Activity-based user modeling in wireless networks

Wireless network research still lacks methods to evaluate the performance that can be expected from application layer protocols. User behavior is the predominant factor affecting network performance on this layer. It has two aspects: user mobility and user network usage. These aspects are not orthogonal, but highly correlated: a user's mobility pattern will influence her usage of the network. Existing approaches, however, reduce the modeling of user behavior to analytical mobility models and network traffic models, thereby separating these intertwined parameters. This paper demonstrates how the use of an integrated view based on the users' real-world activity can explain network-relevant parameters both with respect to mobility and to network usage and thereby allows a more natural modeling of user behavior. The evaluation within a campus scenario shows that such an activity based model captures the motion and service usage much more realistically than existing models.