Motorcycle crash prediction model for signalised intersections

In this paper, a mathematical model to predict the occurrence of motorcycle crashes at four-arm signalised intersections is established. Using data of motorcycle crashes at fifty-two intersections in the Southwestern part of Singapore over the period from 1992 to 1999, the relationship between the likelihood of occurrence of these crashes and the various geometric elements as well as traffic factors of the intersections is investigated. The Poisson regression model and the negative binomial regression model were explored. The negative binomial model was found to be more suitable. An evaluation of model parameters showed that heavy approach traffic volumes, the presence of uncontrolled left-turn lane, larger approach road width, existence of surveillance camera, number of phases per cycle and high imposed approach speed limits will increase the likelihood of motorcycle crashes. On the other hand, a higher number of bus bays, the presence of an acceleration section or exclusive right-turn lane and the average cycle time and the adaptive signal control will decrease the likelihood of crashes.