打印

作者

Ling Cai;Jun Xu;Ju Liu;Ting Ma;Tao Pei;Chenghu Zhou

来源

CITIES,Vol.93,Issue1,Pages 31-42

语言

英文

关键字

Spatial function;Urban dynamics;Spatial-temporal pattern;Tencent location big data;Tensor factorization

作者单位

State Key Laboratory of Resources and Environmental Information System, Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China;College of Resources and Environment, University of Chinese Academy of Sciences, Beijing 100049, China;State Key Laboratory of Resources and Environmental Information System, Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China;College of Resources and Environment, University of Chinese Academy of Sciences, Beijing 100049, China

摘要

Urban spaces have multiple functions, and the main functions of these space change with human activities during a day; thus, there are dynamic semantics of spaces in a city. Knowing the dynamic semantics of urban spaces, which are implied in spatiotemporal patterns of human activities, can help urban planners and managers understand how a city performs over time and space. The very large amount of multidimensional user-generated data makes it possible to disclose the spatiotemporal patterns of human activities from multiple perspectives. In this paper, using Beijing as a case study, we extract the dynamic semantics of urban spaces through the spatiotemporal patterns of human activities discovered from crowdsourced positioning data. A high-order decomposition method, tensor factorization, is used to explore the crowdsourced positioning data. The decomposition results reveal five hourly patterns, four daily patterns and six spatial patterns of urban dynamics in Beijing, showing that urban dynamics in Beijing vary noticeably over different hours, days and space. The human activities implicated by hourly and daily patterns are inferred through empirical knowledge, and the activity semantics of spatial patterns are further disclosed by using the interaction relations among three dimensions stored in the core tensor. The k-means clustering method is executed to aggregate similar spatial units into one group. Five clusters of regions with similar activity semantics are discovered, the function semantics of clusters are clarified with point of interest (POI) data.