Presenting an integrated model for predicting urban accidents based on spatial structure parameters of the road network (case study approach)

Today, road safety is a great concern for traffic engineers, because accidents have imposed extensive impacts on the quality of people’s life. Therefore, it is very important to predict accident-prone sections of the road network to prevent future accidents and improve traffic safety. One of the most important factors that cause accidents is the spatial structure of the road network. The spatial structure of the road network is related to the arrangement and layout of the road network components, which as a spatial constraint is very influential on urban flows. The novelty of this study is to present a new combination approach to determine the effective spatial structure parameters in predicting accident-prone sections (District 3 of Tehran city). In this regard, the combination of geographically weighted regression (exponential and bi-square kernels) and binary particle swarm optimization algorithm was used. The recommended combination method is suitable for spatial regression problems, because it is compatible with two unique properties of spatial data, i.e. spatial autocorrelation and spatial non-stationarity. The best value of the fitness function (1-R2) for exponential and bi-square kernels was obtained 0.064 and 0.003, respectively. It also found that spatial structure parameters had a significant impact on predicting accident-prone sections in the study area.