Predicting epidemic threshold in complex networks by graph neural network.

Area of Science:

• Complex Systems Science
• Network Science
• Computational Epidemiology

Background:

• Predicting epidemic thresholds in complex networks is crucial for public health interventions.
• Existing models often struggle to integrate network topology and spreading dynamics effectively.
• Understanding these factors is key to controlling disease outbreaks.

Purpose of the Study:

• To develop a novel method for precise epidemic threshold prediction in complex networks.
• To create a model that incorporates both network topology and spreading dynamics.
• To enhance the accuracy and adaptability of epidemic threshold prediction models.

Main Methods:

• Development of a novel threshold graph neural network (TGNN).
• TGNN integrates network topology and spreading dynamical processes.
• Validation through extensive experiments on synthetic (Erdős-Rényi, scale-free) and real-world networks.

Main Results:

• The TGNN accurately predicts epidemic thresholds in homogeneous networks like Erdős-Rényi random networks.
• The model demonstrates usability and accuracy across altered spreading rate ranges.
• TGNN shows adaptability to diverse network topologies without requiring network-specific retraining.

Conclusions:

• The proposed TGNN offers a precise and adaptable approach to predicting epidemic thresholds.
• This method effectively combines network structure and spread dynamics for improved prediction.
• TGNN's validated performance on various networks highlights its practical applicability.