Integrating machine learning models for real-time pollution monitoring in smart cities

The integration of machine learning (ML) models into smart city infrastructures offers a robust solution for real-time pollution monitoring. This paper presents innovative approaches, including data collection, preprocessing, model deployment, and adaptive learning techniques. Through comprehensive experiments, our study demonstrates the efficacy of various ML models. The Gradient Boosting model achieved the highest accuracy with an R-squared value of 0.93, MAE of 2.9 µg/m³, and RMSE of 4.2 µg/m³. The DBSCAN algorithm effectively identified pollution hotspots with a silhouette score of 0.71 and a Davies-Bouldin index of 0.35. The Deep Q-Network (DQN) optimized pollution control strategies, achieving a cumulative reward of 1400 and a convergence rate of 90%. An ensemble model combining Random Forest and Gradient Boosting achieved an R-squared value of 0.94. These results underscore the potential of ML models to enhance pollution monitoring and management in smart cities, improving urban air quality and public health.