Abstract
Social media service such as Facebook has been enhancing its capability to provide more varied contents to its users. For content publishers, posting article links to Facebook is still a popular way to direct traffic to own websites. However, past studies have shown that only a small number of contents could become popular among Internet users. Being able to predict the performance of a particular piece of content before posting it would give the content publisher a significant competitive advantage.The researchers attempted to build such a predictive model in this study. Features of the social media posts posted by our research partner were collected, including temporal, textual, and monetary ones. The features were used for further statistical and data mining analysis, and the resultant predictive model, albeit with mediocre accuracy, can effectively reduce the number of social media posts needed to achieve the same monetary return. Using the model enables the publisher to post less and earn the same.The contribution of this study is as follows. First, in this study, we were able to build a model to predict advertisement income, which is unique in content popularity and performance studies. Second, this study focuses on the influence of social media posting on web contents performance, which has yet to be looked into in previous studies. Finally, the use of data-mining methods enables us not only to look at the research questions from an academic perspective but also the opportunity to build a predictive model to use in the real-world scenario.