Machine Learning-Based Handover Performance Prediction for Beyond 5G Communication Networks

With the introduction of 5G network systems, massive connectivity, low latency, and high reliability in network systems are provided to ensure better quality of service (QoS) for subscribers. However, some problems may emerge as a result of the introduction of 5G networks and systems besides their advantages. Mobility management issue is one of the problems that emerged beyond 5G (B5G) network systems and has gained significance for providing reliable connection and uninterrupted service in B5G network systems by telecommunication operators. The problem leads to frequent handovers (HO) for a mobile UE that will eventually result in degraded QoS for customers. Thus, the problem needs to be handled by a sophisticated approach to manage and optimize network resources. In this study, we predict the system performance of a mobility scenario between the user equipments (UEs) and base stations (BSs), through machine learning (ML) deployment. We use handover margin (HOM) and time-to-trigger (TTT) values as inputs, and handover ping-pong (HOPP) and handover failure (HOF) as outputs in ML deployments via two cross-validation schemes. According to the obtained results, the ML methods with nested cross-validation scheme achieve better results concerning the conventional scheme in terms of mean absolute error (MAE) and mean square error (MSE) metrics.