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

Abstract

Overview

The objective of this study is to determine the optimal machine learning model for constructing electric field strength maps across urban areas, advancing the field of environmental monitoring. These models incorporate detailed datasets including electromagnetic readings, population density, urbanization levels, and building characteristics.

Methodology and Findings

This approach utilizes explainable AI to elucidate key factors affecting electromagnetic exposure. Through the use of 410 machine learning models, this study covers various methodologies like k-nearest neighbors, neural networks, and decision trees particularly across Paris. A detailed analysis led to the identification of the k-nearest neighbors model as the most reliable, achieving consistent performance metrics.

Significant findings from the study include:

• The kNN model's RMSE was 1.63, with a standard deviation of 0.20, indicating superior performance compared to other tested models.
• A SHAP analysis highlighted the total volume of buildings feature (V) and population density (POP) as crucial predictors in electromagnetic field strength mapping using kNN.

Conclusion

The study successfully demonstrates the ability of kNN models, aided by explainable AI, to accurately predict and map electromagnetic field distribution in urban settings, greatly enhancing urban environmental monitoring.