A New Methodology for User Equipment Trajectory Prediction in Cellular Networks

User mobility prediction can be exploited in cellular networks for different purposes, such as enhancing the handover process, proactive resource allocation, proactive load balancing, etc., in order to improve the network performance. While many works aimed for the prediction of the next cell visited by the User Equipment (UE), the prediction of future UE locations has received less attention. In fact, only a few works deal with the prediction of the next UE location while other few works aim to predict the future direction of UEs arriving at a crossroad. This paper presents a methodology for the prediction of the trajectory followed by the UEs inside the cell. First, UE trajectory patterns are learnt by means of an off-line clustering of historical UE trajectories. Then, the obtained trajectory patterns are used for on-line prediction of the UE trajectory inside the cell, the prediction of the next cell that the UE will visit and an estimation of the time to reach this new cell. A dataset with UE trajectories moving around a large real-life cellular network has been considered.