Social Network Analysis

Main Focus

A social network can be modelled as graph-based visualization (Figure 1) composed of individuals (organizations, company etc.) also called nodes, which are connected by links represent relationships and interactions between individuals, a rich relational interdependency and content for mining.

Figure 1.

An example of social network: individuals on nodes (black square) and links represent relationships (black line)

Social network analysis (SNA) is a mathematical technique developed in modern sociology, in order to understand structure and behaviour between members of social systems, to map relationships and social connections between individuals in social network, also to serve up business intelligence on the ties. SNA assumes that individuals are all interdependent, instead other approaches to business problems assume that what individuals do, think, and feel is independent of who they know. SNA is related to network theory and graph theory, so the network topology helps to determine a network's usefulness to its individuals. It is possible to classify objectives in: static to find community structures, dynamic to monitor community structure evolution and to spot abnormal individuals or abnormal time-stamps. Some typical problems in SNA include discovering groups of individuals sharing the same properties (Schwartz, & Wood, 1993), evaluating the importance of individuals (Kautz, Selman, & Milewski, 1996). Another goals of SNA regards to explain the observed network, identification of subgroup, inferring real-world connections and discovering, labelling, characterizing communities (Adamic & Adar, 2007).

Evaluation of people location in network, that is the centrality of a node, is relevant to understand networks and their participants. These measures provided by social network analysis give us insight into the various roles and groupings: who are the connectors, leaders, bridges, isolates, the clusters existing and who is in them, who is in the kernel of network and who is on the periphery.

SNA is descriptive, because it is built with only a few global parameters, so it is not useful for making prediction of future behaviour of network. This is due to networks availability, few information about each node and lack of data. In this sense, SNA assumes local variables (gender, contacts, etc), network variables depending on business problem, first-order variables (immediate connections behaviours) and second order variables (behaviours of friends of friends). In Web 2.0 age we have very large social networks creating massive quantities of data and we have substantial quantities of information at level of individual nodes suitable to build statistical models of individuals. The relevant difficult regards how to extract social data from a set of very different communication resources (Matsuo, Tomobe, & Nishimura, 2007).