Short-term air quality prediction based on EMD-transformer-BiLSTM.

Area of Science:

• Environmental Science
• Data Science
• Artificial Intelligence

Background:

• Air quality time series data are volatile and non-stationary, posing challenges for accurate prediction.
• Existing methods struggle with the complexity of nonlinear time series data containing noise.

Purpose of the Study:

• To develop and validate a hybrid model for accurate ultrashort-term air quality prediction.
• To address the challenge of predicting nonlinear, noisy air quality time series data.

Main Methods:

• Empirical Mode Decomposition (EMD) to decompose AQI into intrinsic mode functions (IMFs).
• An improved Transformer algorithm based on Bidirectional Long Short-Term Memory (BiLSTM) for predicting IMFs.
• Integration of predicted IMFs using BiLSTM to obtain final AQI predictions.

Main Results:

• The hybrid model achieved low error metrics (RMSE: 5.6853, MAE: 2.8230, MAPE: 2.23%) for 5-hour AQI prediction in Patna, India.
• The model demonstrated scalability and high performance on air quality datasets from multiple cities.

Conclusions:

• The proposed EMD-Transformer-BiLSTM hybrid model offers a robust solution for real-time air quality prediction.
• The model shows significant potential for broad application in environmental monitoring and public health.