Visual mining of web logs with DataTube2

Abstract

We present in this paper a new method for the visual and interactive exploration of Web sites logs. Web usage data is mapped onto a 3D tube which axis represents time and where each facet corresponds to the hits of a given page and for a given time interval. A rearrangement clustering algorithm is used to create groups among pages. Several interactions have been implemented within this visualization such as the possibility to add annotations or the use of a virtual reality equipment. We present results for two Web sites (1148 pages over 491 days, and 107 pages over 625 days). We highlight the actual limits of our system (9463 pages over 153 days) and show that it outperforms similar existing approaches. © 2009 Springer-Verlag Berlin Heidelberg.