Based On User Preference Modeling Personal Interests

With its popularization, the Internet has become an important way for people to share resources and to get access to information. Massive information is not only to enrich people’s life but also bring the problem of overloading information. In this context, at present, the technology of filtering information has been applied widely to the field of personalized recommendation. The key of the personalized recommendation is to build a model of users’ preference, whose purpose is to help users to select interesting information from large number ones. We can calculate the similarity between the users through the users’ evaluation of related resources. Then we can make use of similar users calculated to predict the degree of current users’ preferences. To be able to provide more accurately effective services of information and products for internet users, the exploration on model of users’preferences based on personal interests has a wider application.At present, the representation of users’ preferences has mainly two kinds of methods:the qualitative method and quantitative method. Because the user’s preferences are generally based on certain environment, therefore there are also the studies about preferences based on the context and mixed users. But the expression of existing preferences has many shortcomings, for example:a qualitative model is not a good way to describe directly the relationship of preferences between any two specific users, and the quantitative methods don’t possess the quantization of data, which makes it difficult to construct complex preferences.In order to make up the deficiency of these methods, on the basis of traditional model of users’ preferences, this thesis introduces the concept of personal interest to distinguish the users’ needs of information in different periods by dividing the personal interest into short-term interest and long-term interest. On this basis, the disadvantages of traditional model of users’ preferences are analyzed, and the personal model of preferences with a dynamic feedback is established. This model can describe more adequately users’ interests and demand, and it is suitable especially for rapid changes of personal interests.This thesis works from the following points:(1) This thesis introduces firstly backgrounds, significances and the domestic and foreign current situation of researches about the model users’ preferences.(2) It introduces secondly, from different aspects, the technologies in the construction on models of users’ preferences, including its’ representations, methods and techniques.(3) And then, the construction of models of users’ preferences is described in details, including how to construct the vector about user’s short-term interests and long-term interests, and their interactive relationship and renewing strategy.(4) In the end, there are detailed descriptions about the designed idea of the model of users’ preferences based on personal interests and the relationships between each module in the recommending system. And by comparing the differences of recommending effects with experiments between this model and the traditional one, results show that the new model can better describe the changes of the users’interests, and better provide personalized recommendation.