There has complex relationship between traffic crashes and many influencing factors. This study develops a random forest model for predicting highway crash severity using crash database collected from 11 highways. We extract important factors affecting highway crash severity and classify the crash severity into three types: minor injury, serious injury and fatality. SPSS MODELER software is used to model a total of 5211 collisions occurring from the January of 2017 to the April of 2019. In addition, sensitivity analysis of significant factors is also implemented. The results in this study can provide technical services for relevant management departments to make scientific and reasonable accident response decisions. For future investigation, real-time highway crash severity prediction models will be developed to reduce the severity of the accident.

CICTP