Understanding, modeling and predicting the popularity of online content on social media applications

• Ano de defesa: 2015
• Autor(a) principal: Flavio Vinicius Diniz de Figueiredo
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: Atividades de usuário; Popularidade; Tendências; Mídia social; Séries temporais; Mídia digital; Computação; Redes de relações sociais
• Link de acesso: http://hdl.handle.net/1843/ESBF-9XZFMV

Social media has emerged as the de-facto form of online publishing. Motivated by the success of social media applications, our objectives are threefold. Firstly, we aim at understanding how different textual, content and social features relate with the evolution of popularity of social media content. We achieve this based on a characterization of the YouTube application, as well as a crowdsourced user study. Secondly, we provide popularity prediction methods to predict the popularity trends and values that social media content will achieve at future dates. Here we build upon previous prediction methods, paying attention to the unaccounted issue of remaining interest in content after prediction. Lastly, we present two novel data mining methods to understand how user activities (e.g., viewing and sharing) relate with the popularity evolution of social media content on YouTube, Twitter and LastFM. Our three studies are discussed in light of real world applications (e.g., advertising, provisioning and analytics platforms) that may benefit from our results..