IPOD Abstract for presentation (Poster or Podium)
Intelligent Transportation Systems
Yufei Xu
Graduate research assistant
Georgia Institue of Technology, United States
Yufei Xu
Graduate research assistant
Georgia Institue of Technology, United States
Srinivas Peeta, Ph.D.
Frederick R. Dickerson Chair and Professor
Georgia Institute of Technology
Atalanta, Georgia, United States
Yufei Xu
Graduate research assistant
Georgia Institue of Technology, United States
Accurate traffic incident detection and analysis are essential for effective traffic planning, management, and control. Numerous deep learning techniques have been proposed to identify traffic incidents in terms of their timing, location, and type, using loop detector or social media data. However, this remains a challenging problem due to constraints such as limited incident data, limited availability of loop detectors, and inherent noise in social media data. We propose to integrate domain knowledge with advanced natural language models to provide more precise and informative incident detection and analysis. The proposed model is designed to identify traffic incidents from social media and analyze their potential impact on network performance by leveraging domain knowledge and information extracted from disaggregated social media data. In addition, the proposed model considers the potential presence of fake information in social media, which is crucial for ensuring the accuracy of incident detection and analysis. Numerical experiments illustrate the effectiveness of the proposed approach.