基于TPE-XGBoost模型的高速道路出口车流量预测  

Exit Traffic Flow Prediction of Highway Based on TPE-XGBoost Model

作　　者：洪强 谢建国 周跃琪 姚佳娜 HONG Qiang;XIE Jianguo;ZHOU Yueqi;YAO Jiana(Zhejiang Scientific Research Institute of Transport,Hangzhou 310023,China;Jingning She Autonomous County Transportation Development Center,Lishui 323500,China)

机构地区：[1]浙江省交通运输科学研究院,浙江杭州310023 [2]景宁畲族自治县交通运输发展中心,浙江丽水323500

出　　处：《微型电脑应用》2024年第11期6-9,共4页Microcomputer Applications

基　　金：浙江省重点研发计划(2021C01011);浙江省交通运输厅2022年科技计划项目(202209)。

摘　　要：由于家用轿车数量的快速增长以及货客运业务的发展,高速道路交通车流量日益增大。在海量的数据基础上,高速道路出口车流量预测难度增大。为了精准预测高速道路出口车流量,提出一种基于树结构帕尔赞估计器(TPE)超参数优化的梯度提升树(XGBoost)模型来预测高速道路出口车流量。所提模型使用TPE方法寻求XGBoost模型的超参数最优组合解,可实现超参数的快速优化和对车流量的精准预测。实例验证结果表明,相较于梯度增强决策树(GBDT)、随机森林、支持向量机(SVM)等传统机器学习模型,所提出的模型在真实数据上具有更高的拟合程度和更低的误差,R^(2)可达到0.985,RMSE达到0.95。Due to the rapid growth of the number of family cars and the development of freight and passenger transport business,the traffic flow of highway traffic is increasing.On the basis of huge data,it is more difficult to predict the traffic flow of highway exit.In order to accurately predict the traffic flow at the exit of highway roads,this paper proposes a extreme gradient boosting(XGBoost)model based on tree-structured parzen estimator(TPE)hyperparameter optimization to predict the traffic flow at the exit of highway roads.The model uses TPE method to seek the optimal combination of hyperparameters of XGBoost algorithm,which can achieve fast optimization of hyperparameters and accurate prediction of traffic flow.Example verification results show that compared with traditional machine learning models such as gradient boosting decision tree(GBDT),random forest and support vector machine(SVM),the proposed model has a higher degree of fitting and lower error on the real data,with R^(2)up to 0.985 and RMSE up to 0.95.

关 键 词：TPE超参数优化 XGBoost 高速车流量预测 回归算法 

分 类 号：TP3-0[自动化与计算机技术—计算机科学与技术]