Current Concepts in Engineering Research and Technology Vol. 4

Rapid industrialisation and urban expansion have positioned Bhiwadi, Rajasthan, as one of India's emerging air-pollution hotspots, with frequent surges in particulate matter and gaseous pollutants driven by traffic congestion, industrial activity, and seasonal crop-residue burning. These pollution peaks correspond to increased respiratory complaints, hospital emergency visits, and heightened public concern on social media. Traditional forecasting systems often rely solely on pollutant concentrations or meteorological factors, providing limited insight into population-level health risks or vulnerable areas. This study addresses this limitation by developing an integrated, interpretable early-warning framework using heterogeneous data sources, including air-quality sensors, meteorological APIs, mobility and traffic data, social-media complaints, and hospital emergency room (ER) records collected across four zones of Bhiwadi from 2020 to 2024. For model development, pollution-induced health risk was formulated as a binary classification problem based on predefined pollution and health-risk criteria. The XGBoost-based gradient-boosting model, enhanced by spatial clustering, was employed to predict pollution spikes and their associated health impacts up to 24 hours in advance, thereby identifying high-risk neighbourhoods and vulnerable demographic groups. Among the four zones, Zone 4 exhibited the highest predicted health risk. The model demonstrated strong predictive performance, achieving a ROC-AUC of 0.932, precision of 0.871, recall of 0.860, and an F1-score of 0.865. Explainable Artificial Intelligence (XAI) techniques, particularly SHapley Additive exPlanations (SHAP), provide transparency by revealing the key drivers of the predictions, thereby supporting policymakers in implementing targeted interventions. The proposed multimodal and interpretable approach demonstrates reliable predictive performance and provides a practical basis for developing early-warning systems and supporting evidence-based environmental health management in rapidly urbanising regions.