Underground loop detectors are one of the most common traffic surveillance apparatus on freeways. Traffic data obtained from these detectors are used for various ITS (Intelligent Transportation Systems) applications such as travel time estimation and incident detection. In the recent past, however, researchers have been interested in proactive applications of these data. These proactive applications primarily involve real-time crash risk assessment based on analysis of traffic surveillance data observed prior to historical crashes. In this study, we have analyzed the crash data from five freeway sections in the Utrecht region of the Netherlands to identify the traffic conditions significantly associated with crash occurrences. Random Forest, a data mining methodology employing multiple classification trees, would be used to analyze the data and identify traffic parameters significantly associated with the binary variable representing crash vs. non-crash. It was found that the turbulence in traffic speeds is related to real-time crash likelihood. The results are promising in that they show the potential of transferring the proactive traffic management approach proposed by Pande and Abdel-Aty (2007) for Interstate-4 in Orlando, FL to freeways in the Netherlands.
Available at: http://works.bepress.com/apande/21/