-
摘要: 为实时动态评估交通态势，结合低秩张量补全理论，提出了一种基于自回归正则项与拉普拉斯正则项的交通速度预测模型；为提高模型在全局空间维度的表达能力，构建基于低秩张量补全框架的拉普拉斯卷积正则项表示路段间的关联关系；为提高模型在局部空间维度的表达能力，利用自回归模型的时间序列趋势捕获能力提高模型在时间维度的短时与长时表达能力，更精确地捕获交通数据的时空信息；为提高算法效率，通过时域与频域信号的转换降低了矩阵运算量，并采用截断核范数作为低秩张量逼近模型；使用交替方向乘子法实现高效的低秩拉普拉斯自回归张量补全(LLATC)预测方法；基于出租车行驶速度数据集和高速公路交通速度数据集，分析了LLATC算法在不同缺失率情况下的补全效果，对比了LLATC算法与其他基线预测算法的预测精度。研究结果表明：在交通数据随机缺失模式下，缺失率为20%~70%时，相对于传统的低秩张量补全模型，LLATC算法补全平均绝对误差降低了2%~6%，相比于传统的预测方法，LLATC算法预测平均绝对误差降低了4%~22%；在交通数据非随机缺失模式下，相对于传统的低秩张量补全模型，LLATC算法的平均绝对误差降低了2%~6%，相比于传统的预测方法，LLATC算法的预测平均绝对误差降低了13%~25%。可见，在2种交通数据缺失模式下，改进低秩张量补全方法降低了交通量数据的补全误差，能有效提高交通数据的预测精度，简化了数据处理流程。Abstract: To dynamically evaluate traffic condition in real time, a traffic speed prediction model based on autoregressive regularization terms and Laplacian regularization terms was proposed. To improve the model's expression capability in global dimensions, a Laplace convolutional regularization term based on a low-rank tensor completion framework was introduced to represent the correlations of road segments. To improve the model's expression capability in local spatial dimensions, the time series trend-capturing capabilities of autoregressive models were utilized, and the short- and long-term expression capabilities of the models in the time dimension were improved to capture the spatio-temporal information of traffic data more effectively. The implementation of the truncated kernel norm as the low-rank tensor approximation model and the conversion of time- and frequency-domain signals leaded to improve the computation efficiency. An efficient low-rank Laplacian autoregressive tensor completion (LLATC) prediction method was developed by using the alternating direction multiplier method. Based on taxi speed data set and expressway traffic speed data set, the completion performances of the LLATC algorithm under different missing rates were systematically analyzed, and the prediction accuracy of the LLATC algorithm was compared with other baseline prediction algorithms. Research results show that under the random missing pattern with a missing rate of 20% to 70%, the mean absolute error (MAE) of the LLATC algorithm reduces by 2% to 6% compared to the traditional low-rank tensor completion models, and the MAE reduces by 4% to 22% compared to the traditional prediction methods. Under the non-random missing pattern, the MAE of the LLATC algorithm reduces by 2% to 6% compared to the traditional low-rank tensor completion models, and the MAE reduces by 13% to 25% compared to the traditional prediction methods. The finding indicates that the LLATC algorithm effectively reduces the completion error of traffic volume data, significantly enhances the prediction accuracy of traffic volume data under two kinds of missing data patterns, and simplifies the data processing workflow.
