Distance-based urban function colocation analysis method and system
Through the distance-based urban functional isometric analysis method, combined with the improved Lesile-Kronenfeld isometric algorithm and deep learning model, the problem of difficulty in characterizing spatial heterogeneity and ignoring spatial correlation in the existing technology is solved, and the problem of urban functional zoning methods is realized. The multi-scale portrayal and refined analysis of urban functions have improved the decision-making support capabilities of urban planning and management.
Patent Information
- Application Number
- CN202410740102.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-07
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-06-07
AI Technical Summary
The existing urban functional zoning method is difficult to accurately characterize the spatial heterogeneity and mixed characteristics of urban functions, ignores the spatial correlation and interaction between different functional areas, and lacks the portrayal of multi-scale functional structures.
Adopting distance-based urban functional parametric analysis method, by obtaining POI geographical interest data, calculating geographical distance, building a spatio-temporal POI correlation matrix, applying the improved Lesile-Kronenfeld aliasing algorithm and community discovery algorithm, combining deep learning models, building POI correlation The graph divides and adjusts functional areas, and comprehensively considers urban functional semantics, spatial correlations and multi-scale features.
It realizes the accurate portrayal of the complex functional structure and spatial organization model of the city, improves the accuracy, comprehensiveness and practicality of urban functional analysis, and provides refined spatial decision-making support for urban planning and management.
Smart Images

Figure CN118568189B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to urban function zoning technology, and in particular to a distance-based urban function co-location analysis method and system. Background Art
[0002] With the rapid advancement of urbanization, the functional zoning within cities has become increasingly refined and complex. Rationally planning urban functional zoning and optimizing urban spatial structure are of great significance for improving urban operation efficiency and improving the quality of life of residents. Traditional research on urban functional spatial organization mainly relies on manual experience and qualitative analysis, which has limitations such as strong subjectivity and low precision.
[0003] In the prior art, there are some methods that use mathematical statistics and computer technology to analyze the spatial organization of urban functions. For example, spatial cluster analysis based on land use types clusters urban land use data to identify areas with similar land use characteristics, and then divide urban functional zones. Another example is the functional zoning method based on POI data, which infers the dominant functions of different areas of the city by analyzing the spatial distribution characteristics of urban POIs.
[0004] However, the above methods still have some shortcomings. The analysis method based on land use type is difficult to accurately describe the spatial heterogeneity and mixed characteristics of urban functions; although the method based on POI data takes into account the semantic information of urban functions, it ignores the spatial association and interaction between different functional areas. In addition, the existing methods generally focus on the urban functional zoning at a single scale and lack the description of multi-scale functional structures.
[0005] Therefore, there is an urgent need for an analysis method that can comprehensively consider urban functional semantics, spatial associations, and multi-scale characteristics in order to accurately characterize the complex functional structure and spatial organization pattern within the city and provide refined spatial decision-making support for urban planning and management. Summary of the invention
[0006] The embodiment of the present invention provides a distance-based urban function co-location analysis method and system, which can solve the problems in the prior art.
[0007] According to a first aspect of the embodiments of the present invention,
[0008] A distance-based urban functional co-location analysis method is provided, including:
[0009] The POI geographic point of interest data of the city to be analyzed is obtained, wherein the POI data includes the geographic coordinate information and functional attribute information of the POI; based on the obtained POI data, the geographic distance between each POI and other POIs is calculated to obtain a city POI distance matrix; according to a preset distance threshold, the city POI distance matrix is binarized to obtain a spatiotemporal POI association matrix; based on the spatiotemporal POI association matrix, an improved Lesile-Kronenfeld homonymous quotient algorithm is used to calculate the connection strength matrix between POI points;
[0010] Using the community discovery algorithm, based on the connection strength matrix between POI points, the city POIs are divided into multiple communities, and the POIs in each community are clustered geographically; for each community obtained, the functional attribute information of the internal POIs is counted to obtain the functional attribute distribution of the community;
[0011] A deep learning model is used to obtain the functional attribute representation of the target POI, and a POI association graph is constructed based on the functional attribute representation of the target POI. The functional areas of POIs are merged and adjusted according to the preset graph neural network model to obtain the merged urban community division results and the target functional attributes of each community, thus completing the distance-based urban functional co-location analysis.
