Data Mining Research For Personalized Information Service

With the acceleration of the social information process, people's life has being changed unprecedentely by the Internet. The rapid growth of network information leads to the massive information resources. Owing to lack of search skills and navigation backwardness, users can not make efficient use of network resources. Facing the current situation of network information services, people are seeking a kind of active information service which recommends the information they are interested in, which is called personal information service.The essence of personalized information service is "offering information for users, and serving according to requirements", which means providing specific information service according to specific needs of users, making the information filtering based on information needs of various users. Therefore, how to better explore the demanding of users will be the key point of improving personalized information services.Taking information demands of users into account, this paper did a research on user interest degree based of user's browsing behavior, improved the traditional collaborative filtering recommendation Algorithm, then proposed a personalized recommendation model based on users'browsing behavior. The main research works of this paper are as follows:(1) Analyzing the relationship between the user browsing behavior and the user interest degree, creating user interest model to digitize the data of user's information demands and better support the following data mining process.(2) Considering the deficiency in real-time ability of Collaborative Filtering Recommendation Method (the low speed of nearest neighbor query for the active user), this paper suggested using the K-means clustering method to reduce users space, which increased the speed of nearest neighbor query and got a more typical resultsA grading system was applied to the personalized recommendation model to assess user interest degree based on browsing behavior, which strengthens real-time ability of the recommendation through K-means clustering method. Compared with the traditional personalized recommendation model, this one had higher speed and better accuracy, which provided important guiding value for many personalized information servers.