Mobility Prediction Using Fully-Complex Extreme Learning Machines.
Lahouari Ghouti · 2014
Abstract. Efficient planning and improved quality of service (QoS) in wireless networks call for the use of mobility prediction schemes. Such schemes ensure accurate mobility prediction of wireless users and units which plays a major role in optimized planning and management of the available bandwidth and power resources. In this paper, fully-complex extreme learning machines (CELMs) model and predict the mobility pat-terns of arbitrary nodes in a mobile ad hoc network (MANET). Unlike their real-valued counterparts, CELMs properly capture the existing in-teraction/correlation between the nodes ’ location coordinates leading to more realistic and accurate prediction. Simulation results using stan-dard mobility models and real-world mobility data clearly show that the proposed complex-valued prediction algorithm outperforms many existing real-valued learning machines in terms of prediction accuracy. 1