Temporal Dynamics of COVID-19 Outbreak and Future Projections: A Data-Driven Approach

Area of Science:

• Epidemiology
• Data Science
• Mathematical Modeling

Background:

• Epidemiological models often predict symmetrical infection curves (Gaussian distribution).
• Recent COVID-19 data shows slower infection decay than predicted by traditional models.
• This leads to underestimation of the final epidemic size.

Purpose of the Study:

• To develop improved data-driven models for epidemic forecasting.
• To accurately predict the decay phase of an epidemic.
• To address the underprediction issue of traditional epidemiological models.

Main Methods:

• Proposed two data-driven models: Gaussian and piecewise-linear fits for infection rate during decline.
• For countries not yet in decline, models use epidemiological peak predictions and adjust for slow decline.
• Comparative analysis of data-driven versus epidemiological models.

Main Results:

• Data-driven models offer improved accuracy in forecasting epidemic decay.
• The proposed methods provide more realistic projections compared to standard models.
• Demonstrated effectiveness on data from highly affected countries.

Conclusions:

• Data-driven approaches are crucial for accurate epidemic forecasting, especially during the decline phase.
• The proposed models enhance the reliability of predicting the final epidemic size.
• These methods offer a valuable tool for public health policy and resource allocation.