[0012] In an optional embodiment,
[0013] Based on the acquired POI data, the geographical distance between each POI and other POIs is calculated to obtain the city POI distance matrix; according to the preset distance threshold, the city POI distance matrix is binarized to obtain the spatiotemporal POI association matrix, including:
[0014] According to the timestamp information in the acquired POI data, the time axis is divided into several time windows, and the POI data in each time window constitutes a time slice; for each time slice, a multi-level POI association matrix is constructed according to the preset distance threshold, and the distance matrix is binarized to obtain a multi-level POI association matrix reflecting the spatial association strength of POIs at different distance scales;
[0015] For each pair of POIs, based on their functional attribute information, the POI functional attribute similarity matrix is obtained through a similarity calculation algorithm, and the semantic association matrix is obtained by combining the POI association matrix and the POI functional attribute similarity matrix;
[0016] By integrating the multi-level POI association matrix and the semantic association matrix through the weighted average algorithm, a set of time slice POI association matrices reflecting the comprehensive association strength of POIs in different time slices are obtained;
[0017] The POI association matrices of each time slice are spliced in chronological order to obtain a spatiotemporal POI association matrix, where the rows of the matrix represent POIs, the columns represent time slices, and the matrix elements represent the comprehensive association strength of the corresponding POI with other POIs in the corresponding time slice.
[0018] In an optional embodiment,
[0019] The POI function attribute similarity matrix is obtained through the similarity calculation algorithm. The semantic association matrix is obtained by combining the POI association matrix and the POI function attribute similarity matrix, including:
[0020] The calculation formula of semantic association matrix is as follows:
[0021]
[0022] in, represents the semantic association matrix, i represents the i-th time slice, j represents the distance scale, (p, q) represents a pair of POIs, represents the POI association matrix at the i-th time slice and the j-th distance scale, S i represents the POI functional attribute similarity matrix at the i-th time slice;
[0023] The calculation formula of the time slice POI association matrix is as follows:
[0024]
[0025] Among them, R i represents the time slice POI association matrix, m represents the mth distance scale, ω j represents the weight of the POI association matrix at the jth distance scale, v j represents the weight of the semantic association matrix at the jth distance scale.
[0026] In an optional embodiment,
[0027] The community discovery algorithm is applied to divide the city POIs into multiple communities based on the spatiotemporal POI association matrix. The POIs in each community are clustered geographically. For each community obtained, the functional attribute information of the POIs within it is counted to obtain the functional attribute distribution of the community, including:
[0028] Based on the spatiotemporal POI association matrix, a multi-level POI association network is constructed according to different distance thresholds. A multi-level community discovery algorithm is applied to the multi-level POI association network to identify cross-level community structures. A geospatial adjacency matrix is constructed and introduced as a constraint in the community discovery algorithm. The boundaries of each community are optimized and smoothed, and the city POIs are divided into multiple communities.
[0029] Use the spatial clustering algorithm to perform spatial clustering and boundary optimization on the POIs in the community, obtain the community optimized boundary, calculate the mean and standard deviation of the geographic coordinates of the POIs in the community, and determine the degree of spatial aggregation of the POIs;
[0030] The kernel density estimation algorithm is used to identify high-density POI clustering areas, count the number and proportion of POIs of different functional types in the community, identify the dominant functions and functional mixed characteristics of the community, and obtain the functional attribute distribution of each community.
[0031] In an optional embodiment,
[0032] The geospatial adjacency matrix is introduced as a constraint in the community discovery algorithm to optimize and smooth the boundaries of each community, including:
[0033] The objective function of the multi-level community discovery algorithm after introducing geographic space constraints is as follows:
[0034]
[0035] Where L represents the number of layers of the POI association network, s represents the sth POI, t represents the tth POI, l represents the lth layer POI association network, N represents the number of POIs, represents the adjacency matrix of the POI association network at the lth layer, represents the probability that the s-th POI and the t-th POI belong to the same community in the l-th layer network, represents the marginal probability distribution of the s-th POI in the l-th layer network, represents the marginal probability distribution of the t-th POI in the l-th layer network, λ represents the weight coefficient of the geographic space constraint term, S st represents the geospatial adjacency matrix, represents the community belonging vector of the s-th POI in the l-th layer network, represents the community belonging vector of the t-th POI in the l-th layer network;
[0036] Use the spatial clustering algorithm to perform spatial clustering and boundary optimization on the POIs in the community to obtain regular and compact community boundaries, calculate the mean and standard deviation of the geographic coordinates of the POIs in the community, and determine the degree of spatial aggregation of the POIs, including:
[0037] For each community, the mean and standard deviation of the geographic coordinates of the POIs within it are calculated, and the spatial clustering function is used to determine the degree of spatial aggregation of the POIs. The calculation formula is as follows:
[0038]
[0039] Among them, K k (r) represents the spatial aggregation degree of POIs in the community, C k Represents community, |C k | indicates community C k The number of POIs in the system, I(·) represents the exponential function, μ k Indicates community C k The density of POIs within the network.
