Developing Crash Modification Factors for Interchange Influence Areas on Urban Four-Lane Freeways Using Multivariate Adaptive Regression Splines

摘要

Crash modification factors (CMFs) are used to measure the safety impacts from changesin specific geometric characteristics. Their development has gained much interest following theadoption of CMFs by the recently released Highway Safety Manual (HSM) and SafetyAnalysttool. This paper aims at developing CMFs for interchange influence areas on urban four-lanefreeways in the state of Florida. The need to model freeway interchange areas apart from theirbasic mainlines has been recognized by SafetyAnalyst. Different methods for developing CMFscan be found in the literature. This paper introduces a promising data mining method known asmultivariate adaptive regression splines (MARS) for CMF development. MARS is characterizedby its capability to accommodate the nonlinearity in crash predictors, which can be a mixture ofcontinuous and categorical variables. In addition, MARS allows the impact of more than onegeometric variable to be simultaneously considered when estimating CMFs. Separate CMFswere developed for inside and outside shoulder widths within the interchange influence areas.The results showed that MARS produced CMFs with higher confidence than those using thenegative binomial model. The influential variables identified included the inside shoulder width,outside shoulder width, traffic volume, shoulder type, and speed limit. It was also found thatinside shoulders less than 4 ft and 4 to 6 ft experienced safety improvement in relative to the 4-ftwidth.