Advanced AI-Driven Approaches for Predicting Air Quality: A Comprehensive Review
Keywords:
Artificial Intelligence, Air Quality Prediction, Deep Learning, Hybrid Models, Ensemble Techniques, PM2.5, PM10, NOx, Ozone, Performance Metrics, Data Integration, Model Scalability, Computational Efficiency, Interpretability, Literature ReviewAbstract
This comprehensive review examines the application of Artificial Intelligence (AI) techniques in predicting air quality parameters, focusing on studies published from 2019 to 2024. The review aims to provide a thorough understanding of how AI models, including deep learning, hybrid approaches, and ensemble techniques, are utilized for forecasting air quality indicators such as PM2.5, PM10, NOx, and ozone. A systematic selection criterion was employed to filter relevant studies from high-impact journals and conferences. Key aspects evaluated include the types of AI models used, dataset characteristics, and performance metrics such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and R-squared (R²). The review also identifies critical gaps in the current literature, such as limitations in model scalability, data integration challenges, and issues with computational efficiency and interpretability. By mapping these gaps against recent advancements in the field, this review highlights how subsequent research has addressed these challenges and proposes future research directions. The findings underscore the evolving landscape of AI in air quality prediction and provide insights into emerging trends and methodologies that can enhance predictive accuracy and applicability across diverse geographical and temporal scales.
