A Machine Learning Approach to Predicting Air Quality and Health Risks for Construction Worker Safety

Abstract Air quality at construction sites significantly impacts workers’ health and safety. This study investigates the application of machine learning models to monitor and predict air quality while assessing health risks associated with construction activities. Data was collected indoor during peak construction phases from three sites: a 35-story apartment building, a two-floor villa, and a townhouse for comparison and evaluation. Advanced air quality monitors measured key pollutants. The data was pre-processed by addressing missing values, removing outliers, and normalizing to ensure it accurately reflects the air quality. Decision Tree regression and other classification models were used to predict pollutant trends and assess associated health risks. The regression models demonstrated high predictive accuracy, with R-squared values exceeding 0.95 for different pollutants across used datasets. The classification models achieved over 91% accuracy in estimating health risks based on pollutant thresholds. The Emirati Air Quality Index (EAQI) identified poor air quality especially at the high-rise site. Test Data was collected from a new site, townhouse construction site was used to test the model, confirming its robustness with prediction accuracies of 88% for PM1.0 and 96% for PM2.5, however variability was observed in the values for CO2 and PM10. Pollutant levels were higher at high-rise sites emphasizing the need for targeted mitigation. This study confirms the feasibility and importance of utilizing predictive models for effective air quality monitoring and managing health risks at construction sites. Graphical Abstract