Explainable Machine Learning-Based Electric Field Strength Mapping for Urban Environmental Monitoring

Abstract

Overview

This study aims to find the optimal machine learning model for constructing electric field strength maps in urban areas, significantly advancing environmental monitoring efforts. Distinguished by its unique dataset, the research goes beyond standard electromagnetic measurements to include details such as population density, varying urbanization levels, and specific building characteristics.

Findings

• Integration of explainable AI—specifically SHAP analysis—reveals the most influential factors contributing to electromagnetic field strengths in cities.
• The research team analyzed data using 410 machine learning models over Paris, assessing k-nearest neighbors (kNN), neural networks, and decision trees.
• The kNN approach proved superior, achieving an RMSE of 1.63 and SD of 0.20, offering robust, accurate mapping in diverse urban contexts.
• Models enable not only static snapshots, but also dynamic tracking of electromagnetic pollution over time, assessment of mitigation efforts, and deeper insight into distribution patterns.
• The volume of buildings near antennas is the strongest predictive factor, closely followed by local population density, both of which are linked to increased urban electromagnetic field exposure.

Conclusion

The research demonstrates that explainable machine learning models can accurately and dynamically map urban electric field strengths. Importantly, these findings highlight key urban features that predict higher levels of electromagnetic exposure, which are relevant for EMF safety policy and monitoring. Enhanced understanding of these factors supports public health efforts to reduce potential risks associated with electromagnetic fields in densely populated environments.