An Efficient Short-Term Traffic Speed Prediction Model Based on Improved TCN and GCN
Abstract
:1. Introduction
- According to the temporal and spatial characteristics of traffic speed data, we replaced the one-dimensional dilated convolution in TCN with the two-dimensional dilated convolution to form a two-dimensional Temporal Convolutional Network (2DTCN), which can extract the local spatial features whilst also extracting the temporal features.
- In GCN, we added the distance information between nodes to the adjacency matrix to form a weighted adjacency matrix, to extract the global spatial correlation between nodes and improve the prediction accuracy of the model.
- We developed the SPTMN model by combining 2DTCN and GCN. The model uses 2DTCN to achieve feature enhancement and extract temporal and local spatial features, and uses GCN to obtain the topology between road nodes, to model the spatial dependence. Finally, the geographical structure features of the road are integrated into the model.
- We evaluated the proposed model on real traffic speed datasets. The experimental results show that compared with the existing baseline models, the SPTMN model obtains the best prediction performance, and the SPTMN model is more stable and efficient.
2. Related Research
3. Methodology
3.1. Problem Definition
3.2. SPTMN Model Framework
3.2.1. Feature Enhancement Module
3.2.2. Graph Convolution Neural Network Module
3.2.3. Parameter Module
4. Experiments and Results
4.1. Datasets
4.2. Experimental Parameters and Evaluation Metrics
4.3. Baseline Method
- LSTM [39]: LSTM is a special RNN, which can solve the problems of gradient disappearance and gradient explosion in the process of long sequence training through a gating mechanism.
- GRU [40]: Gate Recurrent Unit, which filters and processes data through updated and reset gates but has fewer parameters than LSTM.
- GCN [36]: Graph convolution neural network, which can extract the feature of graph structure through a Laplace matrix and spectral transformation.
- TGCN [27]: Temporary graph revolutionary, which combines GRU and GCN to realize traffic prediction.
- STDN [26]: Spatial-Temporal Dynamic Network, which consists of CNN and LSTM.
- STGCN [28]: Spatial-Temporal Graph Convolutional Network, which is composed of graph revolution and gated temporary revolution, and realizes the prediction of traffic flow.
- ASTGCN [30]: Attention Based Spatial-Temporal Graph Convolutional Network, which consists of GCN, standard convolution, and spatial-temporal attention mechanism.
4.4. Experimental Results and Analysis
4.5. Case Study
4.6. Ablation Study
5. Conclusions
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
- Zambrano-Martinez, J.L.; Calafate, C.T.; Soler, D.; Cano, J.C.; Manzoni, P. Modeling and characterization of traffic flows in urban environments. Sensors 2018, 18, 2020. [Google Scholar] [CrossRef] [PubMed] [Green Version]
- Kiersztyn, A.; Gandzel, A.; Koczan, L.; Celiński, M. Travel time forecasting based on fuzzy patterns. Adv. Sc. Technol. Res. J. 2021, 15, 224–232. [Google Scholar] [CrossRef]
- Irawan, K.; Yusuf, R.; Prihatmanto, A.S. A survey on traffic flow prediction Methods. In Proceedings of the 6th ICIDM 2020, Bandung, Indonesia, 14–15 December 2020; pp. 1–4. [Google Scholar]
