基于XGBoost模型的路段交通流量短时预测  

Short-term Traffic Flow Forecasting of Road Based on XGBoost Model

作　　者：蒋源 陈小鸿 胡松华 JIANG Yuan;CHEN Xiaohong;HU Songhua(Chengdu Institute of Planning&Design,Chengdu 610041,China;Institute of Rail Transit,Tongji University,Shanghai 201804,China;Department of Civil&Environmental Engineering,University of Maryland,College Park,Maryland 20742,USA)

机构地区：[1]成都市规划设计研究院,成都610041 [2]同济大学铁道与城市轨道交通研究院,上海201804 [3]马里兰大学帕克分校土木环境工程系,马里兰州20742

出　　处：《武汉理工大学学报（交通科学与工程版）》2024年第1期25-30,36,共7页Journal of Wuhan University of Technology（Transportation Science & Engineering）

基　　金：国家自然科学基金(71734004)。

摘　　要：文中利用上海杨浦区雷达设备采集的城市道路流量数据,基于XGBoost模型对路段流量进行预测.考虑城市道路交通流量的复杂性与随机性,选用包括整体特征、时间相关特征、空间相关特征等31个特征变量,并通过格网搜索对模型主要参数进行调整.结果显示:在不同时间粒度上,XGBoost模型的RMSE精度皆优于其余五个对比模型,且在效率上也具有优势.以5 min为时间粒度时,RMSE值为14.22,MAPE值为0.153,耗时23.84 s.此外,XGBoost具有较高可解释性.通过对不同特征变量的组合预测及特征变量重要度分析发现,以时间粒度为单元,1、2、3阶滞后流量及彼此间的差值可明显提高模型预测精度,随时间粒度增大,流周期性增强,随机性减弱.Based on XGBoost model,the urban road traffic data collected by radar equipment in Yangpu District,Shanghai was used to predict the road traffic.Considering the complexity and randomness of urban road traffic flow,31 characteristic variables including overall characteristics,time-related characteristics and space-related characteristics were selected,and the main parameters of the model were adjusted by grid search.The results show that the RMSE accuracy of XGBoost model is better than the other five comparative models in different time granularity,and it also has advantages in efficiency.When the time granularity is 5 minutes,the RMSE value is 14.22 and the MAPE value is 0.153,which takes 23.84s s.In addition,XGBoost is highly interpretable.Through the combination prediction of different characteristic variables and the analysis of the importance of characteristic variables,it is found that the 1st,2nd and 3rd order lag flows and their differences can obviously improve the prediction accuracy of the model.With the increase of time granularity,the periodicity of flow increases and the randomness decreases.

关 键 词：路段流量 短时预测 机器学习 XGBoost模型 

分 类 号：U412.6[交通运输工程—道路与铁道工程]