APPLICATION OF MACHINE LEARNING AND COMPUTER VISION FOR AIR POLLUTION INDICATORS FORECASTING AS A FACTOR OF SOCIAL WELFARE

Abstract

A methodology for integrating computer vision technologies with ensemble machine learning methods has been developed for atmospheric air pollution forecasting based on automated traffic flow analysis. The proposed approach addresses critical environmental challenges in urban areas where transportation contributes significantly to air quality degradation, directly impacting public health and social welfare. The system architecture employs the YOLO11s model for automated recognition and classification of six vehicle categories: private cars, taxis, commercial vehicles, medium trucks, heavy trucks, and buses in complex urban traffic conditions. The ExtraTrees ensemble algorithm ensures high-precision forecasting of PM2.5 and PM10 fine particulate matter concentrations in ambient air through multi-output regression methodology. The research approach incorporates spatio-temporal traffic flow analysis, lag variable utilization for capturing temporal dependencies, and systematic hyperparameter optimization through GridSearchCV with cross-validation techniques. Data integration combines real-time traffic monitoring from computer vision systems with environmental sensor networks, creating comprehensive datasets for machine learning model training and validation. Empirical validation conducted on more than 50,000 observations demonstrates high forecasting accuracy with R² values of 0.972 for PM10 and 0.776 for PM2.5, confirming the methodology's stability and suitability for automated environmental monitoring systems in urban territories. The developed approach provides significant advantages over traditional monitoring methods, including scalability, real-time data processing capabilities, reduced operational costs, and independence from additional sensor infrastructure deployment. Research findings contribute to evidence-based environmental policy formulation and urban mobility management strategies, providing a robust foundation for sustainable urban development and environmental protection in metropolitan areas.