基于异构数据特征的城市轨道交通OD客流短时预测方法

陈喜群; 沈楼涛; 李俊懿; 李传家

doi:10.3963/j.jssn.1674-4861.2024.02.016

基于异构数据特征的城市轨道交通OD客流短时预测方法

doi: 10.3963/j.jssn.1674-4861.2024.02.016

1.
浙江大学建筑工程学院杭州 310058
2.
浙江大学工程师学院杭州 310015

基金项目:

国家自然科学基金项目 72171210

详细信息

作者简介:
陈喜群（1986—），博士，教授. 研究方向：交通运输管理、共享出行等

中图分类号: U491
计量
- 文章访问数: 44
- HTML全文浏览量: 26
- PDF下载量: 2
- 被引次数: 0
出版历程
- 收稿日期: 2023-05-29
- 网络出版日期: 2024-09-14

A Short-term Prediction for OD Passenger Flow in Urban Rail Transit Based on Heterogeneous Data Feature Extraction

1.
Institute of Intelligent Transportation Systems, College of Civil Engineering and Architecture, Zhejiang University, Hangzhou 310058, China
2.
Polytechnic Institute, Zhejiang University, Hangzhou 310015, China

摘要

摘要: 城市轨道交通起讫点（origin-destination，OD）客流短时预测在智能交通系统中意义重大，它为交通管控策略实施以及出行者出行选择提供了重要的决策依据。卷积神经网络被广泛用于交通数据空间相关性提取，但其平移不变性与局部敏感性导致该方法更重视局部特征而忽视全局特征。本研究构建了基于注意力机制的异构数据特征提取机模型（heterogeneous data feature extraction machine，HDFEM）以实现OD矩阵空间相关性的全局感知。该模型从时空特征和用地属性特征出发，构造异构数据OD时空张量与地理信息张量，依托模型张量编码层对异构数据张量进行分割与编码，通过注意力机制连接各张量块特征，提取OD矩阵中各个部分间的空间相关性。该方法不仅实现了异构数据与OD客流数据的融合，还兼顾了卷积神经网络模型未能处理的OD矩阵远距离特征，进而帮助模型更全面地学习OD客流的空间特征。对于OD时序特性，该模型使用了长短时记忆网络来处理。在杭州地铁自动售检票系统（auto fare collection，AFC）数据集上的实验结果表明：HDFEM模型相对于基于卷积神经网络的预测模型，其均方误差、平均绝对误差与标准均方根误差分别下降了4.1%，2.5%，2%，验证了全局OD特征感知对于城市轨道交通OD客流预测的重要性。
- 智能交通 /
- OD客流预测 /
- 异构数据融合模型 /
- 深度学习 /
- 注意力机制 /
- 城市轨道交通
Abstract: As an important basis for rail transit operations and travel choices, prediction for origin-destination (OD) passenger flow in urban rail transit is of great significance in intelligent transportation systems. The conventional convolutional neural network (CNN) mostly focuses on local OD features due to their translation invariance and local sensitivity. To improve its global perception capacity in OD matrix modeling, a heterogeneous data feature extraction machine (HDFEM) model is proposed based on attention mechanism. The model constructs a heterogeneous data OD spatio-temporal tensor and a geographic information tensor from the perspective of spatio-temporal characteristics and land use attributes. It segments and encodes heterogeneous data tensors via a tensor coding layer to obtain the features of tensor blocks in heterogeneous data tensors. Then, it connects the features of each tensor block through the attention mechanism to extract the spatial correlation among various OD matrix parts. This approach not only realizes multi-source heterogeneous data fusion, but also extracts remote features of OD matrix. Meanwhile, the model uses long short-term memory (LSTM) network to deal with the OD temporal feature. Compared with the convolutional neural network-based prediction model, the results on the Hangzhou metro auto fare collection (AFC) dataset show that the mean square error, mean absolute error, and normalized root mean square error of the HDFEM model decreases by 4.1%, 2.5%, and 2%, respectively. The importance of extracting whole spatial features for OD passenger flow prediction of urban rail transit is verified.
- intelligent transportation /
- OD passenger flow prediction /
- heterogeneous data fusion model /
- deep learning /
- attention mechanism /
- urban rail transit

HTML全文

图 1 OD时空张量与地理信息张量示意图

Figure 1. Schematic diagram of OD spatio-temporal tensor and geographic information tensor

