Machine Learning-Based Fifth-Generation Network Traffic Prediction Using Federated Learning

Guardado en:

Detalles Bibliográficos
Publicado en:	International Journal of Advanced Computer Science and Applications vol. 16, no. 1 (2025)
Autor principal:	PDF
Publicado:	Science and Information (SAI) Organization Limited
Materias:	Wireless networks Machine learning Performance evaluation 5G mobile communication Root-mean-square errors Quality of service architectures Communications traffic Resource allocation Network latency Transmission rate (communications) Data transmission Federated learning Predictions Data collection Deep learning Computer science Regression analysis Communication Traffic flow Privacy Time series Neural networks Computer engineering Subscriptions
Acceso en línea:	Citation/Abstract Full Text - PDF
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Descripción
Resumen:	The rapid development and advancement of 5G technologies and smart devices are associated with faster data transmission rates, reduced latency, more network capacity, and more dependability over 4G networks. However, the networks are also more complex due to the diverse range of applications and technologies, massive device connectivity, and dynamic network conditions. The dynamic and complex nature of the 5G networks requires advanced and accurate traffic prediction methods to optimize resource allocation, enhance the quality of service, and improve network performance. Hence, there is a growing demand for training methods to generate high-quality predictions capable of generalizing to new data across various parties. Traditional methods typically involve gathering data from multiple base stations, transmitting it to a central server, and performing machine learning operations on the collected data. This work suggests a hybrid model of Long Short Term Memory (LSTM), Gated Recurrent Unit (GRU), and federated learning applied to 5G network traffic prediction. The model is assessed on one-step predictions, comparing its performance with standalone LSTM and GRU models within a federated learning environment. In evaluating the predictive performance of the proposed federated learning architecture compared to centralized learning, the federated learning approach results in lower Root Mean Square error (RMSE) and Mean Absolute Errors (MAE) and a 2.25 percent better Coefficient of Determination (R squared).
ISSN:	2158-107X 2156-5570
DOI:	10.14569/IJACSA.2025.0160130
Fuente:	Advanced Technologies & Aerospace Database