Ambient air pollutants concentration prediction during the COVID-19: A method based on transfer learning

Area of Science:

• Environmental Science
• Data Science
• Public Health

Background:

• The COVID-19 pandemic heightened interest in air pollution correlations.
• Predicting ambient air pollutant concentration (APCP) during COVID-19 remains challenging due to data distribution shifts.
• Existing machine learning models struggle with pre- and post-pandemic data discrepancies.

Purpose of the Study:

• To develop an accurate APCP model for the COVID-19 era.
• To address the limitations of traditional machine learning in handling distribution shifts.
• To improve the reliability of air quality forecasting during public health crises.

Main Methods:

• Proposed an improved transfer learning model integrating machine learning.
• Utilized Gaussian mixture and optimization algorithms to create a similar source domain.
• Trained and transferred machine learning models from the new source to the target domain for APCP.

Main Results:

• The transfer learning approach significantly improved APCP accuracy.
• Experimental results on real-world data confirmed the method's effectiveness.
• The model successfully predicted ambient air pollutant concentrations.

Conclusions:

• The improved transfer learning model offers a robust solution for APCP during COVID-19.
• The findings provide valuable insights into air quality management during pandemics.
• Managerial insights on influential factors and their importance were derived from the model.