A Two-Phase Feature Selection Method for Identifying Influential Spreaders of Disease Epidemics in Complex Networks

Area of Science:

• Network epidemiology
• Computational social science
• Infectious disease modeling

Background:

• Identifying influential spreaders is vital for understanding epidemic dynamics and control.
• Traditional centrality measures have limitations in universality and computational efficiency for large networks.
• Machine learning approaches improve spreader identification but can be computationally intensive.

Purpose of the Study:

• To develop a computationally efficient, two-phase feature selection method for identifying influential spreaders.
• To determine optimal feature combinations for spreader identification across different network types.
• To reduce feature dimensionality while maintaining accuracy in identifying key individuals.

Main Methods:

• A two-phase feature selection approach was employed.
• Optimal feature combinations were identified based on network topology (Barabasi-Albert, Erdős-Rényi, Watts-Strogatz) and dataset imbalance.
• The method was validated on both synthetic and real-world network data.

Main Results:

• For Barabasi-Albert (BA) networks, a combination including two-hop neighborhood centrality is fundamental.
• For Erdős-Rényi (ER) networks, degree centrality is essential.
• Feature selection was found to be unnecessary for Watts-Strogatz (WS) networks, and selected features on real-world networks showed high similarity to synthetic network findings.

Conclusions:

• The proposed two-phase feature selection method effectively reduces feature dimensions for identifying influential spreaders.
• The optimal feature combinations are network-dependent, offering tailored strategies for different network structures.
• This approach provides a novel and efficient pathway for identifying superspreaders to enhance epidemic control.