基于图深度学习和模糊聚类的深圳市通勤生活圈空间范围与内部结构研究  

Examining the spatial extent and internal structures of commuting-based life circles in Shenzhen based on an integrated model of graph deep learning and fuzzy clustering

作　　者：邹逸 陈梓烽 ZOU Yi;CHEN Zifeng(School of Geography and Planning,Sun Yat-sen University,Guangzhou 510006,China)

机构地区：[1]中山大学地理科学与规划学院,广州510006

出　　处：《地理研究》2024年第9期2392-2410,共19页Geographical Research

基　　金：国家自然科学基金项目(42301222);广东省自然科学基金项目(2023A1515011389)。

摘　　要：在“以人为本”的发展理念下,城市发展越来越强调城市生活空间的优化,作为城市生活空间的重要组成部分的通勤生活圈备受关注。以往研究多从基于地方的形态指标和基于流动的功能指标来对通勤生活圈进行范围划定,但前者较难体现通勤行为带来的流动性,后者较难详尽刻画生活圈内部形态结构。本文提出融合形态与功能视角的方法,利用人工智能技术构建“编码器-解码器”的无监督学习模型(GCN-FCM:Graph Convolution Network-Fuzzy Cluster Model),对深圳市通勤生活圈进行范围划定与内部结构分析。研究发现:①深圳市可划分为10个通勤生活圈。②中心城区和郊区的通勤生活圈的内部结构分别呈现出“核心显著且聚集,外围副中心初现”“核心薄弱且分散,整体相对破碎”的特征。③识别的结果能很好回应学者提出的生活圈理想模型以及印证现实规划中城市空间结构发展目标,并加深了对特大城市通勤生活圈的理解。④与经典的两类方法相比(基于地方的自下而上聚类和基于流动的社区检测),新方法能同步进行通勤生活圈范围划定和内部结构的识别。总体上,新方法体现了人工智能技术支撑下的时空间行为方法论能有效融合空间与行为、融合形态与功能,展现出在通勤生活圈及其他时空间规划中的良好应用前景。The concept of"people-oriented"planning emphasizes the need to understand the interaction between behavior and space and accurately portray urban spatial and structural characteristics from the perspective of micro-residents'behavior,and the commuting-based life circle is one of the most important objects to identify and analyze.Previous studies have identified and portrayed the circles from two approaches:place-based and flow-based.The morphological perspective makes it more difficult to consider the mobility brought by commuting behaviors,while the functional perspective makes it difficult to portray the internal structure of the life circles.Therefore,this paper proposes an artificial intelligence method that integrates the placebased and flow-based methods,which takes into account the characteristics of the built environment on the supply side and the commuting behaviors on the demand side.The method constructs an unsupervised model of"encoder-decoder"(GCN-FCM:Graph Convolution Network-Fuzzy Cluster Model)to delineate the scope and analyze the internal structure of the commuting-based life circles of Shenzhen,and the study finds that:(1)The model divides Shenzhen into 10 major commuting-based life circles,which on the whole shows the structural characteristics of"large scale in the central and western life circles and small scale in the eastern life circles,and dense spatial connection in the western life circles and relatively weak in the eastern life circles";(2)The commuting-based life circles can be classified into two categories according to the differences in their internal structure.One is the southern and central life circles with a"significant and concentrated core and the emergence of the peripheral subcenters",and the other is the northern and eastern suburban life circles with a"weak and dispersed core and a relatively fragmented overall situation";(3)The results can well respond to the ideal model of life circles as well as corroborate the development goal of urban spatial structure in real plann

关 键 词：通勤生活圈范围 城市空间结构 人工智能方法 图卷积网络 时空间行为 

分 类 号：K901[历史地理—人文地理学] C913.32[经济管理]