共享单车行程识别算法性能研究  

Performance Study of Trip Identifcation Algorithm for Shared Bike

作　　者：孟瑶 MENG Yao(School of Transportation and Logistics,Southwest Jiaotong University,Chengdu 611756,China)

机构地区：[1]西南交通大学,交通运输与物流学院,成都611756

出　　处：《综合运输》2024年第10期131-137,共7页China Transportation Review

摘　　要：受限于共享单车行程数据的获取难度,越来越多研究者开始从共享单车位置数据中推测行程,然而,这种推测的准确性尚未得到充分研究。本文基于成都市摩拜共享单车行程数据,构建位置数据并应用行程识别算法,比较两种常用算法的性能及相关影响因素。结果显示,随着网格尺寸增大,行程识别精度提高,但网格尺寸过小时会导致明显差异。采样间隔延长会降低识别精度,但基于空间的算法受影响较小。建议在需要划分区域的研究中选择较大网格尺寸以减小误差,同时在使用较长采样间隔的位置数据时选择基于空间的算法以获得更准确的行程数据。这些发现对共享单车位置数据的分析提供了有益的洞见,突出了行程识别算法在不同参数下的性能特征。Due to the challenges in obtaining shared bike trip data,an increasing number of researchers have begun to infer trips from shared bike location data.However,the accuracy of such inferences has not been fully investigated.This paper constructs location data based on Mobike shared bike trip data in Chengdu and applies trip identification algorithms to compare the performance of two commonly used algorithms and related infuencing factors.The results show that as the grid size increases,the accuracy of trip identifcation improves,but excessively small grid sizes can lead to signifcant discrepancies.Extending the sampling interval reduces identifcation accuracy,but space-based algorithms are less afected.It is recommended to choose a larger grid size in studies requiring regional division to reduce errors,and to select space-based algorithms when using location data with longer sampling intervals to obtain more accurate trip data.These fndings provide valuable insights into the analysis of shared bike location data and highlight the performance characteristics of trip identifcation algorithms under diferent parameters.

关 键 词：共享单车 位置数据 行程识别 时空分布 数据采样 

分 类 号：U491.1[交通运输工程—交通运输规划与管理]