基于多尺度时空特征与软注意力机制的交通流预测方法

doi:10.19678/j.issn.1000-3428.0068659

摘要/Abstract

摘要：

交通流预测在规划交通系统、优化道路资源和缓解交通拥堵等方面具有重要意义。针对交通流预测中时间周期性特征提取不充分的问题, 提出一种基于多尺度时空特征和软注意力机制的交通流预测方法MSTFSA。首先, 利用图交谈注意力网络(GTHAT)提取空间数据的非欧几里得结构特征, 通过分配动态权重表征不同时间相邻道路交通流的影响程度; 其次, 利用双向增强注意力门控循环单元(Bi-EAGRU)结构提取时间数据的连续性关联特征, 增强每个时刻的时间特征与上下时刻的联系; 然后, 基于软注意力机制融合周周期、日周期和近邻时间3个尺度下的相似交通流趋势, 实现对时间周期性特征的充分提取, 最后, 结合高速公路数据集PeMS04和PeMS08验证MSTFSA的预测精度。实验结果表明, MSTFSA的交通流预测精度表现出良好效果, 与基线模型STSGCN和ASTGCN相比, 在PeMS04数据集上的预测均方根误差(RMSE)分别降低7.15%和3.8%, 平均绝对误差(MAE)分别降低7.79%和3.99%。MSTFSA能较充分地提取并融合交通数据的多时间尺度时空特征, 在交通流预测精度提升方面表现出一定的优势。

关键词: 交通流预测, 时空域联合, 图注意力网络, 软注意力机制, 双向门控循环单元

Abstract:

Traffic flow prediction has considerable value in design of transportation systems, optimization of road resources, and mitigation of traffic congestion. To address the issue of limited prediction accuracy due to insufficient extraction of temporal periodic features in traffic flow forecasting, in this study, a multi-scale spatio-temporal features soft attention mechanism method MSTFSA is proposed for traffic flow forecasting. The method is based on multi-scale spatial and temporal features and a soft attention mechanism. First, the Graph Talking Head Attention Network (GTHAT) is used to extract the non-Euclidean structural features of the spatial data. The dynamic weights are calculated to represent the impact of traffic flow on adjacent roads at different times. Second, a Bidirectional Enhanced Attention Gated Recurrent Unit (Bi-EAGRU) is utilized to capture the continuity correlation features of temporal data, thereby enhancing the temporal features of each moment and the continuity between adjacent moments. Subsequently, similar traffic flow trends at three scales of periodicity: weekly, daily, and nearest-neighbor time are fused based on soft attention to implement the comprehensive extraction of temporal periodic features. Finally, the prediction accuracy of MSTFSA is verified on the highway datasets PeMS04 and PeMS08. The experimental results demonstrate that MSTFSA provides distinct advantages in terms of traffic flow prediction accuracy. Compared with the baseline methods of Spatio-Temporal Synchronous Graph Convolutional Network (STSGCN) and Attention-based Spatio-Temporal Graph Convolutional Network (ASTGCN), MSTFSA not only reduces the Root Mean Square Error (RMSE) by 7.15% and 3.8% but also decreases the Mean Absolute Error (MAE) by 7.79% and 3.99% on PeMS04 dataset, respectively. In summary, MSTFSA can efficiently extract and merge the multi-temporal and spatial attributes of traffic data, thereby considerably improving the prediction accuracy of traffic flow.

Key words: traffic flow prediction, spatio-temporal union, graph attention network, soft attention mechanism, Bidirectional Gated Recurrent Unit (Bi-GRU)

史昕, 曹凤腾, 纪艺, 马峻岩. 基于多尺度时空特征与软注意力机制的交通流预测方法[J]. 计算机工程, 2024, 50(12): 346-357.

