Identifying Accident-Prone Locations Using Fuzzy Pattern Recognition

Abstract

This paper describes a method to identify accident-prone locations (APLs) based on an assessment of factors that contribute to accidents. Current methods to identify APLs make no distinction between accidents that result from road- and nonroad-related factors. Combining accidents that are treatable and nontreatable by road improvements can be misleading and may lead to a misallocation of funds by road authorities. This paper presents a computerized procedure that uses safety experts' knowledge on classifying accidents into a finite set of categories. In practice, the categories can include any one or a combination of the three basic highway system components: the driver, the vehicle, and the road environment. Realizing the complex interaction of these components within the accident environment, the procedure employs fuzzy pattern recognition techniques for the classification process. Accidents that do not belong to the road environment category are excluded from the identification of APLs. The method is tested using data from the accident database of the British Columbia Ministry of Transportation and Highways. The method and results are described.