基于时空相关性的短时交通流预测模型

Short term traffic flow forecasting model based on temporal-spatial correlation

在智能交通系统的诱导、控制和管理中,实时准确的短时交通流量预测具有重要意义。为提高预测精度,充分分析交通流特性和外部空间关联对预测结果的影响,提出一种短时交通流预测模型——SARIMA-RF模型。利用SARIMA模型良好的线性拟合能力,提取交通流数据中的周期性特征;利用随机森林模型较强的泛化能力,分析交通流的时空相关性,得出预测结果。实验结果表明,该组合模型与单一模型相比具有更高的预测精度,是一种有效的预测方法。

Real-time short-term traffic flow forecast is of great significance in the induction,control and management of intelligent transportation system(ITS).To improve prediction accuracy and fully analyze the impact of traffic flow characteristics and external spatial correlation on prediction results,a short-term traffic flow prediction model,namely SARIMA-RF model was proposed.The periodic features in traffic flow data were extracted using the good linear fitting ability of SARIMA model.The spatio-temporal correlation of traffic flow was analyzed and the forecast result was obtained using the strong generalization ability of stochastic forest model.Experimental results show that the combined model has higher prediction accuracy than single model and it is an effective prediction method.