SHI Xin, CAO Fengteng, JI Yi, MA Junyan. Traffic Flow Prediction Method Based on Multi-Scale Spatio-Temporal Features and Soft Attention Mechanism[J]. Computer Engineering, 2024, 50(12): 346-357.

https://www.ecice06.com/CN/Y2024/V50/I12/346

图/表 15

图1 交通流预测任务示意图

Fig.1 Schematic diagram of traffic flow prediction task

图2 1号传感器周一至周三交通流变化曲线

Fig.2 Traffic flow change curves of sensor No. 1 from monday to wednesday

图3 相邻3个传感器的交通流情况

Fig.3 Traffic flow of adjacent three sensors

图4 MSTFSA模型整体架构

Fig.4 Overall architecture of MSTFSA model

图5 多时间尺度划分示例

Fig.5 Partitioning example of multiple time scale

图6 时空特征提取模块内部结构

Fig.6 Internal structure of spatio\|temporal feature extraction module

图7 软注意力机制网络结构

Fig.7 Structure of soft attention mechanism network

图8 PeMS传感器分布情况与PeMS04邻接矩阵

Fig.8 PeMS sensor distribution and PeMS04 adjacency matrix

图9 超参数变化对性能的影响

Fig.9 Effect of hyperparameter variation on performance

图10 不同模型在PeMS04上各预测时长的性能比较

Fig.10 Performance comparison of different models for different prediction durations on PeMS04

图11 不同模型在PeMS08上不同预测时长的性能比较

Fig.11 Performance comparison of different models for different prediction durations on PeMS08

图12 MSTFSA模型在PeMS04数据集上预测值和实际值的比较

Fig.12 Comparison of predicted and actual values of MSTFSA model on PeMS04 dataset

图13 MSTFSA模型在PeMS08数据集上预测值和实际值的比较

Fig.13 Comparison of predicted and actual values of MSTFSA model on PeMS08 dataset

