@Rank: Personalized Centrality Measure for Email Communication Networks
[ 1 ] Instytut Informatyki, Wydział Informatyki, Politechnika Poznańska | [ P ] pracownik
2014
rozdział w monografii naukowej
angielski
EN Email communication patterns have been long used to derive the underlying social network structure. By looking at who is talking to who and how often, the researchers have disclosed interesting patterns, hinting on social roles and importance of actors in such networks. Email communication analysis has been previously applied to discovering cliques and fraudulent activities (e.g. the Enron email network), to observe information dissemination patterns, and to identify key players in communication networks. In this chapter we are using a large dataset of email communication within a constrained community to discover the importance of actors in the underlying network as perceived independently by each actor. We base our method on a simple notion of implicit importance: people are more likely to quickly respond to emails sent by people whom they perceive as important. We propose several methods for building the social network from the email communication data and we introduce various weighting schemes which correspond to different perceptions of importance. We compare the rankings to observe the stability of our method. We also compare the results with an a priori assessment of actors’ importance to verify our method. The resulting ranking can be used both in the aggregated form as a global centrality measure, as well as personalized ranking that reflects individual perception of other actors’ importance.
209 - 225