Research On The Modeling And Application Of Urban Functional Regions

Modelling and applied analysis urban functional regions is a crucial factor in promoting urban development,and serves as an effective means of gaining deeper insights into the laws governing urban development and supporting planning decisions.However,current challenges in modelling and applied analysis urban functional regions include limited feature expression ability,insufficient objectivity in regionalization scales,and isolated functional units.Furthermore,due to the complex nature of urban spatial structures,research in this area faces significant difficulties,and there is an urgent need for improvements in related technologies and methods.In this dissertation,I conduct a series of exploratory studies on the modeling and applied analysis of urban functional regions.This work uses multiple sources of geographic data within the city to carry out research work from four aspects: "functional feature engineering,functional regionalization,functional attribute classification,and urban vitality correlation analysis," and constructs a methodological framework for research on urban functional regions.The main content of this work includes the following four aspects:i.Addressing the issue of differential expression of functional semantic features,this thesis proposes an improved universal language model based on Point of Interest(POI)data.Previous research has used a general language model for POI feature modeling,ignoring differences in POI category semantics such as functional intensity and service category.Based on this,this thesis constructs an improved POI topic word vector model that takes into account the "polysemy" of POI semantics,and uses analogy reasoning to compare urban spatial elements with natural language processing corpus elements,combined with functional themes and word vector technology,to train topic word vector representations of POI types and achieve clustering analysis of urban functional regions.ii.Addressing the issue of inability to reconcile multiple features of functional region,this thesis proposes a fine-scale functional regionalization method that combines road network structure and group travel interaction features.At present,the spatial embedding network model has limitations such as difficult-to-determine scales and unclear delimitation,while the road network abstraction model is limited to the expression of road network topological features,ignoring the group travel interaction among residents within the city.Based on this,this thesis constructs a spatial embedding network model that combines road network structure and resident travel trajectory,maps the road network into a graph structure,trains word vector representations of road nodes using travel trajectory data,and achieves hierarchical community structure mining based on traffic relevance.iii.Addressing the issue of neglecting spatial correlation in functional attribute modeling,this thesis proposes a graph neural network-based urban functional recognition method that considers spatial correlation characteristics.Current mainstream research on urban functional recognition usually assumes that urban functional units are independent samples,ignoring their spatial correlation characteristics.Based on this,this thesis proposes a street-level urban functional recognition model based on taxi trajectory traffic flow information and graph convolutional neural network,extracts intermediate traffic flow features using massive taxi trajectory data,models them using graph convolutional neural network,aggregates traffic context information of road segments,and improves the accuracy of street-level functional attribute classification.iv.Addressing the inadequate analysis of urban functional structure element correlation,this thesis proposes an application of functional region modeling in urban vitality correlation analysis.Previous research on the correlation between urban functional structure and urban vitality still has shortcomings such as limited spatial scale and insufficient data sources,and the method framework for analyzing the correlation between urban functional structure elements and urban vitality is not yet complete.Based on this,this thesis proposes a method framework for analyzing the correlation of urban vitality based on the construction of functional regions,the expression of functional features,and the evaluation of correlations,and conducts a deep analysis of the overall characteristics of urban vitality and the correlation relationship between urban functional structure elements in the central urban area of Wuhan.In this dissertation,I innovatively propose a methodological framework for modeling and applied analysis of urban functional regions using related technologies from multiple disciplines such as geographic information science,urban planning,complex network theory,and deep learning,which can be promoted in urban planning practice and land use pattern optimization applications.The research methods and results of this thesis can provide detailed information on urban functional regions for urban planning and management,urban sustainable development assessment,and provide scientific basis for macro decision-making,playing an important role in monitoring urban construction and assisting urban planning.