Estimating crowd sizes through social media

Abstract

Social media is of great significance in modern society and the analysis of Twitter data has shown an increasing trend towards supporting different research efforts, from understanding society to marketing and analysis of organizations. The ubiquity of Twitter with over 2.9million Tweeters in Australia alone means that it can be used for many (potentially unforeseen) activities. In this paper we consider how it can be used for estimation of crowd sizes. This is useful for calculating public event attendance as well as for potential safety concerns. In order to establish the relationship between social media usage and crowd sizes we focus on a range of sporting events, where attendances are known. From this we create a prediction model that can be used to estimate unknown crowd sizes based on social media use alone. We show how it is possible to benchmark the accuracy of the crowd size estimation by introducing other information. Finally, we apply the prediction model to estimate the attendance at the Melbourne marathon event where unknown numbers of attendees were present.