Design And Implementation Of Web Log Mining Prototype System

Web Log Mining aims to use data mining techniques for the data which are uservisiting Log in the Web server to explore the hidden rules and models which behind thelog data. We can extract useful knowledge we need from a large number of Web loginformation, and then improve the Web site structure and increase the site’s servicequality, improve site performance. And thus we can provide users with personalizedservices.Based on the analysis of progress and development trend，this paper focus on userclustering、association rules and frequent access path which are in the Web Log Mining.Based on the previous studies, we made some improvements on the algorithm, and thenon the VC6.0, designed and implemented a Web log mining prototype system.On the User clustering,we use the page access time as the user interest standardmeasure, and use fuzzy clustering method to get user clustering from UserID-URLassociated matrix. On the association rules, we propose a method to scan the itemdatabase to replace the transaction database, and according to features of Web log wedelete the home and the second page items, improve the efficiency of the algorithmsignificantly, Quickly find sets of frequently accessed pages. On the frequent accesspath Mining. Based on the MFR, we achieve the Sim_Apriori method that is similar ofApirori, and propose a multi-tree M_tree mining algorithms, effectively improve thespeed of the mining user access frequently path.Finally, summarized the research results and proposed the future further research.