-
图 5 非随机和随机缺失下各算法的均方根误差
Figure 5. Root-mean-square errors of each algorithm under non-random and random missing
图 6 基于SZ数据集，非随机缺失率为20%时，路段3、28、32、44的1 d预测值与原始值对比
Figure 6. Based on SZ data set, comparison of predicted values and original values for road sections 3, 28, 32 and 44 at non-random missing rate of 20% in 1 d
图 7 基于SZ数据集，随机缺失率为20%时，路段3、28、32、44的1 d预测值与原始值对比
Figure 7. Based on SZ data set, comparison of predicted values and original values for road sections 3, 28, 32 and 44 at random missing rate of 20% in 1 d
图 8 基于SZ数据集，非随机缺失率为20%时，路段3、28、32、44的1 h预测值与原始值对比
Figure 8. Based on SZ data set, comparison of predicted values and original values for road sections 3, 28, 32 and 44 at non-random missing rate of 20% in 1 h
图 9 基于SZ数据集，随机缺失率为20%时，路段3、28、32、44的1 h预测值与原始值对比
Figure 9. Based on SZ data set, comparison of predicted values and original values for road sections 3, 28, 32 and 44 at random missing rate of 20% in 1 h
图 10 基于LP数据集，非随机缺失率为20%时，传感器5、14、19、31的1 d预测值与原始值对比
Figure 10. Based on LP data set, comparison of predicted values and original values for sensors 5, 14, 19 and 31 at non-random missing rate of 20% in 1 d
图 11 基于LP数据集，随机缺失率为20%时，传感器5、14、19、31的1 d预测值与原始值对比
Figure 11. Based on LP data set, comparison of predicted values and original values for sensors 5, 14, 19 and 31 at random missing rate of 20% in 1 d
图 12 基于LP数据集，非随机缺失率为20%时，传感器5、14、19、31的1 h预测值与原始值对比
Figure 12. Based on LP data set, comparison of predicted values and original values for sensors 5, 14, 19 and 31 at non-random missing rate of 20% in 1 h
图 13 基于LP数据集，随机缺失率为20%时，传感器5、14、19、31的1 h预测值与原始值对比
Figure 13. Based on LP data set, comparison of predicted values and original values for sensors 5, 14, 19 and 31 at random missing rate of 20% in 1 h
表 1 随机缺失下各方法数据补全精度(E1/E2)
Table 1. Data completion accuracy of each method for random missing (E1/E2)
数据集 缺失率/% 不同算法的数据补全精度/(km·h-1) CP-ALS HaLRTC BGCP LRTC_TNN BTTF LLATC SZ 20 3.78/5.44 3.73/5.41 3.76/5.42 3.60/5.26 3.78/5.44 3.54/5.14 30 3.76/5.43 3.77/5.45 3.76/5.41 3.63/5.28 3.78/5.44 3.57/5.20 40 3.79/5.46 3.81/5.51 3.76/5.42 3.65/5.32 3.80/5.46 3.62/5.27 50 3.80/5.48 3.87/5.57 3.78/5.45 3.68/5.35 3.79/5.45 3.64/5.29 60 3.82/5.51 3.95/5.66 3.80/5.46 3.72/5.37 3.80/5.46 3.68/5.33 70 3.87/5.58 4.05/5.79 3.82/5.48 3.76/5.43 3.85/5.52 3.72/5.38 80 3.97/5.76 4.20/6.00 3.85/5.53 3.81/5.49 3.87/5.55 3.80/5.45 LP 20 3.55/5.73 3.32/8.13 3.97/6.53 2.66/4.31 3.55/5.73 2.65/4.26 30 3.57/5.75 3.47/8.48 3.99/6.55 2.74/4.46 3.56/5.75 2.74/4.43 40 3.58/5.76 3.61/8.86 4.00/6.57 2.83/4.61 3.57/5.77 2.84/4.58 50 3.59/5.79 3.77/9.29 4.00/6.56 2.94/4.81 3.58/5.79 2.96/4.76 60 3.61/5.83 3.96/9.83 4.01/6.59 3.09/5.05 3.60/5.83 3.09/5.01 70 3.65/5.89 4.18/10.46 4.04/6.62 3.29/5.39 3.63/5.88 3.28/5.35 80 3.73/6.00 4.48/11.33 4.06/6.66 3.63/5.90 3.70/5.97 3.60/5.91 
下载: 导出CSV
表 2 非随机缺失下各方法数据补全精度(E1/E2)
Table 2. Data completion accuracy of each method for non-random missing (E1/E2)
数据集 缺失率/% 不同算法的数据补全精度/(km·h-1) CP-ALS HaLRTC BGCP LRTC_TNN BTTF LLATC SZ 20 3.95/5.62 3.82/5.48 3.95/5.62 3.62/5.23 3.95/5.62 3.58/5.17 30 3.97/5.66 3.89/5.59 3.97/5.66 3.66/5.30 3.97/5.66 3.62/5.27 40 3.95/5.66 3.96/5.68 3.95/5.66 3.68/5.33 3.96/5.66 3.64/5.27 50 3.95/5.65 4.06/5.79 3.95/5.65 3.70/5.35 3.95/5.65 3.67/5.32 60 3.97/5.67 4.22/5.99 3.97/5.67 3.75/5.41 3.97/5.67 3.72/5.36 70 3.96/5.67 4.59/6.47 3.96/5.66 3.79/5.45 3.96/5.66 3.77/5.43 80 3.98/5.67 5.61/7.90 3.97/5.66 3.86/5.52 3.97/5.67 3.89/5.62 LP 20 3.51/5.73 3.42/5.04 3.91/6.52 2.82/4.70 3.50/5.74 2.79/4.57 30 3.51/5.75 3.81/5.41 3.93/6.51 2.94/4.89 3.51/5.73 2.89/4.75 40 3.48/5.70 4.33/5.81 3.90/6.46 3.00/4.99 3.48/5.69 2.99/4.96 50 18.22/115.26 5.84/7.21 3.95/6.51 3.21/5.33 3.52/5.74 3.17/5.27 60 108.12/292.34 11.16/13.86 4.00/6.61 4.26/9.19 3.59/6.42 4.18/7.30 70 229.93/609.04 21.60/26.37 4.34/6.94 8.93/19.69 6.42/11.30 5.60/8.97 80 155.33/521.23 37.14/40.70 17.48/30.72 16.13/29.04 22.04/36.81 6.98/10.71
