ABSTRACT
The utility of social network analysis has long been recognized as important for the fields of epidemiology, counterterrorism, and business applications. The importance of having accurate classification techniques in these areas cannot be understated. By utilizing social network analysis, we are able to perform classification tasks upon nodes in the graph, allowing us to determine whether an individual may be infected with a virus, a potential terrorist, or a likely customer of an industry. Social network analysis techniques are generally based on the principle of homophily; that is, individuals who are in near proximity to one another in the social graph will form relationships to one another. This property is a reflection of the adage “birds of a feather flock together.” Through application of this property, modeling of social networks typically relies on only the first-degree neighbors of a node (Macskassy and Provost, 2007). However, medical studies have determined that smoking cessation and obesity networks may benefit from an extended analysis (Christakis and Fowler, 2007, 2008; Rosenquist et al., 2010). Instead, there appear to be nodes that influence those around them, and this influence spreads throughout the nodes in a near, nondirect relation to them.
