Identification of Driver and Vehicle Characteristics through Data Mining the Highway Crash Data

A crash can be thought of as a system composed of several elements, including drivers and vehicles that continually interact with each other, while a crash database is a record of the errors attributable to different components of the crash system. Learning from mistakes (errors) is important if crashes are to be avoided. With more than one hundred variables related to the drivers, occupants, crash sites and vehicles involved in crashes, the General Estimates System database contains crucial information about the phenomena of crash occurrence. This information can be used to develop crash countermeasures at all levels, including drivers, vehicles and roadways. One of the ways to achieve this objective is to explore the data for any patterns that exist among drivers, vehicles, and roadways. In this study, we identify driver and vehicle characteristics that contributed to their crash involvement. Preliminary analysis was conducted for selection of crash variables that were relevant to drivers and vehicles involvement in crashes. One of the data mining techniques called “principal components analysis” was further used to identify ageand gender-based groups of drivers and body types of vehicles by highlighting their relation with the crash variables. Some of the variables that were considered in this study included distraction, drinking, speeding etc. (at driver level), and vehicle contributing factors, vehicle’s control and the path prior to its initial involvement in the crash (at vehicle level). This in turn helped in identifying the hidden characteristics that may have adversely influenced the driving behavior of drivers and/or running of vehicles, eventually resulting in crashes.