An air quality index prediction model based on CNN-ILSTM

Area of Science:

• Environmental Science
• Computer Science
• Data Science

Background:

• Air quality index (AQI) is vital for assessing air pollution and its health impacts.
• Accurate AQI prediction is significant for public health and environmental monitoring.
• Existing prediction models may have limitations in accuracy and efficiency.

Purpose of the Study:

• To develop and validate a novel Convolutional Neural Network-Improved Long Short-Term Memory (CNN-ILSTM) model for enhanced AQI prediction.
• To improve the efficiency and accuracy of AQI forecasting by leveraging the strengths of CNN and ILSTM.
• To compare the performance of the proposed CNN-ILSTM model against various established prediction models.

Main Methods:

• The study introduces an Improved Long Short-Term Memory (ILSTM) by modifying LSTM gates and adding a Conversion Information Module (CIM).
• Convolutional Neural Networks (CNN) are employed for effective feature extraction from air quality data.
• The CNN-ILSTM model was trained and validated using air quality data from Shijiazhuang City, China (2017-2021), and compared with eight other models.

Main Results:

• The CNN-ILSTM model achieved a Mean Absolute Error (MAE) of 8.4134 and a Mean Squared Error (MSE) of 202.1923.
• The model demonstrated a high coefficient of determination (R²) of 0.9601, indicating strong predictive power.
• The CNN-ILSTM model exhibited a significantly reduced training time of 85.3 seconds compared to other models.

Conclusions:

• The proposed CNN-ILSTM model offers a highly accurate and efficient solution for air quality index prediction.
• The integration of CNN for feature extraction and ILSTM for time-series analysis proved effective.
• The model's superior performance suggests its potential for real-world air quality monitoring and management systems.