Personlization-based User Interest Modeling And Its Applying Study

With Web2.0 time arriving, people set the higher and higher request to the information acquisition method and the efficiency. The traditional Internet service pattern is gradually transforming into driving type, personalization, and high efficiency. The personalization technology appeared to solve in certain degree the contradiction between the magnanimous growth of Internet information and the relative crude of user approach to information. As the core question of personalization service, user interest modeling technology concern on effective expression, updating, store as well as the computation. This article bases on a complete Web personalized information recommendation system - C&P, introducing following aspects of our research work:①Further research on user behavior analysis. According to the drives theory in psychology user's behavior of browsing web pages be able to reflect his (or her) interest in some degree. Applying these to the personalized service we may discover some kind of relations between user behavior and the interest user have to the page he(or she) browsed. The traditional method is linear regression, carrying on the regression analysis to the user behavior, however, this article has made the improvement in view of the deficiency of this method, putting forward the Logistic model and the hyperbolic model as tools for analyzing and fitting different kinds of browsing behaviors, namely non-linear user behavior analysis.②Working out a new kind of user profile. Proposing the idea of classified vector for describing user feature based on the content, and introducing above non-linear user behavioral analysis for describing user feature based on the behavior, and further more using the NIC set to filter to the noise which brought in by training process. Basing on these three improvements we get a multi- vectors tree as the logical architecture of our new user profile.③Working out a personalized recommendation system which is basing on the new user profile. This system takes the new user profile model as a core, fully displays the advantages of new user profile such as new multi- vectors feature description, the non-linear behavioral analysis and the NIC set filtering noise.④Massive experiments for the improved user behavior analysis to confirm and revise the mathematical model fitting the user interest, as well as for user description ability and the personalized recommendation ability of our personalized information recommendation system which is based on the new user profile. Our experiments confirmed that the new user profile has improved the service quality of personalization system.Nowadays, the personalized service gradually becomes a hot spot in both the scholarly research and commercial use. The user profile studied and proposed by in this article may use in different fields such as the user personalized information service, the customer information management, the electronic commerce, as well as the data mining.