To reduce the probability of roadside accidents in small curve sections of highways

eight roadside accident risk factors including road geometric design indexes (horizontal curve radius

hard shoulder width

longitudinal slope

superelevation slope

and widen value of curve)

pavement condition (adhesion coefficient)

and traffic characteristics (running speed and vehicle type) were chosen to carry out PC-crash simulation test

and a total of 12 800 accident data sets were collected. Chi-squared automatic interaction detection (CHAID) decision tree technique was employed to identify significant risk factors

and the comprehensive influence of the interaction of various factors on roadside accidents was discussed. These factors were then chosen as predictors of probability of roadside accidents in Bayesian network analysis to establish the probabilistic prediction model of roadside accidents. Finally

according to probabilistic prediction results

the identification method for roadside accidents black spots was proposed and verified through tests. Results show that running speed had the greatest effect on the occurrence of roadside accidents

followed by horizontal curve radius

vehicle type

adhesion coefficient

and hard shoulder width. When 80 km/h