PhD Thesis Proposal (Comprehensive Examination)
The recent global surge in the use of technologies such as social media, smart phones and GPS-enabled devices has provided abundant resources to understand dynamical processes on complex networks and a unique chance to characterize geospatial and temporal distribution of real time social events. Meanwhile, the easy accessibility to bibliographic data and geographical database allow better understanding of scholarly networks and in charting the creation of knowledge globally. Moreover, the availability of large-scale communication datasets presents new opportunities to study information dissemination on social networks. In this thesis we focus on the diffusion processes on complex networks aggregated from these data. First, we investigate geospatial and temporal features of a publication dataset. We characterize the knowledge diffusion pattern between worldwide urban areas and its temporal evolution, and identify the key cities in the scientific research in physics. Second, we propose to detect and predict seasonal flu epidemics in countries of interest with geolocalized Twitter signals. In the early stage of a flu season, tweets containing information related with influenza-like illness indicate the spatial distribution of possible initial infected cases. Modeling disease and spreading dynamics with these initial seeds can provide the possibility of forecasting ILI cases in the coming month. Beyond geospatial information, we also investigate the role of other network properties play on information diffusion process. For a human-to-human communication network, we survey different definitions of weak ties and develop a novel link property importance to characterize the strength of ties. Controlling weak ties defined by importance can more efficiently confine the information spreading within a small community than controlling weak ties under other definitions. Last but not least, we find a phase transition between absorbing and active states of the classic Maki-Thompson rumor spreading model on random networks. The parameters of the contagion process as well as the network architecture determine whether the rumor will spread globally or whether it will be confined within a small neighborhood.
- Thesis Advisor
- - Alessandro Vespignani , Distinguished Professor, College of Computer and Information Science & Department of Physics & Bouve' College of Health Science, Northeastern University
- Committee Members
- - David Lazer , Professor, College of Computer and Information Science & Department of Political Science, Northeastern University
- - Alan Mislove , Assistant Professor, College of Computer and Information Science, Northeastern University
- External Member
- - Marta C. González, Assistant Professor, Department of Civil and Environmental Engineering, MIT
Justification for thesis committee composition: Prof. David Lazer is an authority in social network analysis, and he has published several papers on the diffusion of information among interest groups. Prof. Alan Mislove is an expert in networks and distributed systems with a focus on using social networks to solve system problems. Prof. Marta González's research focuses on integrating methods of complex systems with statistical physics approaches, computational sciences, geographic information systems and network theory to characterize and model human dynamics.