Hit and run crash analysis using association rules mining

Abstract

Hit-and-run crashes have drawn growing attention because of the severe consequence of the delaying emergency assistance to victims. However, the number of related studies is still limited due to the lack of relevant adequate data. The objectives of this research are to, (1) identify the crash and geometric features, which contribute to hit-and-run crashes, (2) discover how those measures change in the case of the segment and intersection-related crashes. The study applied market basket analysis to mine associations between the crash and geometric features of hit-and-run crashes. Based on the generated rules, the results show single-vehicle crashes are the first common factor of hit-and-run crashes and dark lighting is the second factor. The combination of these two factors was found to clearly associate with more severe crashes. The study also found hit-and-run crashes mostly occurred in urban areas. The rules also show segment-related crashes have higher fatality rates than intersection-related crashes. These findings suggest that improvements such as roadway markings, lighting, and installation of cameras at intersections could help to reduce hit-and-run crashes or detect the hit and run offenders.