Predicting Highway Crash Frequency: A Review of Statistical Regression Models

Riana Tanzen, Jawad Mahmud Hoque and Md. Sultanul Islam

Mitigating the losses of human lives owing to highway crashes needs better insights to the roadway locations as well as the performance of various safety measures. All these require the analysis of existing crash data for various roadway segments, both deemed “safe” and “unsafe” with different safety measures installed. Statistical models project the safety standards and their suitability into the future. Various tools and modeling techniques or regression models, primarily based on Poisson Regression, are employed to analyze crash data. This paper discusses some common methodological issues which are constraining the effectiveness of various existing models, then assesses the strengths and weaknesses of these models, followed by introducing more promising and advanced approaches that predict results with much more accuracy than the traditional statistical models.