Statistical User Model for the Internet Access

Abstract

A new statistical based model approach to characterize a user's behavior in an Internet access link is presented. The real patterns of Internet traffic in a heterogeneous Campus Network are studied. We find three clearly different patterns of individual user's behavior, study their common features and group particular users behaving alike in three clusters. This allows us to build a probabilistic mixture model, that can explain the expected global behavior for the three different types of users. We discuss the implications of this emergent phenomenology in the field of multi-agent complex systems.