主管：中国科学院
主办：中国优选法统筹法与经济数学研究会
　　　中国科学院科技战略咨询研究院
中国管理科学

中国管理科学 ›› 2022, Vol. 30 ›› Issue (10): 210-223.doi: 10.16381/j.cnki.issn1003-207x.2020.0635

• 论文 • 上一篇    下一篇

基于TPCBoost模型的新型交通服务定价研究—以纽约网约车为实例

赵雪峰1, 吴伟伟2, 吴德林1, 时辉凝3, 廉莹4, 赵德从5   

  1. 1.哈尔滨工业大学深圳经济与管理学院，广东 深圳518000；2.哈尔滨工业大学经济与管理学院，黑龙江 哈尔滨150001；3.广东外语外贸大学会计学院，广东 广州510006；4.沈阳师范大学信息技术学院，辽宁 沈阳110000；5.安徽大学电子信息工程学院，安徽 合肥230000
  • 收稿日期:2020-04-09 修回日期:2020-09-10 出版日期:2022-10-20 发布日期:2022-10-12
  • 通讯作者: 吴伟伟(1978-)，男(汉族)，河北卢龙人，哈尔滨工业大学经济与管理学院，教授，博士生导师，研究方向：技术管理与创新管理，Email:wuweiwei@hit.edu.cn. E-mail:wuweiwei@hit.edu.cn
  • 基金资助:
    国家自然科学基金资助项目(71472055)；国家自然科学基金资助面上项目(72072047);国家社会科学基金资助重点项目(16AZD0006)；教育部人文社科资助项目(20YJC630090)；黑龙江省哲学社会科学研究规划资助项目(19GLB087)

Research on the Pricing of New Transportation Services Based on TPCBoost Model——Take New York City for Example

ZHAO Xue-feng1, WU Wei-wei2, WU De-lin1, SHI Hui-ning3, LIAN Ying4, ZHAO De-cong5   

  1. 1. School of Economics and Management, Harbin Institute of TechologyShenZhen, Shenzhen518000, China;2. School of Economics and Management, Harbin Institute of Techology, Harbing 150001, China; 3. School of Accounting, Guangdong University of Foreign Studies, Guangzhou 510000, China;4. School of information technology, Shenyang Normal University, Shenyang 110000, China; 5. School of electronic information engineering, Anhui University, Hefei 230000, China
  • Received:2020-04-09 Revised:2020-09-10 Online:2022-10-20 Published:2022-10-12
  • Contact: 吴伟伟 E-mail:wuweiwei@hit.edu.cn

PDF (PC)

316

摘要： 当前新型交通服务定价一般基于单个特征为前提进行定价探讨，缺乏对不同特征及总体定价的宏观研究。本文集成优化多个CART树得到TPCBoost模型，同时利用纽约市网约车搭乘数据为基础，训练、测试TPCBoost模型，并利用模型分析不同特征的关系及对定价的影响，不仅验证了模型具有更强的鲁棒性及稳定性，同时发现：(1)特征非线性影响定价，短距离、短时间、少乘客时定价稳定，长距离、长时间、多乘客时定价波动剧烈；(2)当特定搭乘距离与特定搭乘时间进行组合时，会出现定价敏感期，在定价敏感期时市场竞争白热化，此时特定搭乘时间抑制定价上涨；(3)搭乘人数不直接影响定价，但与其他特征进行组合时会间接影响定价；(4)搭乘距离正影响定价，但在定价敏感期时不直接影响定价；(5)每日搭乘时间周期性影响定价，特别地，每日会出现若干定价转折时间点，其中波峰定价时间点一般多于波谷定价时间点。本文提出的TPCBoost模型经实际数据验证符合定价规律，可以为营运公司、监管部门及乘客的交通决策提供有益参考。

关键词: TPCBoost模型；新型交通服务；定价敏感期；CART树

Abstract: Present studies of the pricing of new transportation services mainly focus on a single feature, and there is a lack of a overall pricing model considering different features, and the effects of the interaction among features on the pricing of new transportation services. Several CART trees are integrated and optimized to construct the TPCBoost model. Then the ride-hailing data of New York City from Google Cloud and Coursera,including the time, distance and number of passengers, are used to train and test the TPCBoost model. The TPCBoost model is also used to analyze the relationships between different features and their influences on pricing. The robustness and stability of the TPCBoost model are verified, and it is found that:(1) features nonlinearly affects pricing, short distance, short time, and few passengers make the price stable, but pricing is volatile under the conditions of long distances, long hours, and many passengers; (2) When a specific ride distance is combined with a specific ride time, a pricing sensitivity period occurs, and market competition heats up in the pricing sensitive period, and the specific ride time inhibits the price rise; (3) The number of riders does not directly affect pricing, but its combination with other features can indirectly influence pricing; (4) Ride distance positively affects the pricing, but it does not directly affect the pricing during the pricing sensitive period; (5) Daily ride time periodically affects pricing. In particular, there are several pricing transitions each day,when the peak pricing time pointsare generally more than the trough pricing time points. For China, online taxi-hailing service has its own features, including fierce competition, large market space, and intelligence. Therefore, according to the research results of this paper, it is shown that Chinese companies can lower their prices by introducing preferential measures when taking special trips, such as time or distance, to show their attraction to passengers.Since there exist pricing trough stage and pricing peak stage, and the peak stage of pricing presents fierce market competition, Chinese enterprises need to adjust pricing accurately and timely at the peak stage; the trough stage of pricing means market competition slows down, and thus Chinese enterprises can properly improve the pricing to obtain revenue profits. When the travel distance and travel time are combined, the new transportation service market in China enters the stage of fierce competition, and Chinese regulators need to monitor pricing fluctuations in the market in real time and prevent wide pricing fluctuations, which leads to the phenomenon that vicious competition destroys the market.The TPCBoost model developed in this paper is verified by the actual data to conform to the pricing law, and can provide a useful reference for the transportation decision-making of operating companies, regulatory departments and passengers.It also contributes to the new transportation services pricing literature by providing a tool with high accuracy and robustness, and by revealing the comprehensive effects of distance, time, and number of passengers.

Key words: TPCBoost model; new transport services; pricing sensitive period; CART trees

中图分类号: