Enhanced forecasting of air quality index through an integrated deep learning framework

Area of Science:

• Environmental Science
• Data Science
• Artificial Intelligence

Background:

• Accurate air quality forecasting is crucial for public health and policy.
• Traditional models struggle with complex air quality data dependencies.

Purpose of the Study:

• To develop a hybrid deep learning model for enhanced air quality prediction.
• To integrate multiple neural network architectures and ensemble methods for improved accuracy.

Main Methods:

• Hybrid deep learning model integrating Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN), and Graph Neural Networks (GNN).
• Ensemble learning techniques (stacking and boosting) for fusing model outputs.
• Evaluation using real-world air quality monitoring datasets.

Main Results:

• Achieved R² scores exceeding 0.92 and reduced prediction errors compared to baseline models.
• Demonstrated superior performance on Beijing dataset (MAE 11.30, R² 0.89) versus standalone models.
• Showcased robustness and transferability with cross-dataset evaluation (Beijing to Los Angeles).

Conclusions:

• The hybrid model offers significant improvements in air quality forecasting accuracy and consistency.
• The proposed framework exhibits strong generalizability and interpretability for environmental monitoring.
• This approach provides a promising tool for decision-making in pollution-sensitive regions.