TRACING TOPIC DISCUSSIONS WITH THE EVENT-DRIVEN SIR MODEL FOR ONLINE FORUMS

Author: 

Jiyoung Woo
Sung Ho Ha
Hsinchun Chen

Abstract: 

Members of online communities on social media (e.g., Web forums) disseminate and receive information byinteracting with one another, a situation that provides a good opportunity to examine information diffusion on socialmedia. This study models topic-level opinion diffusion in Web forums through an epidemic model, namely theSusceptible, Infective, and Recovered (SIR) model, which has been used to analyze disease outbreaks and knowledgediffusion. In addition, this study proposes an event-driven SIR model that reflects the effect of a given event on opiniondiffusion. The proposed model incorporates the effects of news postings on social media in terms of the growth innumbers of potential authors, the increase in the infection rate for a given topic, and the acceleration of the transitionfrom potential authors to active ones. This study evaluates the proposed model based on a large longitudinal data setfrom a Web forum of a major company. The analysis results show that the event-driven SIR model performs well interms of the estimation of the number of authors who discuss a given topic, and outperforms the baseline SIR modelin terms of model fit and the forecasting of major hot topics that reflect outbreaks of author participation. This studyhas some practical business implications: Web forums are places where corporate brands and reputations are formed,as evidenced by forum posts; and Web forums are advertising platforms for contacting audiences and tracking usersin order to hone marketing messages and encourage discussions and reviews on a firm’s products and services.

Key Word: 

Published Date: 

May, 2016

Full File: