AI-Enhanced air quality assessment and prediction in industrial cities: A case study of Kryvyi Rih, Ukraine

Abstract

Kryvyi Rih, Ukraine, a city marked with high mining, metallurgical, and automobile activities is such a case, and lacks capability with predictive soundness and real-time anomaly identification. This framework proposes an AI-based air quality monitoring system that combines traditional air quality monitoring data (2021–2023) with machine learning models. The developed system utilizes XGBoost for pollutant concentration prediction and Iso- lation Forest for anomaly detection of critical pollutants such as CO, NO₂, SO₂, hydrocarbons, and benzene. Data from fixed monitoring stations placed around busy junctions was filtered and combined into a supervised and unsupervised learning model. The XGBoost model provided high accuracy (R² > 0.84), while the Isolation Forest algorithm was able to detect pollution spikes with high precision (F1-scores > 0.80). The comparison of traditional data validated the system’s reliability in determining hotspot regions and trending changes over time. The research suggests some policy interventions relating to air quality management systems and frameworks that can be ad- justed to other industrial cities themes of environmental integrity. The combination of AI/ML achieves the required response time, improves ecological monitoring, assistance guided sustainable urban development.