考虑最优时滞因子时空模型的高速公路短时交通流预测研究

针对高速公路短时交通流预测问题,提出了考虑最优时滞因子的高速公路时空模型。首先通过相关系数计算确定了空间状态变量的个数,随后计算获得每个变量的最优时滞次数,完成空间模型部分的数据预处理,然后分别从连续性时间序列和周期性时间序列两个方面建立时间状态变量,最后分别构建以经典的线性回归和随机森林算法为基学习器的集成预测器LR_Adaboost和RF_Adaboost,利用训练完毕的模型对兰海高速贵阳北收费站路段的实时交通流量进行预测验证,结果证明考虑最优时滞因子的时空模型相对基本的时空模型有更好的预测效果。

Study on time - space model considering optimal delay factor in short - term traffic flow prediction of expressway

For the prediction of short-term traffic flow of the expressway, we propose a spatio-temporal model of optimal delay factor to consider. Firstly, we determine the number of spatial macroscopic parameters through the correlation coefficient, so we have got the optimal delay times for each variable, which completes the data preprocessing of spatial model. Then, we build the temporal macroscopic parameter from two aspects of continuous time series and the periodic time series. Finally, the integrated predictor LR _ AdaBoost and RF _ AdaBoost based on classical linear regression and random forest algorithm are constructed respectively. The model has been used to predict the real-time traffic flow of the Guiyang North toll station at the Lan Hai expressway, which shows that the spatio-temporal model considering the optimal delay factor has better prediction accuracy than the general space-time model.