Predictive Mobility-Based Call Admission Control

The increasing demand for mobile data services places strong pressure on cellular networks to use limited radio resources efficiently while maintaining reliable service during user mobility. Conventional Call Admission Control (CAC) typically makes admission decisions based on the current network state. It does not explicitly consider how cell load may change in the near future or whether users are likely to move into neighboring cells through handover. This thesis addresses this limitation by developing a Predictive Mobility-Based Call Admission (PM-CAC) framework for LTE networks using operational network measurements and user-equipment mobility data. The proposed framework consists of three integrated components. Model A predicts the total downlink Physical Resource Block (PRB) demand of each LTE cell one hour ahead using Random Forest regression. On the chronologically held-out test set, it achieves a mean absolute error of 4.736 PRBs, a root mean squared error of 6.812 PRBs and an R2 of 0.866, outperforming a persistence baseline. Model B uses a two-stage probabilistic approach to estimate handover occurrence and rank likely target cells. The handover-occurrence model achieved a ROU-AUC of 0.844 and a PR-AUC of 0.842, while the final target-ranking model achieved an exact-cell ROU-AUC of 0.918 and a Top-3 accuracy of 0.8235% Model C combines the future-load predictions from Model A with the joint handover probabilities from Model B to perform probability-weighted PRB reservation and admission control. Integer PRB reservations are assigned using the Webster/Saint-Laguë method. A paired Monte Carlo evaluation of 10,000 common scenarios selected a final reservation-strength parameter of α = 1.00, reducing handover-failure probability from 0.024621 under conventional CAC to 0.23072 while maintaining a lower call-blocking probability of 0.0258 compared with 0.0277 for conventional CAC. The results show that combining short-term traffic prediction with probabilistic mobility information supports more proactive, resource-aware admission control in LTE networks.