Abstract:
|
The Center for Advanced Public Safety for the Alabama Department of Transportation reported that 21.2% of the crashes in 2012 across the state of Alabama were intersection related. To mitigate intersection related crashes, it is critical to identity influential intersection factors on traffic accidents and build a more accurate safety performance function (SPF).
Through this paper, a question is addressed -- does the choice of SPF (i.e., statistical or data mining model) impact the perceived effect of a factor on accident frequency? If the answer is no, the simplest and most interpretable model is preferred for crash modeling. Otherwise, practitioners must apply further investigation on which SPF to use for accident modeling. In this paper, factor importance and interaction effects of intersection factors are studied for non-signalized data in the state of Alabama. Various statistical and data mining models are investigated and their predictive performance is discussed. Finally suggestions are made on modeling intersection-related accident data.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.