下载: 导出CSV
表 3 随机缺失下各方法数据预测精度(E1/E2)
Table 3. Prediction accuracy of each method for random missing(E1/E2)
数据集 缺失率/% 不同算法的数据预测精度/(km·h-1) HA ARIMA LSTM GRU T-GCN LLATC SZ 0 1.69/2.42 2.18/2.98 1.60/2.07 1.60/2.06 2.34/3.01 1.38/1.75 20 1.63/2.35 2.18/2.97 1.57/2.01 1.65/2.08 2.23/2.88 1.38/1.77 30 1.59/2.31 2.17/2.96 1.62/2.11 1.55/1.99 1.75/2.32 1.38/1.79 40 1.57/2.25 2.17/2.96 1.53/2.05 1.45/1.91 1.83/2.37 1.41/1.82 50 1.51/2.13 2.17/2.96 1.57/2.07 1.56/2.01 2.19/2.75 1.43/1.84 60 1.51/2.09 2.17/2.96 1.59/2.07 1.47/1.93 1.67/2.21 1.43/1.85 70 1.64/2.27 2.17/2.96 1.61/2.03 1.51/1.93 1.50/1.98 1.43/1.85 80 10.76/10.93 2.17/2.95 1.46/1.92 1.51/1.96 1.87/2.41 1.46/1.87
下载: 导出CSV
表 4 非随机缺失下各方法数据预测精度(E1/E2)
Table 4. Prediction accuracy of each method for non-random missing(in E1/E2)
数据集 缺失率/% 不同算法的数据预测精度/(km·h-1) HA ARIMA LSTM GRU T-GCN LLATC SZ 0 1.69/2.42 2.18/2.98 1.60/2.07 1.60/2.06 2.34/3.01 1.35/1.75 20 1.56/2.18 2.17/2.97 1.59/2.06 1.57/2.02 2.10/2.69 1.33/1.73 30 1.52/2.13 2.18/2.97 1.59/2.02 1.54/1.97 2.20/2.84 1.34/1.73 40 1.52/2.13 2.17/2.97 1.59/2.03 1.53/1.99 2.32/2.95 1.33/1.75 50 1.60/2.23 2.17/2.96 1.58/2.03 1.54/2.01 2.18/2.84 1.35/1.81 60 1.59/2.22 2.17/2.96 1.60/2.05 1.58/2.04 2.25/2.86 1.38/1.83 70 1.56/2.15 2.17/2.95 1.54/1.96 1.52/1.95 2.52/3.13 1.39/1.81 80 1.53/2.15 2.17/2.95 1.51/1.94 1.51/1.93 1.82/2.44 1.93/2.44
下载: 导出CSV
-
[1] VELIČKOVIĆ P, CUCURULL G, CASANOVA A, et al. Graph attention networks[C]//ICLR. 6th International Conference on Learning Representations. Vancouver: ICLR, 2018: 149806. [2] LI Qin, TAN Hua-chun, WU Yuan-kai, et al. Traffic flow prediction with missing data imputed by tensor completion methods[J]. IEEE Access, 2020, 8: 63188-63201. doi: 10.1109/ACCESS.2020.2984588 [3] CUI Zhi-yong, KE Rui-min, PU Zi-yuan, et al. Stacked bidirectional and unidirectional LSTM recurrent neural network for forecasting network-wide traffic state with missing values[J]. Transportation Research Part C: Emerging Technologies, 2020, 118: 102674. doi: 10.1016/j.trc.2020.102674 [4] YANG Fu-ning, LIU Guo-liang, HUANG Li-ping, et al. Tensor decomposition for spatial-temporal traffic flow prediction with sparse data[J]. Sensors, 2020, 20(21): 6046. doi: 10.3390/s20216046 [5] ZHONG Wei-da, SUO Qiu-ling, JIA Xiao-wei, et al. Heterogeneous spatio-temporal graph convolution network for traffic forecasting with missing values[C]//IEEE. 2021 IEEE 41st International Conference on Distributed Computing Systems. New York: IEEE, 2021: 707-717. [6] TAN Hua-chun, WU Yuan-kai, SHEN Bin, et al. Short- term traffic prediction based on dynamic tensor completion[J]. IEEE Transactions on Intelligent Transportation Systems, 2016, 17(8): 2123-2133. doi: 10.1109/TITS.2015.2513411 [7] 杨军, 侯忠生. 一种基于灰色马尔科夫的大客流实时预测模型[J]. 北京交通大学学报, 2013, 37(2): 119-123, 128.YANG Jun, HOU Zhong-sheng. A grey Markov based on large passenger flow real-time prediction model[J]. Journal of Beijing Jiaotong University, 2013, 37(2): 119-123, 128. (in Chinese) [8] 赵阳阳, 夏亮, 江欣国. 基于经验模态分解与长短时记忆神经网络的短时地铁客流预测模型[J]. 