Abstract:Governments and healthcare institutions are increasingly recognizing the value of leveraging social media data to address disease outbreaks. This is due to the rapid dissemination and rich content of social media data, which includes real-time reactions and calls for help from people. However, current research on which social media data can be utilized for information support, as well as the underlying reasons why social media data can be utilized for information support, remains limited. This study aims to make up for the aforementioned limitation by investigating which social media information is more likely to reflect the severity of an outbreak through empirical and prediction models, while also elucidating why social media data has the ability to reflect pandemic through content analysis. The COVID-19 outbreak was utilized as a case example in this study because it has the advantage of enhancing the universality of results and promoting the validation of the model with multiple waves of data. The empirical model results indicate that social media activity from public users is more likely to reflect the ground truth during pandemic. In particular, it was found that negative sentiment expressed in blog posts by public users during pandemic aligns more closely with the severity of disease outbreak. Then, a prediction model was proposed to further validate these findings of the empirical model. Finally, a content analysis was conducted based on the conclusions drawn from empirical model and prediction model. The content analysis revealed that the predictive capability of social media data for pandemic originates from individual self-reporting of illness. This study provides contributions and insights into which types of information can be used for pandemic monitoring and forecasting. The findings of our research have significant implications for governments and healthcare institutions in leveraging social media data for pandemic monitoring and forecasting.

An Event-Centric Prediction System for COVID-19

Forecasting the Long-Term Trend of COVID-19 Epidemic Using a Dynamic Model

Predicting the Epidemics Trend of COVID-19 Using Epidemiological-Based Generative Adversarial Networks.

Exploring the Effect of Social Media and Spatial Characteristics During the COVID-19 Pandemic in China

Leveraging Human Mobility Data for Efficient Parameter Estimation in Epidemic Models of COVID-19

COVID-19 Surveiller: toward a robust and effective pandemic surveillance system basedon social media mining

Cooperation or Confrontation? —— Prediction of COVID-19 Pandemic Situation Via Deep Learning

#StayHome or #Marathon? Social Media Enhanced Pandemic Surveillance on Spatial-temporal Dynamic Graphs

COVID-19 Surveiller: Toward a Robust and Effective Pandemic Surveillance System Based on Social Media Mining

An Intelligent Early Warning System of Analyzing Twitter Data Using Machine Learning on COVID-19 Surveillance in the US.

A Social Media Infodemic-Based Prediction Model for the Number of Severe and Critical COVID-19 Patients in the Lockdown Area

Development of the Second Version of Global Prediction System for Epidemiological Pandemic

Event Detection from Social Media for Epidemic Prediction

Leveraging Social Media Data for Pandemic Detection and Prediction

MobCovid: Confirmed Cases Dynamics Driven Time Series Prediction of Crowd in Urban Hotspot

Be Aware of the Hot Zone: A Warning System of Hazard Area Prediction to Intervene Novel Coronavirus COVID-19 Outbreak

Visual Data Analysis and Simulation Prediction for COVID-19

Predictive Modeling for Epidemic Outbreaks: A New Approach and COVID-19 Case Study

Adaptive Epidemic Forecasting and Community Risk Evaluation of COVID-19

A data-driven network model for the emerging COVID-19 epidemics in Wuhan, Toronto and Italy

Forecasting New COVID-19 Cases and Deaths Based on an Intelligent Point and Interval System Coupled With Environmental Variables