Analysis and Prediction of Urban Traffic Accident Correlating Factors Based on SARIMAX Model

Ying Zhu; Wei Yu; Hongyan Li; Weiyan Yang; Yue Lin

doi:10.63313/JCSFT.9062

Authors

Ying Zhu College of International Business, Zhejiang Yuexiu Univerisity, Shaoxing, 312069, China Author
Wei Yu College of International Business, Zhejiang Yuexiu Univerisity, Shaoxing, 312069, China Author
Hongyan Li College of International Business, Zhejiang Yuexiu Univerisity, Shaoxing, 312069, China Author
Weiyan Yang College of International Business, Zhejiang Yuexiu Univerisity, Shaoxing, 312069, China Author
Yue Lin College of International Business, Zhejiang Yuexiu Univerisity, Shaoxing, 312069, China Author

DOI:

https://doi.org/10.63313/JCSFT.9062

Keywords:

SARMAX model, accident data analysis, time series forecasting, spatial visualization, seasonal analysis

Abstract

Nowadays, with the acceleration of urbanization and the rapid increase in population and motor vehicles, frequent traffic accidents pose a threat to public safety, forming the background of this study. Predicting accidents and their resulting losses to provide a basis for traffic management decisions is a key issue to be solved by society. Theoretically, in terms of research significance, the introduction of the SARIMAX model broadens the research path of traffic accidents and promotes interdisciplinary integration. In practice, it can help traffic departments optimize resource allocation and prevent accident risks. Visualized results can provide travel warnings for the public, reducing economic and human losses. The research focuses on urban traffic accident data. It first widely collects multi-dimensional raw data, followed by cleaning, seasonal decomposition, and stationarity testing. Exogenous variables such as weather and holidays are introduced to construct the SARIMAX model. Finally, the Gaode Maps API is used to visualize accident-prone areas. The research methods include literature research, empirical research, and interdisciplinary research methods. The literature method sorts out cutting-edge materials to lay a foundation; the empirical method relies on big data to verify according to the process; and the interdisciplinary method integrates multi-disciplinary expertise to solve problems, providing data support for regional control strategies.

References

[1] Luo, G.C., Gao, J.Y., & Cai, W.X. (2021). Seasonal time series prediction based on GRA-LSTM and SARIMA combination model. Intelligent Computer and Applications, 11(6), 195-200.

[2] Sun, Y.X., Shao, C.F., Ji, X., & Zhu, L. (2014). Traffic accident time series prediction based on ARIMA and information granulation SVR combination model. Journal of Beijing Jiaotong University, 54(3), 348-353.

[3] Amin, S.M., et al. (2007). Road accident forecasting models using time series analysis and artificial neural networks. Accident Analysis & Prevention, 39(2), 311-320.

[4] Wang, Y.J., & Sun, Y.X. (2007). Road traffic accident prediction based on the GM(1,1) model. Traffic Technology and Economy, 9(5), 97-98.

[5] Liu, Z.Q., Shan, X.M., Xin, N., et al. (2024). Research on traffic accident prediction based on ES-ARIMA-BP neural network with combined weighting optimization. Modern Electronic Technology, 47(22), 71-76.

[6] Yang, X.M. (2000). Fuzzy neural network prediction model for bus route passenger flow. Highway Traffic Technology, 17(4), 38-41.

[7] Qiao, Y.Q. (2022). Traffic Accident Prediction Based on Deep Learning (Master’s thesis). Xi'an University of Science and Technology.

[8] Zhao, J.B., Deng, W., & Wang, J. (2011). Urban road traffic accident analysis based on Bayesian network. Journal of Southeast University (Natural Science Edition), 41(6), 1300-1306.

[9] Zhang, J., Liu, X.M., He, Y.L., et al. (2007). Application of the ARIMA model in traffic accident prediction. Journal of Beijing University of Technology, 33(12), 1295-1299.

[10] Wu, Y.D. (2023). Traffic Accident Prediction and Analysis Based on an Improved LSTM Neural Network Model (Master’s thesis). Chongqing Jiaotong University.