Research On Personalized News Recommendation Algorithm Of Tripartite Graph Based On Topic Model And Time Weight

With the rapid development of network technology and industry application,the data in the network is increasing day by day,which makes people feel helpless when facing the explosive growth of network data,and it is difficult to find the information they really need and are interested in.Therefore,personalized news recommendation system emerges as the times require.The mission of news recommendation system is to find the information that users are really interested in in the vast ocean of information.However,up to now,news recommendation system has not been able to meet the growing needs of users in terms of recommendation accuracy and personalization,which is manifested in the problems of text feature extraction,data sparsity and user interest drift,etc.How to effectively solve these problems has become an important research direction of news recommendation system.In order to solve these problems,this paper introduces LDA(late Dirichlet allocation)topic model to extract the topic features in news text,deeply studies the ternary relationship based on "user news text news topic",and combined with the temporal context information,makes a lot of work and in-depth research on news topic model and recommendation method.The main research contents and achievements are as follows:(1)This paper proposes a personalized news recommendation algorithm based on topic model.It extracts multiple news topics from a single text by using LDA topic model,and uses these topics and news texts as well as users to build a tripartite graph recommendation model.The initial resource value of news text node is initialized by defining the importance weight of user node and news topic node.Based on this,a weight based material diffusion algorithm is proposed Finally,the accuracy of recommendation result is improved by running the weight based material diffusion algorithm in the tripartite graph model.The simulation results show that the method can alleviate the problem of data sparsity in the recommendation system to a certain extent.(2)In order to solve the problem of recommendation accuracy and user interest drift,by simulating the real scene of the recommender system and studying the influence of time context information on the recommendation accuracy of the recommendation system,a new algorithm is proposed by integrating the time weight into the personalized news recommendation algorithm of tripartite graph.In the "user-news text" side and "news text-news topic" side of the tripartite graph recommendation system,two distinct time attenuation functions are integrated to alleviate the impact of time effect on the recommendation system.The simulation results show that the algorithm can solve the problem of user interest migration and improve the recommendation accuracy.(3)The experimental results show that the two tripartite graph personalized news recommendation algorithms based on topic model and time weight are better than other traditional news recommendation algorithms.This has a certain reference significance for the practical system application of the industry.