Mathematicians from RUDN University have conducted a study comparing machine learning models used for forecasting 5G and 6G traffic. The researchers aimed to identify the strengths and weaknesses of these models in order to optimize network performance and ensure quality of service.
The increasing demand for 5G and the imminent arrival of 6G networks pose significant challenges in terms of network load and resource consumption. To address this, networks need to track current indicators and make accurate predictions to effectively allocate network resources. Machine learning models are commonly employed for this purpose.
The study, published in the journal Future Internet, focused on comparing two time series analysis models: the seasonal integrated autoregressive moving average (SARIMA) model and the Holt-Winter model. Data obtained from a Portuguese mobile operator on traffic volumes for fixed periods was used to build the models.
Associate Professor Irina Kochetkova from the RUDN Institute of Computer Science and Telecommunications explained the significance of the study, stating, 5G and 6G networks will support drones, virtual and augmented reality. Moreover, if the number of connected devices grows, the traffic increases sharply, and network congestion occurs. As a result, the quality of service decreases, network delays and data loss increase. Therefore, the network architecture must adapt to the volume of traffic and take into account several types of traffic with different requirements.
The results of the comparison revealed that both models were suitable for forecasting traffic for the next hour. However, SARIMA performed better in predicting traffic from the user to the base station, with an average error of 11.2%, which was 4% lower than the second model. On the other hand, the Holt-Winter model demonstrated superior performance in predicting traffic from the base station to the user, with an error of 4.17% compared to 9.9% from the SARIMA model.
Associate Professor Kochetkova emphasized the effectiveness of both models in predicting traffic averages. However, she also noted the need for tailoring the approach to each specific data set. Future research will concentrate on combining statistical models with machine learning methods to enhance accuracy in forecasts and detect anomalies.
This study provides valuable insights into the optimization of 5G and 6G networks by comparing and evaluating the performance of different machine learning models for traffic forecasting. By understanding the strengths and weaknesses of these models, network operators and service providers can make informed decisions about resource allocation and ensure an optimal user experience.
