The Application Of Web Log Mining On Personalized Information Recommendaiton

Along with the Internet rapidly expanding, each kind of information is swiftly and violently increasing, too. Facing magnanimous information, peoples are unable to choose and digest frequently, they don't know how to discover their needed information resource more conveniently, more quickly and more effectively. At present, Web systems provide the same services for all users, their typical service mode is promulgating the same information for all users through establishing a Web site, however the Web users'demand is infinitely varied, they hope that the Web system could provide personalized different services according to their characteristics. Web log mining technology applied in the personalized information recommendation service, could recommend the needed information on own initiative for the users according to their access patterns. Its appearance will solve the problem that peoples are difficult to seek for the information they need.In this dissertation we studied personalized information recommendation technology based on the Web log mining, i.e., discovering the preferred browsing patterns from the users'access log files, and providing the recommendation service to the users. First, preprocessing to the primitive log files were carried on to clear the dirty data. Next, based on the Web log preprocessing results and association rules algorithm in Web log mining, using the notion of support-preference degree, an algorithm of discovering preferred browsing paths based on access frequency and browsing time was proposed. Finally, the personalized information recommendation algorithm was proposed based on the algorithm, which provides the personalized recommendation Webpages for the users according to their current access Webpages by establishing sliding window for instructing users' browsing behaviors.The experimental results indicate that the proposed algorithm based on access frequency and browsing time is superior to the algorithm based only on access frequency as browsing attribute, and the F-measure of recommendation Webpages is obviously enhanced.