交通运输工程学报, 2020, 20(4): 194-204. doi: 10.19818/j.cnki.1671-1637.2020.04.016ZHAO Yang-yang, XIA Liang, JIANG Xin-guo. Short-term metro passenger flow prediction based on EMD-LSTM[J]. Journal of Traffic and Transportation Engineering, 2020, 20(4): 194-204. (in Chinese) doi: 10.19818/j.cnki.1671-1637.2020.04.016 [9] TAN Hua-chun, WU Yuan-kai, FENG Guang-dong, et al. A new traffic prediction method based on dynamic tensor completion[J]. Procedia-Social and Behavioral Sciences, 2013, 96: 2431-2442. doi: 10.1016/j.sbspro.2013.08.272 [10] FENG Jian-shuai, WANG Wu-hong, ZHANG Yu-Jin, et al. A tensor-based method for missing traffic data completion[J]. Transportation research Part C: Emerging Technologies, 2013, 28: 15-27. doi: 10.1016/j.trc.2012.12.007 [11] WANG Xu-dong, FAGETTE A, SARTELET P, et al. A probabilistic tensor factorization approach to detect anomalies in spatiotemporal traffic activities[C]//IEEE. 2019 Intelligent Transportation Systems Conference. New York: IEEE, 2019: 1658-1663. [12] XU Ming, WU Jian-ping, WANG Hao-han, et al. Anomaly detection in road networks using sliding-window tensor factorization[J]. IEEE Transactions on Intelligent Transportation Systems, 2019, 20(12): 4704-4713. doi: 10.1109/TITS.2019.2941649 [13] RAN Bin, TAN Hua-chun, WU Yuan-kai, et al. Tensor based missing traffic data completion with spatial-temporal correlation[J]. Physica A: Statistical Mechanics and Its Applications, 2016, 446: 54-63. doi: 10.1016/j.physa.2015.09.105 [14] RAN Bin, SONG Li, ZHANG Jian, et al. Using tensor completion method to achieving better coverage of traffic state estimation from sparse floating car data[J]. PLoS One, 2016, 11(7): e0157420. doi: 10.1371/journal.pone.0157420 [15] LIU Ji, MUSIALSKI P, WONKA P, et al. Tensor completion for estimating missing values in visual data[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35(1): 208-220. doi: 10.1109/TPAMI.2012.39 [16] CHEN Xin-yun, YANG Jin-ming, SUN Li-jun. A nonconvex low-rank tensor completion model for spatiotemporal traffic data imputation[J]. Transportation Research Part C: Emerging Technologies, 2020, 117: 102673. doi: 10.1016/j.trc.2020.102673 [17] SONG Yun, LI Jie, CHEN Xi, et al. An efficient tensor completion method via truncated nuclear norm[J]. Journal of Visual Communication and Image Representation, 2020, 70: 102791. doi: 10.1016/j.jvcir.2020.102791 [18] BENGUA J A, PHIEN H N, TUAN H D, et al. Efficient tensor completion for color image and video recovery: low-rank tensor train[J]. IEEE Transactions on Image Processing, 2017, 26(5): 2466-2479. doi: 10.1109/TIP.2017.2672439 [19] ZHENG Yu- bang, HUANG Ting-zhu, JI Teng- yu, et al. Low-rank tensor completion via smooth matrix factorization[J]. Applied Mathematical Modelling, 2019, 70: 677-695. doi: 10.1016/j.apm.2019.02.001 [20] CHEN Xin-yu, LEI Meng-ying, SAUNIER N, et al. Low-rank autoregressive tensor completion for spatiotemporal traffic data imputation[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(8): 12301-12310. doi: 10.1109/TITS.2021.3113608 [21] 李令先. 基于CNN的轨道交通拥堵预测算法研究[D]. 成都: 成都理工大学, 2019.LI Ling-xian. Research on prediction algorithm of rail transit congestion based on CNN[D]. Chengdu: Chengdu University of Technology, 2019. (in Chinese) [22] 赵建东, 申瑾, 刘麟玮. 多源数据驱动CNN-GRU模型的公交客流量分类预测[J]. 交通运输工程学报, 2021, 21(5): 265-273. doi: 10.19818/j.cnki.1671-1637.2021.05.022 ZHAO Jian-dong, SHEN Jin, LIU Lin-wei. Bus passenger flow classification prediction driven by CNN-GRU model and multi-source data[J]. Journal of Traffic and Transportation Engineering, 2021, 21(5): 265-273. (in Chinese) doi: 10.19818/j.cnki.1671-1637.2021.05.022 [23] 户佐安, 邓锦程, 韩金丽, 等. 