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
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 place-based 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 sub-centers",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 planning,and deepen the knowledge of life circles for megacities;(4)Compared with the place-based approach,the new model can more accurately and rationally classify the commuting-based life circles;and compared with the flow-based approach,the new model can portray the internal structural differences of different life circles.Overall,the GCN-FCM model embodies the integration of spatial and behavioral interactions by the space-time behavioral methodology under the support of artificial intelligence methods,and shows a better application prospect in commuting-based life circles and other spatio-temporal planning.
spatial extent of commuting-based life circlesurban spatial structureartificial intelligence methodsGraph Convolution Networks(GCN)space-time behavior
万方数据
维普
