Short-term passenger flow prediction framework for comprehensive transportation hub based on behavioral representation spatial correlation network
-
摘要: 为提升枢纽内部多类型功能区域客流短时预测的准确性，针对现有综合交通枢纽客流预测方法难以精准捕捉非邻接区域复杂的时空交互关系从而影响预测精度的问题，研究了综合运用虚拟现实技术（VR），改进了多尺度时序特征编码模块，构建了一种基于旅客活动链行为表征的空间关联网络客流短时预测框架（BRSCN）。利用VR技术搭建综合交通枢纽场景，开展旅客行为试验，获取了旅客在虚拟环境中的出行轨迹数据；对轨迹数据进行区域识别、停留判定与链路重构，构建了完整的旅客空间活动链集合，进而通过滑动窗口采样和图嵌入算法生成了各功能区域的特征向量；综合余弦相似度与客流转移频率构建反映了区域间非邻接关联的空间关联图；在预测阶段，使用图注意力网络（GAT）实现了邻接与非邻接区域空间特征的聚合，构建了双向扩展长短期记忆网络（Bi-sLSTM）和动态窗口稀疏注意力Transformer相结合的多尺度时序特征编码模块，自适应捕获枢纽客流复杂的非线性多尺度时空波动特征。试验结果表明：选取的上海虹桥综合交通枢纽高架层真实客流数据，与AGCRN、STTN等既有先进模型相比，BRSCN在均方根误差（RMSE）、平均绝对误差和平均绝对百分比误差3个指标上可分别降低30.2%、21.1%和28.3%；空间关联图的引入使RMSE指标降低16.7%，可显著提升模型对非邻接区域间客流交互的预测能力；动态窗口稀疏注意力机制在不牺牲预测精度的情况下使模型复杂度降低2.9%。所提BRSCN预测框架可有效捕捉综合交通枢纽内非邻接区域间的复杂时空关系，显著提高了客流短时预测的精度与模型泛化性能，可为综合枢纽空间资源优化配置和客流动态管理提供科学决策依据。Abstract: To improve the accuracy of short-term passenger flow prediction for heterogeneous functional areas within the transportation hub, the prediction accuracy issue caused by the limitation of existing prediction methods in accurately capturing complex spatiotemporal interactions among non-adjacent areas was investigated. The comprehensive application of virtual reality (VR) technology was studied and the multi-scale temporal feature encoding module was improved to establish a behavioral representation spatial correlation network (BRSCN) framework for short-term passenger flow prediction. A comprehensive transportation hub scenario was constructed using VR technology. The passenger behavior experiments were conducted to collect trajectory data of passengers in the virtual environment. The trajectory data were processed through regional identification, dwell-time detection, and activity-chain reconstruction to generate a complete set of passenger spatial activity chains. Subsequently, sliding-window sampling and graph embedding techniques were applied to derive feature vectors for each functional area. A spatial correlation graph reflecting non-adjacent inter-area relationships was constructed by jointly considering cosine similarity and passenger flow transition frequencies. During the prediction stage, a graph attention network (GAT) was employed to aggregate spatial features from both adjacent and non-adjacent areas. A multi-scale temporal feature encoding module was then designed by combining a bidirectional extended long short-term memory network (Bi-sLSTM) with a dynamic-window sparse-attention Transformer, enabling adaptive capturing of complex nonlinear and multi-scale spatiotemporal passenger flow fluctuation features within the hub. According to experimental results, compared with state-of-the-art models such as AGCRN and STTN, the real passenger flow data selected from the elevated level of the Shanghai Hongqiao comprehensive transportation hub reduces root mean square error (RMSE), mean absolute error, and mean absolute percentage error by 30.2%, 21.1%, and 28.3%, respectively. The incorporation of the spatial correlation graph further decreases RMSE by 16.7%, significantly enhancing the model's ability to predict passenger flow interactions between non-adjacent areas. Moreover, without lowering prediction accuracy, the dynamic-window sparse attention mechanism reduces model complexity by 2.9%. The proposed BRSCN framework effectively captures complex spatiotemporal relationships among non-adjacent functional areas within comprehensive transportation hubs, substantially improving short-term passenger flow prediction accuracy and model generalization performance. A scientific decision-making basis can be provided for spatial resource allocation optimization and dynamic passenger flow management in comprehensive transportation hubs.
