GANNSTER: Graph-augmented neural network spatio-temporal reasoner for traffic forecasting

Carlos Salort Sánchez, Alexander Wieder, Paolo Sottovia, Stefano Bortoli, Jan Baumbach, Cristian Axenie

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

2 Scopus citations

Abstract

Traffic forecast is a problem of high interest due to its impact on mobility and inherent socio-economic aspects of people’s lives. Particularly for adaptive traffic light systems, the ability to predict traffic throughput in intersections enables fast adaptation, thus reducing traffic jams. In this work, we propose a novel approach for traffic forecasting, termed Graph Augmented Neural Network Spatio-TEmporal Reasoner (GANNSTER), which fuses spatial information, given by the traffic network topology, with temporal reasoning and learning capabilities of recurrent neural networks. Our modelling contribution is supplemented by the public release of a novel real-world dataset containing urban traffic throughput in intersections. We comparatively evaluate GANNSTER against state-of-the-art models for traffic forecast and demonstrate its superior performance.

Original languageEnglish
Title of host publicationAdvanced Analytics and Learning on Temporal Data - 5th ECML PKDD Workshop, AALTD 2020, Revised Selected Papers
EditorsVincent Lemaire, Simon Malinowski, Anthony Bagnall, Thomas Guyet, Romain Tavenard, Georgiana Ifrim
PublisherSpringer Science and Business Media Deutschland GmbH
Pages63-76
Number of pages14
ISBN (Print)9783030657413
DOIs
StatePublished - 2020
Event5th ECML PKDD Workshop on Advanced Analytics and Learning on Temporal Data, AALTD 2020 - Ghent, Belgium
Duration: 18 Sep 202018 Sep 2020

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume12588 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference5th ECML PKDD Workshop on Advanced Analytics and Learning on Temporal Data, AALTD 2020
Country/TerritoryBelgium
CityGhent
Period18/09/2018/09/20

Keywords

  • Deep learning
  • Graph neural network
  • Traffic forecast

Fingerprint

Dive into the research topics of 'GANNSTER: Graph-augmented neural network spatio-temporal reasoner for traffic forecasting'. Together they form a unique fingerprint.

Cite this