Role Detection and Prediction in Dynamic Political Networks


Evans E., Guo W., Gençtav A., Tari S., Domeniconi C., Murillo A., ...More

in: Advances in Data Science, Ilke Demir,Yifei Lou,Xu Wang,Kathrin Welker, Editor, Springer Cham, Zug, pp.233-252, 2021

  • Publication Type: Book Chapter / Chapter Research Book
  • Publication Date: 2021
  • Publisher: Springer Cham
  • City: Zug
  • Page Numbers: pp.233-252
  • Editors: Ilke Demir,Yifei Lou,Xu Wang,Kathrin Welker, Editor
  • TED University Affiliated: Yes

Abstract

The interaction of users, either online or in real-life, can be modeled via a network (or graph). In such networks, users engage in activities and take on specific roles. Users who manifest similar structural and connectivity patterns assume similar roles. The analysis of a network in terms of the component roles can facilitate the discovery of communities. By compressing big and complex networks, roles can also enable the discovery of important patterns, as well as important differences across networks.

While role detection in network data has been recently applied in a variety of domains, limited work has been done on the use of roles for predictive modeling in dynamic networks. In this work, we discuss a methodology to discover roles, and predict future role distributions in dynamic networks. We adapt previously developed feature-based techniques to discover roles associated to nodes. We verify the persistency of roles over time, and develop a strategy to trace them. Finally, we introduce a dynamic system that models roles’ evolution and their dynamics, and use it to predict future roles’ distributions. We demonstrate the effectiveness of our approach on a political network.