Investigating Driver Injury Severity in Traffic Accidents Using Fuzzy ARTMAP

Abstract

This paper applies fuzzy adaptive resonance theory MAP (fuzzy ARTMAP) neural networks to analyze and predict injury severity for drivers involved in traffic accidents. The paper presents a modified version of fuzzy ARTMAP in which the training patterns are ordered using the K–means algorithm before being presented to the neural network. The paper presents three applications of fuzzy ARTMAP for analyzing driver injury severity for drivers involved in accidents on highways, signalized intersections, and toll plazas. The analysis is based on central Florida’s traffic accident database. Results showed that the ordered fuzzy ARTMAP proved to reduce the network size and improved the performance. To facilitate the application of fuzzy ARTMAP, a series of simulation experiments to extract knowledge from the models were suggested. Results of the fuzzy ARTMAP neural network showed that female drivers experience higher severity levels than male drivers. Vehicle speed at the time of an accident increases the likelihood of high injury severity. Wearing a seat belt decreases the chance of having severe injuries. Drivers in passenger cars are more likely to experience a higher injury severity level than those in vans or pickup trucks. Point of impact, area type, driving under the influence, and driver age were also among the factors that influence the severity level.