Traffic congestion is a significant challenge that cities worldwide have to tackle as it poses many potential risks. Building a predictive system to anticipate future congestion would alleviate them. Furthermore, if the system can discover future frequent traffic congestion patterns, authorities can build reaction plans to deal with congestion more effectively. Unfortunately, other works have failed to achieve it. This study proposes a novel dynamic system to address the mentioned problem. It integrates a traffic congestion prediction model and a periodic-frequent pattern discovery algorithm. In particular, we utilize our novel Fusion-3DCNN deep learning model and a periodic-frequent pattern discovery algorithm in the system. The former predicts long-term traffic congestion on citywide mesh codes using multi-modal urban sensing data, while the latter identifies sets of mesh codes that are regularly predicted to have heavy traffic congestion. Experimental results on a real-world dataset collected in Kobe City, Japan, from 2014 to 2015 show that our framework is efficient in terms of accuracy and time.
Insights for Urban Road Safety: A New Fusion-3DCNN-PFP Model to Anticipate Future Congestion from Urban Sensing Data
Periodic Pattern Mining ; Chapter : 14 ; 237-263
2021-10-30
27 pages
Article/Chapter (Book)
Electronic Resource
English
CHARACTERISING CONGESTION ON URBAN ROAD NETWORKS
British Library Conference Proceedings | 2003
|Urban Road Congestion and Seaport Competition
Online Contents | 2013
|Research on Evaluation of Urban Road Congestion
Trans Tech Publications | 2014
|Model-Based Traffic Congestion Control in Urban Road Networks
Transportation Research Record | 2013
|Study on the Model of Urban Road Congestion Charging
ASCE | 2011
|