AI-driven approaches for air pollution modelling: A comprehensive systematic review.

Area of Science:

• Environmental Science
• Computer Science
• Data Science

Background:

• Air quality is a global concern, particularly in urban areas, leading to environmental degradation and health issues.
• Accurate prediction of future air pollution concentration levels is crucial for mitigation efforts.

Purpose of the Study:

• To systematically review Machine Learning (ML) and Deep Learning (DL) techniques for air quality prediction.
• To compare different methodologies, including temporal and spatiotemporal models, and analyze the impact of external features on prediction accuracy.

Main Methods:

• Conducted an extensive systematic literature review of ML and DL models for air pollutant prediction.
• Grouped studies by similar approaches for comparative analysis.
• Analyzed the influence of external data (meteorological, traffic, land usage) on model performance.

Main Results:

• Deep Learning models demonstrate superior performance in air quality forecasting due to their ability to represent complex features and spatio-temporal correlations.
• The inclusion of external features significantly enhances the accuracy of air quality prediction models.
• Both ML and DL approaches have distinct performances and limitations that are explored.

Conclusions:

• Deep Learning models are generally more suitable for air quality forecasting tasks.
• Future research should focus on optimizing model utilization and incorporating diverse external features for improved predictions.
• This review provides insights into the current landscape and future directions for air quality prediction research.