下载: 全尺寸图片幻灯片

图 2 HDFEM模型框架

Figure 2. Framework of HDFEM

下载: 全尺寸图片幻灯片

图 3 张量编码示意图（以二维张量为例说明）

Figure 3. Schematic diagram of tensor embedding (illustrated by a two-dimensional tensor)

下载: 全尺寸图片幻灯片

图 4 基于异构数据的注意力机制示意图

Figure 4. Diagram of attention mechanism based on heterogeneous data

下载: 全尺寸图片幻灯片

图 5 异构数据特征提取机示意图

Figure 5. Illustration of HDFEM

下载: 全尺寸图片幻灯片

图 6 杭州地铁站点分布图

Figure 6. Distribution of Hangzhou metro stations

下载: 全尺寸图片幻灯片

图 7 部分OD客流预测值与真实值对比图

Figure 7. Comparison of predicted and the ground truth value of some OD flow

下载: 全尺寸图片幻灯片

表 1 HDFEM模型参数寻优

Table 1. Parameter tuning of HDFEM

层数	头数
层数	1	2	4	8
1	11.85	11.85	11.97	12.02
2	11.95	11.97	11.82	12.10
4	12.19	11.94	11.85	11.87

下载: 导出CSV

表 2 模型预测性能比较

Table 2. Comparison of model prediction performance

模型	评价指标
模型	MAE	MSE	NRMSE/%
HA	2.39	31.29	52.77
XGBoost	2.04	19.37	41.52
LASSO	2.07	18.21	40.26
GBDT	2.00	16.90	38.78
RF	1.92	16.23	38.00
LSTM	1.69	15.85	37.55
Hybrid-TCAE-MGM	1.57	10.01	29.85
HDFEM	1.53	9.60	29.24

下载: 导出CSV

参考文献(22)

