ADVANCING TRAFFIC SAFETY THROUGH CAPRNN: A HYBRID CAPSULE-RECURRENT NEURAL NETWORK FOR ACCIDENT FORECASTING

Rahul Verma; Madan Mohan Agarwal

doi:10.30572/2018/KJE/160419

Authors

Rahul Verma Research Scholar, Department of Computer Science & Engineering, Birla Institute of Technology, Mesra, Jaipur Campus, India https://orcid.org/0009-0003-6184-7627
Madan Mohan Agarwal Assistant Professor, Department of Computer Science & Engineering, Birla Institute of Technology, Mesra, Jaipur Campus, India https://orcid.org/0000-0003-4275-7485

DOI:

https://doi.org/10.30572/2018/KJE/160419

Keywords:

Capsule Recurrent Neural Network (CapRNN), road accident prediction, deep learning, accident forecasting, capsule networks

Abstract

One of the primary causes of death globally has been traffic accidents involving vehicles on public roads. To address this issue, we introduce Capsule Recurrent Neural Network (CapRNN), a new deep learning architecture designed specifically to improve road accident prediction performance. CapRNN uses capsule networks for capturing spatial relationships and recurrent neural networks for capturing temporal patterns. A series of experiments has been performed with CapRNN on a large-scale dataset of U.S. traffic accidents from 2016 to 2023 and compared it with other typical models. The result was the 95.82% accuracy of CapRNN, which was the highest one. The standard LSTM and Bi-LSTM models could only achieve 93.58%, respectively. Also, the performance of CapRNN was higher than that of XGBoost (95.73%) and Random Forest (95.02%). Besides accuracy, recall was also high for CapRNN (95.82%) and precision was also high (95.24%), i.e., it was stable at different time resolutions. This finding supports our claim that hybrid deep learning models like ours have the potential to effectively solve the spatiotemporal nature problem of road accident data. CapRNN's improved prediction accuracy could be used directly to make public safety policies more efficient by enabling them to better allocate the resources devoted to traffic management and urban planning. We, therefore, propose further steps: testing the model on other data sets, making it more interpretable, and, finally, incorporating real-time data to strengthen predictions

Downloads

Download data is not yet available.

References

Agarwal, M. M., Verma, R., & Berwal, K. (2025, May). Building the Framework for Road Accident Prediction for India using Enhanced SVM Techniques. In 2025 IEEE International Conference on Computer, Electronics, Electrical Engineering & their Applications (IC2E3) (pp. 1-6). IEEE.

Asad, F., & Hadi, S. (2024). Modelling The Factors Affecting Crash Occurrence and Frequency Resulted From Mobile Phone Use While Driving: Evidence From Al-Najaf, Iraq. Kufa Journal of Engineering, 15(2), 74–92. https://doi.org/10.30572/2018/KJE/150206

Asad, F., & Saeed, M. (2024). Investigating The Risk Factors Affecting The Occurrence, Frequency, And Severity Of Large Truck Accidents In Al-Najaf Governorate, Iraq. Kufa Journal of Engineering, 15(1), 30–46. https://doi.org/10.30572/2018/KJE/150103

Bhuyan, M. K., Mohapatra, D. P., & Sethi, S. (2016). Prediction Strategy for Software Reliability Based on Recurrent Neural Network (pp. 295–303). https://doi.org/10.1007/978-81-322-2731-1_27

Du, X., Ding, X., & Tao, F. (2023). Network Security Situation Prediction Based on Optimized Clock-Cycle Recurrent Neural Network for Sensor-Enabled Networks. Sensors, 23(13), 6087. https://doi.org/10.3390/s23136087

Fan, J., Mu, D., & Liu, Y. (2019). Research on Network Traffic Prediction Model Based on Neural Network. 2019 2nd International Conference on Information Systems and Computer Aided Education (ICISCAE), 554–557. https://doi.org/10.1109/ICISCAE48440.2019.221694

