Predicting the Retweeting of Social Media Content during Disasters

The spread of disaster-related information in social media is critical for hazard control as social media platform is widely adopted in emergency communicating during disasters, such as epidemics, earthquakes, and floods. However, it is difficult to diffuse essential information to target users effectively as the posts in social media are massive, complex, and highly distributed. Prior related studies overlooked the impacts of disaster-related textual features of posts in terms of predicting information retweeting during disasters. Thus, this paper proposes a XGBoost-based retweetability prediction model by leveraging disaster-related features together with conventional indicators. To validate our model, this study collects social media data of 2013 Colorado flood from Twitter and compares the performance of proposed model with baselines. The statistical results demonstrate the effectiveness of the proposed model in retweeting prediction