Development of Artificial Neural Network Models to Predict Driver Injury Severity in Traffic Accidents at Signalized Intersections

Abstract

The relationship between driver injury severity and driver, vehicle, roadway, and environment characteristics was examined. The use of two well-known neural network paradigms, the multilayer perceptron (MLP) and fuzzy adaptive resonance theory (ART) neural networks, was investigated. The use of artificial neural networks can lead to greater understanding of the relationship between the aforementioned factors and driver injury severity. Accident data for 1997 for the Central Florida area, which consists of Orange, Osceola, and Seminole Counties, were used. The analysis focuses on two-vehicle accidents that occurred at signalized intersections. The MLP neural network has a better generalization performance of 65.6 and 60.4 percent for the training and testing phases, respectively. The performance of the MLP was compared with that of an ordered logit model. The ordered logit model was able to correctly classify only 58.9 and 57.1 percent for the training and testing phases, respectively. A simulation experiment was then carried out to understand the MLP neural network model. Results show that rural intersections are more dangerous in terms of driver injury severity than urban intersections. Also, female drivers are more likely to experience a severe injury than are male drivers. Speed ratio increases the likelihood of injury severity. Drivers at fault are less likely to experience severe injury than are those not at fault. Wearing a seat belt decreases the chance of sustaining severe injuries. Vehicle type plays a role in driver injury severity. Drivers in passenger cars are more likely to experience a greater injury severity level than are drivers of vans or pickup trucks. Finally, drivers exposed to impact at their side experience greater injury severity than those exposed to impact elsewhere.