In this paper, we use an innovative mathematical tool, rough set analysis (RSA), combined with logistic regression modeling to understand the key factors associated with hit and run collisions in Hawaii. After describing the nature of the problem in Hawaii and providing some background on the RSA, we apply the methods to a comprehensive database of police-reported accidents over the period 2002-2005. RSA is utilized to extract the key determinants of hit and run collisions. Using the information from the RSA, we build a logistic regression model to explain the key factors associated with hit and run crashes in Hawaii. We find that factors such as being a male, tourist, intoxicated, and driving a stolen vehicle are strong predictors of hit and run crashes. In addition to the obvious human factors associated with these crashes, there are also some interesting roadway features such as horizontal alignment, weather, and lighting which are also significantly related to hit and run crashes. Some suggestions for reducing hit and run crashes as well as opportunities for additional research are identified.
Abstract