Model-Based Clustering of RF-EMF Monitoring Data to Analyze Time Variability

Human exposure to electromagnetic fields (EMFs) has increased over the years because of the significant evolution of cellular network technologies. The growing interest in studying the time behavior of exposure levels requires employing innovative monitoring systems and data analysis techniques. Since 2017, the Republic of Serbia has used the national EMF RATEL network to monitor EMFs continuously over its territory. To enhance the available information on EMF, this paper uses mixture distributions in a model-based clustering approach to analyze the time variability of EMF data and determine if the field strength has any time pattern, focusing on a kindergarten as a case study of a sensitive area. Results show that Log-Normal Mixture Model (LNMM) usually performs better than its Dirichlet-process extension. For workdays, peak and trough values are grouped separately and two additional intermediate clusters are identified; for holidays, only two clusters are identified, possibly for the smaller range of values and flatter behavior over time.