User Access Pattern Mining And Adaptive Web

Now the internet application based on the B/S is widespread day by day, the function and service that each kind of websites provide is more and more rich, however it's insufficient in the intellectualized aspect. For instance, the website structure lacks the flexibility; the distribution of service or information is unreasonable that has not considered the need of users fully; the access way of information hasn't considered the importance and the universal attention and so on. If the websites can study themselves and improve their organization and presentation automatically, as is called adaptive web site, they will be more convenient and more effective for the users.Based on the analysis of structure and meaning of Web log records, combined with the website of Graduate School of Shandong University at Weihai and its partial logs, this thesis gives a detailed description about the whole course of Web user access patterns mining, and proposes three constructive methods.This thesis focus on the user access patterns mining based on the graph structure and the application on the adaptive websites. Based on the analysis of structure of websites and combination of technique of data mining, database and WEB, adjusting and improving the existed websites' structure and its web pages to enable them to be self-adapting and optimizing through the data mining of user access patterns, so as to provide users with information and service in a more efficient way. First, it analyses the data structure of the log and data preprocessing such as the data cleaning, user identification, and session identification and so on before web log mining. Second, it presents a method to mine user access patterns based on the topology of a web site (graph structure) on the user session file, this method is different from the common one based on the tree structure, and reduce the number of the candidate in the period of the candidate generation and pruning, so it improves the precision of the patterns and mining efficiency, and avoid the "corrupted transaction". At the same time it also considers the problem of the incremental log mining technology. Finally, adjusting and improving the structure and pages of the exist website through the rules and patterns from web log mining, including changing the link position, increasing ordeleting the link of pages and recommending the pages which have similar subject to the user group with similar browsing patterns and so on, so it realizes the optimization of the website organization structure.The method is of a high value for research and application, which considers many aspects about data mining analysis so that it can improve the website organization structure dynamically and make the performance and quality of the service better.But it's insufficient both in theory and in practice, which use the user access patterns to adjust and improve the structure and pages of websites. It will be significant that how to combine the results and get the access patterns by dispersing the whole web log mining task to every web sever.