Crash Density and Severity Prediction Using Recurrent Neural Networks Combined with Particle Swarm Optimization
- 274 Downloads
Predicting traffic crashes has been an important topic of traffic safety research for the past many years. This paper investigates the data from police crash reports provided by the Washington State Department of Transportation. The data consists of records of four years from January 2011 to December 2014 for three main interstate highways (including I-5, I-90, and I-405). A deep learning model using a recurrent neural network (RNN) combined with particle swarm optimization (PSO) is developed and employed to predict the crash density in different severity levels such as property damage only (PDO) and fatal-injury crashes, based on 48,154 crash records that have occurred. All the crash records are randomly divided into training set, validation set, and test set with the proportion ratio of 70, 15, and \(15\%\). The cross-validation is employed to prevent the model from over fit during the training period. A normalized probability-based PSO is designed for optimizing the identified significant factors which can improve the prediction accuracy. The weighted mean squared error (MSE) of the prediction result is employed to measure the performance of the developed model. Nine explanatory variables are selected from fifteen contributing factors. The proposed model is compared with generalized nonlinear model-based mixed multinomial logit approach (GNM-based mixed MNL). The results show that the new model has lower fatal-injury and PDO MSEs. Sensitivity analysis on the selected variables demonstrates the capability of the new model for generating interpretable parameters. The findings of this study provide new insights into the prediction of crash density and severity from the perspective of using roadway segment-based crash records.
KeywordsCrash density and severity Deep learning Recurrent neural network Particle swarm optimization Prediction
This research was supported by the Youth Program of National Natural Science Foundation of China (Grant No. 71501137), the Humanities and Social Sciences Programs of Ministry of Education in China (Grant No. 17XJC790016), the project of Research Center for System Sciences and Enterprise Development (Grant No. Xq16B05), Sichuan University (Grant No. skqy201647), Soft Science Project of Sichuan Science and Technology Department (Grant No. 2019JDR0167), and Sichuan Social Science Planning Project (Grant No. SC18TJ014). The authors would like to give our great appreciation to the editors and anonymous referees for their helpful and constructive comments and suggestions, which have helped to improve this article.
- 1.Sameen, M., Pradhan, B.: Severity prediction of traffic accidents with recurrent neural networks. Appl. Sci. Basel 7(6):AN, 476 (2017)Google Scholar
- 2.World Health Organization: Global status report on road safety. http://www.who.int/violence_injury_prevention/road_safety_status/2015/zh/ (2015)
- 4.Fan, W.D., Gong, L., Washing, E.M., Yu, M., Haile, E.: Identifying and quantifying factors affecting vehicle crash severity at highway-rail grade crossings: models and their comparison. Transportation Research Board 95th Annual Meeting, Washington DC (2016)Google Scholar
- 6.Ye, F., Lord, D.: Comparing three commonly used crash severity models on sample size requirements: multinomial logit, ordered probit and mixed logit models. Accid. Anal. Prev. 1, 72–85 (2014)Google Scholar
- 24.Donahue, J., Anne Hendricks, L., Guadarrama, S., Rohrbach, M., Venugopalan, S., Saenko, K., Darrell, T.: Long-term recurrent convolutional networks for visual recognition and description. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Boston, MA, USA, vol. 7–15, pp. 2625–2634 (2015)Google Scholar
- 25.Kennedy, J., Eberhart, R.C.: Particle swarm optimization. In Proceedings of the IEEE International Conference on Neural Networks (ICNN 95), Perth, Australia, 27 Nov–1 Dec, no. 1–6, pp. 1942–1948 (1995)Google Scholar
- 30.Mostaghim, S., Teich, J.: Strategies for finding good local guides in multi-objective particle swarm optimization (MOPSO). In: IEEE 2003 Swarm Intelligence Symposium, pp. 26–33 (2003)Google Scholar