- Wang, Z.M.; Li, Y.; Zhang, R. Action recognition method based on features of autocorrelation function. Appl. Res. Comput. 2018, 35, 6. [Google Scholar]
- Zhao, P.; Li, L. Prediction of urban rail transit station inflows Based on ARIMA model. J. Chongqing Jiaotong Univ. 2020, 39, 1. [Google Scholar]
- Emami, A.; Sarvi, M.; Bagloee, S.A. Using Kalman filter algorithm for short-term traffic flow prediction in a connected vehicle environment. J. Mod. Transport. 2019, 27, 222–232. [Google Scholar] [CrossRef] [Green Version]
- Van Hinsbergen, C.P.; Schreiter, T.; Zuurbier, F.S.; Van Lint, J.W.C.; Van Zuylen, H.J. Localized extended kalman filter for scalable real-time traffic state estimation. IEEE Trans. Intell. Transp. Syst. 2012, 13, 385–394. [Google Scholar] [CrossRef]
- Zhu, Z.Y.; Liu, L.; Cui, M. Short-term traffic flow forecasting model combining SVM and Kalman filtering. Comput. Sci. 2013, 40, 248–251. [Google Scholar]
- Tian, Y.; Gelernter, J.; Wang, X. traffic sign detection using a multi-scale recurrent attention network. IEEE Intell. Transp. 2019, 20, 12. [Google Scholar] [CrossRef]
- Lv, Z.Q.; Li, J.B.; Dong, C.H. DeepPTP: A deep pedestrian trajectory prediction model for traffic intersection. KSII Trans. Internet Inf. Syst. 2021, 15, 2321–2338. [Google Scholar]
- Otter, D.W.; Medina, J.R.; Kalita, J.K. A survey of the usages of deep learning for natural language processing. IEEE Trans. Neural Netw. Learn. Syst. 2020, 32, 604–624. [Google Scholar] [CrossRef] [Green Version]
- Lea, C.; Vidal, R.; Reiter, A.; Hager, G.D. Temporal Convolutional Networks: A unified approach to action segmentation. In Proceedings of the ECCV 2016 WORKSHOPS, PT III, Amsterdam, The Netherlands, 8–10 October 2016; pp. 47–54. [Google Scholar]
- Lv, Z.Q.; Li, J.B.; Li, H.R. Blind travel prediction based on obstacle avoidance in indoor scene. Wirel. Commun. Mobile Comput. 2021, 14, 5536386. [Google Scholar]
- Ahmed, M.S.; Cook, A.R. Analysis of freeway traffic time-series data by using box-jenkins techniques. Transp. Res. Rec. J. Transp. Res. 1979, 773, 1–9. [Google Scholar]
- Cai, C.J.; Yao, E.J.; Wang, M.Y.; Zhang, Y.S. Prediction of urban railway station’s entrance and exit passenger flow based on multiply ARIMA model. J. Beijing Jiaotong Univ. 2014, 38, 135–140. [Google Scholar]
- Zhou, X.; Tang, Y.Z.; Liu, Q. Research on road average speed prediction model based on Kalman filter. J. Zhe Jiang Univ. Technol. 2020, 48, 393–396. [Google Scholar]
- Ren, S.; Han, L.; Li, Z.X.; Veeravalli, B. Spatial-temporal traffic speed bands data analysis and prediction. In Proceedings of the IEEE International Conference on Industrial Engineering and Engineering Management (IEEM), Bali, Indonesia, 4–7 December 2017; pp. 10–13. [Google Scholar]
- Wang, Y.H.; Gao, H.; Chen, X.L. Traffic prediction method using structure varying dynamic Bayesian networks. J. Front. Comput. Sci. Technol. 2017, 11, 528–538. [Google Scholar]
- Liang, Y.P.; Mao, Z.Y.; Zou, W.B.; Xu, R. Short-term traffic flow prediction based on similar data aggregation and KNN with varying K-value. J. Geo-Inf. Sci. 2018, 20, 1403–1411. [Google Scholar]
- Jia, Y.; Wu, J.; Ben-Akiva, M.; Seshadri, R.; Du, Y. Rainfall-integrated traffic speed prediction using deep learning method. IET Intell. Transp. Syst. 2017, 11, 531–536. [Google Scholar] [CrossRef]
