In our previous posts along this thread, I have suggested that graph analytics provides benefits in identifying actionable knowledge inherent in the relationships between and among entities, as opposed to typical analyses that focus on characterizing individual entities. I have to admit, that suggestion is a little bit misleading. What I really mean is that the process of analyzing the relationship among entities allows you to infer characteristics about entity behaviors that you might not have been able to identify otherwise.

A good example involves analyzing the level of influence that one individual exerts over others in a social network. Analyzing the sequence of transactions that a single customer makes over time provides one view of that person’s value as a customer. By examining the purchase patterns in the past, you can estimate (or perhaps even predict) what that customer’s projected consumption of products will be in the future.

However, through graph analysis, you may be able to recognize that every time a specific customer buys a product and likes that product, she posts a review on Facebook. After the Facebook review is posted, on the average, four of her Facebook friends proceed to buy the same product. In this case, that person’s value as a customer is enhanced, since her positive reviews influence additional product purchases.

Within the domain of social networks, this is one example of influence characteristics that are determined through specific graph metrics such as (among many others):

• Distance, which measures the number of edges to be traversed to connect any two specific vertices. This measures how closely connected any two individuals are.
• Betweenness centrality, which is the number of shortest paths from all vertices to any other vertex that pass through a selected entity’s vertex. This looks for individuals who intersect smaller sets of communities and can influence individuals among those different communities.
• Closeness centrality, which looks at how well-connected any specific vertex is to parts of the network with many connections. This looks for people well-known or popular within a group and who can influence people within that group.
• Bridges/connectors, which identifies vertices that provide the only link between two other sub graphs. This is used to identify key people needed for transmitting information across different parts of a social network.

There are many other metrics used in analyzing social networks, and each of the measures can be used in making those inferences about influence among the parties within the network. And despite the value of these methods for social network analysis, graph analytics algorithms are not limited to the social scene. In the next post, we will look at some other examples of using graph models and analyses.