A Data-driven Study of Popularity and Engagement in Online Videos

The share of videos in the internet traffic has been growing, therefore understanding how videos capture attention on a global scale is also of growing importance for research and practice. People can interact with online videos via various behaviors -- there are behaviors of active participation (liking, commenting, and sharing), and that of passive consumption (viewing). Overall, this research takes a data-driven approach to study aggregate user behaviors for understanding the allocation of human attention in online videos. First, we summarize our recent work on measuring and predicting engagement on YouTube videos. We propose a novel metric of engagement for online videos, which correlates closely with content quality, is stable over lifetime, and predictable at the time of upload. Next, we explore the correlated popularity bursts within a recommendation graph and model content diffusion across different platforms. Findings from this research can enable content producers to create engaging videos, and hosting site to optimize advertising strategies, recommender systems, and many more online applications.