The access patterns of the users of a web-site are traditionally analyzed in order to facilitate the user access to the site's information. In this study, however, a systematic approach is introduced in order to analyze the users' navigation paths to the advantage of the web-site owner. Brie y, as users navigate through a web-site, it seems that they are transparently lling a questionnaire designed a priori by the web-site owner. To achieve this, we rst cluster the users who navigate similar paths employing the Path Mining algorithm. Next, the correlation between a set of target questions and the structure of the WWW-site is quanti ed. This has been done by borrowing the concept of channel from information theory. Finally, we compute the probability of a user's certain answer to each question, given the user's navigation path. The accuracy of the computed probabilities highly depends on the channel parameters. Hence, we describe a learning process based on a set of...
Amir M. Zarkesh, Jafar Adibi, Cyrus Shahabi, Reza