Bi-Directional Isometric GRU for Urban Travel Time Prediction Using Deep Learning on Synthetic Global Positioning System Trajectories

Riki Riki; Setyawan Widyarto

Authors

Riki Universitas Buddhi Dharma
Setyawan Widyarto Universiti Selangor

Keywords:

Travel time prediction, Bi-Directional Isometric Gated Recurrent Unit (BDIGRU), Temporal Forecasting, Synthetic Traffic Data

Abstract

Accurate short-term travel time prediction is crucial for optimizing urban transportation systems and enabling responsive Intelligent Transportation Systems (ITS). The research introduces a novel Deep Learning (DL) architecture, the Bi-Directional Isometric Gated Recurrent Unit (BDIGRU), to enhance predictive accuracy and resilience under congested traffic and variable conditions. BDIGRU uses isometric memory alignment to process temporal data in both forward and backward directions, enabling symmetric learning and balanced context integration. This is unlike traditional sequence models that are unidirectional. The researchers construct a highresolution synthetic traffic dataset generated through stochastic simulation techniques to evaluate the model and replicate real-world urban traffic patterns in Jakarta, including variability in speed, volume, and congestion cycles. The dataset spans over 4 years, with more than 437,000 records captured at 5-min intervals. It includes multivariate traffic features such as average speed, vehicle volume, lane occupancy, and derived travel time. The performance of BDIGRU is benchmarked against five baseline models using four standard evaluation metrics. The results show that BDIGRU outperforms all the comparative models, achieving the lowest prediction errors (e.g., Mean Absolute Percentage Error (MAPE) = 7.2%, Symmetric Mean Absolute Percentage Error (SMAPE) = 8.0%). It also demonstrates superior robustness during high-variance traffic periods and congestion spikes. These findings underscore the architectural advantages of bidirectional and isometric temporal modeling for short-term traffic forecasting. Overall, BDIGRU offers a scalable and efficient framework suitable for real-time ITS deployment and highlights the value of synthetic datasets in developing and validating DL models for transportation analytics.

Dimensions

Author Biographies

Riki, Universitas Buddhi Dharma

Information System Department, Faculty of Science and Technology,

Setyawan Widyarto, Universiti Selangor

Computing Department, Faculty of Communication, Visual Art and Computing

References

[1] The World Bank, “Planning, connecting & financing cities–Now,” 2013. [Online]. Available: https://bit.ly/4cBQH4r

[2] Y. An, M. Li, X. Lin, F. He, and H. Yang, “Space-time routing in dedicated automated vehicle zones,” Transportation Research Part C: Emerging Technologies, vol. 120, 2020.

[3] D. Rezaei, I. Aghayan, and F. Hadadi, “Studying perturbations and wave propagations by lane closures on traffic characteristics based on a dynamic approach,” Physica A: Statistical Mechanics and its Applications, vol. 566, 2021.

[4] R. Deshmukh, D. Sun, K. Kim, and I. Hwang, “Temporal logic learning-based anomaly detection in metroplex terminal airspace operations,” Transportation Research Part C: Emerging Technologies, vol. 126, 2021.

[5] D. Liu, Z. Wang, Y. Liu, and F. E. Alsaadi, “Extended Kalman filtering subject to random transmission delays: Dealing with packet disorders,” Information Fusion, vol. 60, pp. 80–86, 2020.

[6] H. Rao, S. Xu, X. Hu, J. Cheng, and B. Hu, “Augmented skeleton based contrastive action learning with momentum LSTM for unsupervised action recognition,” Information Sciences, vol. 569, pp. 90–109, 2021.

[7] A. Sarraf and S. Khalili, “An upper bound on the variance of scalar multilayer perceptrons for log-concave distributions,” Neurocomputing, vol. 488, pp. 540–546, 2022.

[8] J. B. Florindo, Y. S. Lee, K. Jun, G. Jeon, and M. K. Albertini, “VisGraphNet: A complex network interpretation of convolutional neural features,” Information Sciences, vol. 543, pp. 296–308, 2021.

[9] V. Singh, S. K. Sahana, and V. Bhattacharjee, “A novel CNN-GRU-LSTM based deep learning model for accurate traffic prediction,” Discover Computing, vol. 28, pp. 1–25, 2025.

[10] J. Tang et al., “Deep spatio-temporal dependent convolutional LSTM network for traffic flow prediction,” Scientific Reports, vol. 15, pp. 1–15, 2025.

