A combined forecasting structure based on the L₁ norm: Application to the air quality

Area of Science:

• Environmental Science
• Data Science
• Atmospheric Chemistry

Background:

• Air pollution poses significant risks to industrial production and public health.
• Existing prediction methods often overlook the intrinsic structural characteristics of pollutant data.
• Accurate air quality forecasting is crucial for public guidance and mitigation strategies.

Purpose of the Study:

• To develop a novel combined forecasting structure for air pollution monitoring and analysis.
• To improve the accuracy of air quality predictions by incorporating data structure.
• To provide a reliable framework for assessing and communicating air quality levels.

Main Methods:

• Data decomposition and phase space reconstruction to analyze structural characteristics.
• A weighted combination forecasting module based on the L1 norm.
• Multi-tracker optimization algorithm for parameter tuning and fuzzy evaluation for qualitative analysis.

Main Results:

• The proposed L1 norm-based combined forecasting structure demonstrated effectiveness and efficiency in predicting daily pollution sources.
• The method successfully integrates data decomposition, phase space reconstruction, and optimized forecasting.
• Comprehensive fuzzy evaluation provided qualitative insights into air quality.

Conclusions:

• The developed combined forecasting structure shows significant potential for practical application in air quality prediction.
• This approach offers a more robust method for air pollution analysis compared to traditional techniques.
• The study highlights the importance of considering data structure in environmental forecasting models.