基于GBDT算法的高速公路分车型交通流短时预测模型  

Short Term Prediction Model of Expressway Traffic Flow by Vehicle Type Based on GBDT Algorithm

作　　者：张耀方 陈坚[1] ZHANG Yao-fang;CHEN Jian(School of Traffic Transportation,Chongqing Jiaotong University,Chongqing 400074,China)

机构地区：[1]重庆交通大学交通运输学院,重庆市400074

出　　处：《公路》2022年第1期221-227,共7页Highway

基　　金：重庆市教育委员会人文社会科学研究规划项目,项目编号19SKGH208。

摘　　要：为了准确预测高速公路短时交通流量,以制定科学合理的运营管理方案,运用高速公路联网收费数据和外界环境天气数据,实现数据清洗、预处理,挖掘得到日期、时段、车型等有效特征,构建基于GBDT算法的交通流短时预测模型。以成渝高速公路短时交通流预测为实例分析对象。结果表明,预测误差较BP神经网络模型、RF模型、SVM模型分别降低4.43%、0.32%、1.01%,表明模型具有较好的可靠性和有效性。In order to accurately predict the short-term traffic flow of expressways, to formulate scientific & reasonable operation management scheme, in this paper a short-term traffic flow prediction method is put forward based on GBDT algorithm. Through data cleaning and preprocessing, effective features such as week, time period and vehicle type are explored and combined with weather data to predict short-term traffic flow by vehicle type at Chengdu Station of Chengdu-Chongqing Expressway. The results show that, compared with BP neural network model, RF model and SVM model, the results show that the MAPE decreases by 4.43%, 0.32% and 1.01% respectively, which verifies the reliability and effectiveness of this method.

关 键 词：高速公路 交通流量 联网收费数据 预测模型 GBDT算法 

分 类 号：U491.14[交通运输工程—交通运输规划与管理]