STG-Net: A COVID-19 prediction network based on multivariate spatio-temporal information.

Area of Science:

• Epidemiology
• Data Science
• Network Science

Background:

• Urban population density and rapid movement facilitate COVID-19 spread.
• Existing models inadequately integrate spatio-temporal data and trend analysis for infectious disease prediction.
• Time-series data alone is insufficient for accurate epidemic forecasting.

Purpose of the Study:

• To develop a robust model for predicting daily new COVID-19 cases.
• To enhance infectious disease forecasting by incorporating multivariate spatio-temporal information.
• To address limitations in current cross-domain transmission prediction models.

Main Methods:

• Proposed the spatio-temporal graph network (STG-Net) model.
• Introduced Spatial Information Mining (SIM) and Temporal Information Mining (TIM) modules.
• Utilized slope feature method and Gramian Angular Field (GAF) for enhanced feature extraction.
• Integrated multi-source spatio-temporal data for prediction.

Main Results:

• STG-Net demonstrated superior prediction performance across datasets from five countries.
• Achieved an average decision coefficient R2 of 98.23%.
• Exhibited strong long-term and short-term prediction capabilities with good robustness.

Conclusions:

• STG-Net effectively integrates spatio-temporal data for accurate COVID-19 case prediction.
• The model offers improved forecasting accuracy and robustness compared to existing methods.
• This approach provides a valuable tool for managing and responding to infectious disease outbreaks.