[11] W. Guo, Q. Zeng, H. Duan, W. Ni, and C. Liu, “Process-extraction-based text similarity measure for emergency response plans,” Expert Systems with Applications, vol. 183, 2021.

[12] M. Schuster and K. K. Paliwal, “Bidirectional recurrent neural networks,” IEEE transactions on Signal Processing, vol. 45, no. 11, pp. 2673–2681, 1997.

[13] L. L. Zhang, D. U. Gang, W. U. Jun, and M. A. Yujie, “Joint production planning, pricing and retailer selection with emission control based on Stackelberg game and nested genetic algorithm,” Expert Systems with Applications, vol. 161, 2020.

[14] C. Y. T. Chen, E. W. Sun, M. F. Chang, and Y. B. Lin, “Enhancing travel time prediction with deep learning on chronological and retrospective time order information of big traffic data,” Annals of Operations Research, vol. 343, no. 3, pp. 1095–1128, 2024.

[15] W. Xiao, H. Liu, Z. Ma, and W. Chen, “Attentionbased deep neural network for driver behavior recognition,” Future Generation Computer Systems, vol. 132, pp. 152–161, 2022.

[16] M. Seppecher, L. Leclercq, A. Furno, D. Lejri, and T. V. Da Rocha, “Estimation of urban zonal speed dynamics from user-activity-dependent positioning data and regional paths,” Transportation Research Part C: Emerging Technologies, vol. 129, 2021.

[17] H. Han, J. Teng, J. Xia, Y. Wang, Z. Guo, and D. Li, “Predict high-frequency trading marker via manifold learning,” Knowledge-Based Systems, vol. 213, 2021.

[18] Y. Kim, H. y. Tak, S. Kim, and H. Yeo, “A hybrid approach of traffic simulation and machine learning techniques for enhancing real-time traffic prediction,” Transportation research part C: emerging technologies, vol. 160, 2024.

[19] M. S. Ansari, “A single-layer asymmetric RNN with low hardware complexity for solving linear equations,” Neurocomputing, vol. 485, pp. 74–88, 2022.

[20] F. Orozco, H. Wen, P. P. De Gusm, J. Wahlstr¨om, and M. Luo, “Federated learning for traffic flow prediction with synthetic data augmentation,” IEEE Open Journal of Intelligent Transportation Systems, vol. 7, pp. 445–459, 2026.

[21] A. Zhang, Z. C. Lipton, M. Li, and A. J. Smola, “Dive into deep learning.” [Online]. Available: https://d2l.ai/index.html

[22] V. Bonato and C.-S. Bouganis, “Class-specific early exit design methodology for convolutional neural networks,” Applied Soft Computing, vol. 107, 2021.

[23] W. Shu, K. Cai, and N. N. Xiong, “Research on strong agile response task scheduling optimization enhancement with optimal resource usage in green cloud computing,” Future Generation Computer Systems, vol. 124, pp. 12–20, 2021.

[24] V. Cerqueira, L. Torgo, and I. Mozetiˇc, “Evaluating time series forecasting models: An empirical study on performance estimation methods,” Machine Learning, vol. 109, pp. 1997–2028, 2020.

[25] P. Manandhar, H. Rafiq, E. Rodriguez-Ubinas, and T. Palpanas, “New forecasting metrics evaluated in prophet, random forest, and long shortterm memory models for load forecasting,” Energies, vol. 17, no. 23, pp. 1–30, 2024.

[26] G. Li, Y. Chen, Y. Wang, P. Nie, Z. Yu, and Z. He, “City-scale synthetic individual-level vehicle trip data,” Scientific Data, vol. 10, pp. 1–18, 2023.

[27] T. Thomas, W. Weijermars, and E. Van Berkum, “Variations in urban traffic volumes,” European Journal of Transport and Infrastructure Research, vol. 8, no. 3, pp. 251–263, 2008.

[28] Y. Wang et al., “Exploiting topic-based adversarial neural network for cross-domain keyphrase extraction,” in 2018 IEEE International Conference on Data Mining (ICDM). Singapore: IEEE, Nov. 17–20, 2018, pp. 597–606.

[29] J. B. Sheu, “Microscopic traffic behaviour modelling and simulation for lane-blocking arterial incidents,” Transportmetrica A: Transport Science, vol. 9, no. 4, pp. 335–357, 2013.