基于经验模态分解与支持向量机的道路交通状态预测

摘要

In order to accurately predict the traffic state ,proposed prediction model is based on empirical mode decomposition and support vector machine to predict traffic multidimensional speeds .Chongqing city taxi GPS data as an example ,the use of EMD in solving nonlinear ,non‐stationary signals advantage on the velocity data is decomposed to obtain a set of intrinsic mode component and a margin .Then ,for each component separately SVM prediction .Finally ,the predicted results of the respective layers are fused to give a final prediction value .The simulation showed that ,based on the space‐time traffic speed forecast to get more accurate results higher EMD and SVM combination forecasting model precise than SVM model predictions .%为准确预测道路交通状态，提出以经验模态分解和支持向量机组合的预测模型，对道路多维车流的速度进行预测。以重庆市出租车GPS数据为例，利用EMD解决非线性、非平稳信号上的优势，将速度数据进行分解，得到一组本征模态分量和一个余量；然后，利用SVM对各分量进行单独预测；最后，将预测的各层结果进行融合，得到最终的预测值。仿真结果表明：对基于时空的车流速度进行预测能得到更准确的结果，EMD和SVM 组合预测模型比单一SVM模型的预测精确度更高。