Towards Mathematical Programming Methods for Predicting User Mobility in Mobile Networks

摘要

Motivated by optimal orchestration of virtual machines in mobile cloud computing environments to support mobile users, we face the problem of retrieving user trajectories in urban areas, when only aggregate information on user connections and trajectory length distribution is given. We model such a problem as that of finding a suitable set of paths-over-time on a time-dependent graph, proposing extended mathematical programming formulations and column generation algorithms. We experiment on both real-world and synthetic datasets. Our approach proves to be accurate enough to faithfully estimate mobility on the synthetic datasets, and efficient enough to tackle real world instances.