Hybrid framework for improved PM2.5 prediction based on seasonal-trend decomposition and tailored component...

Area of Science:

• Environmental Science
• Atmospheric Science
• Data Science

Background:

• Accurate prediction of PM2.5 concentrations is critical for environmental management and public health.
• PM2.5 data exhibits complex, non-linear, and non-stationary characteristics influenced by various factors.
• Existing models often fail to capture these multifaceted patterns or optimally utilize decomposition techniques.

Purpose of the Study:

• To develop a novel hybrid framework for precise PM2.5 concentration prediction.
• To address the limitations of existing methods in capturing complex temporal dynamics.
• To improve the accuracy and robustness of PM2.5 forecasting.

Main Methods:

• Proposed HISTCP, a hybrid framework utilizing Seasonal-Trend decomposition using LOESS (STL).
• STL decomposes PM2.5 series into trend, seasonal, and residual components.
• Applied component-specific processing techniques based on informational characteristics.

Main Results:

• HISTCP demonstrated superior performance and robustness in rigorous experiments.
• Evaluated on PM2.5 datasets from five diverse Chinese cities.
• Outperformed baseline and state-of-the-art models across MSE, MAPE, MAE, and R² metrics.

Conclusions:

• The component-specific modeling strategy in HISTCP is effective for PM2.5 forecasting.
• The proposed framework offers significant improvements in prediction accuracy.
• HISTCP provides a valuable tool for environmental management and public health protection.