Citation

Agrawal V, Chatterjee S, Mitra S. J. East Asia Soc. Transp. Stud. 2019; 13: 2614-2629.

Copyright

(Copyright © 2019, Eastern Asia Society for Transportation Studies)

DOI

10.11175/easts.13.2614

PMID

unavailable

Abstract

Various parametric models such as logistic regression or linear discriminant analysis have been most commonly used to explore factors contributing to the severity of crashes. These models assume a functional form and learn the coefficients for the function from the training data. If the assumptions are breached the models can lead to an incorrect prediction of the crash severity. In such cases, Non-parametric models seek to be the best fit for the training data in constructing the mapping function while maintaining some ability to generalize the unseen data. In this study, three non-parametric machine-learning methods viz. Classification and Regression Tree (CART), Extreme Gradient Boosting (XGBoost) and Support Vector Machines (SVM) have been utilised to identify the critical factors affecting the severity of traffic crashes on the State Highways of India. Among the three studied methods, the XGBoost was found to be the best performing. The results show that the presence of pedestrian facility, type of collision, weather conditions, intersection proximity, type of traffic control and speed limits are the critical factors affecting crash severity on the State Highways.

Language: en

Keywords

Crash severity; Data Mining; Traffic Crashes; XGBoost