Hierarchical Graph Structures for Congestion and ETA Prediction

November 21, 2022 · Entered Twilight · 🏛 arXiv.org

Repo contents: .flake8, .github, .gitignore, .gitlab-ci.yml, .pre-commit-config.yaml, Dockerfile, LICENSE, Main-image-web-08.png, README.md, README_DATA_SPECIFICATION.md, README_competition.md, baselines, blogposts, data_pipeline, environment.yaml, exploration, img, install-extras-torch-geometric.txt, install-requirements.txt, setup.cfg, setup.py, t4c20logo.png, t4c22, vulture_whitelist.py

Authors Florian Grötschla, Joël Mathys arXiv ID 2211.11762 Category cs.LG: Machine Learning Cross-listed cs.AI Citations 4 Venue arXiv.org Repository https://github.com/floriangroetschla/NeurIPS2022-traffic4cast ⭐ 3 Last Checked 3 months ago

Abstract

Traffic4cast is an annual competition to predict spatio temporal traffic based on real world data. We propose an approach using Graph Neural Networks that directly works on the road graph topology which was extracted from OpenStreetMap data. Our architecture can incorporate a hierarchical graph representation to improve the information flow between key intersections of the graph and the shortest paths connecting them. Furthermore, we investigate how the road graph can be compacted to ease the flow of information and make use of a multi-task approach to predict congestion classes and ETA simultaneously. Our code and models are released here: https://github.com/floriangroetschla/NeurIPS2022-traffic4cast