Modeling social media contagion using Hawkes processes

TL;DR

This paper introduces a moments-based estimation method for Hawkes processes to model social media contagion, specifically retweet cascades, and validates it with real data analysis.

Contribution

It develops a novel moments method for calibrating Hawkes models in social media contagion, addressing the challenge of parameter estimation.

Findings

Effective estimation of Hawkes process parameters demonstrated

Numerical analysis on real social media data supports the method

Improved understanding of retweet cascade dynamics

Abstract

The contagion dynamics can emerge in social networks when repeated activation is allowed. An interesting example of this phenomenon is retweet cascades where users allow to re-share content posted by other people with public accounts. To model this type of behaviour we use a Hawkes self-exciting process. To do it properly though one needs to calibrate model under consideration. The main goal of this paper is to construct moments method of estimation of this model. The key step is based on identifying of a generator of a Hawkes process. We perform numerical analysis on real data as well.