This study focuses on developing non-linear models for estimating crash frequency of different types on urban arterials with partial access control. These multilane arterials consist of mid-block segments joined together with signalized and unsignalized intersections (or access points). Crashesincluded in the analysis are of three major types; rear-end, angle and head-on crashes. Each crash type is further categorized into mutually exclusive categories based on the roadway element responsible for the crashes: mid-block segment related, signalized intersection related, and access point related crashes. The methodology adopted for predicting crash frequency is genetic programming (GP). GP, which is primarily based on genetic algorithms (GA), uses the concept of evolution to develop models through the processes of crossover and mutation. The GP modeling approach gives the independence for model development without restrictions on distribution of data. The models developed were compared to the basic negative binomial (NB) models. Some of the more interesting findings include that morning and afternoon peak periods are observed to have fewer occurrences of rear-end crashes at all the roadway elements. Higher traffic volume results in increased number of angle crashes. Instances of angle crashes have increased at signalized intersections even at lower maximum posted speed. Higher averagetruck factor increases the instances of head-on crashes on mid-block segments and at signalized intersections.
Abstract