基于支持向量机的交通安全预测模型及仿真研究

Research on Simulation and Prediction Model of Traffic Safety Using Support Vector Machine

分析了交通安全预测中指标和影响因子的选择,确定了指标体系和影响因子集合,利用支持向量机(svm)建立交通安全预测模型并进行实例仿真验证。将1953年-2006年全国交通安全相关样本数据分为训练集和测试集,通过训练SVM得到交通安全预测模型参数值,对测试集指标进行预测,计算预测误差,并与BP神经网络预测模型进行对比。仿真结果表明支持向量机建立的交通安全预测模型学习速度快,泛化能力强,有着比神经网络预测模型更高的运算速度与预测精度。

Indexes and influence factors in the prediction of traffic safety were analyzed, at the same time index system and the set of influence factors were decided, then the prediction model of traffic safety and simulate actual examples were estimated by support vector machine （SVM）. During the course of predicting, the samples from 1953 to 2006 were divided into the training set and the testing set, the prediction model of traffic safety was obtained by training SVM, and then index values of testing samples were predicted and predictive errors were calculated, then the BP model was compared with. The experimental simulation illustrates that SVM model has excellent learning ability and good generalization, and compared with the model based on BP neural network it has lower computational cost and higher predictive accuracy.