User Browsing Interest Prediction And Personalization Recommendation Strategies Based On WEB Usage Mining

Along with the widespread popularization of computers and the rapid development of Internet technology, more and more consumers began to pay close attention to and choose the burgeoning online shopping, enjoying an express and convenient shopping experience just at home. On the other hand, for e-commerce sites, the rapid expansion of online shopping also brought huge challenge, not only the requirement for their commercial standing and merchandise quality but the needs of their website development and platform construction, which had become the key decisive factors, in a sense, determining their survival and further development.In this context, the design and implementation of personalized recommendation system based on web dining become a hot issue researched by scholars. It can not only help e-commerce platform response to individual needs of consumers quickly but also reduce users'consumption cost and increase consumers'satisfaction by appropriate and accurate recommendation. Recently, many scholars regard their study focused on the key technology of personalized recommendation system, which is the prediction of users'browsing behavior. They have also done a lot of useful analysis and research on how to improve the accuracy of prediction and the response time. Accurate prediction can complete web page prefetching and improve the performance of site. Besides, it can conduct recommendation of users'interested page and provide personalized service. Although Markov model has been widely used due to its high prediction accuracy, the conflict issue of computing speed and prediction accuracy between first-order and multi-order Markov chain still needs further study.Therefore, this paper mainly focus on the design of user browsing behavior predictive models which is based on Web usage mining, and a mixed Markov model combined with the dynamic clustering has been presented on the basis of referential review of domestic and foreign bibliography. Taking the influence of users browsing behavior and page features on the users browse interest into account, the paper firstly introduced user browsing interest as clustering parameter, and built a Logistic forecasting model of user browsing interested, grounded on the results of factor analysis. Then established Markov transition matrix for different categories of users, and the validity of the model was verified by Matlab7.0 platform.