交通运输系统工程与信息

交通运输系统工程与信息 ›› 2024, Vol. 24 ›› Issue (6): 145-158.DOI: 10.16097/j.cnki.1009-6744.2024.06.013

• 系统工程理论与方法 • 上一篇    下一篇

考虑拥堵指数的共享单车出行分布预测模型

胡宝雨1，孙钰莹1，苑少伟2，程国柱*1   

  1. 1. 东北林业大学，土木与交通学院，哈尔滨150040；2.广州市交通规划研究院有限公司，广州510030
  • 收稿日期:2024-06-27 修回日期:2024-07-29 接受日期:2024-08-13 出版日期:2024-12-25 发布日期:2024-12-18
  • 作者简介:胡宝雨(1987- )，男，黑龙江宾县人，副教授。
  • 基金资助:
    中国博士后科学基金 (2023M740558)；中央高校基本科研业务费专项资金 (2572023CT21-04)；黑龙江省自然科学基金 (YQ2022E003)。

Travel Distribution Prediction Model for Bike Sharing Considering Congestion Index

HUBaoyu1,SUNYuying1,YUAN Shaowei2,CHENG Guozhu*1   

  1. 1. College of Civil Engineering and Transportation, Northeast Forestry University, Harbin 150040, China; 2. Guangzhou Transportation Planning and Research Institute Co Ltd, Guangzhou 510030, China
  • Received:2024-06-27 Revised:2024-07-29 Accepted:2024-08-13 Online:2024-12-25 Published:2024-12-18
  • Supported by:
    ChinaPostdoctoralScienceFoundation (2023M740558)； Fundamental Research Funds for the Central Universities (2572023CT21-04)； Natural Science Foundation of Heilongjiang Province (YQ2022E003)。

PDF

摘要： 准确的共享单车出行分布预测对城市非机动车交通规划和单车运营调度至关重要。本文从居民的出行目的地决策行为出发，提出考虑POI(兴趣点)的共享单车出行分布单因素预测模型，在此基础上，建立考虑拥堵指数的双因素预测模型及其改进模型。首先，基于深圳市福田区单车骑行OD数据，分析出行分布网络中的聚集现象，并运用社区发现算法将福田区划分为4个交通分区。之后，探究POI、拥堵指数和出行距离对单车出行的影响发现，POI数量与共享单车出行量呈线性正相关关系；同时，拥堵指数与单车出行量表现出显著的正相关关系，特别是在出行量较大的区域，每当拥堵指数增加0.1，单车出行比例会增加6%~7%；出行距离显示出长尾对数分布特征。预测结果显示，在休息日期间，本文所建立的双因素改进模型在4个交通分区内的准确率分别达到了81.2%、79.5%、80.1%和78.9%；在工作日期间，准确率分别为78.7%、76.3%、80.8%和75.5%。相较于辐射模型，预测准确率最高提升了51.1%。

关键词: 城市交通, 出行分布预测模型, 拥堵指数, 共享单车, 社区发现

Abstract: Accurate prediction of bike-sharing trip distributions is critical for urban non-motorized traffic planning and bike-sharing operation scheduling. This paper takes residents' travel destination decision-making behavior as a starting point and proposes a single-factor prediction model for bike-sharing trip distribution considering POI. Based on this model, a two-factor prediction model considering congestion index and its improvement model are developed. Based on bicycle riding data from Futian District, Shenzhen, this study analyzes the clustering phenomenon within the travel OD network. It utilizes a community detection algorithm from complex networks to partition Futian District into four traffic analysis zones (TAZ). The influence of POI, congestion index and travel distance on bike sharing are then analyzed, and it is found that the number of POIs shows a linear positive correlation with the amount of bike sharing trips. At the same time, the congestion index shows a significant positive correlation with the amount of bike sharing trips, especially in the areas with larger travel volumes, whenever the congestion index increases by 0.1, the proportion of bike sharing trips will increase by 6%~7%. The travel distance shows a long-tailed logarithmic distribution characteristics. The prediction results show that during weekends, the accuracy of the two-factor improvement model developed in this paper is 81.2%, 79.5%, 80.1%, and 78.9% for the four TAZs, respectively. During weekdays, the accuracy rates were 78.7%, 76.3%, 80.8%, and 75.5%, respectively. Compared to the radiation model, the prediction accuracy was improved by a maximum of 51.1%.

Key words: urban traffic, travel distribution prediction model, congestion index, bike sharing, community detection

中图分类号: