Personalized News Recommendation Based On User's Interest Model

Common news recommendation algorithms are divided into two categories: content-based recommendation and collaborative filtering recommendation.In order to better utilize the advantages of different algorithms,enhance the diversity of recommendation results and effectively avoid the cold start problem,this paper decides to adopt a hybrid recommendation algorithm based on content-based and collaborative filtering.Firstly,the keyword set is refined based on the text content,and the vector space model representing the news text is constructed.Secondly,the user's browsing time,the last browsing time and the user interest are considered to construct a user-interested interest model;and on the other hand,extract a news text that may be of interest to the target user in a browsing record of a similar user to construct a user potential interest model.Then construct a user-weighted interest model by considering the user's existing interest model and the user's potential interest model.Finally,the news to be recommended is divided into two categories according to the release time being less than the current time,and the release time is greater than the current time;the former matches the user's existing interest model,and the latter matches the user's weighted interest model.The real data indicates this article mentions the user interest model is in line with the user's real interest preference,and the news recommendation algorithm based on this model is superior to the conventional algorithm in the performance of multiple indicators.