A Guide for Choosing Community Detection Algorithms in Social Network Studies: The Question Alignment Approach

Area of Science:

• Social Network Analysis
• Public Health Interventions
• Epidemiology

Background:

• Community detection is crucial for identifying subgroups in social networks for prevention research.
• Existing guidance on community detection methods lacks clear operationalization for aligning methods with research questions.
• The Question Alignment approach was developed to enhance the quality and applicability of community detection in research.

Purpose of the Study:

• To introduce and demonstrate the Question Alignment approach for selecting appropriate community detection methods.
• To evaluate the effectiveness of the Question Alignment approach in a hypothetical public health intervention scenario.
• To compare the performance of different community detection algorithms in identifying high-risk groups.

Main Methods:

• Discussed six community detection methods: Walktrap, Edge-Betweenness, Infomap, Louvain, Label Propagation, and Spinglass.
• Applied the Question Alignment approach to a hypothetical case study focused on hand hygiene intervention for influenza prevention.
• Utilized real-world network data collected in 2013 for the case study, conducted hypothetically in 2019.

Main Results:

• The Walktrap method, guided by the Question Alignment approach, best fit the hypothetical intervention requirements.
• Walktrap-derived communities differed significantly in size and number from those identified by other methods.
• The Question Alignment approach demonstrated potential for generating more useful community detection results for intervention targeting.

Conclusions:

• The Question Alignment approach can improve the theoretical meaningfulness and practical utility of community detection in prevention research.
• As social network analysis use grows in prevention, this approach offers a structured way to enhance results.
• Further research is needed to confirm if the Question Alignment approach leads to improved intervention outcomes.