October 8, 2007: Seminar: Matthias Trier:"Exploring Patterns of dynamic Networks with event-based Visualization and Analysis"
Exploring Patterns of dynamic Networks with event-based Visualization and Analysis
Institute for Business Informatics
Technical University Berlin, Germany
Tuesday, October 9
10:00 a.m., Room 1000 SEO
A current method to study interactions among actors is Social Network Analysis (SNA). However, its methods model the observed network as a cumulative aggregation of interactions. While this approach is powerful for analyzing the cumulative and stable micro-organization of kinship, affection or power structures, it is less appropriate in domains where the actual timing of events is having an impact on the final social network structures. With the increased ability to collect large sets or streams of data, currently much research effort is dedicated to extend SNA with methods for dynamic network analysis. They include deductive mathematical modeling approaches but also inductive empirical methods for analysis. Taking the latter perspective, I will introduce a recent method of empirical analysis of dynamic network data. It builds on a dynamic data model which moves from capturing actors and relationships towards modeling actors and (timed) events. Based on this model, the empirical analysis is a combination of exploring a dynamic visual representation of the network's structural evolution with a time-window based measurement. Due to the computational effort, the method is implemented as a comprehensive software application. Using a variety of datasets of electronic communication and co-authorship, I will demonstrate recent studies including an analysis of the impact of external events on a corporate communication structure, the identification of actors' contributions over time compared to their centrality, an analysis of patterns of actors to become central over time, and visual insights into the spread of topical concepts throughout a communication network. The empirical patterns of network formation, which can be uncovered with this method, contribute to our understanding of dynamic processes in a large variety of network domains.
Host: Professor Tanya Berger-Wolf