Traffic Incident Analytics

FSR

Crash frequency is probably the most commonly used absolute measure in transportation research to quantify traffic safety. However, it suffers from high variability that is caused by the randomness of crash incidents and the density distribution of the crash incidents observed by clustering algorithms over a varying number of clusters in a large spatial domain—we can call this variability a traffic safety sensitivity. This paper presents a quantitative measure—called fatal severity ratio (FSR)— that reduces the traffic safety sensitivity problem significantly. A new fatal-point concept is first introduced and used for normalizing the crash frequency in the proposed FSR measure of traffic safety. An extensive empirical study is conducted to validate and evaluate the fatal-point concept and the FSR measure using several clustering techniques. The 2015 North Carolina fatal crash data set of Fatality Analysis Reporting System is also adopted in this study. The experimental analysis shows that the traffic safety sensitivity can be significantly reduced and the FSR measure can quantify traffic safety better than the crash frequency measure by managing cluster variabilities.