Evaluation of Injury Severity for Pedestrian–Vehicle Crashes in Jordan Using Extracted Rules

Abstract

Pedestrian safety is a major concern throughout the world because pedestrians are considered to be the most vulnerable roadway users. This paper sought to identify the main factors in pedestrian-vehicle crashes that increase the risk of a fatality or severe injury. Pedestrian-vehicle crashes which occurred in urban and suburban areas in Jordan between 2009 and 2011 were investigated. Extracted rules from Bayesian networks were used to identify factors related to severity of pedestrian-vehicle crashes. To obtain as much information as possible about these factors, three subsets were used. The first and second subsets contain all types of collisions (pedestrian and nonpedestrian), in which the first subset used collision type as a class variable and the second subset used injury severity. The third subset contains pedestrian collisions only and used injury severity as the class variable. The results indicate that when using collision type as the class variable, better performance was obtained and that the following variables increase the risk of fatality or severe injury: roadway type, number of lanes, speed limit, lighting, and adverse weather conditions.