基于多维时空参数的道路短期交通状态预测

Short-Term Traffic Flow Forecasting Based on Multi-Dimensional Parameters

城市道路交通状态会同时受到时间、空间多维因素的影响.为对城市道路短期交通状态进行比较准确的预测,本文在分析多维时空参数的基础上,构造了基于支持向量机(SVM)的不同维数的道路短期交通状态预测模型,并通过贵阳市中心城区的出租车GPS数据对各种模型的预测精度进行了检验,分析各时空参数对道路交通状态的影响程度.结果表明,基于目标路段先前流量数据及下游路段交通状况的SVM模型具有较高的预测精度.为了进一步分析该模型的性能,将其与线性回归模型和ARMA模型进行了比较,实验结果显示,本文提出的SVM模型具有较好的预测效果,表明该方法是进行道路短期交通状态预测的有效手段.

The traffic condition of urban roads is affected by multi-dimensional factors,such as time and space.To accurately predict the short-term traffic condition of urban roads,this paper analyzes these multi-dimensional temporal and spatial parameters and develops the short-term road traffic prediction models of different dimensions based on the support vector machine（SVM）.Then,the GPS data of taxis in Guiyang is used to test the prediction accuracy of these proposed models and to analyze the impact of each temporal and spatial parameter on traffic condition.The results show higher prediction accuracy of the SVM model based on previous traffic flow of target segment and the traffic condition of downstream segment.To further analyze its performance,the results are compared with those of linear regression and ARMA models.The proposed SVM model is proven to be an effective tool for forecasting short-term traffic condition of urban roads.