etc.) was assigned to the closest point on the segment, relative to the original crash location. Using this methodology, approximately 75% of the potentially fixable on-system crashes were relocated. Location Assignment Crashes were assigned to roadways based on both spatial and attribute criteria. First, all crashes were joined to the closest road within 50 ft, but only if the street name reported in the crash data and street name of the roadway were similar (a Levenshtein distance of <= 10). This was to ensure that if a crash was physically close to multiple streets, it would be assigned to the one that it most likely occurred on. Any crash that was not assigned a road from the previous step was then assigned to the closest road within 50 ft, regardless of other criteria. Crashes were assigned to intersections based on both spatial and attribute criteria. For this assignment, only crashes that were identified as intersection or intersection-related were considered. Crashes were joined to the closest intersection within 150 ft, but only when the roadway segment that was assigned in the previous step participated in the intersection (as determined by name). This requirement was used so that crashes were not assigned to adjacent, non-connected intersections. Finally, overall location type was determined based on being an intersection crash or not. Intersection/intersection-related crashes were assigned as ‘Intersection’ and all other crashes were assigned as ‘Segment’.
53
Made with FlippingBook - Share PDF online