User Model-Based Personalized Recommendation Algorithm for News Media Education Resources

Traditional recommendations for news and media education resources usually ignore the importance of sequential patterns in user check-in behavior and fail to effectively capture the complex and dynamically changing interests of users. As a result, this study provides a recommendation model for news and media education materials based on a user model. To capture changes in users’ interests, the model can represent and fuse short-term and long-term preferences separately. For short-term preferences, a long- and short-term memory network incorporating spatiotemporal contextual information is proposed to learn complex sequential transfer patterns in users’ check-in behaviors and further extract short-term preferences accurately through a goal-based attention mechanism. A user attention-based approach is utilized to capture fine-grained links between users and interest points for long-term preferences. Finally, experimental simulations are conducted on two datasets, Foursquare and Gowalla. The results show that the proposed user model-based recommendation model for news media education resources has better performance compared with the mainstream recommendation methods on different evaluation criteria, which validates the effectiveness of the proposed model.