ML-based Detection of Radiation Azimuth Changes of Base Station

Despite the benefits of mobile services and technology of cellular networks, a certain level of concern exists, regarding exposure to electromagnetic fields (EMFs) radiated from antennas of base stations (BSs), which are mandatory in the infrastructure of those networks. Besides the supreme concerns on health effects trIGGered by EMF exposure, the population's common worries are related to possible non-transparent changes in technical parameters of BS, particularly radiation azimuths and radiated power. The appearance of EMF monitoring networks, such as the Serbian EMF RATEL network, which uses an approach of continuous EMF monitoring, has helped people to get a real picture of usual EMF levels in their surroundings. However, the EMF data acquired by monitoring sensors, also contain information about changes of BS's radiation parameters. This paper proposed Machine Learning (ML)-based detection of radiation azimuth changes of BS. Preliminary results have been verified on a case study of an EMF sensor installed in the University of Novi Sad campus, for which the specific BS acts as a dominant EMF source in its neighboring. The achieved accuracy and Matthews correlation coefficient (MCC) of the proposed ML model are of high value − 98.80% and 0.96 respectively.