

I would however like to generate a query which returns the monthly total and the 12 month moving average in the next column.

year        month        total        average
2012        2            701474       9258089
2012        1            877535       9386664
... etc

(The factor of 6 is because the data represents instantaneous power recorded every 10 minutes, dividing the total by 6 gives the total energy)

Answer 1

Try:

SELECT YEAR(GDTE) AS year, 
       MONTH(GDTE) AS month, 
       SUM(case when i=0 then Active end)/6 as total,
       SUM(Active)/(MAX(i+1)*6) as avg
FROM
(select s.*, DATE_ADD(DTE, INTERVAL m.i MONTH) GDTE, m.i
 FROM saturne s
 cross join (select 0 i union select 1 union select 2 union select 3 union 
             select 4 union select 5 union select 6 union select 7 union 
             select 8 union select 9 union select 10 union select 11) m
) sq
WHERE GDTE <= curdate()
GROUP BY YEAR(GDTE), MONTH(GDTE)