Modeling cellular user mobility using a leap graph

User mobility prediction can enable a mobile service provider to optimize the use of its network resources, e.g., through coordinated selection of base stations and intelligent content prefetching. In this paper, we study how to perform mobility prediction by leveraging the base station level location information readily available to a service provider. However, identifying real movements from handovers between base stations is non-trivial, because they can occur without actual user movement (e.g., due to signal fluctuation). To address this challenge, we introduce the leap graph, where an edge (or a leap) corresponds to actual user mobility. We present the properties of leap based mobility and demonstrate how it yields a mobility trace more suitable for mobility prediction. We evaluate mobility prediction on the leap graph using a Markov model based approach. We show that prediction using model can substantially improve the performance of content prefetching and base station selection during handover.