QoS and Capacity Prediction for 5G Network Slicing

This technical report presents a hybrid forecasting framework aimed at enhancing resource allocation for 5G network slicing while ensuring Quality of Service (QoS). The document outlines the challenges in accurately forecasting cell utilization due to dynamic factors such as user mobility and fluctuating traffic loads. It introduces a novel approach that combines machine learning-based time series prediction with radio channel modeling derived from live network measurements. The framework estimates service-specific radio resource demand and supports slice admission control through proactive feasibility assessments. Key contributions include the development of a forecasting pipeline that integrates interference predictions with radio modeling to derive spectral efficiency and PRB demand estimation. The report details an empirical evaluation using data from a European Tier-1 operator, demonstrating significant accuracy in predictions. Furthermore, it discusses the methodology for data collection, spectral efficiency estimation, and the implementation of machine learning models for forecasting uplink interference and PRB utilization, thereby addressing the limitations of conventional resource estimation techniques.