Consistently higher number of traffic collisions was observed on Friday as compared to other days of the week in Edmonton, Canada. The collision data was investigated and a collision prediction model was formulated based on three contributing factors: 1) number of daylight hours, 2) number of snowfall hours, and; 3) visibility on daily basis. A negative binomial (NB) model was formulated to establish a relation between the collision frequency and the independent variables. The model accuracy test results were found good. Statistical measures indicate that increased traffic collision is significantly related with increased number of snowfall hours. The proposed NB model could provide a realistic estimate of expected collision frequency for a particular day as a function of number of hours of daylight, number of hours of snowfall and visibility. It is expected that predicted collision frequencies will help the decision maker to quantify the traffic safety of Edmonton.
展开▼