Proactive Handover Prediction in Mobile Networks Using Machine Learning Techniques

Mobility management is a major challenge in cellular networks, where frequent handovers can reduce service quality, especially for moving users. Traditional handover methods are mostly reactive and are triggered only after signal reduction happens. This paper introduces a proactive machine learning-based framework to predict handover events before they occur. This early prediction helps the network prepare resources in advance and reduce service interruption. Overall, the framework supports more stable communication, especially in high-mobility scenarios. The suggested approach uses time-ordered user equipment (UE) measurement data, including radio signal and mobility features such as RSRP, SNR, and UE speed. A Random Forest classifier is trained to estimate the probability of an upcoming handover using a multi-horizon labeling method, with a main focus on a 10 -sample prediction window. The predicted probability is compared with an adjustable decision threshold to allow early handover triggering. The framework is evaluated using classification metrics and proactive lead-time analysis. The results show that the suggested method can accurately predict handover events several samples early, providing reliable early predictions and better timing for handover decisions.