MODEL FOR THE PREDICTION OF ACCIDENTS FOR DETERMINING SAFETY STANDARDS OF ROADS

As a commission from the National Swedish Road Administration, the National Swedish Road and Traffic Research Institute has developed a model for the prediction of road accidents, primarily designed for determining safety standard of roads. Empirical data were obtained from the Swedish Road Data Bank and the Accident Register for about 5,500 km of the total main network of approximately 25,00 km. Data on road alignment (degree of curvature and gradient) were available for the investigated part of the main network, which did, however, not include roads with climbing lanes, central reserves, or those parts where overtaking is prohibited. The main type of wearing course is bituminous. Data on motorways and roads marked with the E17 signal were not included. Although the investigation is based on some 7,000 accidents reported to the police. In its present form, the prediction model is only valid for accidents between intersections. The model considers the influence on the number of accidents of road alignment, paved width, speed limit, traffic flow and geographical region. The predicted number of accidents P(O sub T) on a given road section is derived by the model as where i is speed limit, j class of alignment and k paved width while a sub ijk represents model coefficients (estimated accident rates) and T sub ijk represents vehicle mileage. Dividing the predicted number of accidents by the total vehicle mileage for a particular road section, the predicted accident rate is obtained, and by dividing by the length of the road section, the predicted accident density is obtained. The model may be gradually expanded to include additional factors describing the road and traffic environment and also to include intersections. /Author/