-
图 1 基于实际综合交通枢纽场景数据的VR平台试验框架
Figure 1. VR platform experimental framework based on real data in comprehensive transportation hub scenario
图 2 基于受试者运动轨迹生成旅客站内出行游走链路
Figure 2. Passenger mobility chain within the station based on the subjects' movement trajectories
图 3 枢纽客流分布预测网络BRSCN整体框架
Figure 3. Overall BRSCN framework of hub passenger flow distribution prediction network
图 4 上海虹桥枢纽高架层布局以及区域划分方案
Figure 4. Layout and area division scheme of elevated level at Shanghai Hongqiao hub
图 5 基于人流量识别接口进行枢纽内各功能区域客流量统计
Figure 5. Passenger flow statistics in functional areas within hub based on passenger flow recognition interface
图 6 预测均方根误差在不同空间关联权重下的演化结果
Figure 6. Evolution results of prediction RMSE for different values of spatial correlation weights
图 7 枢纽功能区域空间关联强度矩阵热力图
Figure 7. Heatmap of spatial correlation intensity matrix of hub functional areas
图 8 枢纽内不同功能区域客流量实际值与预测值对比
Figure 8. Comparison of actual and predicted passenger flow data in different functional areas within the hub
图 9 BRSCN框架及其变体模型迭代损失变化曲线
Figure 9. Iteration loss variation curves of the BRSCN framework and its variant models
图 11 基于K折交叉验证和RMSE指标的模型预测效果验证
Figure 11. Validation of model prediction effect based on K-fold cross-validation and RMSE metrics
表 1 BRSCN与其他基准模型的预测误差指标平均值对比
Table 1. Comparison of average values of prediction error indexes between BRSCN and other baseline models
模型 误差指标 RMSE MAE MAPE/% ARIMA 28.50 20.86 23.73 BPNN 20.05 18.23 17.39 LSTM 16.44 14.59 16.83 Transformer 8.65 7.24 9.31 ConvLSTM 11.34 11.10 12.45 AGCRN 6.52 4.64 7.75 STTN 5.97 4.04 5.93 ST-MAN 4.72 3.13 5.35 BRSCN 3.29 2.66 3.69 
下载: 导出CSV
表 2 BRSCN及其变体模型的预测误差指标平均值对比
Table 2. Comparison of average prediction error metrics between BRSCN and its variants models
模型 误差指标 Wilcoxon检验 RMSE MAE MAPE/% BRSCN-SA 4.21 3.04 4.10 p < 0.05 BRSCN-G 3.95 2.93 4.36 p < 0.05 BRSCN/GAT 5.26 3.56 4.68 p < 0.05 BRSCN-LSTM 4.43 3.37 5.15 p < 0.05 BRSCN 3.29 2.66 3.69
下载: 导出CSV
表 3 BRSCN框架及其变体模型训练效率指标汇总
Table 3. Summary of training efficiency metrics for the BRSCN framework and its variant models
模型 迭代次数（损失值小于0.001） 参数量/106 每秒浮点运算次数/109 BRSCN_no GAT 89 1.71 0.13 BRSCN-SA 56 2.42 26.62 BRSCN-LSTM 122 1.68 26.61 BRSCN 29 2.35 27.76
下载: 导出CSV
-
[1] WEN K Y, ZHAO G T, HE B S, et al. A decomposition-based forecasting method with transfer learning for railway short-term passenger flow in holidays[J]. Expert Systems with Applications, 2022, 189: 116102. doi: 10.1016/j.eswa.2021.116102 [2] 马鑫俊, 姜嘉伟, 张勤. 基于时间序列算法的城市轨道交通客流量预测[J]. 交通技术, 2022(5): 413-422.MA Xin-jun, JIANG Jia-wei, ZHANG Qin. Passenger flow prediction of urban rail transit based on time series algorithm[J]. Open Journal of Transportation Technologies, 2022(5): 413-422. [3] WANG X M, ZHANG N, ZHANG Y L, et al. Forecasting of short-term metro ridership with support vector machine online model[J]. Journal of Advanced Transportation, 2018, 2018: 3189238. [4] TAN Y F, LIU H X, PU Y, et al. Passenger flow prediction of integrated passenger terminal based on K-means-GRNN[J]. Journal of Advanced Transportation, 2021, 2021(1): 1055910. [5] SUN S L, ZHANG C S, YU G Q. A Bayesian network approach to traffic flow forecasting[J]. IEEE Transactions on Intelligent Transportation Systems, 2006, 7(1): 124-132. doi: 10.1109/TITS.2006.869623 [6] 朱才华, 孙晓黎, 李培坤, 等. 融合车站分类和数据降噪的城市轨道交通短时客流预测[J]. 