Personalized Recommendation Technology Research And Application Of Network Resources

With the development of the technology of Web2.0and e-commerce, the Internet has already deeply affected people’s production and every aspects of life. The paper is supported by the "13115" scientific and technological innovation program of Shaanxi Province: document digitization and resource sharing platform. Facing the demand of the user for resources of individuality and diversification, comparing and analyzing the advantages and disadvantages of various personalized recommendation algorithm, the paper puts forward am improved cooperative filtering algorithm, which based on cloud model and time weighted, and puts its application into the program. The main research includes following parts:This paper illustrates the definitions and the system structure of the recommend system in detail; introduces the traditional collaborative filtering system and analyzes its strength and its weakness; formulates the collaborative filtering algorithm which based on users and items; improves the traditional collaborative filtering recommendation algorithm by introducing Web logging model, user-interesting model, cloud model, etc. The improved collaborative filtering algorithm based on cloud model and the time weighted, shorted for CT-CF, is revealed.In this paper, we calculate the interest degree matrix through analyzing the Web log, combined the explicit resources ratings with implicit score of users browsing-time. The paper introduces interest-theme-threshold factor in the extraction of resource thematic characteristics. In the process of recommendation, at first, the CT-CF calculates similarity based on cloud model; and secondly finds the target user’s neighbor set through the nearest neighbors set to predict user rating for resources, which are not browsed by users; at last, the algorithm generates the recommendation lists combined the user’s interested theme with the resources score.According to concrete business logic characteristics of document digitization and resources sharing platform, the paper design the detailed data structure, the modules functions and the main process to fit the personalized network resources recommendation system. The paper applies the CT-CF to project, and the actual experiment results show that this algorithm does improve the accuracy of user recommendation as well as avoids the negative effects of sparsity and largely increases the quality and accuracy of recommendation.