Abstract
Quantitative understanding of mechanism in complex systems is a common “difficult” problem across many fields such as physical, biological, social and economic sciences. Investigation on underlying dynamics of complex systems and building individual-based models have recently been fueled by big data resulted from advancing information technology. This thesis investigates complex systems in social science, focusing on civil unrests on streets and relevant activities online. Investigation consists of collecting data of unrests from open digital source, featuring dynamical patterns underlying, making predictions and constructing models. A simple law governing the progress of two-sided confrontations is proposed with data of activities at micro-level. Unraveling the connections between activity of organizing online and outburst of unrests on streets gives rise to a further meso-level pattern of human behavior, through which adversarial groups evolve online and hyper-escalate ahead of real-world uprisings. Based on the patterns found, noticeable improvement of prediction of civil unrests is achieved. Meanwhile, novel model created from combination of mobility dynamics in the cyberworld and a traditional contagion model can better capture the characteristics of modern civil unrests and other contagion-like phenomena than the original one.