An Event-Centric Prediction System for COVID-19

Abstract

As COVID-19 evolved into a pandemic, a lot of effort has been made by scientific community to intervene in its spread. One of them was to predict the trend of the epidemic to provide a basis for the decision making of both the public and private sectors. In this paper, a system for predicting the spread of COVID-19 based on detecting and tracking events evolution in social media is proposed. The system includes a pipeline for building Event-Centric Knowledge Graphs from Twitter data streams about COVID-19, and uses the graph statistics to obtain a more accurate prediction based on the simulation of epidemic dynamic models. Experiments of 128 countries or regions conducted on the data set released by Johns Hopkins University on COVID-19 confirmed the effectiveness of the system. At the same time, the guidance our system provided to the plan of return-to-work for an enterprise has attracted the attention of and reported by top influential media.