查看原文
其他

【综述专栏】交通预测中的图神经网络

在科学研究中,从方法论上来讲,都应“先见森林,再见树木”。当前,人工智能学术研究方兴未艾,技术迅猛发展,可谓万木争荣,日新月异。对于AI从业者来说,在广袤的知识森林中,系统梳理脉络,才能更好地把握趋势。为此,我们精选国内外优秀的综述文章,开辟“综述专栏”,敬请关注。

来源:知乎—夏哈哈

地址:https://zhuanlan.zhihu.com/p/437058060


01

文章信息

标题:Graph Neural Network for Traffic Forecasting: A Survey

来源:Arxiv 2021 Weiwei Jiang Tsinghua University

paper link:

https://arxiv.org/abs/2101.11174

code link:
https://github.com/jwwthu/GNN4Traffic


02

摘要
CNN,RNN等方法被用于交通预测问题中来对时间和空间相关性进行建模。近年来为了对交通系统中的图结构和环境信息(contextual information)进行建模,引入了入图神经网络(GNN)并取得了很多任务的SOTA效果。本文中回顾了各种GNN(例如GCN,GAT等)在不同交通预测问题(例如 道路交通流预测、速度预测,人流量预测,打车供需预测等)中的应用。
同时提出了一系列开源的数据和模型实现代码,并探讨了交通预测中存在的问题和下一步研究方向。

03

介绍

数据来源:

sensors installed on roads, subway and bus system transaction records, traffic surveillance videos, smartphone GPS,...
难点:大规模高维数据;多种动态(包括交通事故等紧急情况)
传统线性时间序列模型(如ARIMA等)无法有效处理时空预测问题,故引入了机器学习方法(ML)和深度学习技术来改性预测性能,如文献[1]将城市建模为网格(grid)后再用CNN来处理。但基于CNN的方法不是处理具有图形式的时空预测问题的最优方法。
GNN优点:能捕获时空预测问题中利用非欧图结构展示的空间关联。
现有的交通预测领域综述性文件[5-19][2]均未能像本文一样从图神经网络角度进行综述。

本文贡献如下:

1. 全面的回顾:提供了交通预测领域最全面的基于图的解决方法(2018-2020)
2. 资源搜集:提供了最近全面的开源数据集和代码资源,用于工作的复现和比较
3. 未来方向:讨论了该领域一些挑战和未来可能的研究方向


04

相关研究
该节主要是对不同的交通预测方法进行分类。

依据交通状态(traffic state)进行分类:traffic flow, traffic speed, traffic demand,others;之后再利用road-level, region-level, station-level进行细分


05

图和图神经网络

交通图

Graph Construction:G=(V,E,A),where V is the set of vertices or nodes, E is the set of edges between the nodes, and A is the adjacency matrix Element aij of A represents the “edge weight” between nodes i and j.
   , 其中    为外部特征(例如天气,节假日,事件等)


06

挑战和下一步研究方向

挑战

1. 异构数据(Heterogeneous Data)
2. 多任务的性能(Multi-task Performance)
3. 实际实现(Practical Implementation)

下一步研究方向

1. 集中式数据存储库(Centralized Data Repository)
2. 和其它技术结合(Combine with Other Techniques)
数据增强(Data Augmentation)
迁移学习(Transfer Learning)
元学习(Meta Learning)
生成对抗网络(Generative Adversarial Network (GAN))
自动机器学习(Automated Machine Learning (AutoML))
贝叶斯网络(Bayesian Network)

07

其他

单词:

encompass:vt. 包含,包括,涉及;包围,围绕;
e-scooter:n. 电动摩托车;智能踏板车 scooter: n.小型摩托车;滑板车
intervention:n. 干涉,干预
surveillance : n. 监控

句式:

to be continued...

参考

1. Geospatial data to images: A deep-learning framework for traffic forecasting
2. [5] X. Shi and D.-Y. Yeung, “Machine learning for spatiotemporal sequence forecasting: A survey,” arXiv preprint arXiv:1808.06865, 2018. [6] D. Pavlyuk, “Feature selection and extraction in spatiotemporal traffic forecasting: a systematic literature review,” European Transport Research Review, vol. 11, no. 1, p. 6, 2019. [7] X. Yin, G. Wu, J. Wei, Y. Shen, H. Qi, and B. Yin, “A comprehensive survey on traffic prediction,” arXiv preprint arXiv:2004.08555, 2020. [8] M. Luca, G. Barlacchi, B. Lepri, and L. Pappalardo, “Deep learning for human mobility: a survey on data and models,” arXiv preprint arXiv:2012.02825, 2020. [9] X. Fan, C. Xiang, L. Gong, X. He, Y. Qu, S. Amirgholipour, Y. Xi, P. Nanda, and X. He, “Deep learning for intelligent traffic sensing and prediction: recent advances and future challenges,” CCF Transactions on Pervasive Computing and Interaction, pp. 1–21, 2020. [10] A. Boukerche and J. Wang, “Machine learning-based traffic prediction models for intelligent transportation systems,” Computer Networks, vol. 181, p. 107530, 2020. [11] E. L. Manibardo, I. Lana, and J. Del Ser, “Deep learning for ˜ road traffic forecasting: Does it make a difference?” arXiv preprint arXiv:2012.02260, 2020. [12] J. Ye, J. Zhao, K. Ye, and C. Xu, “How to build a graph-based deep learning architecture in traffic domain: A survey,” arXiv preprint arXiv:2005.11691, 2020. [13] K. Lee, M. Eo, E. Jung, Y. Yoon, and W. Rhee, “Short-term traffic prediction with deep neural networks: A survey,” arXiv preprint arXiv:2009.00712, 2020. [14] P. Xie, T. Li, J. Liu, S. Du, X. Yang, and J. Zhang, “Urban flow prediction from spatiotemporal data using machine learning: A survey,” Information Fusion, vol. 59, pp. 1–12, 2020. [15] S. George and A. K. Santra, “Traffic prediction using multifaceted techniques: A survey,” Wireless Personal Communications, vol. 115, no. 2, pp. 1047–1106, 2020. [16] A. K. Haghighat, V. Ravichandra-Mouli, P. Chakraborty, Y. Esfandiari, S. Arabi, and A. Sharma, “Applications of deep learning in intelligent transportation systems,” Journal of Big Data Analytics in Transportation, vol. 2, no. 2, pp. 115–145, 2020. [17] A. Boukerche, Y. Tao, and P. Sun, “Artificial intelligence-based vehicular traffic flow prediction methods for supporting intelligent transportation systems,” Computer Networks, vol. 182, p. 107484, 2020. [18] D. A. Tedjopurnomo, Z. Bao, B. Zheng, F. Choudhury, and A. Qin, “A survey on modern deep neural network for traffic prediction: Trends, methods and challenges,” IEEE Transactions on Knowledge and Data Engineering, 2020. [19] V. Varghese, M. Chikaraishi, and J. Urata, “Deep learning in transport studies: A meta-analysis on the prediction accuracy,” Journal of Big Data Analytics in Transportation, pp. 1–22, 2020.

本文目的在于学术交流,并不代表本公众号赞同其观点或对其内容真实性负责,版权归原作者所有,如有侵权请告知删除。


“综述专栏”历史文章


更多综述专栏文章,

请点击文章底部“阅读原文”查看



分享、点赞、在看,给个三连击呗!

您可能也对以下帖子感兴趣

文章有问题?点此查看未经处理的缓存