Exploring Potential of ML-aided Mobile Traffic Prediction for Energy-efficient Optimization of Network Resources Using Real World Dataset

Abstract

To meet the extremely stringent but diverse requirements of Beyond Fifth-Generation (B5G) networks, traffic-aware adaptive utilization of network resources is becoming essential. To cope with that, a detailed traffic data analysis enables opportunities for mobile network operators to improve the Quality of Service (QoS) in the next-generation mobile communication systems. This paper presents a comprehensive analysis of the real world data collected from an operator’s 4G+ and 5G infrastructure during a seven-month campaign. Efficient Machine Learning (ML) based network traffic predictions are presented together with a statistical model to develop optimal resource allocation strategies by using the data gathered during the pandemic, an era when the data volume, as well as the bandwidth requirements and the end users’ expectations, were significantly elevated in terms of QoS, given the huge shift to the online world. Data analysis confirmed the assumption that there are traffic changes during the day and the whole week, which helped us to find new research directions regarding resource allocation optimization of next-generation mobile networks. Furthermore, we introduce the Predictive Energy Saver for Baseband Units (PESBiU) algorithm, which utilizes traffic prediction and power consumption analysis to manage the power states (sleep or active) of BBUs in a network. The PESBiU algorithm utilizes the results from ML predictions to effectively balance energy efficiency and network performance, demonstrating its potential for practical deployment in future mobile communication networks by transitioning BBUs to sleep mode during low-traffic periods, thereby achieving significant power savings.To meet the extremely stringent but diverse requirements of Beyond Fifth-Generation (B5G) networks, traffic-aware adaptive utilization of network resources is becoming essential. To cope with that, a detailed traffic data analysis enables opportunities for mobile network operators to improve the Quality of Service (QoS) in the next-generation mobile communication systems. This paper presents a comprehensive analysis of the real world data collected from an operator’s 4G+ and 5G infrastructure during a seven-month campaign. Efficient Machine Learning (ML) based network traffic predictions are presented together with a statistical model to develop optimal resource allocation strategies by using the data gathered during the pandemic, an era when the data volume, as well as the bandwidth requirements and the end users’ expectations, were significantly elevated in terms of QoS, given the huge shift to the online world. Data analysis confirmed the assumption that there are traffic changes during the day and the whole week, which helped us to find new research directions regarding resource allocation optimization of next-generation mobile networks. Furthermore, we introduce the Predictive Energy Saver for Baseband Units (PESBiU) algorithm, which utilizes traffic prediction and power consumption analysis to manage the power states (sleep or active) of BBUs in a network. The PESBiU algorithm utilizes the results from ML predictions to effectively balance energy efficiency and network performance, demonstrating its potential for practical deployment in future mobile communication networks by transitioning BBUs to sleep mode during low-traffic periods, thereby achieving significant power savings.