AQI time series prediction based on a hybrid data decomposition and echo state networks

Area of Science:

• Environmental Science
• Data Science
• Artificial Intelligence

Background:

• Accurate Air Quality Index (AQI) forecasting is crucial for public health and environmental management.
• Traditional time series models often struggle with the complex, non-linear dynamics of AQI data.
• Developing robust prediction models is essential for timely environmental interventions.

Purpose of the Study:

• To propose a novel hybrid model for enhanced AQI time series prediction.
• To improve the accuracy and reliability of AQI forecasting using advanced decomposition and feature selection techniques.
• To evaluate the model's performance on real-world AQI data from major cities.

Main Methods:

• A hybrid model combining Empirical Wavelet Transform-Singular Entropy-Variational Mode Decomposition (EWT-SE-VMD) for secondary data decomposition.
• Imperialist Competitive Algorithm (ICA) for optimal feature selection from decomposed subseries.
• Echo State Network (ESN) neural network for constructing the AQI prediction model.

Main Results:

• The proposed EWT-SE-VMD secondary decomposition method outperformed other decomposition techniques in data processing.
• The hybrid model, integrating EWT-SE-VMD, ICA, and ESN, demonstrated significant accuracy in AQI prediction.
• Experimental results on Beijing, Tianjin, and Shijiazhuang AQI data validated the model's effectiveness.

Conclusions:

• The developed hybrid AQI prediction model offers a promising approach for accurate time series forecasting.
• The EWT-SE-VMD secondary decomposition and ICA feature selection significantly enhance prediction performance.
• The model shows considerable application prospects and research value in the field of AQI prediction.