This paper presents a machine-learning method to estimate ground-level electromagnetic radiation (electric field strength) in the near field of 5G base stations, using multiple technical and environmental input parameters. The authors report experimental performance with a mean absolute percentage error of about 5.89% and suggest the approach can reduce costs compared with on-site measurements. The work is positioned as supporting exposure management and base-station placement, while noting the need for careful EMF management due to potential health-risk links.