如果要多天,则需要GROUP BY子句中的日期和小时.目前,它是对几天内同一小时的所有值进行平均. (今天的11,昨天的11等的值相加得出平均值)
我没有一种简单的方法来测试它,但是类似：
SELECT
TRUNCATE(AVG(TIME_TO_SEC(TIMEDIFF(END_DATE, START_DATE))/60), 2) as elapsed,
EXTRACT(DAY FROM END_DATE) as day,
EXTRACT(HOUR FROM END_DATE) as hour
FROM
TIME_INFO
WHERE
EXTRACT(HOUR FROM END_DATE) BETWEEN 9 AND 18 AND
DATE(END_DATE) > CURDATE() - INTERVAL 3 DAY
GROUP BY
EXTRACT(DAY FROM END_DATE),
EXTRACT(HOUR FROM END_DATE)
ORDER BY
END_DATE DESC
如果您希望它跨月/年边界运行,则可能需要DATE(END_DATE)而不是EXTRACT(DAY FROM END_DATE),在这种情况下,输出可能类似于：
elapsed date hour
-------------------------
2.5 2011/03/25 11
1.7 2011/03/25 10
2.4 2011/03/25 9
1.9 2011/03/24 18
2.4 2011/03/24 17
4.0 2011/03/24 16