Generalized additive model integrating multi-source data for short-term influenza forecasting in Shenzhen, China...

Area of Science:

• Epidemiology
• Public Health
• Data Science

Background:

• COVID-19 pandemic altered respiratory disease patterns, complicating influenza forecasting.
• Existing models face limitations in data integration, interpretability, and capturing nonlinear relationships.
• Accurate influenza prediction is crucial for effective public health interventions.

Purpose of the Study:

• To develop and validate a multisource data-integrated generalized additive model (GAM) for influenza forecasting in Shenzhen, China.
• To improve prediction accuracy and practical utility compared to traditional methods.
• To provide a robust tool for public health decision-making.

Main Methods:

• Developed a GAM incorporating local and Hong Kong influenza surveillance data, cross-boundary mobility, meteorological factors, and Baidu Search Index.
• Utilized data from 2023 to 2025 for model development and validation.
• Compared GAM performance against the Seasonal Autoregressive Integrated Moving Average with Exogenous Variables (SARIMAX) model using RMSE, MAPE, and R².

Main Results:

• The multisource GAM demonstrated high predictive accuracy for short-term influenza forecasting.
• Achieved R² values of 0.85 (1-week), 0.80 (2-week), and 0.74 (3-week) ahead forecasts.
• Outperformed the SARIMAX model in overall predictive accuracy.

Conclusions:

• The integrated GAM provides robust and stable influenza forecasts up to 3 weeks in advance for Shenzhen.
• This approach supports cross-boundary public health collaboration between Hong Kong and Shenzhen.
• The model can serve as a reference for regional public health strategies.