Social Network Analysis of Facebook : How to Reveal Social Circles
"We are on the cusp of a new way of doing social science. Our predecessors could only dream of the kind of data we now have." Nicholas Christakis
This is an analysis of my personal Facebook network which is the social media site that I use most often. My Facebook account has only been active for two months which explains the relatively small number of vertices or nodes in the network. This analysis allowed me to reexamine the relationships in a social network that I am especially familiar with. It generated unpredicted results such as the existence of three unconnected vertices (isolates) with no edges (shown in the lower left of the graph) and an unexpected range of betweenness centralities. Remember, betweenness centrality measures how "important" someone is within the network giving a higher value for those vertices that bridge the various clusters.[1] The betweenness centrality metrics for my network were 0.00 for minimum, 1.41 for maximum, and 0.38 for the average.
Figure 1: Facebook Social Network
NodeXL [2] found three distinct groups in my Facebook network: family, coworkers, and team members from a scavenger hunt. When I generated the social network graph, I was surprised to see three unconnected vertices (isolates) with no edges to connect to any group. The first unattached vertex belongs to a sister-in-law who naturally I would have classified in the family group. I realized this was the only family member from my husband's side of the family as opposed to relatives on my side of the family which comprised the family group. No other connection existed between my sister-in-law and the family group and so NodeXL did not recognize her as a part of that group. Ironically of all the vertices that belong to family members the only one I share a last name with is not included in the family group. This is simply a reflection on NodeXL's focus on relational data as opposed to attribute data. The remaining unattached vertices belonged to two friends who are not connected to any of the groups.
Another unexpected result was the variation in betweenness centralities between the family and coworker groups and the scavenger hunt group. Betweenness centralities were mapped to vertex size making variations in this metric easy to identify. The vertices in the family and coworker groups ranged from small to large while each vertex in the scavenger hunt group was sized identically. Looking at the graph and the degree metric I realized that the scavenger hunt group was unique in that each vertex connected to every other vertex in that group. This meant that any information sent by one vertex automatically reached all other vertices. This is not true of the family and coworker groups. Despite their low betweenness centralities the vertices in the scavenger hunt group shared the highest closest centralities of any of the vertices in my network meaning that information could flow quickest through that group.
This analysis of my Facebook network showed me the unexpected results that occur even with a network that I was familiar with. I learned how different a network can look when my vertex is removed from the graph and how computing metrics can reveal the strengths and weaknesses of the groups in that network. This type of analysis is useful for anyone wishing to examine a network they are a part of.
-----
[1] Betweenness Defined -http://en.wikipedia.org/wiki/Social_network_analysis