Postprint version. Published in 10th World Conference on Transport Research, Istanbul, Turkey, July 4, 2004.
Copyright © 2004 Elsevier.
NOTE: At the time of publication, the author Anurag Pande was not yet affiliated with Cal Poly.
Despite of the recent advances in traffic surveillance technology and ever-growing concern over traffic safety, there have been very few research efforts establishing Jinks between the real-time traffic flow parameters and crash occurrence. This study aims at the identification of the patterns in the freeway loop detector data, which potentially precede traffic crashes. This would have impOltant implications for Advanced Traffic Management Centers (ATMC). ATMCs could then be able to predict the potential for crashes on freeways and take action to reduce this hazard by warning drivers or introducing variable speed limits. Solution approach to this research problem essentially involves classification of traffic speed patterns emerging from the loop detector data. Historical crash and loop detector data from Interstate-4 corridor in Orlando metropolitan area has been used for this study. The classification methodology adopted here is the probabilistic neural network (PNN): neural network implementation of well-known Bayesian-Parzen classifier. The PNN, not being a training based classifier, has strong statistical basis. The inputs to the model found best suitable for classification were logarithms of the coefficient of variation in speed obtained from three stations, namely, station of the crash (i.e. station nearest to the crash location) and two stations immediately preceding it in the upstream direction, during 5 minute time slice of 10-15 minutes prior to the crash time. The results showed that about 70% of the crashes on the evaluation dataset could be identified using the classifiers developed here.
Civil and Environmental Engineering