The aim of the work is to highlight road traffic accident patterns in the context of interrelations between road characteristics and a traffic safety threat. The actual data concerning multi-vehicle accidents without pedestrians on non-urban roads in a chosen region of Poland was the subject of the research. The roadway and roadside data at the accident site have been combined with the crash data that define the roadway hazard, i.e. driver’s behaviour, type and accident severity. The data were subject to multivariate segmentation by means of such conceptual grouping techniques as the K-means clustering algorithm and competitive artificial neural networks. The Ward’s method was used as a supporting tool in establishing the final number of accident profiles. Six distinct accident patterns have been recognised, quantified and labelled, where the first, second and third one are typical of rural areas, the fourth and fifth – of built-up areas, and the last one – of intersections. The analysis indicates that apart from threat factors, the following road related features play an important role in road accident profiling tasks: area type and area development level, roadway surface condition, intersection indicator, shoulder type, and also to some extent: lighting conditions, shoulders’ width, and horizontal curve radius. (Author/publisher)
Samenvatting