Predicting mass incidents from weibo

Abstract

The outbreak of mass incidents severely affects the stability of society. If we can predict mass incidents in advance, we may find the solution to avoid the confliction in time. Some of the existing approaches rely on emotional modeling. Much research has been conducted on microblog incident detection using statistical models, like LASSO regression method, Dynamic Query Expansion (DQE) and so on. In this paper, we propose to combine sentiment analysis and statistical methods, and uses LASSO regression method for mass incidents prediction. Experiments on Qingdao demonstrated that our proposed approach achieves a good performance.