Research On The Prediction Model Of Traffic Safety Using Support Vector Machine

Recently our country basic installation construction approached perfection. Substantial increase of automobile possessive quantity and highway mileage impelled the economy development, while brought some negative impacts such as the problem that traffic accident rate hold a high level.Generalizing and analyzing the related data of traffic safety in former years and then finding the internal laws of the traffic accident occurrence, at last forecasting the development tendency of traffic accident in existing road condition had important and practical significance to establish the decision of traffic safety planning.This paper analyzed the data characteristics in the traffic safety prediction, described the advantages of support vector machine in the traffic safety prediction, chose the index set and affected factor set and then established the model of traffic safety prediction by the use of support vector machine and least square support vector machine. During the course of predicting, the related data of nationwide traffic safety from 1953 to 2006 was divided into the training set and the testing set and was simulated by the use of MATLAB7.0. The specific parameters of two models were gained by the training of set data based on traffic safety prediction models of support vector machine and least square support vector machine receptively. The prediction results which were gained by predicting testing set compared with practical value and the prediction results of BP neural network prediction model.Simulation results represented that the model of prediction of traffic safety based support vector machine and least square support vector machine had fast learning, strong generative ability and had lower computational cost and higher predictive accuracy than neural network prediction model.