Development of highway safety policies by discriminating freeway curve alignment features

In the Republic of Korea, severe crashes with injuries and fatalities have most frequently occurred in freeway sections with combinations of horizontal and vertical curves. To address this issue related to geometric complexness, this study quantitatively verified the adequacy of Korean highway safety policy and identified supplemental strategies required for the freeway sections with horizontal and vertical curve combinations. Based on in-depth road geometric data, this study combined K-means clustering with binomial logit regression to identify the effects of contributing factors on cluster-based crash severity levels. Resultantly, six clusterbased crash severity estimation models were produced. Single vehicle to roadside protective facility and multivehicle crash types were significant to decrease the likelihood of injury crashes for both full-and three cluster-based regression models. Compared with full data model, driver’s age, action, crash occurrence time, and traffic management related factors were significantly identified in only single cluster. The resultant findings show that multidirectional highway safety policies are further required for the current Korean highway safety policy associated with curve alignment features. This study will assist in researchers and field practitioners’ decision-making for future cost-effective highway safety policies provision related to freeway sections with geometric complexness.