Real-World Popularity Estimation from Community Structure of Followers on SNS

Abstract

In this paper, we propose methods of estimating the offline real-world popularity of users of online social network services (SNSs). Because their followers on an SNS are biased sampling from their offline real-world fans, we cannot estimate their real-world popularity simply by the number of their online followers. Our methods are based on the following hypothesis: SNS users with followers more distributed over many communities are likely to have more real-world popularity. We developed four methods, three of which use variations of the clustering coefficients of the followers to measure how much they are distributed, and one of which uses a metric we newly designed. Through the evaluation of our methods on the data from nine Ms/Mr university competitions, we validated our hypothesis.