Visualizing and Mining Social Media Data for Smart Emergency Management

Abstract. The ability of a community to prepare for, absorb, recover from, and more successfully adapt to disastrous events is defined as disaster resilience. Disaster resilience can be better understood by investigating human behaviors during the four phases of emergency management – preparedness, response, recovery, and mitigation. However, a major challenge is that data describing communities’ behaviors in different phases of emergency management are difficult to access through traditional databases. Social media such as Twitter is increasingly being used as an effective platform to observe human behaviors in disastrous events. These responses and behaviors could be better understood by analyzing real-time social media data through categorizing them into different phases of the emergency management.

This research studies the Twitter use during 2012 Hurricane Sandy and 2017 Hurricane Harvey, which struck the U.S. northeast and south coasts, respectively. The objectives are fourfold: (1) to develop a Twitter data mining and visualization framework and a set of indexes for emergency management and resilience analysis; (2) to visualize the spatial-temporal patterns of disaster-related Twitter activities during the two hurricane events; (3) to examine and compare the social-geographical disparities of disaster-related Twitter activities during Sandy and Harvey; and (4) to build applications using social media data for smart management, including surveying human behaviors and emergency rescue.