This systematic review contrasts traditional forecasting methods with hybrid models for air quality, highlighting advantages and limitations.
Key Points
Air quality prediction improves with hybrid models that combine traditional and machine learning techniques, enhancing robustness.
Machine learning methods like Random Forests and LSTM provide better spatial and temporal accuracy compared to traditional methods such as ARIMA.
Traditional models like MLR and GAM are easy to interpret but face limitations in handling nonlinear interactions and long-range dependence.
The review indicates that while hybrid frameworks improve forecasting, they do not significantly reduce computational burdens or enhance interpretability.