User Interest Mining And Personalized Recommendation Based On Socialized Annotation

With the rapid development of Internet technology,bringing great convenience to people's life,the explosive growth of network information resources has led to the "information overload" phenomenon,and the application of personalized recommendation technology makes the personalized needs of people with different interests and preferences are met,so there are more and more researches related personalized recommendation.With the development of Web2.0 used in social tagging system,users freely tag the cyber source,the tag not only reflects the user's preferences,but also reflects the characteristics of attributes and potential resources,this advantage is very suitable to make personalized recommendation.Firstly this paper analyzes the domestic and foreign research situation based on social tagging and application of personalized recommendation,summing up the results of previous studies and theoretical analysis,the main problems existing in the current research include :(1)the semantic labels with its own ambiguous problems,affect the quality of personalized recommendation;(2)the current research on the label characterization of user interested feature,just considering the relationship between the user and the label from the number and structure,unable to accurately express the user interest feature;(3)with the increase of user annotation activity,the increase of computational scale reduces the efficiency of recommendation system.To solve the above problems,considering the characteristics of this tagging system in society,based on socialized annotation for user interest mining,a personalized recommendation algorithm based on user interest model is constructed,and the improved algorithm proposed in this paper are verified by experimental research.The main work completed in this paper is as follows:(1)Tagging latent semantic topic mining by fusing social relations.Analysis of the social tagging latent semantic relations between users,resources,tags,and the influence of social relationship on its user tagging behavior,the LDA model is applied to mining user tags based on latent semantic topics,this method can solve the fuzzy semantic label,provide a good basis for the recommendation user interest modeling and personalized high quality.(2)Multi user interest model construction based on tag topic.Through the analysis of characteristics of user interest model tagging system,construct the user interest model of a multilayer,multi label theme based on the dimensions of the components,and realize the dynamic update of user interest model based on the forgetting factor,the expression of the user's interest and preference the model better,to ensure the real-time and effectiveness of the recommended.(3)Personalized recommendation algorithm based on user interest model.With the help of user interest model,the similarity between users is calculated,the computational dimension is reduced,and the sparsity is reduced;meanwhile,the recommendation resource is searched in the range of user interest,and the quality and efficiency of the recommendation algorithm are guaranteed and improved.Finally,collecting the data from social tagging system Cite Ulike site,and verify that the improved algorithm proposed in this paper accurately expresses user interest preference,and effectively improves the accuracy of personalized recommendation.