Collision Type Categorization Based on Crash Causality and Severity Analysis

The purpose of this paper was to present an empirical inquiry into the categorization of collision types based on contributing factors and severity distribution. This study used Connecticut crash data from selected two-lane roads originated from police reports from 1996 to 2001. K-means cluster analysis methodology was conducted to categorize 10 collision types into 4 groups according to the similar pattern of their contributing factors. The severity distribution of the collision types was then considered to further divide up or combine the categories. The result of this analysis offers an analytical way at categorizing collisions to relate crash risk to causalities and driver’s misbehaviors. It also provides a crash categorization that can lead to more accurate and specific severity prediction.