(eng): This paper deals with the data analysis of traffic accidents. Traffic accidents can be caused by different reasons, e.g., by watchfulness of a driver, failure of a vehicle, bad structural arrangements, etc. The aim of this paper is to investigate seriousness of incidents in dependence on different circumstances of an accident. Description of these circumstances leads to the use of a high number of different variables (about 50 variables), which are mostly discrete. The majority of statistical methods dealing with discrete variables use a frequency table. This is not suitable for traffic data because of a huge dimension. In this paper, several methods are proposed for solution to the problem with high-dimensional traffic data.