Web Usage Mining Inter-Sites: Analyse du comportement des utilisateurs à impact immédiat

Analyzing access log les is usually called Web Usage Mining and provides very useful knowledge for performance enhancement, restructuring a Web site and customer targeting in electronic commerce. These techniques are however limited to a single Web server log le analysis (Intrasite Web Usage Mining). We propose in this paper the Web Usage Mining Inter-Sites notion in order to study the behaviours of users shared by several partner Web sites. The provided knowledge is thus used to dynamically improve the hypertext organization of a Web site, with two main advantages over the classic approach: more accurate frequent behaviours are found and the result allows an immediate impact. Finally, we prove that there exists a correlation between the properties of the frequent patterns, at the intersection of the considered data, and the user's minimum support. Through this study, we are able to prove that a reduction from the Web Usage Mining Inter-Sites problem to the incremental data mining problem, allows the knowledge discovery phase to be optimized.