0

我最近开始使用 GCP，并且正在学习如何使用公共数据集。

我正在尝试使用 SQL 识别 N 天的非活动用户。

您可以在此处找到数据集：https ://developers.google.com/analytics/bigquery/web-ecommerce-demo-dataset

通过文档我发现了以下代码。

/**
 * Builds an audience of N-Day Inactive Users.
 *
 * N-Day inactive users = users in the last M days who have not logged one  
 * event with event param engagement_time_msec > 0 in the last N days 
 *  where M > N.
 */

 
SELECT
  COUNT(DISTINCT MDaysUsers.user_id) AS n_day_inactive_users_count
FROM
  (
    SELECT
      user_id
    FROM
      /* PLEASE REPLACE WITH YOUR TABLE NAME */
      `YOUR_TABLE.events_*` AS T
    CROSS JOIN
      T.event_params
    WHERE
      event_params.key = 'engagement_time_msec' AND event_params.value.int_value > 0
      /* Has engaged in last M = 7 days */
      AND event_timestamp >
          UNIX_MICROS(TIMESTAMP_SUB(CURRENT_TIMESTAMP(), INTERVAL 7 DAY))
      /* PLEASE REPLACE WITH YOUR DESIRED DATE RANGE */
      AND _TABLE_SUFFIX BETWEEN '20180521' AND '20240131'
  ) AS MDaysUsers
-- EXCEPT ALL is not yet implemented in BigQuery. Use LEFT JOIN in the interim.
LEFT JOIN
  (
    SELECT
      user_id
    FROM
      /* PLEASE REPLACE WITH YOUR TABLE NAME */
      `YOUR_TABLE.events_*`AS T
    CROSS JOIN
      T.event_params
    WHERE
      event_params.key = 'engagement_time_msec' AND event_params.value.int_value > 0
      /* Has engaged in last N = 2 days */
      AND event_timestamp >
          UNIX_MICROS(TIMESTAMP_SUB(CURRENT_TIMESTAMP(), INTERVAL 2 DAY))
      /* PLEASE REPLACE WITH YOUR DESIRED DATE RANGE */
      AND _TABLE_SUFFIX BETWEEN '20180521' AND '20240131'
  ) AS NDaysUsers
  ON MDaysUsers.user_id = NDaysUsers.user_id
WHERE
  NDaysUsers.user_id IS NULL;

但是，当用 user_pseudo_id 替换 user_id、用 bigquery-public-data.ga4_obfuscated_sample_ecommerce 替换 YOUR_TABLE 值并更正日期时，我没有得到任何结果。

有没有办法将此代码应用于示例数据集？

4

1 回答 1

1

我可以看到你必须删除行，_TABLE_SUFFIX因为数据集bigquery-public-data.ga4_obfuscated_sample_ecommerce只有一个事件表。

我所做的是获取最大时间戳日期以将其替换为 CURRENT_TIMESTAMP() 以更正日期。

SELECT MAX(TIMESTAMP_MICROS(event_timestamp)) as time
FROM `bigquery-public-data.ga4_obfuscated_sample_ecommerce.events_*

user_id之后，我通过替换user_pseudo_id 并更正了时间和表名来更改您所说的值。

考虑以下方法：

SELECT
 COUNT(DISTINCT MDaysUsers.user_pseudo_id) AS n_day_inactive_users_count
FROM
 (
   SELECT
     user_pseudo_id
   FROM
     /* PLEASE REPLACE WITH YOUR TABLE NAME */
     `bigquery-public-data.ga4_obfuscated_sample_ecommerce.events_*` AS T
   CROSS JOIN
     T.event_params
   WHERE
     event_params.key = 'engagement_time_msec' AND event_params.value.int_value > 0
     /* Has engaged in last M = 7 days */
     AND event_timestamp >
         UNIX_MICROS(TIMESTAMP_SUB("2021-01-31 23:59:55.412363", INTERVAL 7 DAY))
     /* PLEASE REPLACE WITH YOUR DESIRED DATE RANGE */
 ) AS MDaysUsers
-- EXCEPT ALL is not yet implemented in BigQuery. Use LEFT JOIN in the interim.
LEFT JOIN
 (
   SELECT
     user_pseudo_id
   FROM
     /* PLEASE REPLACE WITH YOUR TABLE NAME */
     `bigquery-public-data.ga4_obfuscated_sample_ecommerce.events_*`AS T
   CROSS JOIN
     T.event_params
   WHERE
     event_params.key = 'engagement_time_msec' AND event_params.value.int_value > 0
     /* Has engaged in last N = 2 days */
     AND event_timestamp >
         UNIX_MICROS(TIMESTAMP_SUB("2021-01-31 23:59:55.412363", INTERVAL 2 DAY))
     /* PLEASE REPLACE WITH YOUR DESIRED DATE RANGE */
 ) AS NDaysUsers
 ON MDaysUsers.user_pseudo_id = NDaysUsers.user_pseudo_id
WHERE
 NDaysUsers.user_pseudo_id IS NULL;

它显示的结果是下一个：

在此处输入图像描述

于 2021-11-11T19:38:48.883 回答