实时更新问题

[Otaku-gsf]

1. 因为是离线训练，所以是不是无法对用户的反馈实时增量更新推荐内容，必须等待下次训练才会更新？如果缩短训练时间到分钟级，比如 1～5 分钟，在用户量和行为数据相对较多的情况下会产生什么问题？如何配置才能在数据量大的条件下做到近似实时的增量更新？
2. 我看文档中有个不活跃用户更新周期的配置，这个配置的周期如果卡在某次训练前，是不是会出现不活跃用户相当于新用户只能使用兜底的推荐数据？

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

1. Because it is offline training, is it impossible to incrementally update the recommended content in real time based on user feedback? Do I have to wait for the next training to update? If the training time is shortened to the minute level, such as 1 to 5 minutes, what problems will occur when the number of users and behavior data is relatively large? How to configure it to achieve near real-time incremental updates under the condition of large data volume?
2. I see that there is a configuration for the update cycle of inactive users in the document. If the cycle of this configuration is stuck before a certain training, will there be inactive users equivalent to new users who can only use the recommended data?

[zhenghaoz]

实时推荐需要等后续更新

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

Real-time recommendations need to wait for subsequent updates

[Otaku-gsf]

实时推荐需要等后续更新

好的，目前有规划什么时候能上吗？

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

Real-time recommendations need to wait for subsequent updates

Okay, do you have any plans for when it will be available?

[Otaku-gsf]

目前我能不能通过缩短训练时间来达到近似实时的推荐效果呢？这在用户量和行为数据相对较多的情况下会产生什么问题呢？

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

Currently, can I achieve near-real-time recommendation effects by shortening the training time? What problems will this cause when the number of users and behavioral data is relatively large?

[zhenghaoz]

没有任何问题，这是数据库压力会大一些（读写压力变大）

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

There is no problem. This is because the database pressure will be greater (the reading and writing pressure will increase)

[Otaku-gsf]

没有任何问题，这是数据库压力会大一些（读写压力变大）

可以帮我看下这个配置目前可用吗？

• 现有数据：一百万的用户，四十万个物品，五千万的反馈。
• 训练周期: 10 分钟。
• 活跃用户: 7 天（168h）(基于这个有个问题，非活跃用户之前训练的推荐数据会被清除吗)。

[Issues-translate-bot]

Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿

---

There is no problem. This is because the database pressure will be greater (the reading and writing pressure will increase)

Can you help me see if this configuration is currently available?

• Current data: 1 million users, 400,000 items, 50 million feedbacks.
• Training period: 10 minutes.
• Active users: 7 days (168h) (Based on this, there is a question, will the recommendation data previously trained by inactive users be cleared).