Explainability of the COVID-19 epidemiological model with nonnegative tensor factorization.

Area of Science:

• Epidemiology and Public Health
• Computational Modeling
• Data Science

Background:

• The COVID-19 pandemic necessitated diverse global response measures.
• Understanding patterns in these responses is crucial for future pandemic preparedness.
• Interpreting complex epidemiological models often requires significant manual effort.

Purpose of the Study:

• To develop a novel, explainable method for analyzing epidemiological model outputs.
• To identify and interpret global patterns of country response behaviors during the early pandemic.
• To cluster countries based on their pandemic response strategies.

Main Methods:

• Utilized a stochastic epidemiological model (SEM) to simulate disease spread and response effects.
• Represented SEM outputs as a tensor model.
• Applied nonnegative tensor factorization (NTF) for pattern identification and country clustering.

Main Results:

• Successfully identified distinct patterns in global country response behaviors.
• Clustered countries based on these identified patterns, revealing similarities in containment strategies.
• Demonstrated the effectiveness of NTF for analyzing complex epidemiological data.

Conclusions:

• The proposed NTF method offers an effective approach to explain epidemiological model outputs.
• The identified country clusters provide valuable insights into diverse pandemic response characteristics.
• This approach enhances our understanding of global strategies for infectious disease containment.