Data mining to discover trends on social media

点击翻译按钮获取中文摘要

Content on social media platforms such as YouTube is often user-generated and is now the predominant form of information consumption and sharing. The ability of many millions of users to generate content leads to information glut that requires new methods for searching and filtering content that is available on these platforms. On the other hand, some content becomes extremely popular and have a large audience and a majority of the content receives little attention. The central question that we would like to begin answering through this study can be stated as follows: "What makes certain social media content dramatically popular?" Related to this major question are a few other questions: -- What factors affect the evolution of a social media object with time? -- What is the impact of the social context within which an object is created and made available? The social context includes the author, the author's social neighborhood, and the nature of the content itself. -- Are there specific interventions - including metadata modifications (when possible) - that affect object popularity? By answering these questions, we hope to identify guidelines for selecting objects that are likely to become popular as well as actions that can be taken to improve the popularity of objects. These observations can guide content creators as well as groups such as our partner organization, BBTV, in marketing social media content.