Identifying Key Contributors to Regional Ozone Levels: A Machine Learning Approach Integrating Environmental and Socioeconomic Factors

Abstract

Ozone levels exhibit significant regional variability, influenced by a combination of socioeconomic
factors, emissions, and meteorological conditions. This study aims to identify the primary industrial and
socioeconomic contributors to ozone pollution, as well as the impact of other emissions such as NO2 and
CO. Utilizing explainable machine learning techniques, specifically Random Forest, XGBoost and SHAP
(SHapley Additive exPlanations) values, we quantify and identify the major contributors affecting ozone
levels in different regions. By analyzing industry data and various economic sectors—including mining,
agriculture, scientific services, arts, and real estate—alongside other relevant emissions, we uncover
potential reasons and sources for ozone levels unique to each area. These insights provide actionable
information for policymakers and environmental agencies, enabling targeted interventions to manage and
reduce ozone pollution. These insights offer actionable information for policymakers and environmental
agencies, enabling targeted interventions to manage and reduce ozone pollution.