ONE METHOD FOR ESTIMATION AUDIENCE OVERLAP IN SOCIAL MEDIA

Abstract

In this paper we provided the definition of the Audience overlap network, as well as proposed a simple algorithm to compute overlap between two users on social media based on public data about their followers. There was proposed an alternative approach for computing overlaps based only on public data about users. This approach allows to include content overlap and activity patterns signals to be incorporated into more general statistical models featuring other covariates such as influencers’ direct engagement in shared conversations; relative influencer sizes and histories and links to similar third-party content to recover otherwise censored network structures and properties. For validate results there was designed a calibration process which utilizes Evolution Strategies algorithm to find a set of conditions which will make Audience overlap network built using similarity measures structurally equivalent to the Audience overlap network build on full information about followers.