基于Adaboost-CART模型的动卧列车客座率预测  

Seating Rate Prediction for EMU Sleeping Train Based on Adaboost-CART Model

作　　者：王煜 方伟 王亮 薛冰 WANG Yu;FANG Wei;WANG Liang;XUE Bing(Computing Technology Research Institute,China Academy of RailwaySciences Corporation Limited,Beijing 100081,China;China State Railway Group Co Ltd,Beijing 100844,China)

机构地区：[1]中国铁道科学研究院集团有限公司电子计算技术研究所,北京100081 [2]中国国家铁路集团有限公司,北京100844

出　　处：《中国铁路》2019年第10期34-38,共5页China Railway

基　　金：中国铁路总公司科技研究开发计划项目（N2018X009、2018F012）

摘　　要：由于动卧列车运行距离较长,主要竞争对手为同区间航空运输,航空票价水平和动态浮动会对动卧列车客流产生影响,因此从航空票价角度,研究动卧列车客座率。选取CART模型作为弱学习器,通过Adaboost集成学习算法将弱学习器训练为强学习器,即采用Adaboost-CART模型实现对动卧列车客座率的预测。以京沪高铁动卧列车为例,对该方法进行验证,结果表明:利用Adaboost-CART模型能够较好地对动卧列车客座率进行预测,且精度优于单一CART模型和多元回归模型等传统预测方法,验证了Adaboost-CART模型的有效性和可靠性。Due to the long operating length of EMU sleeping train, airline has become its main competitor for the same section. The air fare and its fluctuation will naturally affect the passenger flow of EMU sleeping trains. Therefore, this paper mainly focuses on the seating rate of EMU sleeping trains from the perspective of air fare. In this paper, CART algorithm is selected as a weak learner, and the weak learner is trained as a strong learner through Adaboost integrated learning algorithm to predict the seating rate of EMU sleeping trains. The results show that the Adaboost-CART model could better predict the seating rate of EMU sleeping train on BeijingShanghai HSR, and it’s more accurate than the traditional prediction models such as single CART algorithm and multiple regression, which verifies the validity and reliability of the Adaboost-CART model.

关 键 词：动卧列车 客座率 航空票价 Adaboost-CART模型 集成学习 学习器 

分 类 号：U293.13[交通运输工程—交通运输规划与管理]