Cellular Network Traffic Analysis and Forecasting By Using LGBM

The rapid growth of cellular traffic due to the rapid expansion of smartphone devices and multimedia applications necessitates accurate traffic forecasting to provide QoS. In this work, a new ML system is introduced aimed at predicting cellular traffic loads efficiently with minimal computations and data utilization. Unlike traditional models like LightGBM, SVR, and LR, the study employs a Random Forest Regressor to achieve low-computation cost high-accuracy predictions. The design is based on a vast dataset with performance metrics like user throughput, latency, and resource utilization from real cellular networks. Performance is estimated by MSE and compared to other approaches to present the proposed approach's capability of enhancing traffic prediction, optimization of network resource, and improvement in QoS for dynamic networks.