Algorithm 3 Collaborative filtering-based personalized recommendation algorithm

Require:  User historical behavior; Results of the food public opinion prediction model Ensure:  Personalized recommendation list  1.Calculate the similarity matrix between user u and other users based on the historical ratings of all users for different news items.  2.Select the user $vi$ most similar to user u, and based on the most similar user’s preference for the target news, predict and calculate the specific rating value of user u for the target news item.  3.Calculate the early warning index $ϵ$ for all news items based on the sentiment prediction model, and compute the final user recommendation matrix.  4.Recommend news based on the rating values from highest to lowest.