铁道科学与工程学报, 2022, 19(8): 2182-2192.ZHU Cai-hua, SUN Xiao-li, LI Pei-kun, et al. Short-term urban rail transit passenger flow prediction based on incorporating station classification and data noise reduction[J]. Journal of Railway Science and Engineering, 2022, 19(8): 2182-2192. [7] JEONG Y S, BYON Y J, CASTRO-NETO M M, et al. Supervised weighting-online learning algorithm for short-term traffic flow prediction[J]. IEEE Transactions on Intelligent Transportation Systems, 2013, 14(4): 1700-1707. doi: 10.1109/TITS.2013.2267735 [8] LIU L B, CHEN J W, WU H F, et al. Physical-virtual collaboration modeling for intra-and inter-station metro ridership prediction[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(4): 3377-3391. doi: 10.1109/TITS.2020.3036057 [9] DAI Z C, LI D W, FENG S Q. Attention mechanism with spatial-temporal joint deep learning model for the forecasting of short-term passenger flow distribution at the railway station[J]. Journal of Advanced Transportation, 2024, 2024(1): 7985408. doi: 10.1155/2024/7985408 [10] CAO B X, LI Y X, CHEN Y Y, et al. A CNN-LSTM model for short-term passenger flow forecast considering the built environment in urban rail transit stations[J]. Journal of Transportation Engineering, Part A: Systems, 2024, 150(11): 04024072. doi: 10.1061/JTEPBS.TEENG-8579 [11] MA X L, ZHANG J Y, DU B W, et al. Parallel architecture of convolutional bi-directional LSTM neural networks for network-wide metro ridership prediction[J]. IEEE Transactions on Intelligent Transportation Systems, 2019, 20(6): 2278-2288. doi: 10.1109/TITS.2018.2867042 [12] XU W J, MIAO L X, XING J J. Short-term passenger flow forecasting of the airport based on deep learning spatial-temporal network[C]//ACM. 2022 The 9th International Conference on Industrial Engineering and Applications. New York: ACM, 2022: 77-83. [13] LI P, WANG S, ZHAO H T, et al. IG-net: An interaction graph network model for metro passenger flow forecasting[J]. IEEE Transactions on Intelligent Transportation Systems, 2023, 24(4): 4147-4157. doi: 10.1109/TITS.2023.3235805 [14] 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[J]. IEEE Transactions on Intelligent Transportation Systems, 2020, 21(11): 4883-4894. doi: 10.1109/TITS.2019.2950416 [15] WU J X, HE D Q, JIN Z Z, et al. Learning spatial-temporal pairwise and high-order relationships for short-term passenger flow prediction in urban rail transit[J]. Expert Systems with Applications, 2024, 245: 123091. doi: 10.1016/j.eswa.2023.123091 [16] YANG Y J, ZHANG J L, YANG L X, et al. Network-wide short-term inflow prediction of the multi-traffic modes system: An adaptive multi-graph convolution and attention mechanism based multitask-learning model[J]. Transportation Research Part C: Emerging Technologies, 2024, 158: 104428. doi: 10.1016/j.trc.2023.104428 [17] ZHAO L, SONG Y J, ZHANG C, et al. T-GCN: A temporal graph convolutional network for traffic prediction[J]. IEEE Transactions on Intelligent Transportation Systems, 2020, 21(9): 3848-3858. doi: 10.1109/TITS.2019.2935152 [18] CHANG Y C, ZONG M Y, DANG Y T, et al. Multi-step passenger flow prediction for urban metro system based on spatial-temporal graph neural network[J]. Applied Sciences, 2024, 14(18): 8121. doi: 10.3390/app14188121 [19] 周浪雅, 王亦乐, 谢余晨, 等. 站城融合背景下高速铁路综合枢纽短时客流预测研究[J]. 铁道学报, 2023, 45(4): 1-7.ZHOU Lang-ya, WANG Yi-le, XIE Yu-chen, et al. Prediction of short-term passenger flow of high-speed railway integrated passenger hub under station-city integration[J]. Journal of the China Railway Society, 2023, 45(4): 1-7. [20] ZHANG J L, CHEN F, CUI Z Y, et al. Deep learning architecture for short-term passenger flow forecasting in urban rail transit[J]. IEEE Transactions on Intelligent Transportation Systems, 2021, 22(11): 7004-7014. doi: 10.1109/TITS.2020.3000761 [21] VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[J]. arXiv, 2017, DOI: 10.48550/arXiv.1706.03762. [22] 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[J]. Transportation Research Part C: Emerging Technologies, 2015, 54: 187-197. doi: 