摘要
提出了一种融合多重时空上下文特征的共享单车站点功能区自动识别方法,解决了共享单车OD数据缺乏站点所在区域类型描述信息的问题,从而更好地支撑出行模式分析与规律挖掘。该方法从多源数据中提取共享单车站点周围区域的社会经济、城市形态、骑行流量变化等特征,综合运用规则判定和机器学习方法构建分类模型,最终实现对站点区域功能类型的识别。采用纽约市Citi Bike数据进行实验,结果表明基于反向传播神经网络模型的居住区、商业区、工业区站点分类精度达到89.37%。将该训练模型应用于芝加哥Divvy Bike和华盛顿特区Capital Bikeshare数据,分类精度均高于85%,表明该方法具有较好的泛化能力。根据分类结果,剖析了上述3个城市两个不同时期的出行模式变化和可能的成因。
We proposed an automatic identification method for functional zones of bike-sharing stations by integrating multiple spatio-temporal contextual features.This approach aims to address the lack of description information of areas surrounding bike-sharing stations in origin-destination data,thereby better supporting the analysis and discovery of travel patterns.The method extracts features from various data sources,including socioeconomic characteristics,urban morphology,and cycling traffic variations around bike-sharing stations,and combines rule-based methods with machine learning to construct a classification model for identifying functional zones around these stations.Experiment using Citi Bike data from New York demonstrated that the classification accuracy for residential,commercial,and industrial zones based on the back propagation neural network model reached 89.37%.The trained model was further applied to Divvy Bike data from Chicago and Capital Bikeshare data from Washington,D.C.,achieving a classification accuracy of over 85%,indicating strong generalization ability.Based on the function identification of areas around bike-sharing stations,we analyzed the changes in travel patterns during two different periods in the three cities and explored the possible underlying reasons.
作者
张羿祺
黄浩然
信睿
杨敏
ZHANG Yiqi;HUANG Haoran;XIN Rui;YANG Min(School of Resource and Environmental Sciences,Wuhan University,Wuhan 430079,China;College of Geodesy and Geomatics,Shandong University of Science and Technology,Qingdao 266590,China)
出处
《地理空间信息》
2025年第11期27-32,共6页
Geospatial Information
基金
国家自然科学基金资助项目(42101452)。
关键词
共享单车站点
功能区分类
上下文特征
机器学习
出行模式
bike-sharing station
functional zone classification
contextual feature
machine learning
travel pattern