[1]	张艺铭, 陈明明, 石磊, 等. 基于IGWO-BP算法的轨道交通短时客流预测[J]. 交通信息与安全, 2021, 39(3): 85-92. doi: 10.3963/j.jssn.1674-4861.2021.03.011 ZHANG Y M, CHEN M M, SHI L, et al. A forecast of short-term passenger flow of rail transit based on IGWO-BP algorithm[J]. Journal of Transport Information and Safety, 2021, 39(3): 85-92. (in Chinese) doi: 10.3963/j.jssn.1674-4861.2021.03.011
[2]	蔡昌俊, 姚恩建, 王梅英, 等. 基于乘积ARIMA模型的城市轨道交通进出站客流量预测[J]. 北京交通大学学报, 2014, 38(2): 135-140. CAI C J, YAO E J, WANG M Y, et al. Prediction of urban railway station's entrance and exit passenger flow based on multiply ARIMA model[J]. Journal of Beijing Jiaotong University, 2014, 38(2): 135-140. (in Chinese)
[3]	谢臻, 郭建媛, 秦勇. 基于支持向量回归的地铁进站客流短时预测模型[J]. 都市快轨交通, 2020, 33(2): 82-86. doi: 10.3969/j.issn.1672-6073.2020.02.014 XIE Z, GUO J Y, QIN Y. Short-term prediction model of subway entry passenger flow based on support vector regression[J]. Urban Rapid RailTransit, 2020, 33(2): 82-86. (in Chinese) doi: 10.3969/j.issn.1672-6073.2020.02.014
[4]	帅春燕, 谢亚威, 单君, 等. 基于SSA-SVR模型的城市轨道交通短时进站客流预测[J]. 都市快轨交通, 2022, 35(5): 76-83. SHUAI C Y, XIE Y W, SHAN J, et al. Prediction of short-term inbound passenger flow of urban rail transit based on the singular spectrum analysis and support vector regression model[J]. Urban Rapid Rail Transit, 2022, 35(5): 76-83. (in Chinese)
[5]	YANG X, XUE Q, DING M, et al. Short-term prediction of passenger volume for urban rail systems: A deep learning approach based on smart-card data[J]. International Journal of Production Economics, 2021, 231: 107920. doi: 10.1016/j.ijpe.2020.107920
[6]	崔洪涛, 陈晓旭, 杨超, 等. 基于深度长短期记忆网络的地铁进站客流预测[J]. 城市轨道交通研究, 2019, 22(9): 41-45. CUI H T, CHEN X X, YANG C, et al. Forecast of subway inbound passenger flow based on DLSTM recurrent network[J]. Urban Mass Transit, 2019, 22(9): 41-45. (in Chinese)
[7]	施俊庆, 李睿, 程明慧, 等. 基于动态时空神经网络模型的地铁客流预测[J]. 交通运输系统工程与信息, 2023, 23(2): 139-147. SHI J Q, LI R, CHENG M H, et al. Metro passenger flow prediction based on dynamic spatio-temporal neural network model[J]. Journal of Transportation Systems Engineering and Information Technology, 2023, 23(2): 139-147. (in Chinese)
[8]	HAO S, LEE D H, ZHAO D. Sequence to sequence learning with attention mechanism for short-term passenger flow prediction in large-scale metro system[J]. Transportation Research Part C: Emerging Technologies, 2019, 107: 287-300. doi: 10.1016/j.trc.2019.08.005
[9]	LIU Y, LIU Z, JIA R. DeepPF: a deep learning based architecture for metro passenger flow prediction[J]. Transportation Research Part C: Emerging Technologies, 2019, 101: 18-34. doi: 10.1016/j.trc.2019.01.027
[10]	DJUKIC T, VAN LINT J W C, HOOGENDOORN S P. Application of principal component analysis to predict dynamic origin-destination matrices[J]. Transportation Research Record, 2012, 2283(1): 81-89. doi: 10.3141/2283-09
[11]	BASAGOITI R, BEAMURGIA M, PETERS R, et al. Origin destination matrix estimation and prediction in vertical transportation[C]. 2^nd Symposium on Lift and Escalator Technologies, Northampton, UK: LESET, 2012.
[12]	LEE C Y, BOTTOU L, BENGIO Y, et al. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 1998, 86(11): 2278-2324. doi: 10.1109/5.726791
[13]	徐新颖. 基于卷积神经网络的城市轨道交通短时客流预测方法研究[D]. 福州: 福州大学, 2023. XU X Y. Research on short-term passenger flow forecasting method of urban rail transit based on convolutional neural network[D]. Fuzhou: Fuzhou University, 2023.
[14]	王博文, 王景升, 王统一, 等. 基于卷积神经网络与门控循环单元的交通流预测模型[J]. 重庆大学学报, 2023, 46(8): 132-140. WANG B W, WANG J S, WANG T Y, et al. Multivariable traffic flow prediction model based on convolutional neural network and gate recurrent unit[J]. Journal of Chongqing University, 2023, 46(8): 132-140. (in Chinese)
[15]	SHEN L, SHAO Z, YU Y, et al. Hybrid approach combining modified gravity model and deep learning for short-term forecasting of metro transit passenger flows[J]. Transportation Research Record, 2021, 2675(1): 25-38.
[16]	ZHANG J, CHE H, CHEN F, et al. Short-term origin-destination demand prediction in urban rail transit systems: a channel-wise attentive split-convolutional neural network method[J]. Transportation Research Part C: Emerging Technologies, 2021, 124: 102928.
[17]	CHU K F, LAM A Y S, LI V O K. Deep multi-scale convolutional LSTM network for travel demand and origin-destination predictions[J]. IEEE Transactions on Intelligent Transportation Systems, 2019, 21(8): 3219-3232.
[18]	NOURSALEHI P, KOUTSOPOULOS H N, ZHAO J. Dynamic origin-destination prediction in urban rail systems: a multi-resolution spatio-temporal deep learning approach[J]. IEEE Transactions on Intelligent Transportation Systems, 2021, 23(6): 5106-5115.
[19]	李若怡. 基于改进时空LSTM模型的城市轨道交通系统OD客流短时预测[D]. 北京: 北京交通大学, 2020. LI R Y. Short-term prediction of origin-destination passenger flow for urban rail transit based on improved spatio-temporal LSTM model[D]. Beijing: Beijing Jiaotong University, 2020.
[20]	DOSOVITSKIY A, BEYER L, KOLESNIKOV A, et al. An Image is worth 16×16 words: transformers for image recognition at scale[EB/OL]. (2021-6-3)[2023-6-19]. https://doi.org/10.48550/arXiv.2010.11929.
[21]	VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[J]. Advances in Neural Information Processing Systems, 2017, 30: 5998-6008.
[22]	HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780.