[Question] about: Solve user cold-start problems

[phamkhactu]

Thanks for great repo. When I read the readme, it shows that: "model can face with poor user", but with your example, it contains the users - items and interaction with my define: User cold-start problems: When there is almost no information available about the user and all are new for system, the user cold-start problem arises. How to solve it if use Your Resys Thank you

[wangzhen38]

Thx for attention, for your problem, I can give the following advice:

• Provide non-personalized recommendations.
• Using user registration information.
• Choose the right item to trigger the user's interest.
• Use the content information of the item.
• Labeled by experts.
• Use the user's mobile phone and other interests to perform cold start.
• Use data that users have deposited elsewhere for cold start.

And if you want to combine our repo to solve the cold start problem, may be you can try the MetaHeac model, a kind of meta-learning, which can learn the connection between different tasks, so that it can quickly adapt to new categories of content promotion tasks.