基于图卷积门控循环网络的个体地铁出行预测

doi:10.3969/j.issn.1674-0696.2022.09.01

重庆交通大学学报（自然科学版） ›› 2022, Vol. 41 ›› Issue (09): 1-8.DOI: 10.3969/j.issn.1674-0696.2022.09.01

• 交通+大数据人工智能 • 下一篇

基于图卷积门控循环网络的个体地铁出行预测

翁小雄，覃镇林，张鹏飞

(华南理工大学土木与交通学院，广东广州 510630)

收稿日期:2021-03-01 修回日期:2021-07-11 发布日期:2022-09-30
作者简介:翁小雄(1958—)，女，浙江杭州人，教授，博士，主要从事大数据挖掘方面的研究。E-mail:cxxweng@qq.com 通信作者：覃镇林(1995—)，男，广西梧州人，硕士研究生，主要从事大数据挖掘，人类行为分析等方面的研究。E-mail：408011458@qq.com
基金资助:
国家自然科学基金项目(51578247)

Individual Metro Travel Prediction Based on Graph Convolution Gated Recurrent Network

WENG Xiaoxiong, QIN Zhenlin, ZHANG Pengfei

(School of Civil Engineering and Transportation, South China University of Technology, Guangzhou 510630, Guangdong, China)

Received:2021-03-01 Revised:2021-07-11 Published:2022-09-30

摘要/Abstract

摘要： 针对个体地铁出行受到地理位置的限制，对地铁出行预测任务提出一种时空特性建模的方法，解决过去的研究中只对出行序列的时间特征建模，而不能充分挖掘数据中的空间特性的问题。提出的方法先对个体出行序列的空间特征进行构建，通过图卷积神经网络提取空间特征，使用门控的方法将表示时空特性的嵌入向量进行融合，然后使用门控循环网络对时空特征进行学习。通过广州地铁羊城通刷卡数据构建的个体出行序列数据集进行验证，该模型的预测准确率高于只对时间特性建模的模型。最后通过构造具有显著个体出行空间特性的人造数据与真实数据进行不同比例混合，验证该模型能有效学习个体出行的空间特性，具有更强的鲁棒性。

关键词: 交通运输工程；个体出行预测；个体出行时空特性；图卷积神经网络；门控循环单元

Abstract: Aiming at the restriction of individual metro travel caused by geographic location, a spatiotemporal characteristic modelling method was proposed for the metro travel prediction task, which solved the problem that only the time characteristics of travel series were modeled, and the spatial characteristics in the data could not be fully mined in the past research. The proposed method firstly constructed the spatial features of individual travel series, extracted spatial features through graph convolutional neural network and fused the embedded vectors representing the spatiotemporal characteristics by using gating. And then, the gated recurrent networks were used to learn spatiotemporal features. The individual travel sequence data set constructed by Guangzhou Metro Yangchengtong card swiping data was verified, and the prediction accuracy of the proposed model was higher than that of the model that only modeled the time characteristics. Finally, by constructing artificial data with significant individual travel spatial characteristics and mixing them with real data in different proportions, it is verified that the proposed model can effectively learn the spatial characteristics of individual travel and has better robustness.

Key words: traffic and transportation engineering; individual travel prediction; spatiotemporal characteristics of individual travel; graph convolution neural network; gated recurrent unit

中图分类号:

U491.1+4

翁小雄，覃镇林，张鹏飞. 基于图卷积门控循环网络的个体地铁出行预测[J]. 重庆交通大学学报（自然科学版）, 2022, 41(09): 1-8.

WENG Xiaoxiong, QIN Zhenlin, ZHANG Pengfei. Individual Metro Travel Prediction Based on Graph Convolution Gated Recurrent Network[J]. Journal of Chongqing Jiaotong University(Natural Science), 2022, 41(09): 1-8.

