Faculty Candidate Talk: Bruno Ribeiro
Mining and Predicting a Complex Networked World: Theory, Models and Applications
Recent rapid advances in computational data-driven information and knowledge discovery have helped solve important technological problems and answer big scientific questions, with tremendous economic and societal impacts. Networks are at the heart of this revolution—from biological, ecological, and transportation networks, to online social networks, media streaming and online retail services that reach over three billion users and are valued at well over one trillion dollars.
How can we use this vast amount of data to answer fundamental questions, such as how does one forecast the growth or demise of online social network companies? And, more specifically, why did four large online social networks—namely MySpace, Hi5, Friendster and Multiply—synchronously go from growth and stability to their sudden downfall in the summer of 2008? MySpace was valued at $2 billion at its peak in 2008, and yet sold for a tiny fraction of that three years later. MySpace’s seemingly inexplicable sudden death still negatively impacts valuations of Internet companies today.
In my talk, I argue that such elusive problems require crossing disciplinary boundaries, especially those between epidemiology, population ecology, information theory, statistics, and psychology. I specifically look at how modeling limited user attention predicts the growth and death of Internet startups and how it offers an explanation for the 2008 mysterious demise of Facebook competitors. Finally, in the last part of my talk I turn to the more general problem of forecasting the evolution of complex network processes. Using information theory tools I unveil an apparent paradox, where larger online social networks entail more user data but also less analytic and forecasting capabilities.
Part of the content of this talk was featured in both specialized and popular media outlets including CNET, ACM TechNews, and The Pittsburgh-Post Gazette.
Dr. Bruno Ribeiro is a Postdoctoral Fellow at the Carnegie Mellon University School of Computer Science. Dr. Ribeiro completed his Ph.D. at University of Massachusetts Amherst in 2010. Between 2012 and 2013 Dr. Ribeiro was a visiting researcher at the Department Physics at Northeastern University. His research interests are in the areas of data science, data mining, complex system modeling, and computer networks.