This study aims to use the multinomial logit (MNL) model and the random forest (RF) to predict the severity of crash injuries on highways with or without the involvement of traffic hazards. The prediction models are built with highway crash data from January 2017 to April 2019. For crashes occurring on the highways, involving traffic hazards and those, not involving traffic hazards, the RF predicts severity driver injury more accurately than the MNL model. The prediction accuracy in this study is based on variables such as precision, recall, and F1 score. Additionally, the RF is applied to determine the importance of each variable in comparison to the others. The results showed that time the of day is the most important factor affecting severity of driver injury in crashes involving traffic hazards while vehicle type is the most important factor affecting severity of driver injury in crashes involving no traffic hazards.
Predicting Crash Injury Severity for the Highways Involving Traffic Hazards and Those Involving No Traffic Hazards
20th COTA International Conference of Transportation Professionals ; 2020 ; Xi’an, China (Conference Cancelled)
CICTP 2020 ; 4195-4206
2020-08-12
Conference paper
Electronic Resource
English
Taylor & Francis Verlag | 2021
|Factors impacting injury severity of crashes involving traffic barrier end treatments
Taylor & Francis Verlag | 2021
|Examining traffic crash injury severity at unsignalized intersections
Online Contents | 2010
|