Citation

Sun Z, Xing Y, Wang J, Gu X, Lu H, Chen Y. Safety Sci. 2022; 150: e105682.

Copyright

(Copyright © 2022, Elsevier Publishing)

DOI

10.1016/j.ssci.2022.105682

PMID

unavailable

Abstract

Recently, vulnerable road user to motor vehicle (VRU-MV) crashes have been the focus of much attention. Data from Shenyang, China were used for crashes analysis and significant differences existed in VRU-MV crashes across seasons. Therefore, this study analysed three crash datasets (whole, "spring & summer", "fall & winter") by using a hybrid method integrating random parameter logit (RP-logit) model and Bayesian network (BN). First, RP-logit model was established to find out significant factors and heterogeneity considering those three datasets. Second, significant factors identified from RP-logit model were utilized to establish a BN to investigate statistical associations between injury severity and explanatory attributes.

RESULTS showed seven significant factors were found in "spring & summer" dataset while five significant factors in "fall & winter" dataset, four hidden factors were found by comparative analysis of those three datasets. Besides, five factors were found to be random and normally distributed.

RESULTS of BN indicated that some factors could significantly increase high possibility of fatality when combined with other factors. For example, functional zone (ZON) in "spring & summer" dataset and motor vehicle type (MVT) in "fall & winter" dataset. The proposed hybrid method demonstrated both the consistence of methods (RP-logit model & BN) and the differences across seasons for VRU-MV crashes analysis. Three personalized factors including physical isolation (PI), crash type (CT) and motor vehicle type (MVT) made great differences across seasons.

Language: en

Keywords

Bayesian network; Injury severity; Random parameter logit model; Vulnerable road user