SMA4TD: A social media analysis methodology for trajectory discovery in large-scale events

Abstract

The widespread use of social media platforms allows scientists to collect huge amount of data posted by people interested in a given topic or event. This data can be analyzed to infer patterns and trends about people behaviors related to a topic or an event on a very large scale. Social media posts are often tagged with geographical coordinates or other information that allows identifying user positions, this way enabling mobility pattern analysis using trajectory mining techniques. This paper describes SMA4TD, a methodology for discovering behavior and mobility patterns of users attending large-scale public events, by analyzing social media posts. The methodology is demonstrated through two case studies. The first one is an analysis of geotagged tweets for learning the behavior of people attending the 2014 FIFA World Cup. The second one is a mobility pattern analysis on the Instagram users who visited EXPO 2015. In both cases, a very high correlation (Pearson coefficient 0.7–0.9) was measured between official attendee numbers and those produced by our analysis. This result shows the effectiveness of the proposed methodology and confirms its accuracy.