A new system for discovering usage profiles

Abstract

The exponential growth of information available on the web, makes need to intelligence systems can give required information immediately, more than before. Web Usage Mining (WUM) is one of the web mining techniques, that extracts useful web usage patterns from user's accessed web pages. In web personalization based on WUM, a set of objects like text, product, link and etc. with respect to a user's interests and preferences, recommend to active user. The operation is done with adoption active session of user with discovered usage profiles. Clustering is one of the usage profile discovery methods. In the distance based clustering methods, the precision of distance measure is very important. To effectively provide sage profiles, we have proposed a system that process log files of web servers then extract user sessions and prepare them for clustering and discovering usage profiles. We have introduced a new hybrid distance measure that involved both syntactic similarities between URL's of session and time connectivity between all pairs of URL's in a session. Our experimental results on music machine dataset show that by using new distance measure in proposed system, cluster coherency and accuracy of the usage profiles are increased.