FORECASTING LOAD IN FIFTH GENERATION TELECOMMUNICATION NETWORKS BASED ON BIG DATA

Abstract

The relevance of this study is driven by the rapid growth of traffic volumes, service heterogeneity, and increased requirements for latency, reliability, and quality of service in fifth generation telecommunication networks (hereinafter referred to as 5G). Under such conditions, traditional reactive approaches to network resource management prove to be insufficiently effective, which makes the use of Big Data technologies relevant as a tool for proactive load forecasting and for enhancing the resilience of network infrastructure operation. The purpose of the article is the scientific substantiation and development of approaches to load forecasting in 5G telecommunication networks using Big Data technologies in order to improve the efficiency of network resource management and to ensure stable quality of service. The research methods are based on a system analysis of contemporary approaches to network load forecasting, the generalization of methods for processing large volumes of heterogeneous network data, as well as numerical experiments using time series of network cell load. A comparative analysis of reactive and forecast oriented resource management based on intelligent models is applied. The research results indicate that load in 5G networks exhibits a nonlinear spatiotemporal nature, which should be taken into account through the use of Big Data approaches. It is established that the integration of forecasting models into resource management loops makes it possible to reduce the frequency of overloads, increase the average level of resource utilization, and decrease the number of corrective management actions. It is demonstrated that even relatively simple intelligent models trained on historical data provide a quantitatively measurable positive effect in practical network operation scenarios. The conclusions indicate that the application of Big Data technologies represents an appropriate and effective basis for load forecasting and proactive resource management in 5G telecommunication networks. Key forecasting challenges related to data quality and heterogeneity, model scalability, processing delays, and limited interpretability of results are identified. Prospects for further research are associated with the development of hybrid forecasting models, the improvement of spatiotemporal load analysis, the enhancement of result interpretability, and the integration of such approaches with the concept of digital twins of telecommunication networks.