- Essien, A.; Petrounias, I.; Sampaio, P.; Sampaio, S. improving urban traffic speed prediction using data source fusion and deep learning. In Proceedings of the 2019 IEEE International Conference on Big Data and Smart Computing (BIGCOMP), Kyoto, Japan, 27 February–2 March 2019; pp. 331–338. [Google Scholar]
- Wu, P.; Huang, Z.; Pian, Y.; Xu, L.; Li, J.; Chen, K. A combined deep learning method with attention-based LSTM model for short-term traffic speed forecasting. J. Adv. Transp. 2020, 4, 8863724. [Google Scholar]
- Wang, J.; Gu, Q.; Wu, J.; Liu, G.; Xiong, Z. Traffic speed prediction and congestion source exploration: A deep learning method. In Proceedings of the IEEE 16th International Conference on Data Mining (ICDM), Barcelona, Spain, 12–25 December 2016; pp. 499–508. [Google Scholar]
- Ma, X.; Dai, Z.; He, Z.; Ma, J.; Wang, Y.; Wang, Y. Learning traffic as images: A deep convolutional neural network for large-scale transportation network speed prediction. Sensors 2017, 17, 818. [Google Scholar] [CrossRef] [Green Version]
- Yu, H.; Wu, Z.; Wang, S.; Wang, Y.; Ma, X. Spatiotemporal recurrent convolutional networks for traffic prediction in transportation networks. Sensors 2017, 17, 1501. [Google Scholar] [CrossRef] [Green Version]
- Yao, H.; Tang, X.; Wei, H.; Zheng, G.; Li, Z. Revisiting spatial-temporal similarity: A deep learning framework for traffic prediction. In Proceedings of the 33rd AAAI Conference on Artificial Intelligence, Honolulu, HI, USA, 27 January–1 February 2019; pp. 5668–5675. [Google Scholar]
- Zhao, L.; Song, Y.; Zhang, C.; Liu, Y.; Wang, P.; Lin, T.; Deng, M.; Li, H. T-GCN: A Temporal graph convolutional network for traffic prediction. IEEE Trans. Intell. Transp. Syst. 2020, 21, 3848–3858. [Google Scholar] [CrossRef] [Green Version]
- Yu, B.; Yin, H.; Zhu, Z. Spatio-temporal graph convolutional networks: A deep learning framework for traffic forecasting. In Proceedings of the 27th International Joint Conference on Artificial Intelligence, Stockholm, Sweden, 9–15 July 2018; pp. 3634–3640. [Google Scholar]
- Feng, N.; Guo, S.; Song, C.; Zhu, Q.; Wan, H. Multi-component spatial-temporal graph convolution networks for traffic flow forecasting. J. Softw. 2019, 30, 759–769. [Google Scholar]
- Guo, S.; Lin, Y.; Feng, N.; Song, C.; Wan, H. Attention based spatial-temporal graph convolutional networks for traffic flow forecasting. In Proceedings of the AAAI Conference on Artificial Intelligence, Honolulu, HI, USA, 27 January–1 February 2019; pp. 922–929. [Google Scholar]
- Bai, S.; Kolter, J.; Koltun, V. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling. arXiv 2018, arXiv:1803.01271. [Google Scholar]
- Lv, Z.; Li, J.; Dong, C.; Li, H.; Xu, Z. Deep learning in the COVID-19 epidemic: A deep model for urban traffic revitalization index. Data Knowl. Eng. 2021, 135, 101912. [Google Scholar] [CrossRef]
- Simonovsky, M.; Komodakis, N. Dynamic edge-conditioned filters in convolutional neural networks on graphs. In Proceedings of the 30th IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA, 21–26 July 2017; pp. 29–38. [Google Scholar]
- Defferrard, M.; Bresson, X.; Vandergheynst, P. Convolutional neural networks on graphs with fast localized spectral filtering. Adv. Neural Inf. Process. Syst. 2016, 29, 3844–3852. [Google Scholar]
- Hammond, D.K.; Vandergheynst, P.; Gribonva, R. Wavelets on graphs via spectral graph theory. Appl. Comput. Harmon. Anal. 2011, 30, 129–150. [Google Scholar] [CrossRef] [Green Version]
