Abstract
Knowledge and prediction of system occupancy in wireless cellular systems such as UNITS can play an important role in the optimization of resources allocation procedure. In fact, a resource allocation scheme can make use of such knowledge to anticipate system overloads. On the other hand, studies show that a user's movements in a cellular network can be predictable. Thus, handover prediction methods can be used to improve resource allocation scheme. Two fundamentally different approaches can be found in handover prediction: cell approaches that make use of global knowledge of user traffic flows in a particular cell; and user approaches that rely on the knowledge of user mobility in short or long term. Short-term methods that predict user speed and location are not studied here. Long-term methods are based on the learning of users' behaviors by computing profiles for each user. In this paper, we describe two handover predictions methods: cell and user profile based. We also propose a simulation model that enables a comparison of the two approaches.
