Research On Personalized Recommendation Of Microblog News Based On Probabilistic Topic Model

As the Internet getting popular, traditional media is not the only way for information spreading. Online media has become the mainstream in the Information-digitalized Times. Following high speed, wide coverage, high interactive, microblog, as a new social networking platform, stands out. Microblog has become an important channel for releasing news and reading, exchanging and expressing views. However, because of unlimited resources and limited length, microblog news has very big problems in fragmentation and being noisy. And it is hard for netizen to find news on assigned topic and integrate from the variety of news from microblog. Therefore, the main issue we need to study and solve through this article is: How to filtering and analyzing microblog news effectively and recommending news which the users are interest in.On the base of traditional methods of text topic mining and the characteristics of the topic’s evolution over time, this paper proposes DHPAM, an improved hierarchical probabilistic topic model, to find and modeling topic of microblog news. At the same time,considering the impact of time and hot microblog on user interest, this paper defines the generating and updating mechanism of user preference model and suggests a personalized recommendation which applies microblog news.The detail of the research in this paper including:① Combined with the related research background of microblog news, analyze present status and development trend of probability-topic model and illustrate the questions and practical significance of this paper.② Study algorithms about derived, modeling, evaluation on topic model and recommendation. Related technology research of probabilistic topic model and personalized news recommendation that this paper involves are introduced to offer the theoretical foundation for the later research work.③ Propose an improved model of DHPAM, which is on the basis of time evolution, describe how it forms and works and design the algorithm that applies for microblog news based on modeling. Analyze and evaluate the model’s performance through contrast experiment.④ Classify the user’s preference, combined with the effects of hot topics and time. Improve the author topic model ULLDA and update algorithm about prediction on users’ interest and test its feasibility.⑤ Make experiments on the real data sets of Sina microblog, propose a personalized recommendation and set up off-line experiments to prove that the model proposed in this paper can accurately extract topics of microblog news and realize their multi-level classification.