All are welcome to join Trent University's Applied Modelling and Quantitative Methods Graduate Program for the M.Sc. thesis defence of Bethany Hunt. “Finding Influential Actors in Social Networks Using Centrality Correlations”

Co-Supervisors: Dr. Sabine McConnell and Dr. Richard Hurley External Examiner: Dr. Mariana Kant, Department of Computer Science and Engineering, York University Chair: Dr. Laura Summerfeldt, Department of Psychology

ABSTRACT Centrality measures are commonly used in social network analysis to find the most important or most central person in the network. Since there is no single metric best representing centralization of a network, the linear relationship between two of the most common centralities in literature is examined. Betweenness centrality (BTWC) measures the criticality of a node in the flow of information throughout the rest of the network, acting as a gatekeeper between social circles. Eigenvector centrality (EVC), which finds the best connected individuals, is the most successful centrality measure for influence detection. The first model, Residual Key Actor (RKA), finds the key actor set with the maximum residual distance from the correlation’s best fit line. As an improvement to the first, the new MKA model using the Mahalanobis distance metric locates the actors with the greatest distance from the means of the correlated variables. Results are compared with published literature and an aggregated centrality score (ACS) model. It is expected that the key player set discovered with the new model will simultaneously discover the network’s critical gatekeepers and some leaders, and whose removal will affect the network the greatest. Results were consistent with expectations in that the MKA and ACS models were comparable and both significantly more effective at disrupting the networks than RKA. This result was also found for social networks specifically. For networks in general, a dependency was also discovered between method of key player identification and disruption index. Future work would benefit from a weighted MKA model with dynamic capabilities. 

Location: Science Complex Room 215

For more information please contact: gcollins@trentu.ca

Posted on Wednesday, May 22, 2013.