Crash Avoidance Analysis Using Classification Trees and Random Forests

摘要: Under critical traffic situations leading to motor vehicle accidents, drivers can either “attempt” or “not attempt” avoid crashes. This study investigates the drivers, vehicles, and environments’ characteristics associated with crash avoidance maneuvers (i.e. no avoidance) help develop countermeasures that would mitigate number of maneuvers. Rear-end collisions, head-on angle collisions are analyzed separately using decision trees significance variables on binary response variable (avoidance is determined. Moreover, random forests method, a novel technique in safety studies, employed rank importance drivers/vehicles/environments According results, drivers’ visibility obstruction, physical impairment, distraction maneuver all three types accidents. speed limit rear-end maneuvers, type Passenger cars LTV versus trucks large trucks) correlated