Characterizing visitor groups from web data streams

The efficacy of a Web site is not only measured in terms of the ingoing traffic or the number of users but rather with respect to the knowledge of the user profiles who have visited it. In this paper we propose an approach for discovering the profiles of visitor groups of a Web site. Such knowledge could be especially useful for business applications. To this purpose, we begin by mapping user interests into symbolic objects which represent the user navigational behaviour resulting from a successful interaction with the site. We then identify groups of users with similar behavior by means of a dynamic clustering algorithm applying a proximity function, which is an original point of view. The convergence of the algorithm is guaranteed at the best partitions of the symbolic objects in k classes. We have applied our approach to identify visitor groups of a Web site in the educational domain and also to analyze the traces of different user behaviour. This application allows us to validate the proposed procedure and to suggest it as a useful tool in the Web Usage Mining framework.