Time series AQI forecasting using Kalman-integrated Bi-GRU and Chi-square divergence optimization

Area of Science:

• Environmental Science
• Data Science
• Machine Learning

Background:

• Air pollution poses a significant global health risk, necessitating reliable air quality index (AQI) prediction systems.
• Current AQI forecasting models struggle with missing data, high variability, and distributional uncertainty, limiting their effectiveness.
• Accurate AQI forecasting is crucial for public health protection and environmental policy development.

Purpose of the Study:

• To develop a novel deep learning framework for robust AQI time-series forecasting.
• To address limitations in existing models, including missing data, data variability, and distributional uncertainty.
• To improve the accuracy and reliability of AQI predictions for enhanced environmental monitoring.

Main Methods:

• Integration of Kalman Attention with a Bi-Directional Gated Recurrent Unit (Bi-GRU) for dynamic uncertainty handling and temporal feature weighting.
• Incorporation of a Chi-square Divergence-based regularization term to minimize distributional mismatch between predicted and actual pollutant levels.
• Imputation of missing values using pollutant-specific ARIMA models to preserve time-dependent trends.

Main Results:

• The proposed framework demonstrated significant improvements over baseline models (LSTM, CNN-LSTM) in AQI forecasting.
• Achieved a high R-squared value of 0.96794, indicating strong model performance.
• Reported a Mean Squared Error (MSE) of 4.11×10⁻⁵ and Mean Absolute Error (MAE) of 0.000423, signifying high prediction accuracy.

Conclusions:

• The novel deep learning framework effectively addresses key challenges in AQI forecasting, including uncertainty, distributional alignment, and missing data.
• The integrated architecture provides a scalable solution for environmental monitoring and supports evidence-based policy decisions.
• This research advances the field of AQI prediction, offering a more robust and reliable tool for safeguarding public health from air pollution.