Tracking Urban Geo-Topics based on Dynamic Topic Model

Authors: Fang Yao, PhD student, University of Florida, Yan Wang*, Assistant Professor, University of Florida
Topics: Urban and Regional Planning, Spatial Analysis & Modeling, Quantitative Methods
Keywords: Geo-Topics, Social Media, Spatial Trajectory, Dynamic Topic Model, Urban Informatics,
Session Type: Poster
Day: 4/6/2019
Start / End Time: 9:55 AM / 11:35 AM
Room: Lincoln 2, Marriott, Exhibition Level
Presentation File: No File Uploaded


The detection and tracking of urban physical events (e.g. emergencies) can reveal important patterns of human dynamics in complex urban environments. These physical events can be represented by geographically-close and semantically-similar geo-topics. The widespread usage of social networking platforms provides unprecedented scales of geolocated data, and the improvement of urban computing technique also enables the bottom-up understanding of urban dynamics. We propose a new approach to improve urban informatics and address the challenge of tracking smaller-scale geo-topics in modern cities. Our research develops an intelligent data-driven system to identify and track the spatial, temporal and semantic changes of geo-topics with geo-tagged Tweets. We employed a set of data preprocessing procedures to remove bots and normalize the texts. The system used the Dynamic Topic Model by embedding spatial factors to generate geo-topics and track their spatial trajectory over time using radius of gyration and a trajectory pattern mining method. The proposed tracking system has been validated in three different disaster cases (i.e. Hurricane Florence and California Wildfires) in distinct cities in the United States. This validation demonstrates the usefulness and effectiveness of the system in identifying and tracking geo-topics and smaller-scale emergencies during disasters. We also found that patterns of emergencies are different from the other geo-topics especially at the early stage of a disaster. The proposed system has the potential to be used for tracking population dynamics from other spatial datasets and provides the basis for automatic emergency detection.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login