参考文献

［1］ FAINA K, BOAZ L. Human mobility-pattern discovery and next-place prediction from GPS data ［J］. Lecture Notes in Computer Science, 2017, 10183: 24-35.
［2］ JIANG R, FONG X, FAN Z, et al. Deep ROI-based modeling for urban human mobility prediction ［J］. Proc. ACM Interactive, Mobile, Wearable Ubiquitous Technol, 2018, 2(1): 1-29.
［3］ YANG C, YAN F, UKKUSURI S V. Unraveling traveler mobility patt-erns and predicting user behavior in the Shenzhen metro system ［J］. Transportmetrica A: Transport Science, 2018, 14(7): 576-597.
［4］ WANG H, ZHAO J, YE K, et al. A destination prediction model for individual passengers in urban rail transit［C］∥ 2020 International Conference on High Performance Big Data and Intelligent Systems (HPBD IS). IEEE, 2020: 1-6.
［5］ ZHAO Z, KOUTSOPOULOS H N, ZHAO J. Individual mobility predic-tion using transit smart card data ［J］. Transportation Research Part C: Emerging Technologies, 2018, 89: 19-34.
［6］ HARRIS D M, HARRIS S L. Digital Design and Computer Architecture［M］. 2nd ed. Waltham: Morgan Kaufmann, 2012: 129-131.
［7］ WU Y, LIAN D, JIN S, et al. Graph convolutional networks on user mobility heterogeneous graphs for social relationship inference［C］∥ IJCAI International Joint Conference on Artificial Intelligence. San Francisco: Morgan Kaufmann, 2019: 3898-3904.
［8］ WANG Z, WU C H, LI Q B, et al. Encoding text information with graph convolutional networks for personality recognition ［J］. Applied Sciences (Switzerland), 2020, 10(12):4081-4082.
［9］ ZHAO L, SONG Y, ZHANG C, LIU Y, et al. T-GCN: A temporal graph convolutional network for traffic prediction［J］. IEEE Transactions on Intelligent Transportation Systems, 2020, 21(9): 3848-3858.
［10］ WANG S, ZHOU W, JIANG C. A survey of word embeddings based on deep learning ［J］. Computing, 2020, 102(3):717-740.
［11］闫佳庆，邵鹏，张立立，等. 考虑大型车因素的城市路网交通流速度预测［J］. 重庆交通大学学报(自然科学版)，2021, 40(1):7-11，16.
YAN Jiaqin, SHAO Peng, ZHANG Lili, et al.Traffic flow velocity prediction of urban road network considering large-scale vehicles ［J］. Journal of Chongqing Jiaotong University (Natural Science), 2021, 40(1): 7-11，16.

[1]	裴同松1，裴彧2. 基于马尔科夫链-BP神经网络模型对公路运量的预测研究[J]. 重庆交通大学学报（自然科学版）, 2021, 40(02): 35-41.
[2]	邬岚，王彤伟，李鹏举，翟伟. 基于窄马路理念的车道缩减设计方法[J]. 重庆交通大学学报（自然科学版）, 2020, 39(11): 26-32.
[3]	陈沿伊，曹莹，张佳楠，冯睿. 基于DEMATEL-BCC模型的深圳市生态交通发展效率评价[J]. 重庆交通大学学报（自然科学版）, 2020, 39(07): 27-32.
[4]	张兰怡，胡喜生，陈清耀，邱荣祖. 基于PCA-Logistic回归的汽车保有量预测研究[J]. 重庆交通大学学报（自然科学版）, 2017, 36(5): 104-109.
[5]	朱顺应，王红，李关寿. 路段上短时间区段内交通量预测ARIMA模型[J]. 重庆交通大学学报（自然科学版）, 2003, 22(1): 76-77.
[6]	徐建闽1，臧鹏1，首艳芳2. 基于ARIMA模型和K-means聚类分析的动态规划算法[J]. 重庆交通大学学报（自然科学版）, 2022, 41(07): 9-13.
[7]	翁小雄1，任杰1，覃镇林1，罗瑞发2. 基于深度注意力模型的个体出行多步预测研究[J]. 重庆交通大学学报（自然科学版）, 2022, 41(10): 35-40.

基于图卷积门控循环网络的个体地铁出行预测

Individual Metro Travel Prediction Based on Graph Convolution Gated Recurrent Network

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 7

编辑推荐

Metrics