Novel Graph Topology Learning for Spatio-Temporal Analysis of COVID-19 Spread

Area of Science:

• Computational epidemiology
• Network science
• Data analytics

Background:

• Understanding COVID-19 transmission dynamics is crucial for effective public health interventions.
• Identifying influential countries aids in targeted pandemic response strategies.
• Existing methods for analyzing pandemic spread may not fully capture complex inter-country correlations.

Purpose of the Study:

• To develop and validate a new graph-learning technique for accurate COVID-19 data structure inference.
• To reveal correlations in pandemic dynamics across different countries.
• To identify influential countries for pandemic response analysis.

Main Methods:

• Analytically deriving eigenvectors (graph Fourier transform basis) of COVID-19 data.
• Estimating graph Laplacian eigenvalues using convex optimization.
• Analyzing confirmed COVID-19 variant cases in European countries using centrality measures.

Main Results:

• The new graph-learning technique accurately infers COVID-19 data structure.
• Identified influential countries differ from those found by existing techniques.
• The method demonstrated a 33.3% improvement in RMSE and 11.11% in correlation of determination compared to existing methods.
• Validated accuracy by successfully recovering missing COVID-19 test data.

Conclusions:

• The proposed graph-learning method offers a more accurate approach to analyzing COVID-19 spread.
• The identified influential countries provide valuable insights for pandemic response and control.
• This technique contributes to a deeper understanding of global pandemic dynamics and inter-country relationships.