图神经网络在交通预测中的应用综述[J]. 交通运输工程学报, 2023, 23(5): 39-61. doi: 10.19818/j.cnki.1671-1637.2023.05.003 HU Zuo-an, DENG Jin-cheng, HAN Jin-li, et al. Review on application of graph neural network in traffic prediction[J]. Journal of Traffic and Transportation Engineering, 2023, 23(5): 39-61. (in Chinese) doi: 10.19818/j.cnki.1671-1637.2023.05.003 [24] CUI Zhi-yong, 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 [25] GUO Kan, HU Yong-li, QIAN Zhen, et al. Dynamic graph convolution network for traffic forecasting based on latent network of Laplace matrix estimation[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(2): 1009-1018. doi: 10.1109/TITS.2020.3019497 [26] GU Ya-feng, DENG Li. STAGCN: spatial-temporal attention graph convolution network for traffic forecasting[J]. Mathematics, 2022, 10(9): 1599. doi: 10.3390/math10091599 [27] CHANG Meng-meng, DING Zhi-ming, CAI Zhi, et al. Prediction of evolution behaviors of transportation hubs based on spatiotemporal neural network[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(7): 9171-9183. doi: 10.1109/TITS.2021.3091708 [28] WANG Yi, JING Chang-feng, XU Shi-shuo, et al. Attention based spatiotemporal graph attention networks for traffic flow forecasting[J]. Information Sciences, 2022, 607: 869-883. doi: 10.1016/j.ins.2022.05.127 [29] ZHAO Jian-li, LIU Zhong-bo, SUN Qiu-xia, et al. Attention- based dynamic spatial-temporal graph convolutional networks for traffic speed forecasting[J]. Expert Systems with Applications, 2022, 204: 117511. doi: 10.1016/j.eswa.2022.117511 [30] LIAO Lyu-chao, HU Zhi-yuan, ZHENG Yu-xin, et al. An improved dynamic Chebyshev graph convolution network for traffic flow prediction with spatial-temporal attention[J]. Applied Intelligence, 2022, 52(14): 16104-16116. doi: 10.1007/s10489-021-03022-w [31] RUAN Chang, TAB Xian-chao, LIAO Zhuo-fan, et al. STGAT: spatial-temporal graph attention networks for traffic flow prediction[C]//IEEE. 2023 IEEE 29th International Conference on Parallel and Distributed Systems. New York: IEEE, 2023: 913-919. [32] KILMER M E, BRAMAN K, HAO Ning, et al. Third- order tensors as operators on matrices: a theoretical and computational framework with applications in imaging[J]. SIAM Journal on Matrix Analysis and Applications, 2013, 34(1): 148-172. doi: 10.1137/110837711 [33] TAKAYAMA H, YOKOTA T. Fast signal completion algorithm with cyclic convolutional smoothing[C]//IEEE. 2022 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference. New York: IEEE, 2022: 364-371. [34] LI Yuan-yuan, YU Ke, WU Xiao-fei. Efficient tensor completion for Internet traffic data recovery[C]//ACM. Proceedings of the 2nd International Conference on Telecommunications and Communication Engineering. New York: ACM, 2018: 251-257. [35] CHEN Xin-yu, CHENG Zhan-hong, SAUNIER N, et al. Laplacian convolutional representation for traffic time series imputation[J]. arXiv, 2212: 01529. [36] BOYDS, PARIKH N, CHU E, et al. Distributed optimization and statistical learning via the alternating direction method of multipliers[J]. Foundations and Trends© in Machine Learning, 2010, 3(1): 1-122. doi: 10.1561/2200000016 [37] ZHAO Ling, SONG Yu-jiao, ZHANG Chao, 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 -
点击查看大图
计量
- 文章访问数: 73
- HTML全文浏览量: 15
- PDF下载量: 12
- 被引次数: 0
