Prediction of roadway accident frequencies: Count regressions versus machine learning models

Prediction of accident frequency based on trac and roadway characteristicshas been a very signicant tool in the eld of trac management. the accident frequencieson 185 roadway segments of the city of mashhad, iran, for the year 2007, were used todevelop accident prediction models. negative binomial regression, zero inated negativebinomial regression, support vector machine and back-propagation neural networkmodels were used to t the accident data. both tting and predicting abilities of themodels were evaluated through computing error values.results show that the nbr model is the most eective model for predicting thenumber of accidents because of its low prediction and tting error values. although thebpnn model has high tting capability, it does not have the prediction ability of thenbr model. furthermore, the nbr is easily able to develop and interpret the role ofeective variables, in comparison with machine learning models which have a black-boxform. marginal eect values for the nbr and zinbr models, and sensitivity analysisof the svm and bpnn models, reveal that volume to capacity ratio (v=c), vehicle-kilometers travelled (vkt) and roadway width are the most signicant variables. anincrease in v=c and roadway width will decrease the number of accidents, however, anincrease in vkt and permission to park on the right lane of the roadway can increase thecrash frequency.