Research Of Web Usage Mining For Personalized Recommendation System

As E-commerce, Web services and Web-based information system continued growth and development. Growing Internet-based information content and information exchange, it is a serious challenge how to find useful information from this vast amounts of data to provide more personalized service to meet customer needs Thus, Web data mining bred, as an important part of the analysis of user behavior, Web usage mining(also known as log mining) is becoming a research focus. Web usage mining aims gathered from the user click-stream and user data analysis of product design, assess the results page, to optimize the functionality of the Web application, which provides visitors a more personalized content.The server access log data sources using Web usage mining and personalization recommendation related theories, comprehensive analysis of a large number of Web personalized recommendation aspects of the paper, modeling and analysis of Web access behavior to understand online users’information needs, a Web personalized recommendation system consisting of both online and offline system design framework. Which focuses on data preprocessing methods, access mode page clustering, the similarity of the classification of the page to recommend three in-depth study. Specific activities include the following:(1) Necessary to work as a Web log mining, research process and methods of data preprocessing, data cleaning, user identification and session identification technology. According to this system using an appropriate algorithm and apply.(2) Page clustering can be found in the user’s access patterns, according to a previous page clustering (SUGGEST systems) to propose a new clustering algorithm based on graph partition:the establishment of an undirected graph according to the different pages correlation formula, and cluster according using the depth-first search algorithm(DFS).(3) Forecasts and recommended aspects of a calculation of page views sequence similarity based on the recommendation of the program, according to the current user’s active session and the step is stored on the page clustering algorithm LCS similarity matching, come to recommend the page list feedback to the user.(4) Finally, the proposed algorithm to carry out the experiments and analyzing the results of the actual data to prove that the algorithm and design of this paper is practical and effective.