统计活跃间隔对用户分级结果
明确题意:
统计活跃间隔对用户分级后,各活跃等级用户占比,结果保留两位小数,且按占比降序排序
问题分解:
- 
计算每个用户最早最晚活跃日期(作为子表t_uid_first_last): - 按用户ID分组:GROUP BY uid
- 统计最早活跃:MIN(DATE(in_time)) as first_dt
- 统计最晚活跃:MAX(DATE(out_time)) as last_dt
 
- 
计算当前日期和总用户数(作为子表t_overall_info): - 获取当前日期:MAX(DATE(out_time)) as cur_dt
- 统计总用户数:COUNT(DISTINCT uid) as user_cnt
 
- 
左连接两表,即将全表统计信息追加到每一行上:t_uid_first_last LEFT JOIN t_overall_info ON 1 
- 
计算最早最晚活跃离当前天数差(作为子表t_user_info): - 最早活跃距今天数:TIMESTAMPDIFF(DAY,first_dt,cur_dt) as first_dt_diff
- 最晚(最近)活跃距今天数:TIMESTAMPDIFF(DAY,last_dt,cur_dt) as last_dt_diff
 
- 
计算每个用户的活跃等级: CASE WHEN last_dt_diff >= 30 THEN "流失用户" WHEN last_dt_diff >= 7 THEN "沉睡用户" WHEN first_dt_diff < 7 THEN "新晋用户" ELSE "忠实用户" END as user_grade
- 
统计每个等级的占比: - 
按用户等级分组:GROUP BY user_grade 
- 
计算占比,总人数从子表得到,非聚合列避免语法错误加了MAX:COUNT(uid) / MAX(user_cnt) as ratio 
- 
保留2位小数:ROUND(x, 2) 
 
- 
细节问题:
- 表头重命名:as
- 按占比降序排序:ORDER BY ratio DESC;;
完整代码:
SELECT user_grade, ROUND(COUNT(uid) / MAX(user_cnt), 2) as ratio
FROM (
    SELECT uid, user_cnt,
        CASE
            WHEN last_dt_diff >= 30 THEN "流失用户"
            WHEN last_dt_diff >= 7 THEN "沉睡用户"
            WHEN first_dt_diff < 7 THEN "新晋用户"
            ELSE "忠实用户"
        END as user_grade
    FROM (
        SELECT uid, user_cnt,
            TIMESTAMPDIFF(DAY,first_dt,cur_dt) as first_dt_diff, 
            TIMESTAMPDIFF(DAY,last_dt,cur_dt) as last_dt_diff
        FROM (
            SELECT uid, MIN(DATE(in_time)) as first_dt,
                MAX(DATE(out_time)) as last_dt
            FROM tb_user_log
            GROUP BY uid
        ) as t_uid_first_last
        LEFT JOIN (
            SELECT MAX(DATE(out_time)) as cur_dt,
                COUNT(DISTINCT uid) as user_cnt
            FROM tb_user_log
        ) as t_overall_info ON 1
    ) as t_user_info
) as t_user_grade
GROUP BY user_grade
ORDER BY ratio DESC;



 京公网安备 11010502036488号
京公网安备 11010502036488号