Most traffic congestion research focuses on recurring patterns, leaving non-recurring congestion underexplored. Events cause significant congestion but are difficult to detect from traffic data alone. This work is innovative because it applies deep learning to identify event-caused congestion and provides methods to explain network predictions.
This paper presents DxNAT, a deep neural network approach for identifying and explaining non-recurring traffic congestion caused by events. The work converts traffic data to images and applies convolutional neural networks to classify congestion patterns. The methodology demonstrates high accuracy in detecting event-related congestion from traffic sensor data.
The paper achieves 98.73 percent accuracy in identifying non-recurring traffic congestion using deep neural networks. Results demonstrate successful detection of congestion caused by sports events and accidents, enabling better understanding of event impacts on urban traffic.
@inproceedings{Sun2017,
author = {Sun, Fangzhou and Dubey, Abhishek and White, Jules},
booktitle = {2017 {IEEE} International Conference on Big Data, BigData 2017, Boston, MA, USA, December 11-14, 2017},
title = {DxNAT - Deep neural networks for explaining non-recurring traffic congestion},
year = {2017},
pages = {2141--2150},
abstract = {Non-recurring traffic congestion is caused by temporary disruptions, such as accidents, sports games, adverse weather, etc. We use data related to real-time traffic speed, jam factors (a traffic congestion indicator), and events collected over a year from Nashville, TN to train a multi-layered deep neural network. The traffic dataset contains over 900 million data records. The network is thereafter used to classify the real-time data and identify anomalous operations. Compared with traditional approaches of using statistical or machine learning techniques, our model reaches an accuracy of 98.73 percent when identifying traffic congestion caused by football games. Our approach first encodes the traffic across a region as a scaled image. After that the image data from different timestamps is fused with event- and time-related data. Then a crossover operator is used as a data augmentation method to generate training datasets with more balanced classes. Finally, we use the receiver operating characteristic (ROC) analysis to tune the sensitivity of the classifier. We present the analysis of the training time and the inference time separately.},
bibsource = {dblp computer science bibliography, https://dblp.org},
biburl = {https://dblp.org/rec/bib/conf/bigdataconf/SunDW17},
category = {selectiveconference},
contribution = {lead},
doi = {10.1109/BigData.2017.8258162},
file = {:Sun2017-DxNAT-Deep_neural_networks_for_explaining_non-recurring_traffic_congestion.pdf:PDF},
keywords = {traffic congestion, deep learning, anomaly detection, non-recurring congestion, event detection},
project = {smart-transit,smart-cities,cps-reliability},
tag = {ai4cps,transit},
timestamp = {Wed, 16 Oct 2019 14:14:51 +0200},
url = {https://doi.org/10.1109/BigData.2017.8258162}
}