The ubiquity of online social networks and social media nowadays has enabled information to penetrate across many different communities in the society and made the world closer than it ever was. Connecting with people through social media has become a daily lifestyle for many; people post and share content on social networks such as YouTube, Facebook and Twitter everyday. However, out of the numerous content generated on social media, only a few can create a buzz and subsequently get viral. This phenomenon has piqued the interests of researchers and businesses alike. Many studies have been conducted to answer why certain content can go viral; likewise, various parties have turned to social media to advertise, market, or even raise awareness. In such scenarios, knowledge of the o...[ Read more ]

The ubiquity of online social networks and social media nowadays has enabled information to penetrate across many different communities in the society and made the world closer than it ever was. Connecting with people through social media has become a daily lifestyle for many; people post and share content on social networks such as YouTube, Facebook and Twitter everyday. However, out of the numerous content generated on social media, only a few can create a buzz and subsequently get viral. This phenomenon has piqued the interests of researchers and businesses alike. Many studies have been conducted to answer why certain content can go viral; likewise, various parties have turned to social media to advertise, market, or even raise awareness. In such scenarios, knowledge of the outbreak time - the time at which a piece of content can reach a desired number of audience - is of utmost importance, as it allows campaign organizers to adjust their marketing duration, which translates to maximized cost-effectiveness of the campaign.

This thesis aims to predict outbreak time of online content by tracking and quantifying content popularity growth through social cascades. The first part of the thesis borrows a concept from epidemiology and incorporates community information to predict outbreak time. The second part of the thesis addresses limitations in the first part by modeling popularity growth through cascades’ rates of formation and growth. An iterative algorithm is proposed and is evaluated with datasets from real social networks such as Digg and Twitter.