The Research Of User's Interest Model Based On Web Log Mining

E-commerce has become a new business model with the development of the Internet, the competitions of e-commerce are often determined by just clicks of mouse, and how to obtain and keep more relative network customers becomes the focus of competition among various e-commerce. For e-commerce sites can do is to understand the user's interest, and the data source is web server logs. Through analysing and studying the web logs we can find the user's interests, then provide users with personalized services; Through correlation analysis we can identify potential association with the goods, then carry out the "bundling" sales; Also through cluster analysis of the web pages we can provide the basis for structural adjustment for web site.The work described in this thesis is mainly to find a new method to measure the user's interests of web pages. In this thesis, we first analyse the web logs, and then build interest matrix based on web page. Through analysing the interest matrix using the method of cluster analysis, we can get the page clustering and user clustering, and then we can provide recommendations for web site design. As the Web log data is usually large and redundant, the relationship among the page is vague and uncertain, so we use fuzzy clustering method in this thesis for web log analysis. the principal tasks are as follows: (1) Introduce the development and technology and its theoretical foundation of web log mining. (2) In depth study of pre-processing techniques, Analyse the actual data of Web log, and then propose a new measurement method of the user's interest.(3) Propose a two step weight discretization of fuzzy clustering algorithm. Using this algorithm can improve the the degree of association among pages (users), in the paper, the operation of the algorithm is made a detailed and specific description, and supplemented by a example of calculation.(4) On the basis of the work above, combining with a variety of key technologies, we develop a web log mining system, the main functions include data import, data cleaning, keywords Top10 chart shows, as well as pages and users clustering. All the work is to provide practical reference for web site design.