10.1016/j.trc.2015.03.014 [23] YE X, FANG S, SUN F, et al. Meta graph transformer: A novel framework for spatial-temporal traffic prediction[J]. Neurocomputing, 2022, 491: 544-563. doi: 10.1016/j.neucom.2021.12.033 [24] YAN H Y, MA X L, PU Z Y. Learning dynamic and hierarchical traffic spatiotemporal features with transformer[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(11): 22386-22399. doi: 10.1109/TITS.2021.3102983 [25] TANG M, LI Z W, TIAN G D. A data-driven-based wavelet support vector approach for passenger flow forecasting of the metropolitan hub[J]. IEEE Access, 2019, 7: 7176-7183. doi: 10.1109/ACCESS.2019.2890819 [26] YUE M, MA S H. LSTM-based transformer for transfer passenger flow forecasting between transportation integrated hubs in urban agglomeration[J]. Applied Sciences, 2023, 13(1): 637. doi: 10.3390/app13010637 [27] LIN L, LIU X C, LIU X H, et al. A prediction model to forecast passenger flow based on flight arrangement in airport terminals[J]. Energy and Built Environment, 2023, 4(6): 680-688. doi: 10.1016/j.enbenv.2022.06.006 [28] CHENG L, CAI X M, LIU Z, et al. Characterising travel behaviour patterns of transport hub station area users using mobile phone data[J]. Journal of Transport Geography, 2024, 116: 103855. doi: 10.1016/j.jtrangeo.2024.103855 [29] VAN BEEK A, DUIVES D C, FENG Y, et al. Comparison of pedestrian wayfinding behavior between a real and a virtual multi-story building—A validation study[J]. Transportation Research Part C: Emerging Technologies, 2024, 163: 104650. doi: 10.1016/j.trc.2024.104650 [30] DAI Z C, LI D W, FENG Y, et al. A study of pedestrian wayfinding behavior based on desktop VR considering both spatial knowledge and visual information[J]. Transportation Research Part C: Emerging Technologies, 2024, 163: 104651. doi: 10.1016/j.trc.2024.104651 [31] PARK Y, CHOI Y, KIM K, et al. Machine learning approach for study on subway passenger flow[J]. Scientific Reports, 2022, 12: 2754. doi: 10.1038/s41598-022-06767-7 [32] JOHNSSON C, CAMPOREALE R. Exploring space syntax integration at public transport hubs and public squares using drone footage[J]. Applied Sciences, 2022, 12(13): 6515. doi: 10.3390/app12136515 [33] PEROZZI B, AL-RFOU R, SKIENA S. DeepWalk: Online learning of social representations[C]//ACM. Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York: ACM, 2014: 701-710. [34] 百度公司. 百度AI开放平台行人流动态检测接口[EB/OL]. (2024-9-30). https://ai.baidu.com/tech/pedestrian_flow .Baidu. Baidu AI open platform pedestrian flow dynamic detection interface[EB/OL]. (2024-9-30).https://ai.baidu.com/tech/pedestrian_flow .[35] ZHOU T, MA Z Q, WEN Q S, et al. FEDformer: Frequency enhanced decomposed transformer for long-term series forecasting[C]//ICML. 39th International Conference on Machine Learning. San Diego: ICML, 2022: 27268-27286. [36] ZHOU H Y, ZHANG S H, PENG J Q, et al. Informer: Beyond efficient transformer for long sequence time-series forecasting[C]//AAAI. Proceedings of the AAAI Conference on Artificial Intelligence. Washington DC: AAAI, 2021, 35(12): 11106-11115. [37] BAI L, YAO L N, LI C, et al. Adaptive graph convolutional recurrent network for traffic forecasting[J]. Advances in Neural Information Processing Systems, 2020, 33: 17804-17815. [38] XU M X, DAI W R, LIU C M, et al. Spatial-temporal transformer networks for traffic flow forecasting[J]. arXiv, 2001, DOI: 10.485501/arXiv.2001.02908. [39] LIU Y, GUO B, MENG J X, et al. Spatio-temporal memory augmented multi-level attention network for traffic prediction[J]. IEEE Transactions on Knowledge and Data Engineering, 2024, 36(6): 2643-2658. doi: 10.1109/TKDE.2023.3322405 -
计量
- 文章访问数: 11
- HTML全文浏览量: 6
- PDF下载量: 1
- 被引次数: 0
