Big Data Analytics in Online Social Networks to Characterize and Support Smart User Communities

Advancements in Information and Communication Technologies affect all aspects of our lives, such as how we communicate with each other, how we establish social relationships, and even how we organize our professional interactions. This transformation in our daily lives also changes how we engage with the communities we are a member of. These communities are the groups where we meet other people with similar interests. We call these communities "smart" if their members use ICTs to transform their circumstances in a significant way. Recent studies show the importance of understanding the behavior of the smart community members to meet the needs of local communities in terms of services and resources. One of the ways to analyze the behavioral patterns of an individual is to focus on its social relationships. The collection of all the relationships of an individual constitutes their personal network. In this thesis, we exploit the personal networks of online community members based on a well-established framework from evolutionary anthropology, Dunbar's "ego network model". This model stems from the so-called "social brain hypothesis", which postulates that we are limited on how many relationships we can maintain due to the signal-processing capacity of our brains. These relationships have not the same importance for us, and are organized into five concentric "circles" with a decreasing intimacy moving outwards. The thesis consists of three main parts. In the first part, we focus on a specific community on Twitter (journalists from 17 countries), and we study their ego networks. We find that the same behavioral patterns observed for offline social networks also exist for these online community members, with only minimal differences across the countries. In the second part, we propose to exploit the intimacy levels of the personal networks of online community members (video gamers on Twitter, in our case study) to predict their future relationships. In this part, we show that, in the vast majority of cases, leveraging information on the social circles provides significant improvements in the prediction performance. We also validate these findings on generic Twitter users without any community information. In the third part, we bridge the gap between the literature on ego networks and that on multilayer (multicommunity) networks by introducing the concept of multilayer ego networks. Our goal is to assess whether multilayer ego networks feature the same structural regularities observed in single-layers ego networks and to investigate the role that the different layers play in how people interact with each other. Leveraging a Reddit dataset, we show that multilayer ego networks are self-similar and that people accommodate multiple layers by adapting the outermost social circles without significantly affecting the innermost ones.