参考文献 28

1	姚俊峰, 何瑞, 史童童, 等. 基于机器学习的交通流预测方法综述. 交通运输工程学报, 2023, 23 (3): 44- 67. doi: 10.19818/j.cnki.1671-1637.2023.03.003
	YAO J F , HE R , SHI T T , et al. Review on machine learning-based traffic flow prediction methods. Journal of Traffic and Transportation Engineering, 2023, 23 (3): 44- 67. doi: 10.19818/j.cnki.1671-1637.2023.03.003
2	谭娟, 王胜春. 基于深度学习的交通拥堵预测模型研究. 计算机应用研究, 2015, 32 (10): 2951- 2954. doi: 10.3969/j.issn.1001-3695.2015.10.016
	TAN J , WANG S C . Research on prediction model for traffic congestion based on deep learning. Application Research of Computers, 2015, 32 (10): 2951- 2954. doi: 10.3969/j.issn.1001-3695.2015.10.016
3	ALI A , ZHU Y M , ZAKARYA M . Exploiting dynamic spatio-temporal graph convolutional neural networks for citywide traffic flows prediction. Neural Networks, 2022, 145, 233- 247. doi: 10.1016/j.neunet.2021.10.021
4	LI D , LASENBY J . Spatiotemporal attention-based graph convolution network for segment-level traffic prediction. IEEE Transactions on Intelligent Transportation Systems, 2022, 23 (7): 8337- 8345. doi: 10.1109/TITS.2021.3078187
5	GUO S N, LIN Y F, FENG N, et al. Attention based spatial-temporal graph convolutional networks for traffic flow forecasting[C]//Proceedings of the AAAI Conference on Artificial Intelligence. [S. l. ]: AAAI Press, 2019: 922-929. URL
6	SONG C, LIN Y F, GUO S N, et al. Spatial-temporal synchronous graph convolutional networks: a new framework for spatial-temporal network data forecasting[C]//Proceedings of the AAAI Conference on Artificial Intelligence. [S. l. ]: AAAI Press, 2020: 914-921. URL
7	SMITH B L , WILLIAMS B M , KEITH OSWALD R . Comparison of parametric and nonparametric models for traffic flow forecasting. Transportation Research Part C: Emerging Technologies, 2002, 10 (4): 303- 321. doi: 10.1016/S0968-090X(02)00009-8
8	杨立宁, 李艳婷. 基于SVD和ARIMA的时空序列分解与预测. 计算机工程, 2021, 47 (3): 53- 61. doi: 10.19678/j.issn.1000-3428.0056885
	YANG L N , LI Y T . Spatio-temporal sequence decomposition and prediction based on SVD and ARIMA. Computer Engineering, 2021, 47 (3): 53- 61. doi: 10.19678/j.issn.1000-3428.0056885
9	CHANDRA S R , AL-DEEK H . Predictions of freeway traffic speeds and volumes using vector autoregressive models. Journal of Intelligent Transportation Systems, 2009, 13 (2): 53- 72. doi: 10.1080/15472450902858368
10	VLAHOGIANNI E I , KARLAFTIS M G , GOLIAS J C . Statistical methods for detecting nonlinearity and non-stationarity in univariate short-term time-series of traffic volume. Transportation Research Part C: Emerging Technologies, 2006, 14 (5): 351- 367. doi: 10.1016/j.trc.2006.09.002
11	VAN LINT J W C , HOOGENDOORN S P , VAN ZUYLEN H J . Freeway travel time prediction with state-space neural networks: modeling state-space dynamics with recurrent neural networks. Transportation Research Record, 2002, 1811 (1): 30- 39. doi: 10.3141/1811-04
12	ZHAO Z , CHEN W H , WU X M , et al. LSTM network: a deep learning approach for short-term traffic forecast. IET Intelligent Transport Systems, 2017, 11 (2): 68- 75. doi: 10.1049/iet-its.2016.0208
13	WU Q , JIANG Z , HONG K W , et al. Tensor-based recurrent neural network and multi-modal prediction with its applications in traffic network management. IEEE Transactions on Network and Service Management, 2021, 18 (1): 780- 792. doi: 10.1109/TNSM.2021.3056912
14	MA X L , TAO Z M , WANG Y H , et al. Long short-term memory neural network for traffic speed prediction using remote microwave sensor data. Transportation Research Part C: Emerging Technologies, 2015, 54, 187- 197. doi: 10.1016/j.trc.2015.03.014
15	SCHUSTER M , PALIWAL K K . Bidirectional recurrent neural networks. IEEE Transactions on Signal Processing, 1997, 45 (11): 2673- 2681. doi: 10.1109/78.650093
16	ASIF N A , SARKER Y , CHAKRABORTTY R K , et al. Graph neural network: a comprehensive review on non-euclidean space. IEEE Access, 2021, 9, 60588- 60606. doi: 10.1109/ACCESS.2021.3071274
17	YAN H Y , MA X L , PU Z Y . Learning dynamic and hierarchical traffic spatiotemporal features with transformer. IEEE Transactions on Intelligent Transportation Systems, 2022, 23 (11): 22386- 22399. doi: 10.1109/TITS.2021.3102983
18	马愈昭, 张岩峰, 冯帅. 基于神经网络的侧向激光雷达信号去噪算法. 光电工程, 2023, 50 (6): 220341. URL
	MA Y Z , ZHANG Y F , FENG S . A denoising algorithm based on neural network for side-scatter lidar signal. Opto-Electronic Engineering, 2023, 50 (6): 220341. URL
19	CUI Z Y , HENRICKSON K , KE R M , et al. Traffic graph convolutional recurrent neural network: a deep learning framework for network-scale traffic learning and forecasting. IEEE Transactions on Intelligent Transportation Systems, 2020, 21 (11): 4883- 4894. doi: 10.1109/TITS.2019.2950416
20	GUO S N , LIN Y F , LI S J , et al. Deep spatial-temporal 3D convolutional neural networks for traffic data forecasting. IEEE Transactions on Intelligent Transportation Systems, 2019, 20 (10): 3913- 3926. doi: 10.1109/TITS.2019.2906365
21	LI P , CHEN Z K , YANG L T , et al. An improved stacked auto-encoder for network traffic flow classification. IEEE Network, 2018, 32 (6): 22- 27. doi: 10.1109/MNET.2018.1800078
22	TIAN Y, WEI C C, XU D W. Traffic flow prediction based on stack AutoEncoder and long short-term memory network[C]//Proceedings of the IEEE 3rd International Conference on Automation, Electronics and Electrical Engineering (AUTEEE). Washington D. C., USA: IEEE Press, 2020: 385-388. URL
23	WANG X Y, MA Y, WANG Y Q, et al. Traffic flow prediction via spatial temporal graph neural network[C]//Proceedings of the Web Conference 2020. New York, USA: ACM Press, 2020: 1082-1092. URL
24	LI M Z, ZHU Z X. Spatial-temporal fusion graph neural networks for traffic flow forecasting[C]//Proceedings of the AAAI Conference on Artificial Intelligence. [S. l. ]: AAAI Press, 2021: 4189-4196. URL
25	ZHENG C P, FAN X L, WANG C, et al. GMAN: a graph multi-attention network for traffic prediction[C]//Proceedings of the AAAI Conference on Artificial Intelligence. [S. l. ]: AAAI Press, 2020: 1234-1241. URL
26	LV Y S , DUAN Y J , KANG W W , et al. Traffic flow prediction with big data: a deep learning approach. IEEE Transactions on Intelligent Transportation Systems, 2015, 16 (2): 865- 873. doi: 10.1109/TITS.2014.2345663
27	YU B, YIN H T, ZHU Z X. Spatio-temporal graph convolutional networks: a deep learning framework for traffic forecasting[C]// Proceedings of the 27th International Joint Conference on Artificial Intelligence. New York, USA: ACM Press, 2018: 3634-3640.
28	ZHAO Y P, XU Y P, HE X T, et al. Spatiotemporal graph attention networks for urban traffic flow prediction[C]//Proceedings of the IEEE 33rd Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC). Washington D. C., USA: IEEE Press, 2022: 3634-3640. URL