- Kipf, T.N.; Welling, M. semi-supervised classification with graph convolutional networks. arXiv 2017, arXiv:1609.02907. [Google Scholar]
- Li, Q.; Han, Z.; Wu, X. Deeper insights into graph convolutional networks for semi-supervised learning. In Proceedings of the 32nd AAAI Conference on Artificial Intelligence, New Orleans, LA, USA, 2–7 February 2018; pp. 3538–3545. [Google Scholar]
- Tang, W.; Ma, S.; Li, C. LSTM ground temperature prediction method based on sliding window. J. Cheng Du Univ. Technol. 2021, 48, 3. [Google Scholar]
- Hochreiter, S.; Schmidhuber, J. Long short-term memory. Neural Comput. 1997, 9, 1735–1780. [Google Scholar] [CrossRef]
- Cho, K.; Merrienboer, B.V.; Gulcehre, C.; Bahdanau, D.; Bougares, F.; Schwenk, H.; Bengio, Y. Learning phrase representations using RNN encoder-decoder for statistical machine translation. In Proceedings of the Conference on EMNLP 2014, Doha, Qatar, 25–29 October 2014; pp. 1724–1734. [Google Scholar]
Parameter | Parameter Interpretation |
---|---|
Fwy | Highway number |
Dir | The direction of the highway (N: 0, E: 1, S: 2, W: 3) |
District | District where the sensor is located |
County | County where the sensor is located |
State_PM | The distance of a highway across a state |
Abs_PM | The distance to the start of the highway |
Latitude | Latitude of sensor |
Longitude | Longitude of sensor |
Type | Road type (ML:0; HV:1) |
Lanes | Number of lanes |
Parameter | Parameter Value |
---|---|
Batch size | 16 |
Epoch | 100 |
Optimizer | Adam |
The number of residual blocks in TCN | 5 |
Model | 5 min | 10 min | 15 min | |||
---|---|---|---|---|---|---|
MAE | RMSE | MAE | RMSE | MAE | RMSE | |
LSTM | 3.86 | 6.53 | 3.88 | 6.54 | 3.90 | 6.53 |
GRU | 3.24 | 5.86 | 3.24 | 5.88 | 3.25 | 5.90 |
STDN | 3.23 | 5.16 | 3.25 | 5.16 | 3.29 | 5.18 |
GCN | 2.96 | 5.13 | 2.96 | 5.17 | 2.97 | 5.22 |
T-GCN | 2.70 | 4.81 | 2.72 | 4.81 | 2.73 | 4.84 |
STGCN | 1.76 | 2.71 | 1.62 | 2.40 | 1.63 | 2.41 |
ASTGCN | 1.51 | 2.38 | 1.43 | 2.16 | 1.48 | 2.25 |
SPTMN | 1.37 | 2.09 | 1.32 | 2.01 | 1.37 | 2.09 |
Model | 5 min | 10 min | 15 min | Loss | |||
---|---|---|---|---|---|---|---|
MAE | RMSE | MAE | RMSE | MAE | RMSE | ||
SPTMN | 1.39 | 2.12 | 1.33 | 2.02 | 1.36 | 2.07 | 5.42 |
SPTMN without feature enhancement module | 3.37 | 5.48 | 3.41 | 5.46 | 3.47 | 5.51 | 23.71 |
SPTMN without GCN module | 1.70 | 2.51 | 1.99 | 3.02 | 2.27 | 3.48 | 6.31 |
SPTMN without parameter module | 3.29 | 4.99 | 3.28 | 4.99 | 3.29 | 5.02 | 29.76 |
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Hu, Z.; Sun, R.; Shao, F.; Sui, Y. An Efficient Short-Term Traffic Speed Prediction Model Based on Improved TCN and GCN. Sensors 2021, 21, 6735. https://doi.org/10.3390/s21206735
Hu Z, Sun R, Shao F, Sui Y. An Efficient Short-Term Traffic Speed Prediction Model Based on Improved TCN and GCN. Sensors. 2021; 21(20):6735. https://doi.org/10.3390/s21206735
Chicago/Turabian StyleHu, Zhiqiu, Rencheng Sun, Fengjing Shao, and Yi Sui. 2021. "An Efficient Short-Term Traffic Speed Prediction Model Based on Improved TCN and GCN" Sensors 21, no. 20: 6735. https://doi.org/10.3390/s21206735
APA StyleHu, Z., Sun, R., Shao, F., & Sui, Y. (2021). An Efficient Short-Term Traffic Speed Prediction Model Based on Improved TCN and GCN. Sensors, 21(20), 6735. https://doi.org/10.3390/s21206735