Attention-based recurrent neural network for influenza epidemic prediction

Area of Science:

• Epidemiology
• Computational Biology
• Public Health

Background:

• Influenza poses a significant public health threat, necessitating precise real-time forecasting.
• Accurate prediction of influenza outbreaks is crucial for effective public health interventions.

Purpose of the Study:

• To develop a novel deep neural network for real-time forecasting of influenza-like illness rates (ILI%) in Guangzhou, China.
• To enhance the accuracy of influenza outbreak predictions using advanced deep learning techniques.

Main Methods:

• A multi-channel Long Short-Term Memory (LSTM) neural network architecture was designed.
• An attention mechanism was incorporated to improve the model's ability to handle complex data relationships.
• The model leverages the long-term dependencies and diversity inherent in influenza epidemic data.

Main Results:

• The proposed deep neural network structure effectively forecasts real-time influenza-like illness rates.
• The multi-channel LSTM with attention mechanism demonstrated superior performance in handling diverse input data.
• The model accurately captures complex relationships within the influenza epidemic data for Guangzhou.

Conclusions:

• The developed model exhibits strong competitiveness compared to existing methods.
• The deep learning approach provides effective and reliable real-time influenza epidemic forecasting.
• This method offers a valuable tool for public health management of influenza outbreaks.