Referenced in our Newsletter Volume 3, Issue 2 - February 2004
Social Network Analysis (SNA)
An important form of link analysis that is growing in popularity is Social Network Analysis (SNA), sometimes referred to as Organizational Network Analysis. The purpose of SNA is to evaluate network nodes, and the relationships between them, from a "human" perspective. In this form of analysis, the significance of nodes is derived from each node's positional relationship to other nodes.
In SNA, nodes (objects in VisuaLinks) represent people, cities, computers, businesses or any other activity or process. The links between the nodes represent interactions of some form: phone calls, e-mail exchanges, conversations, chance meetings on the street, drug or weapon sales - the variety is limitless.
A key precept of SNA is that people tend to interact with people with whom they are already familiar and they tend not to step outside the confines of their known associates. Additionally, it is accepted that there is inherent value in the various interactions and relationships. This value is referred to as "Social Capital." In social networks, Social Capital influences interactions within a network.
SNA is the study and application of these, and other, concepts to determine nodes (usually people) in networks who are in some way "important" to that network.
With the release of VisuaLinks 3.0, we have implemented some SNA capabilities. Specifically, we have implemented a set of positioning algorithms, based on the concept of Centrality, that allow you to apply visual aspects of SNA to your own network analysis. This Link Chart of the Month looks at these positioning algorithms.
You may remember that we gave an overview of SNA in our May 2003 issue. As that issue provides information about Centrality and related concepts, we will leave further theoretical discussions as a reference exercise for the reader. In this article, we will look at the mechanics of using the SNA features of VisuaLinks 3.0.
We will begin our discussion with a look at telephone tolls. We begin our discussion with a simple query in a phone toll database to extract an item of interest. We then walk the query result a number of times to build a network of related phone calls. The result is shown here:
We drew a box (using VisuaLinks' new Presentation tools) around each successive data walk result. After five walks, we have a network of 17 phones.
There are four positioning algorithms in the Centrality placement: Degree, Relative Closeness, Absolute Closeness and Betweeness. Again, please refer to our May 2003 issue for detailed explanations. In that discussion, "Centrality" and "Degree" (in the VisuaLinks user interface) are synonymous.
Degree
"Degree" refers to an object's "connectedness." The more connections a node has, the more influence or control that node has the potential to wield within the network. When we apply this placement to the network we created above, we get the following:
The first image is displayed in a circular, or centric, pattern. The second is a hierarchical pattern which, like the Weighted placement, places the most-connected object in the upper left. In the circular display, the object closest to the center is the most-connected. The objects above and below that are the next-most connected and the outer ring are the least connected objects. Which layout you choose to use is entirely personal preference.
So what does this tell us? The object labeled "Buster Cardoza" is the most-connected object in this network. This indicates that this node could be highly influential in the network. Additionally, removing this node from the network would have a strong negative effect on the viability of the network.
Next we look at Closeness. Closeness is a measure of any given node's distance to all other nodes in the network. In VisuaLinks, Closeness is computed relative to a range between 0 and 100. This facilitates comparisons between networks of different sizes. The higher the number, the higher an object's Closeness.
When we apply the Relative Closeness, we see that Buster Cardosa and Whitney Shear share the same closeness value of 100.
Nodes with high Closeness are well connected within the network and can react swiftly to changes in the network because they are closest to all other nodes. Because of this, they can move information more quickly through a network as they will require few intermediaries to accomplish the task.
Our final Centrality Placement is Betweenness. Betweenness is a measure of which nodes are on the shortest paths between other nodes. Nodes with high betweenness control the information flow through a network. Such nodes are also sometimes referred to as "gatekeeper" nodes or liaisons. Nodes with high Betweenness can also act as single points of failure because the paths between so many other nodes pass through them.
Again, we see that Mr. Cardoso wins the Betweenness race.
In this small network, we can easily see that Buster Cardoso is likely a highly influential and crucial member of this network. By all measures of Centrality available in VisuaLinks, Mr. Cardoso is a key player in this network and would make a valuable target for an investigation. Also, removing Cardoso from this network would go far toward disabling it.
For more information on Social Network Analysis, please refer to the web references at the end of the May 2003 article.
|
