Cluster Analysis of RF-EMF Exposure to Detect Time Patterns in Urban Environment: A Model-Based Approach

The increase in human exposure to electromagnetic fields (EMFs), driven by advancements in telecommunication systems like the 5G mobile system, highlights the need for continuous EMF monitoring. Advanced techniques for data analysis, based on machine learning like clustering, can decompose daily variations in EMF exposure into distinct patterns, providing a clearer understanding of how exposure fluctuates over time. Although several exposure monitoring systems exist in Europe, only a few studies have thoroughly examined the time variability. This study addresses the gap by applying model-based clustering techniques to analyze the temporal patterns. Specifically, the study focuses on characterizing fluctuations in field strength during workdays and holidays, thereby contributing to a deeper understanding of time-distributed exposure. Continuous monitoring data, collected through the Serbian EMF RATEL network’s sensors installed in Novi Sad, were processed and analyzed using the Log-Normal Mixture Model (LNMM), a model-based clustering algorithm resorting to mixture distributions. The analyses reveal that the LNMM can separate night and day exposure values and identify periods when values persist longer over the day. This suggests that model-based clustering can be useful for understanding the temporal patterns of local EMF exposure.