New Algorithmic Strategies to Forecasting Contagions


Can tweets be really used to track and forecast an epidemic outbreak? Can we monitor a handful of individuals in a city or a social network to forecast virality? The problems we focus on occur in diverse areas: social networks, public health surveillance, and cybersecurity. This talk will comprise two parts. In the first part, we leverage graphical models to infer the progression of the flu from user tweets. We show how we can effectively and accurately track flu trends and peaks using millions of tweets harvested from Latin America. In the second part, we design social network sensors for early detection of contagious phenomena. The robustness of these methods is demonstrated using six massive city-scale datasets of population activity. We demonstrate how contagious phenomena—both information cascades and disease epidemics—can be anticipated with significant lead time. The talk will conclude with future research directions on how we can improve the effectiveness of contagion tracking.


Tozammel Hossain is a research faculty at the MU Institute for Data Science and Informatics. He earned his PhD in computer science at Virginia Tech and has three years of experience as a postdoctoral researcher at the University of Southern California-Information Sciences Institute. His research interests broadly lie in machine learning and data science with an emphasis on solving problems in bioinformatics, computational social sciences, geopolitical event forecasting, and cyber-security. His research has appeared in ACM SIGKDD, IEEE ICDM, AAAI, and ACM BCB. His research has also received press coverage in The Conversation and ACM TechNews.