Research On Correlative Algorithms And Design Of Prototype System Of Web Log Mining

Web mining is a hot research issue which combines various technologies and methods between data mining and WWW. In general, Web mining includes three research fields: Web Content Mining, Web Structure Mining and Web Usage mining. Web usage mining aims to discover the behavior rules while users browse Web site, to improve the site structure and the linkage structure among pages, to enhance the quality of web services, and to provide the decision support on client relationship management of the E-commerce.Having made analysis and researches on web usage mining systemically based on web log records, this thesis improves and proposes three new data mining algorithms, designs a prototype system of Web log mining through applying the improved or proposed algorithms to web usage mining.The main research work in this thesis includes the followings:(1) FAS-Mining algorithm for mining users' frequent accessing patterns is put forward by improving FP-growth algorithm which is used to mining frequent itemsets. FAS-Mining algorithm firstly constructs FAS-tree as the object for mining maximal frequent accessing patterns, and then transfers these patterns into frequent accessing patterns of different depthes on demand. Also, the increment update algorithm of FAS-Mining is analyzed, and finally the effectivity of FAS-Mining is proved by test.(2) A fast algorithm for mining association rules of web pages is proposed, which considers both the web-site organization and information of pages in order to ensure high interesting of association rules and to discover the disparity between the organization of web-site and the association of users' interest.(3) By improving CLOPE algorithm and considering both expanding scope of user and reduce the workload on calculating benefits of clustering, CLOPE-1 algorithm is brought forth. Then the time and space complexity of the CLOPE-1 algorithm are analyzed. Finally, the application of CLOPE-1 algorithm on web log mining is discussed, and a test for CLOPE-1 algorithm is also done on the data with typical structure.(4) A prototype system for web log mining (WLMS) is designed by integrating the existing general-purpose technologies with the improved or proposed algorithms above. Also, the main modules and interfaces of WLMS are described in details.