Overview:
In the recommendation system, cold start is the system recommendation. Because there is no user behavior or relatively detailed information, the user does not know his fine-grained interest points when recommending this. This situation is called cold start;
solution:
1. Based on the user's profile, if users can be grouped, it is best to use the group's behavior information for recommendation; if not, then coarse-grained points of interest can be obtained based on the profile;
2. Based on the rules, you can use the rules to customize and then recommend the corresponding rule data. This data has at least the dimension of ip;
3. Based on hotspots, hotspots include hot-selling lists, hot clicks, etc., the default assumption is that the cold person that everyone is interested in must be interested;
If you have other good ideas, please leave me a message