考虑共享单车起讫点的出行目的推断方法  

Inference method of travel purpose considering the origin and destination points of shared bicycles

作　　者：闻川 成卫[1] 肖海承[1] 蒋旭宏 WEN Chuan;CHENG Wei;XIAO Haicheng;JIANG Xuhong(School of Traffic Engineering,Kunming University of Science and Technology,Kunming 650000,China;Finance Bureau of Panzhihua,Panzhihua 617000,China)

机构地区：[1]昆明理工大学交通工程学院,云南昆明650000 [2]攀枝花市财政局,四川攀枝花617000

出　　处：《铁道科学与工程学报》2022年第12期3590-3599,共10页Journal of Railway Science and Engineering

基　　金：云南省科技厅省级外专引智项目(YNS2020002)。

摘　　要：为弥补以往居民出行目的推断研究对出行时刻与起点环境因素考虑的不足,对共享单车用户出行目的进行有效推断,提出一种基于共享单车数据和POI(Point of Interest)数据的出行目的推断方法,构建基于重力模型、经验分布拟合和贝叶斯概率规则的共享单车用户出行目的推断模型。该方法首先将原始的共享单车数据集进行清洗,并在上车起点和下车终点周围建立用户最大步行半径,创建目的地区域,以此确定候选的POI类型;其次,将采集到的POI数据进行分类,构建出行目的与相关POI类别的映射关系表;最后,结合起讫点目的地区域内POI类型比例、用户出行时间和用户所处环境等因素,对基本重力模型中的参数进行修正,构建GMOD(gravity model considered origin and destination)模型,以最终确定用户的出行目的。为了验证所提方法的有效性及实用性,通过实地调查和网络问卷收集上海市居民共享单车出行数据,对模型进行精度验证。研究结果表明:提出的GMOD模型成功推断了138名共享单车用户的出行目的,准确率为57.26%,相比于基本重力模型1与基本重力模型2分别提高15.36%和7.47%。同时,GMOD模型对换乘、医疗和教育这类POI占比较低但重要的出行目的有着更好的推断精度,相较基本重力模型2分别提升25%,33.33%和17.40%,说明该方法能够应用于实际数据量较小的共享单车出行目的推断,可以作为出行调查的辅助手段。In order to remedy the lack of consideration of travel time and starting point environmental factors on residents’ travel purpose inference in previous studies, and to effectively infer the travel purpose of shared bicycle users, this paper proposes a travel purpose inference method based on shared bicycle data and POI(Point of Interest) data, and an inference model of shared bicycle users’ travel purpose based on a gravity model,empirical distribution fitting, and Bayesian probability rule is established. Firstly, the original shared bicycle data set is cleaned in this method, and the user’s maximum walking radius around the starting point and the ending point of getting off the bus is established, then a destination area was created and the candidate POI type was determined;secondly, the collected POI data is classified and a mapping relationship table between travel purposes and related POI categories was constructed;finally, combined with factors such as the proportion of POI types in the origin-destination-destination area, the user’s travel time, and the user’s environment, the parameters in the basic gravity model are revised to construct a GMOD(gravity model considered origin and destination)model to determine the user’s travel purpose. In order to verify the effectiveness and practicability of the proposed method, the accuracy of the model is verified through the data collected by Shanghai residents’ bicyclesharing trips through field surveys and online questionnaires. The results show that the GMOD model proposed in this paper successfully infers the travel purpose of 138 shared bicycle users, with an accuracy rate of 57.26%,which is 15.36% and 7.47% higher than that of basic gravity model 1 and basic gravity model 2, respectively. At the same time, the GMOD model has better inference accuracy for less important but important travel purposes such as transfer, medical care, and education, which are 25%, 33.33%, and 17.40% higher than the basic gravity model 2, indicating that the method can

关 键 词：城市交通 目的地推断 重力模型 共享单车 兴趣点 

分 类 号：U491.1[交通运输工程—交通运输规划与管理]