Real-Time Prediction of Freeway Rear-End Crash Potential by Support Vector Machine

Xu Qu,Wei Wang,Wenfu Wang,Pan Liu,David A. Noyce

2012-01-01

Abstract:This paper aims at developing new real-time freeway rear-end crash potential predictors using the support vector machine (SVM) technique. To explore the relationship between rear-end crash occurrences and traffic conditions, historical loop detector data were collected on Interstate-894 in Milwaukee, Wisconsin. The extracted loop data were then aggregated over different stations and time levels to produce explanatory features. Afterwards, a feature selection process, which addresses the interaction between SVM classifiers and explanatory features, was conducted to identify the features that significantly associated with rear-end crashes. Then the selected significant explanatory features over three separate time levels were used to train three SVM models. Another three SVM models using unselected feature inputs were also established to examine the impact of the feature selection procedure on SVM performances. The results showed that the proposed feature selection procedure greatly enhanced the accuracy and generalization capability of SVM models. Moreover, the multi-layer perceptron (MLP) artificial neural network models were developed to evaluate the performance of SVM models against other commonly applied real-time crash potential predictors. In comparison with MLP artificial neural networks, SVM models provided better results in terms of crash prediction accuracy and false positive rate, which affirmed the superior performance of SVM technique in rear-end crash potential prediction analysis.