Section: New Results

Keywords : clustering, evolving data, web usage mining.

Analysis of evolving Web usage data

Participants : Alzennyr Da Silva, Yves Lechevallier, F.A.T. de Carvalho.

In the analysis of Web usage traces, taking into account the time factor has become a necessity since the subjacent distribution of the data can evolve over time. A typical example concerns the Web surfer usage profiles. These models can be dependent on certain temporal factors (day of the week, period of promotions, etc). Our aim is to analyze the evolution of these profiles which can be related to the change in the number of cluster elements or to the displacement of clusters over time. In the article [43] , we propose three strategies of clustering based on overlapping sliding windows. During the spring school [42] , our work was presented and discussed with computer science students. Our newest results were published in two Springer books [51] [50] .