Reconstructing secondary data based on air quality, meteorological and traffic data considering spatiotemporal...

Area of Science:

• Environmental Science
• Data Science
• Machine Learning

Background:

• Air quality monitoring generates complex spatiotemporal data.
• Integrating diverse data sources (air quality, meteorological, traffic) is crucial for accurate prediction.
• Existing datasets may not fully capture the spatiotemporal dynamics of air pollution.

Purpose of the Study:

• To introduce a reconstructed dataset for air quality prediction.
• To develop procedures for implementing spatiotemporal air quality analysis.
• To facilitate the application of advanced machine learning models to air quality data.

Main Methods:

• Reconstruction of a spatiotemporal dataset from Madrid City Council's Open Data portal.
• Incorporation of time series data from various monitoring stations into a spatiotemporal dimension.
• Application of grid-based (Convolutional Long Short-Term Memory, Bidirectional Convolutional Long Short-Term Memory) and graph-based (Attention Temporal Graph Convolutional Network) machine learning algorithms.

Main Results:

• A comprehensive dataset suitable for spatiotemporal analysis was created.
• The dataset was successfully used as input for advanced machine learning models.
• Demonstrated the feasibility of using reconstructed data for sophisticated air quality prediction.

Conclusions:

• The reconstructed dataset enhances capabilities for air quality prediction.
• Spatiotemporal data integration is vital for improving predictive accuracy.
• Advanced machine learning models show promise in analyzing complex environmental data.