Inferences in a Virtual Community: Demography, User Preferences, and Network Topology

TL;DR

This paper introduces a computational method to analyze virtual communities, revealing demographic patterns, user preferences, and network topology, including small-world and scale-free properties, using data from an online social network.

Contribution

The paper presents a novel methodology for extracting demographic, preference, and topological data from virtual communities, demonstrated on a real online network.

Findings

More female users than male in the community.

Strong age-based homophily among users.

Network exhibits small-world and scale-free characteristics.

Abstract

This paper presents a computational procedure for extracting demography data, mining patterns of human preferences, and measuring the topology of a virtual network. The network was created from the personal and relationships data of an online Internet-based community, where persons are considered nodes in the network, and relationships between persons are considered edges. A community of Friendster users whose listed hometown is Los Ba\~nos, Laguna was used as a test bed for the methodology. The method was able to provide the following demographic, preferential, and topological results about the test bed: (1) There are more female users (52.34\%) than male (47.66\%); (2) Homophily (i.e., birds-of-a-feather adage) is observed in the preferences of users with respect to age levels, such that they are strongly biased towards being friends with users of a similar age; (3) There is heterophily in gender preference such that friendship among users of the opposite gender occurs more often. (4) It exhibits a small-world characteristic with an average path length of 4.5 (maximum=12) among connected users, shorter than the well-known {\em six degrees of separation}~\cite{travers69}; And (5) The network exhibits a scale-free characteristics with heavily-tailed power-law distribution (with the power $\lambda = -1.02$ and $R^2 = 0.84$) suggesting the presence of many users acting as the network hubs. The methodology was successful in providing important data from a virtual community which can be used by several researchers in the fields of statistics, mathematics, physics, social sciences, and computer science.