Place-based attributes predict community membership in a mobile phone communication network

Area of Science:

• Network Science
• Sociology
• Epidemiology

Background:

• Social networks exhibit modularity, forming communities where information and behaviors spread rapidly.
• Understanding community formation is crucial for public policy, epidemiology, and social sciences.
• Previous studies focused on individual attributes; the role of place-based factors in large-scale networks was unclear.

Purpose of the Study:

• To investigate if place-based attributes can predict community membership in large-scale mobile phone communication networks.
• To assess the relationship between network modularity and the predictive power of geographic context.

Main Methods:

• Analyzed modularity in a mobile phone communication network in the Dominican Republic.
• Employed linear discriminant analysis (LDA) to test the predictive ability of place-based attributes.
• Correlated modularity scores with LDA's predictive accuracy.

Main Results:

• Place-based attributes (sugar cane production, urbanization, airport proximity, wealth) predicted community membership for over 70% of mobile phone towers.
• A strong positive correlation (r = 0.97) was found between modularity score and LDA predictive ability.
• Geographic context accurately represents the processes driving modularity in these networks.

Conclusions:

• Place-based attributes are strong predictors of community membership in large-scale social networks.
• Geographic context can be used to infer social network structure and predict community membership without direct social data.
• This approach has significant implications for understanding large-scale social dynamics and resource allocation.