Short-Term Transit Decision Support System Using Multi-task Deep Neural Networks

Fangzhou Sun, Abhishek Dubey, Chinmaya Samal, Hiba Baroud, Chetan Kulkarni

2018 IEEE International Conference on Smart Computing, SMARTCOMP 2018, Taormina, Sicily, Italy, June 18-20, 2018 2018

PDF DOI

Why This Matters

Public transportation riders need accurate delay predictions to make informed decisions about departure times and route choices, but short-term prediction is challenging due to highly variable traffic patterns influenced by numerous contextual factors. This work is innovative because it applies multi-task deep learning to leverage shared patterns across route segments while accounting for event-specific impacts, enabling accurate predictions despite limited historical data for individual segment-event combinations.

What We Did

This paper develops a short-term transit delay prediction system using multi-task deep neural networks that predict arrival times for bus route segments considering contextual information including scheduled events, weather conditions, and historical transit data. The approach uses shared route segment networks and event feature vectors to reduce overfitting while improving prediction accuracy. The system addresses data sparsity and generalization challenges in transit prediction through multi-task learning architecture.

Key Results

The multi-task neural network achieved high recall (76%) and F1 scores (55%) in predicting transit delays, effectively capturing relationships between multiple contextual features. Compared to single-task networks, the multi-task approach reduced overfitting and improved generalization to new events and routes. The system successfully identified which route segments were most affected by specific events and forecast corresponding delays.

Full Abstract

Cite This Paper

@inproceedings{Sun2018,
  author = {Sun, Fangzhou and Dubey, Abhishek and Samal, Chinmaya and Baroud, Hiba and Kulkarni, Chetan},
  booktitle = {2018 {IEEE} International Conference on Smart Computing, {SMARTCOMP} 2018, Taormina, Sicily, Italy, June 18-20, 2018},
  title = {Short-Term Transit Decision Support System Using Multi-task Deep Neural Networks},
  year = {2018},
  acceptance = {40},
  pages = {155--162},
  abstract = {Unpredictability is one of the top reasons that prevent people from using public transportation. To improve the on-time performance of transit systems, prior work focuses on updating schedule periodically in the long-term and providing arrival delay prediction in real-time. But when no real-time transit and traffic feed is available (e.g., one day ahead), there is a lack of effective contextual prediction mechanism that can give alerts of possible delay to commuters. In this paper, we propose a generic tool-chain that takes standard General Transit Feed Specification (GTFS) transit feeds and contextual information (recurring delay patterns before and after big events in the city and the contextual information such as scheduled events and forecasted weather conditions) as inputs and provides service alerts as output. Particularly, we utilize shared route segment networks and multi-task deep neural networks to solve the data sparsity and generalization issues. Experimental evaluation shows that the proposed toolchain is effective at predicting severe delay with a relatively high recall of 76\% and F1 score of 55\%.},
  bibsource = {dblp computer science bibliography, https://dblp.org},
  biburl = {https://dblp.org/rec/bib/conf/smartcomp/SunDSBK18},
  category = {selectiveconference},
  contribution = {lead},
  doi = {10.1109/SMARTCOMP.2018.00086},
  file = {:Sun2018-Short-Term_Transit_Decision_Support_System_Using_Multi-task_Deep_Neural_Networks.pdf:PDF},
  keywords = {transit prediction, deep learning, multi-task learning, delay prediction, transportation},
  project = {smart-transit,cps-reliability,smart-cities},
  tag = {ai4cps,transit},
  timestamp = {Wed, 16 Oct 2019 14:14:54 +0200},
  url = {https://doi.org/10.1109/SMARTCOMP.2018.00086}
}
Quick Info
Year 2018
DOI 10.1109/SMARTCOMP.2018.00086
Keywords
transit prediction deep learning multi-task learning delay prediction transportation
Research Areas
transit ML for CPS
Search Tags

Short, Term, Transit, Decision, Support, System, Multi, task, Deep, Neural, Networks, transit prediction, deep learning, multi-task learning, delay prediction, transportation, transit, ML for CPS, 2018, Sun, Dubey, Samal, Baroud, Kulkarni