[1]	杜海军, 余粟. 基于时空图注意力网络的服务机器人动态避障[J]. 计算机工程, 2024, 50(2): 105-112.
[2]	詹泽慧, 钟煊妍, 邹萱萱, 骆丽霞. 基于BERT-HAN增强人机对话的计算思维评估模型[J]. 计算机工程, 2024, 50(12): 110-123.
[3]	郭尚伟, 刘树峰, 李子铭, 欧阳德强, 王宁, 向涛. 基于融合模型的网络安全态势感知方法[J]. 计算机工程, 2024, 50(11): 1-9.
[4]	唐卓然, 柳毅. 基于词汇融合和依存关系的中文命名实体识别[J]. 计算机工程, 2024, 50(10): 145-153.
[5]	周莎, 申国伟, 郭春. 基于安全知识图谱与逆向特征的弱点信息补全[J]. 计算机工程, 2024, 50(1): 145-155.
[6]	王志宝, 江树涛, 李菲, 高俊涛, 马强, 杨彬. 基于多邻域感知的石油数据资产图谱实体对齐[J]. 计算机工程, 2024, 50(1): 339-347.
[7]	张家熔, 苑津莎, 许珈宁, 罗志宏. 基于多元信息嵌入与协同神经网络的力学实体识别算法[J]. 计算机工程, 2023, 49(7): 125-134.
[8]	戎珂瑶, 熊贇. 基于多维度异质图结构的代码注释自动生成[J]. 计算机工程, 2023, 49(4): 240-248.
[9]	刘金硕, 刘宁. 面向招标文件的半结构化文本自动生成[J]. 计算机工程, 2023, 49(3): 67-72.
[10]	蔡瑞初, 张盛强, 许柏炎. 基于结构感知混合编码模型的代码注释生成方法[J]. 计算机工程, 2023, 49(2): 61-69.
[11]	张晓晖, 马慧芳, 王文涛, 高子皓. 基于跨会话知识图谱的图注意力网络推荐方法[J]. 计算机工程, 2023, 49(2): 136-142,149.
[12]	葛宇然, 付强. 基于时空联合学习的城市交通流短时预测模型[J]. 计算机工程, 2023, 49(1): 270-278.
[13]	郭可翔, 王衡军, 白祉旭. 融合多通道CNN与BiGRU的字词级文本错误检测模型[J]. 计算机工程, 2022, 48(9): 63-70.
[14]	梁登玉, 刘大明. 融合多粒度信息与外部知识的短文本匹配模型[J]. 计算机工程, 2022, 48(8): 129-135,143.
[15]	田乔鑫, 孔韦韦, 滕金保, 王照乾. 基于并行混合网络与注意力机制的文本情感分析模型[J]. 计算机工程, 2022, 48(8): 266-273.

选择文件类型/文献管理软件名称

选择包含的内容