A Deep CNN-LSTM Model for Particulate Matter (PM_2.5) Forecasting in Smart Cities

Area of Science:

• Environmental Science
• Data Science
• Computational Science

Background:

• Air pollution, particularly Particulate Matter (PM2.5), poses significant risks to human health and the environment.
• PM2.5 particles can lead to severe respiratory and cardiovascular diseases, including asthma and lung cancer.

Purpose of the Study:

• To develop and evaluate a novel deep learning model for accurate PM2.5 concentration forecasting.
• To assess the performance of a hybrid CNN-LSTM model against other machine learning methods for air quality prediction.

Main Methods:

• A hybrid deep neural network model, APNet, integrating Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) architectures was developed.
• The model utilized historical data including rainfall, wind speed, and PM2.5 concentrations for training and forecasting.
• Performance was evaluated using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Pearson correlation coefficient, and Index of Agreement (IA).

Main Results:

• The proposed CNN-LSTM model (APNet) achieved the highest forecasting accuracy compared to other machine learning methods.
• Experimental results verified the feasibility and practicability of the CNN-LSTM model for PM2.5 concentration forecasting.
• The model demonstrated strong performance based on the applied measurement indexes.

Conclusions:

• The developed APNet model, integrating CNN and LSTM, is highly effective for PM2.5 forecasting.
• This study validates the practical application of deep learning for air quality monitoring and prediction.
• The findings can contribute to future strategies for PM2.5 pollution prevention and control.