Abstract:

In road safety, the process of organizing road infrastructure network data into homogenous entities is called segmentation. Segmenting a road network is considered the first and most important step in developing a safety performance function (SPF). This article aims to study the benefit of a newly developed network segmentation method which is based on the generation of accident groups applying K-means clustering approach. K-means algorithm has been used to identify the structure of homogeneous accident groups. According to the main assumption of the proposed clustering method, the risk of accidents is strongly influenced by the spatial interdependence and traffic attributes of the accidents. The performance of K-means clustering was compared with four other segmentation methods applying constant average annual daily traffic segments, constant length segments, related curvature characteristics and a multivariable method suggested by the Highway Safety Manual (HSM). The SPF was used to evaluate the performance of the five segmentation methods in predicting accident frequency. K-means clustering-based segmentation method has been proved to be more flexible and accurate than the other models in identifying homogeneous infrastructure segments with similar safety characteristics.

Keywords: