Room: Paris
Sponsor(s)
Â
Session Description
Jiahao Wang, a PhD student at the University of Toronto, will demonstrate iROAM and Transit Pulse:
iROAM-
Transit route operations frequently experience bus bunching, crowding, and idling, leading to service inefficiencies. iROAM is an integrated toolbox for anomaly detection, prediction, data preprocessing, and visualization, combining AVL, APC, and GTFS data. By integrating deep learning predictive modeling, iROAM provides proactive alerts, helping transit agencies mitigate disruptions and improve operational efficiency.
Transit Pulse-
High volumes of daily social media posts on public transit offer valuable insights for service improvement. However, manual review is inefficient, and traditional methods—such as TF-IDF—lack nuance by segmenting topics and sentiments rather than examining their interplay. To address these challenges, we introduce a pipeline that leverages a Large Language Model (Llama 3) for integrated topic, sentiment, and sarcasm detection, alongside the identification of unusual system issues and location extraction. Our approach operates without predefined topic labels, further enriched by Retrieval-Augmented Generation (RAG) to incorporate external knowledge sources—specifically GTFS data from the Toronto Transit Commission (TTC)—for more accurate information extraction.
Speaker(s) / Presenter(s)
Jiahao Wang – University of Toronto
Jiahao Wang is a PhD student at the University of Toronto Department of Civil & Mineral Engineering. He received an MSc in computer science from University of Ottawa, and a BEng in computer science from Sichuan University. His current research focuses on transit management optimization using deep learning and reinforcement learning methods. He enjoys skiing and playing guitar.