Fan, X., Chen, L., Xu, X., Yan, C., Fan, J., & Li, X. (2023). Land Cover Classification of Remote Sensing Images Based on Hierarchical Convolutional Recurrent Neural Network. Forests, 14(9), 1881. https://doi.org/10.3390/f14091881

Kalbaliyev, E., & Szegedi, G. (2021). Long Term Stock Market Prediction with Recurrent Neural Network Architectures (pp. 1084–1093). https://doi.org/10.1007/978-3-030-71187-0_100

Kim, Y., Wang, P., & Mihaylova, L. (2019). Scalable Learning With a Structural Recurrent Neural Network for Short-Term Traffic Prediction. IEEE Sensors Journal, 19(23), 11359–11366. https://doi.org/10.1109/JSEN.2019.2933823

Lu, M., & Xu, X. (2024). TRNN: An efficient time-series recurrent neural network for stock price prediction. Information Sciences, 657, 119951. https://doi.org/10.1016/j.ins.2023.119951

Sain, S. K., Khandelwal, K., & Agarwal, M. M. (2024). Smart Vehicales Surveillance in Foggy Conditions Using Enhanced Deep Learning Algorithms. 45(3), 914–929.

Shaik, M. E., Islam, M. M., & Hossain, Q. S. (2021). A review on neural network techniques for the prediction of road traffic accident severity. Asian Transport Studies, 7, 100040. https://doi.org/10.1016/j.eastsj.2021.100040

Shi, C., & Zhang, X. (2023). Recurrent neural network wind power prediction based on variational modal decomposition improvement. AIP Advances, 13(2). https://doi.org/10.1063/5.0135711

Sivalingan, H (2024). Cloud-Smart Surveillance: Enhancing Anomaly Detection In Video Streams With Df-Convlstm-Based Vae-Gan. Kufa Journal of Engineering, 15(4), 125–140. https://doi.org/10.30572/2018/KJE/150409

Verma, R., & Agarwal, M. M. (2024). Enhancing Road Accident Prediction : An Advanced Information Retrieval Approach with Enhanced Focal Loss. Library of Progress-Library Science, Information Technology & Computer , 44(3), 13679–13690.

Verma, R., & Agarwal, M. M. (June 2025). A systematic review on road accident prediction: A special attention towards machine learning and deep learning approaches. Journal of Circuits, Systems and Computers. https://doi.org/10.1142/S0218126625300090

Verma, R., & Agarwal, M. M. (March 2025). Enhancing Road Accident Classification with Capsule Recurrent Neural Network and Improved Reptile Search Algorithm. IETE Journal of Research, 71(5), 1822–1841. https://doi.org/10.1080/03772063.2025.2464239.

Verma, R., and Agarwal, M. M. (April 2025). Road Accident Severity Prediction using Adaptive Custom Weight Initialization and Enhanced Focal Loss Integration Technique. IETE Journal of Research, 1-13. https://doi.org/10.1080/03772063.2025.2487628

Yang, C., Wang, W., Zhang, X., Guo, Q., Zhu, T., & Ai, Q. (2021). A Parallel Electrical Optimized Load Forecasting Method Based on Quasi-Recurrent Neural Network. IOP Conference Series: Earth and Environmental Science, 696(1), 012040. https://doi.org/10.1088/1755-1315/696/1/012040

Yang, J., Li, J., Wei, L., Gao, L., & Mao, F. (2022). ST-AGRNN: A Spatio-Temporal Attention-Gated Recurrent Neural Network for Traffic State Forecasting. Journal of Advanced Transportation, 2022, 1–17. https://doi.org/10.1155/2022/2806183

Zhang, T., Ding, W., & Xing, M. (2023). Attention-based Dynamic Graph Convolutional Recurrent Neural Network for Traffic Flow Prediction in Highway Transportation. http://arxiv.org/abs/2309.07196