Injury severity prediction model for two-wheeler crashes at mid-block road sections

Abstract

Motorised two-wheelers (TW) have the highest proportion among vehicles in Chennai district of Tamil Nadu, and they are involved in a large number of fatal traffic crashes every year. We develop a machine learning model to predict injury severity of TW drivers involved in crashes at mid-block road sections, and thereby identify factors contributing to the severity. We used 7654 TW crash cases that occurred in Chennai from 2016 to 2018. We study the performance of two machine learning models random forest (RF) and Conditional inference forest (Cforest), in injury severity prediction and compared their performance with ordered probit (OP) model. Cforest outperforms both RF and OP models in predicting injury severity. We identify significant variables based on variable importance factor measure. Out of considered variables, type of colliding vehicle has the highest influence on crash severity followed by collision type, driver age, and visibility of the road. The Cforest model captures interaction effects that are missed by the other two models.

Keywords:

• Two-wheeler driver
• injury severity
• conditional inference forest
• random forest
• ordered probit model

Acknowledgements

The authors would like to acknowledge Tamil Nadu police department and Greater Chennai City Traffic Police for providing the crash data for this study.

Disclosure statement

No potential conflict of interest was reported by the authors.