Abstract

This study investigates the application of deep learning to forecast the quality of service (QoS) of 5G networks, with a specific focus on the Berlin V2X dataset. The Multilayer Perceptron (MLP) model surpassed standard linear approaches, effectively managing non-linearities and intricate patterns. Our approach involves meticulous data preprocessing, deliberate feature selection utilizing Ant Colony Optimization, and a focus on error analysis. This sequence of actions creates a model that exhibits both a high level of accuracy in making predictions and a process of continuous learning and improvement. The post-optimization findings demonstrate that the model has achieved improved performance, with accuracy comparable to sophisticated techniques. This highlights the significant potential of deep learning for accurately predicting quality of service (QoS) in practical 5G scenarios, particularly in varied network conditions. This study establishes novel benchmarks and serves as a foundation for forthcoming investigations, tackling the dynamic obstacles in 5G communications.