Forecasting Epidemics Through Nonparametric Estimation of Time-Dependent Transmission Rates Using the SEIR Model

Area of Science:

• Epidemiology
• Mathematical Biology
• Computational Science

Background:

• Epidemic forecasting using early case data is crucial but challenging due to data limitations and parameter instability.
• Previous methods often assume specific functional forms for transmission rates, limiting flexibility.
• Accurate estimation of time-dependent transmission rates is key for reliable outbreak predictions.

Purpose of the Study:

• To develop a novel nonparametric approach for reconstructing time-dependent transmission rates in epidemic forecasting.
• To compare the effectiveness of different regularization algorithms for improving forecasting reliability with limited data.
• To enhance the accuracy of predicting future epidemic incidence cases.

Main Methods:

• Utilized Legendre polynomials for nonparametric reconstruction of time-dependent transmission rates.
• Employed variational (Tikhonov's) regularization, truncated singular value decomposition (TSVD), and modified TSVD.
• Applied the methodology to simulated data and real-world epidemic data (1918 influenza, 2014-2015 Ebola).

Main Results:

• The nonparametric approach effectively forecasts future incidence cases.
• Comparison of regularization algorithms identified the most effective stabilizing strategy for limited data.
• Demonstrated the robustness of the method across different simulated and real epidemic scenarios.

Conclusions:

• The novel nonparametric method offers a significant advantage over traditional approaches for epidemic forecasting.
• Legendre polynomial projection combined with regularization provides a reliable strategy for estimating transmission rates.
• This approach enhances the ability to forecast epidemic outbreaks accurately using early incidence data.