Along with two of my graduate students we have had two book chapters accepted in the upcoming book entitled “Mining and Analyzing Social Networks” which is part of the book series of studies in Computational Intelligence, Springer-Verlag, Heidelberg Germany, 2010. Social Network Analysis and Visualization will form an aspect of collaborative and emerging visualization research projects within the Human Interface Technology Research Laboratory Australia (HITLAB AU).
These chapters are entitled “Actor Identification in Implicit Relational Data” and “Perception of Online Social Networks” which are detailed below.
Actor Identification in Implicit Relational Data
Michael Farrugia and Aaron Quigley
Abstract
Large scale network data sets have become increasingly accessible to researchers. While computer networks, networks of webpages and biological networks are all important sources of data, it is the study of social networks that is driving many new research questions. Researchers are finding that the popularity of online social networking sites may produce large dynamic data sets of actor connectivity.
Sites such as Facebook have 250 million active users and LinkedIn 43 million active users. Such systems offer researchers potential access to rich large scale networks for study. However, while data sets can be collected directly from sources that specifically define the actors and ties between those actors, there are many other data sources that do not have an explicit network structure defined. To transform such non-relational data into a relational format two facets must be identified – the actors and the ties between the actors. In this chapter we survey a range of techniques that can be employed to identify unique actors when inferring networks from non explicit network data sets.We present our methods for unique node identification of social network actors in a business scenario where a unique node identifier is not available. We validate these methods through the study of a large scale real world case study of over 9 million records.
Perception of Online Social Networks
Travis Green and Aaron Quigley
Abstract.
This paper examines data derived from an application on Facebook.com that investigates the relations among members of their online social network. It confirms that online social networks are more often used to maintain weak connections but that a subset of users focus on strong connections, determines that connection intensity to both connected people predicts perceptual accuracy, and shows that intra-group connections are perceived more accurately. Surprisingly, a user‘s sex does not influence accuracy, and one‘s number of friends only mildly correlates with accuracy indicating a flexible underlying cognitive structure. Users‘ reports of significantly increased numbers of weak connections indicate increased diversity of information flow to users. In addition the approach and dataset represent a candidate ―ground truth‖ for other proximity metrics. Finally, implications in epidemiology, information transmission, network analysis, human behavior, economics, and neuroscience are summarized. Over a period of two weeks, 14,051 responses were gathered from 166 participants, approximately 80 per participant, which overlapped on 588 edges representing 1341 responses, approximately 10% of the total. Participants were primarily university-age students from English-speaking countries, and included 84 males and 82 females. Responses represent a random sampling of each participant‘s online connections, representing 953,969 possible connections, with the average participant having 483 friends. Offline research has indicated that people maintain approximately 8-10 strong connections from an average of 150-250 friends. These data indicate that people maintain online approximately 40 strong ties and 185 weak ties over an average of 483 friends. Average inter-group accuracy was below the guessing rate at 0.32, while accuracy on intra-group connections converged to the guessing rate, 0.5, as group size increased.