Vehicle Accident Risk Prediction Over AdaBoost from VANETs

With the increasing seriousness of traffic safety problems, the threat of vehicle accidents to personal safety is more and more serious, which makes traffic security research have great significance. At the same time, the rapid development of VANETs provides original road safety information for the efficient study of road safety, and has provided us with new ideas for predicting the risk of vehicle accidents. In this paper, for predicting the vehicle accident risk, the trichotomy AdaBoost algorithm was proposed to achieve vehicle accident risk prediction model. Firstly, Synthetic Minority Oversampling Technique (SMOTE) was used to reconstruct the experimental dataset, the minority samples from the study dataset were oversampled and new samples were synthesized, complement missing data and balance sample number of each category in the original dataset. The trichotomy AdaBoost algorithm was respectively used to train a series of weak classifiers from the experimental dataset and then combine them into a strong classifier to achieve the prediction model and the performance of this prediction model was field tested. The test results illustrate that using the prediction model by AdaBoost algorithm in special situation can take the AUC of 0.69 and real-time into account. Through early warning, those provides the ITS and driving safety assistance with theoretical foundation.