Road Accident Risk Modeling Based on the Machine Learning

Abstract

AbstractRoad traffic accident is the serious socio-economic problem all over the world. The main solution is the road safety decision support system, which is based on road accident risk mathematical modeling. The study analyses the road accident risk due to the constant characteristics of the road, using the machine learning. The study was carried out on the road outside the human settlements A-322. It was analyzed 273 km outside large settlements. The research process included three stages. At the first stage, the correlation analysis was carried out. As the result, 12 main constant road characteristics for further research were identified. At the next stage, cluster analysis was realized using the k-means clustering algorithm. It allowed dividing road sections into 3 clusters and determining their accident rate. Analysis of variance confirmed the significance of the cluster difference. Discriminant analysis was implemented on the final stage. As the result, the classification functions were obtained. Therefore, it is possible to determine the accident risk rate depending on the values of constant road characteristics. The resulting model can be used in the development of road safety decision support system for the roads of the 2nd category or similar roads, as well as the roads design. In addition, the research results can be the basis in the development of traffic safety components for Intelligent Transportation Systems or software products.KeywordsRoad accident risk modelingMachine learningCluster analysisTraffic safety