Authors: Qunying Huang*, University of Wisconsin - Madison
Topics: Geographic Information Science and Systems, Hazards, Risks, and Disasters, Cyberinfrastructure
Keywords: Natural hazards, GeoAI, deep learning, social media
Session Type: Paper
Start / End Time: 8:00 AM / 9:40 AM
Room: Washington 6, Marriott, Exhibition Level
Presentation File: No File Uploaded
This paper presents a framework for real-time geographic situational awareness (SA) information analysis and disaster management. Within this framework, a common classifier is developed to automatically categorize the social media messages into different categories during a disaster. Such common classifier could help support real-time disaster management and analysis by monitoring subsequent events while social media data are streaming, and mining useful information. Geo-tagged social media data from different natural hazard events (e.g., the Hurricane Sandy, Harvey, and Irma) will be used to train and build such a common classifier. Commonly used machine learning and deep learning models, such as Naive Bayes, Support Vector Machine, Convolution Neural Network, will be tested to select an optimal model for building the common classifier. Different data processing and optimization strategies will be explored to generate optimal features for developing the classification model, and ensure the classifier built on data from exiting events can extract relevant information for a future event. Finally, a spatial web portal will be developed to analyze and explore the classified SA information for informing decision making. The study addresses the need for incorporating spatial data when using social media in real time disaster management applications.