Machine Learning-Based Prediction of Backhaul Capacity Requirements for Cellular Networks

The accurate prediction of the required backhaul transmission capacity for cellular networks is critical to ensure efficient and reliable network performance, especially with the increasing demand for high-speed data services and the introduction of new radio technologies. This paper presents a framework for predicting the required capacity of backhaul networks based on the base stations’ radio resources utilization and serving radio conditions. The proposed framework utilizes machine learning techniques to accurately estimate the required backhaul capacity by analyzing the base stations’ traffic patterns and performance indicators. The proposed framework also utilizes a hybrid feature engineering approach to select the most relevant features from the network Key Performance Indicators (KPIs) and domain experience to accurately estimate the required backhaul capacity. The proposed framework is evaluated on a real-life dataset obtained from measurements of an operational LTE cellular network. The results indicate high accuracy and effectiveness in predicting the required backhaul capacity with an error as low as 2.5 % in estimating the required backhaul capacity.