摘要
针对现有的地理流双变量异常聚类方法忽视了时间维度的问题,提出面向地理流的双变量时空扫描统计方法。先构建面向地理流的多尺度时空扫描窗口;通过伯努利模型下的扫描统计量检测窗口中是否存在异常流簇,采用蒙特卡洛模拟方法检验扫描统计量的统计显著性;筛选一系列时空分布无重叠的异常流簇。应用该方法识别厦门市网约车流和巡游车流的时空异常流簇,以发现两类出租车竞争模式的时空格局。结果表明:巡游车流占优簇常发生在凌晨,分布在娱乐、餐饮、住宿等场所;网约车流占优簇常发生在上午或傍晚,分布在办公地点与居住地之间。该方法挖掘的结果能够发现异常流簇准确的时空分布特征,可为城市交通规划提供支持。
In response to the problem that existing bivariate anomaly clustering methods for geographic flows overlook the temporal dimension,this article proposes a bivariate spatiotemporal scan statistics method for geographical flows.Firstly,multi-scale spatiotemporal scanning windows for geographical flows are constructed.Secondly,the scan statistics of Bernoulli model are used to detect anomalous flow clusters in the spatiotemporal scanning window.Thirdly,the Monte Carlo simulation method is used to test the statistical significance of the scan statistics.Finally,a series of bivariate anomalous flow clusters with non-overlapping spatiotemporal distributions are screened.The method proposed in this paper is applied to the detect spatiotemporal anomalous flow clusters of ride-hailing flows and taxi flows in Xiamen City.The results show that the proposed method can reveal the spatiotemporal pattern of the competition mode between taxi flows and ride-hailing flows.The spatiotemporal anomalous clusters in which taxi flows occupy competitive advantages often occur in entertainment,catering and homestays places in the wee hours;The spatiotemporal anomalous clusters in which ride-hailing flows occupy competitive advantages often occur in offices and residences in the morning or evening.The presented method can identify the accurate spatiotemporal distribution of the anomalous clusters,which can provide support for optimizing the allocation of traffic resources.
作者
王钰辉
阳孟杰
周梦杰
周楷淳
WANG Yuhui;YANG Mengjie;ZHOU Mengjie;ZHOU Kaichun(School of Geographical Sciences,Hunan Normal University,Changsha 410081,China;Hunan Key Laboratory of Geospatial Big Data Mining and Application,Changsha 410006,China;Key Laboratory for Urban-Rural Transformation Processes and Effects at Hunan Normal University,Changsha 410081,China)
出处
《测绘科学》
CSCD
北大核心
2024年第1期204-215,共12页
Science of Surveying and Mapping
基金
国家自然科学基金项目(41901314)
湖南省自然科学基金(2023JJ40447)
湖南省教育厅科学研究项目(22C0018,23B0093)
关键词
地理流
双变量异常流聚类
时空扫描统计
时空分布差异
geographical flows
bivariate anomalous flow clusters
spatiotemporal scan statistics
spatiotemporal distribution differences
作者简介
王钰辉(1999—),男,天津人,硕士研究生,主要研究方向为时空数据挖掘。E-mail:wyh@hunnu.edu.cn;通信作者:周梦杰,副教授,E-mail:mengjiezhou@hunnu.edu.cn
