Time Aware Web Users Clustering

Web users clustering is a crucial task for mininginformation related to users needs and preferences. Up to now,popular clustering approaches build clusters based on usagepatterns derived from users’ page preferences. This paper emphasizesthe need to discover similarities in users’ accessing behaviorwith respect to the time locality of their navigational acts. Inthis context, we present two time aware clustering approachesfor tuning and binding the page and time visiting criteria. Thetwo tracks of the proposed algorithms define clusters with usersthat show similar visiting behavior at the same time period, byvarying the priority given to page or time visiting. The proposedalgorithms are evaluated using both synthetic and real datasetsand the experimentation has shown that the new clusteringschemes result in enriched clusters compared to those createdby the conventional non-time aware users clustering approaches.These clusters contain users exhibiting similar access behaviornot only in terms of their page preferences but also of their accesstime.