A semi-mechanistic modeling strategy for infectious diseases forecasting: Error correction and probabilistic...

Area of Science:

• Epidemiology
• Computational Biology
• Data Science

Background:

• Pandemic threats are rising due to climate change and technology.
• Existing infectious disease models struggle with complex transmission dynamics and uncertainty.
• Accurate forecasting is crucial for effective public health interventions.

Purpose of the Study:

• To develop a novel hybrid methodology for enhanced infectious disease forecasting.
• To integrate dynamic modeling with statistical approaches for error correction and probabilistic prediction.
• To improve the accuracy and reliability of outbreak predictions.

Main Methods:

• Developed a semi-mechanistic model integrating dynamic modeling and statistical error correction.
• Utilized a quantile regression long short-term memory (QRLSTM) network to estimate prediction errors.
• Incorporated population mobility and meteorological data to refine forecasts.
• Validated the framework using United States (U.S.) coronavirus disease 2019 (COVID-19) outbreak data.

Main Results:

• The hybrid approach significantly reduces dynamic prediction errors by over 50%.
• Population mobility and meteorological conditions were identified as key factors affecting prediction accuracy.
• The semi-mechanistic model demonstrated superior long-term prediction performance and interpretability compared to pure deep learning methods.

Conclusions:

• The proposed hybrid approach enhances predictive accuracy and reliability for real-world outbreaks.
• Integrating mechanistic modeling with data-driven learning offers effective decision support for public health.
• This methodology provides a more robust tool for navigating complex infectious disease dynamics.