Construction And Application Of User Model In Document Recommendation System

With the rapid development of the Internet technology and continuous enrichment of the information resources, the amount of information explosively grows. Facing the mass of information, how to find the really required information has become one of the most serious problems in the era of big data. Therefore, using personalized recommendation to provide the required information to users is a research topic with high application value.In this paper, a universal user model is constructed based on the research of user model combining with the LDA topic model, and it is applied to the document recommendation systems to meet the demand of personalized service. Firstly, this paper proposes a method to construct a user topic model and a user behavior model. Based on the analysis of users' interest in reading documents, the user model is constructed by using LDA in two ways. On the one hand, in the perspective of topic interest, a user topic model is constructed by selecting the user interested documents to reflect the user's interest in topic; On the other hand, in the perspective of behavior, a user behavior model is constructed by selecting the score of user's recent documents to reflect the user's interest in behavior.Secondly, this paper proposes a document recommendation method by combining user topic interest and user behavior. According to the user topic model and user behavior model, a document recommendation method based on the user topic model and a document recommendation method based on the user behavior model are respectively constructed. Meanwhile, in order to more comprehensively describe user's interests, this paper adopts the linear weighted method to combine the recommendation methods mentioned above. And the experiment results show that the proposed method is superior to the collaborative filtering recommendation method and the content-based recommendation method.Finally, we design and implement a document recommendation system based on user model. The system can record the user's browsing behavior automatically, and also can construct a user topic model and a user behavior model for the users according to the user's browsing behavior data, so as to recommend the document that the user may be interested in.