Poster: A Diffusion Model for Predicting Spectrum Efficiency in 5G Networks

This study proposes a regression diffusion model for predicting spectrum efficiency (SE) in 5G networks, addressing limitations of traditional analytical models in dense urban environments. Using 5G user equipment (UE) measurement data collected in Seoul, we developed a model that leverages SINR (Signal-to-Interference-plus-Noise Ratio) and TBS (Transport Block Size) as input variables with classifier-free guidance and hybrid loss functions. Experimental results demonstrate improved performance with 26.37% MAPE compared to alternative methods. The model accurately captures SE distribution characteristics and temporal variations, contributing to enhanced simulation accuracy for 5G network management in dense urban areas.