Enhancing Downlink Throughput Prediction in 4G LTE Networks Using Ensemble Learning Techniques

Abstract

Forecasting downlink throughput in 4G LTE networks is essential for optimizing resource distribution, guaranteeing Quality of Service (QoS), and enhancing overall network efficacy. Conventional models, including statistical methods, frequently do not adequately represent the dynamic and nonlinear attributes of contemporary cellular networks. This paper examines the application of machine learning and ensemble learning methodologies to improve throughput prediction. We evaluate the efficacy of individual models, such as SVM, Decision Tree, MLP, and LSTM, alongside ensemble techniques, including Bagging, Boosting, and Stacking. Our experimental findings indicate that ensemble learning methods substantially surpass solitary models, with Bagging attaining the minimal RMSE. The enhanced efficacy of these methods is ascribed to their capacity to integrate multiple base learners, thereby mitigating overfitting and augmenting generalization in intricate network settings. The results indicate that ensemble learning methods provide a reliable means of improving throughput prediction in 4G LTE networks.