Exploring the influential factors of roadway departure crashes on rural two-lane highways with logit model and association rules mining

Abstract

• Simultaneous presence of multiple factors results in roadway departure crashes. • Animal, no streetlight, drug, alcohol influence rural two-lane road departure. • Associated factors are – animal presence & no streetlight, male driver & 12am – 6am. • Other associations are – male driver & no passenger, use of both drug & alcohol. • Association rule mining facilitates strategic countermeasure development. Roadway departure (RwD) crashes are a major contributor of rural two-lane (R2L) highway crashes and fatalities. For targeted reduction of crashes and fatalities due to roadway departure, a thorough understanding of factors associated with RwD crashes is necessary. This study quantitatively assessed the available pre-crash factors that might influence the RwD crashes by developing a logit model comparing roadway, crash environment, and the vehicle and driver-related characteristics of 122,978 crashes that occurred in Louisiana over thirteen years. With a high prediction accuracy (81.9% area under the receiver operating characteristics curve), the model presented significant individual associations across crash characteristics with RwD crashes on R2L highways, for example – animals on roadways, snow/sleet/hail, 50–55 mph speed limit, AADT of 1,001 to 5,000 vehicles per day, drug intoxication, motorcycles, driving during 12 am to 6 am, curve radius of 501–1,000 ft., absence of streetlight, alcohol intoxication. Investigation on these top factors using association rules mining reveals findings such as – a higher likelihood of RwD crashes can be strongly associated animal presence coupled with the absence of streetlights, male drivers during the early morning (12 am to 6 am), male drivers driving with no passengers, drivers being intoxicated by both drugs and alcohol, etc. Findings from this study are expected to help highway safety specialists not only in identifying and predicting RwD crashes but also in an improved understanding of associated contributing factors leading to the application of proper countermeasures for the strategic reduction of RwD crashes.