Predicting injury severity levels in traffic crashes: A modeling comparison
Abbreviated Journal Title
J. Transp. Eng.-ASCE
traffic accidents; injuries; neural networks; comparative studies; models; ACCIDENT SEVERITY; Engineering, Civil; Transportation Science & Technology
This paper investigates the use of two well-known artificial neural network (ANN) paradigms: the multilayer perceptron (MLP) and fuzzy adaptive resonance theory (ART) neural networks in analyzing driver injury severity. The objective of this study is to investigate the viability and potential benefits of using the ANN in predicting driver injury severity conditioned on the premise that a crash has occurred. The performance of the ANN was compared to a calibrated ordered probit model. Modeling results showed that the testing classification accuracy was 73.5% for the MLP, 70.6% for the fuzzy ARTMAP, and 61.7% for the ordered probit model. This result indicates a more accurate prediction capability of injury severity for ANN (particularly the MLP) over other traditional methods. The results of the models showed that gender, vehicle speed, seat belt use, type of vehicle, point of impact, and area type (rural versus urban) affect the likelihood of injury severity levels.
Journal of Transportation Engineering-Asce
"Predicting injury severity levels in traffic crashes: A modeling comparison" (2004). Faculty Bibliography 2000s. 4168.