A densely connected causal convolutional network separating past and future data for filling missing PM_2.5 time...

Area of Science:

• Environmental Science
• Data Science
• Artificial Intelligence

Background:

• Air pollution is a global threat requiring accurate monitoring.
• Missing data in air quality datasets hinders reliable analysis and prediction.
• Existing methods struggle with systematic, long-term data gaps.

Purpose of the Study:

• To develop a novel deep learning model for imputing continuous missing PM2.5 data.
• To enhance the accuracy of air quality analysis and prediction.
• To address the challenge of systematic data loss in environmental monitoring.

Main Methods:

• Proposed a Densely Connected Causal Convolutional Network Separating Past and Future Data (DCCN-SPF).
• Utilized densely connected causal convolutional networks to extract features from past and future data.
• Integrated linear interpolation and deep learning for improved prediction accuracy.

Main Results:

• The DCCN-SPF model demonstrated superior performance in predicting PM2.5 concentrations compared to baseline models.
• Achieved significant reductions in Mean Absolute Error (MAE) by 8.7-21.6%.
• Achieved significant reductions in Root Mean Square Error (RMSE) by 7.1-23.5%.

Conclusions:

• The DCCN-SPF model effectively addresses missing PM2.5 data imputation.
• The model offers enhanced accuracy for air quality analysis and prediction.
• This approach provides a valuable tool for environmental monitoring and management.