Predicting a User's Next Cell Using Supervised Learning Based on Channel States

Knowing the next cell of wireless users allows more efficient resource allocation and enables new location aware services. In this work, we propose a machine learning-based prediction system that uses Channel State Information (CSI) to predict the cell an UE will hand-over to. This classification problem is solved by Support Vector Machines (SVM), which are embedded in our efficient pre-processing structure. Simulation results from a Manhattan Grid scenario and from the Frankfurt radio map indicate that our system provides early prediction at high accuracy.