Identifying contributing factors on occurrence traffic accidents applying in-depth studies and Bayesian neural networks

Abstract

It is not enough to analyse the consequences of traffic accidents for quality consideration of traffic safety, but it is necessary to permanently monitor the traffic safety situation and analyse the causes of traffic accidents. Worldwide, the importance of an in-depth study of traffic accidents has been recognized by the end of the 20th century. Different models of in-depth study analysis of traffic accidents have been developed worldwide as the most comprehensive method for determining the impact factors on traffic accidents. The application of the methodology of in-depth study analysis to a particular traffic accident details the impact of road, vehicle, human, and environment on a traffic accident. In the Republic of Serbia, independent assessment of the impact of the road on traffic accident on state roads have been carried out since 2015, and in the course of 2016, in-depth analysis of traffic accidents for the first time carried out at the territory of the City of Belgrade. Due to the complexity and high cost of limiting the use of this tool, a model for identifying influencing factors using a Bayesian neural network has been developed based on the aforementioned analysis and experience from traffic accident expertise. Based on the results obtained, it can be concluded that the applied network is highly reliable for the recognition of influential factors, as it showed the percentage of recognition of influential factors in deeply analyzed accidents of 74.1%. Based on the applied model, it can be concluded that similar models should be developed for analysis of traffic accidents so the contributing factors could be analyzed in a quality way.