Social networks have permeated our daily lives. We transmit ideas, innovations, news, and even diseases through them. They affect the products we buy, the languages we speak and the behaviors we exhibit. Given such implications, an accurate understanding of social networks is crucial. In addition, with most social interactions moving online, researchers have access to unprecedented amounts of detailed data about social interactions. Therefore, we are at a point in history in which both the motivation and the opportunity to study social networks is overwhelmingly strong, attracting researchers from various backgrounds to social networks research. Naturally researchers, whether they are from databases, machine learning or theory background, have a tendency to apply techniques from their fields directly to this new paradigm. However, given the interdisciplinary nature of problems in social networks, one view point is insufficient in capturing the essence of these problems. The main goal of this dissertation is to bring knowledge from various backgrounds to tackle problems relating to social networks research. In addition to relying on diverse research fields, we also leverage the power of big data. The unprecedented amounts of data on online human interactions present great opportunities for the study of social networks. As demonstrated in this thesis, big data can help build better models, algorithms and infrastructures in social networks research.

The entirety of the vast space of problems relating to social networks research is likely too complex to summarize in one thesis. Instead, we focus on problems relating to information diffusion in online social networks. The overreaching goal of this thesis is to develop useful tools for understanding, managing and reporting on information diffusion by leveraging various research areas such as data mining, statistics, data management, theory and social sciences, rather than relying on only one. While identifying influentials in social networks, we leverage data-driven methods. When modeling diffusion of information and user behavior, we rely on statistical methods and theories from social science literature. Given a solid understanding of information diffusion in social networks, we can focus on various applications. Discrete math optimization techniques provide us an optimal direction to limiting the spread of misinformation in social networks. And finally, we rely on data streams solutions for building an informational trend detection framework in social networks. Throughout our studies, we focus on various networks such as Twitter, Digg, Facebook and the Blogosphere.