Forecasting Weekly Influenza Outpatient Visits Using a Two-Dimensional Hierarchical Decision Tree Scheme

Area of Science:

• Public Health
• Epidemiology
• Data Science

Background:

• Influenza poses significant public health risks, causing suffering, death, social disruption, and economic loss.
• Accurate forecasting of influenza outpatient visits is vital for proactive management of medical resources and prevention of shortages.

Purpose of the Study:

• To propose and evaluate a novel two-dimensional hierarchical decision tree scheme for forecasting influenza outpatient visits.
• To assess the scheme's performance against existing models using real-world data.

Main Methods:

• Utilized weekly regional influenza outpatient visit data from Taiwan's national infectious disease statistics system (2005-2020).
• Developed and applied a two-dimensional hierarchical decision tree model for influenza forecasting.
• Compared the proposed model's accuracy against five competing forecasting models.

Main Results:

• The proposed two-dimensional hierarchical decision tree scheme demonstrated superior performance compared to five other models.
• The scheme accurately forecasted influenza outpatient visits one to four weeks into the future.
• Significant predictors for nationwide influenza forecasting in Taiwan included one- and two-time lag information and regional data from Taipei, North, and South.

Conclusions:

• The developed forecasting scheme offers a promising and effective alternative for predicting nationwide influenza outpatient visits in Taiwan.
• The findings highlight the importance of temporal lags and specific regional data in influenza forecasting models.