Identifying significant predictors of injury severity in multi-vehicle crashes using multi-layer perceptron neural network

Articles in Press

Document Type : Original Article

Authors

1 M.Sc student, civil engineering department, Iran university of science and technology, Tehran, Iran

2 Professor, Dept. of Civil Engineering, Iran university of science and Technology, Tehran, Iran.

3 Associate Professor, Dept. of Civil Engineering, Isfahan University of Technology, Isfahan, Iran.

Abstract

Based on the analyzed data, multi-vehicle crashes compose ten percent of crashes in Tehran province in 1390. In order to reduce injuries severity in this kind of crashes, we need to identify significant factors affecting injuries severity. In this paper, multi vehicle crashes in Tehran province has been analyzed by using MLP neural network. In order to increase the modeling accuracy, two models with different level of injuries severity have been developed. The classification accuracy in model one and two was calculated 98.5% and 98.4%. The area under the ROC curve for all different models and severities have been calculated and shows acceptable ability to distinguish between the outcome groups. The result from sensitivity analysis shows that the vehicle type and vehicle model have significant importance on multi vehicle crashes at all level of injuries severity. In the lower level of injuries severity, light condition, accident position and any obstacle in driver’s sight can influence injuries severity and in higher levels of injuries, driver’s age, manner of collision and human fault can affect injuries severity in multi vehicle crashes.

Keywords

Main Subjects

Journal of Transportation Research

Articles in Press, Accepted Manuscript
Available Online from 16 December 2018

Files

Share

How to cite

Statistics