Predicting driver injury severity in freeway rear-end crashes by support vector machine

This study aims at investigating the relationship between driver injury severity level and driver, vehicle, roadway, and environmental factors on the basis of support vector machine (SVM) model. The multi layer perceptron (MLP) artificial neural network model formed the benchmark for evaluating the performance of SVM model. Historical crash data of the Wisconsin State from 1994 to 2009 were used as the data source. The best SVM model provided an overall classification accuracy of 63.4% and 58.6% for the training group datasets and testing group datasets, respectively. By comparing the performance of SVM model with those of MLP models, SVM model demonstrated satisfactory predicting accuracy with less datasets over-fitting, therefore, SVM model is capable of predicting driver injury severity levels in freeway rear-end crash.