[0040] In an optional embodiment,
[0041] A deep learning model is used to obtain the functional attribute representation of the target POI. A POI association graph is constructed based on the functional attribute representation of the target POI. The functional areas of POIs are merged and adjusted according to the preset graph neural network model. The merged urban community division results and the target functional attributes of each community are obtained to complete the distance-based urban functional co-location analysis, including:
[0042] Convolutional neural network and long short-term memory network are used to extract features from text and image data respectively. The text and image features are fused through the attention mechanism, and the importance weights of different modal features are automatically learned to obtain the functional attribute representation of the target POI.
[0043] Based on the functional attribute representation of the target POI and the geographic coordinate information of the POI, a POI association graph is constructed. Each POI is regarded as a node in the graph, and edges are constructed according to the geographic distance and functional attribute similarity between each POI. The attribute information of the POI is used as the initial feature of the POI node.
[0044] According to the preset graph neural network model, a clustering algorithm is used to divide the initial features of POI nodes into communities, and the functional areas of POIs are merged and adjusted according to the division results to obtain the results of urban functional area division and the target functional attributes of each community, completing the distance-based urban functional co-location analysis.
[0045] In an optional embodiment,
[0046] According to the preset graph neural network model, a clustering algorithm is used to divide the initial features of POI nodes into communities, and the functional areas of POIs are merged and adjusted according to the division results, including:
[0047] According to the preset graph neural network model, the POI association graph is encoded, and the embedded representation of the node is learned and updated;
[0048] Based on the updated embedding representation of the nodes, a clustering algorithm is used to divide the nodes into communities. According to the division results, the functional areas of POIs are merged and adjusted to obtain the final urban functional area division results.
[0049] In an optional embodiment,
[0050] The method further comprises:
[0051] Based on the business attributes and metadata of POIs, they are initially classified into basic functional categories;
[0052] Using machine learning and expert systems, we further subdivide the functional categories based on the basic functional categories, and further locate the functions by analyzing the names and frequency characteristics of POIs;
[0053] Based on the determined basic functional categories, isotope quotient values are calculated by a multi-scale isotope quotient calculation engine;
[0054] According to the parity quotient value, a city function parity network diagram is constructed, in which nodes represent POIs of different functional types, and edge weights correspond to parity quotient values, reflecting the mutual influence and dependence between functions;
[0055] According to the community detection algorithm, functional clusters are automatically identified and divided to reveal the modular characteristics of the urban functional spatial structure;
[0056] Provides an interactive visualization interface to display the distribution pattern of functional clusters.
[0057] According to a second aspect of the embodiments of the present invention,
[0058] A distance-based urban function co-location analysis system is provided, comprising:
[0059] The first unit is used to obtain POI geographic interest point data of the city to be analyzed, wherein the POI data includes geographic coordinate information and functional attribute information of the POI, and based on the obtained POI data, calculate the geographic distance between each POI and other POIs to obtain a city POI distance matrix; binarize the city POI distance matrix according to a preset distance threshold to obtain a spatiotemporal POI association matrix; and calculate the connection strength matrix between POI points based on the spatiotemporal POI association matrix using an improved Lesile-Kronenfeld homonymous quotient algorithm;
[0060] The second unit is used to apply the community discovery algorithm, based on the connection strength matrix between POI points, to divide the city POIs into multiple communities, and the POIs in each community are clustered geographically; for each community obtained, the functional attribute information of the internal POIs is counted to obtain the functional attribute distribution of the community;
[0061] The third unit is used to use a deep learning model to obtain the functional attribute representation of the target POI, build a POI association graph based on the functional attribute representation of the target POI, merge and adjust the functional areas of POI according to the preset graph neural network model, obtain the merged urban community division results, and the target functional attributes of each community, and complete the distance-based urban functional co-location analysis.
[0062] According to a third aspect of the embodiments of the present invention,
[0063] An electronic device is provided, comprising:
[0064] processor;
[0065] a memory for storing processor-executable instructions;
[0066] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.
[0067] According to a fourth aspect of the embodiments of the present invention,
[0068] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.
[0069] In this embodiment, by obtaining urban POI data, calculating the geographical distance between POIs, and constructing a spatiotemporal POI association matrix, the spatial distribution characteristics and mutual correlation of POIs in different areas of the city can be accurately characterized. Further application of the community discovery algorithm to cluster POIs can identify POI functional areas that are geographically clustered, revealing the spatial differentiation law of urban functional layout. On the basis of community division of urban POIs, the distribution of functional attributes of POIs within each community is statistically analyzed, and the functional semantic characteristics of different areas of the city can be deeply analyzed. Using a deep learning model to learn POI functional attribute representation can extract high-level functional semantic information from massive heterogeneous POI data and enhance the semantic understanding ability of urban functional analysis. Based on the POI functional attribute representation, a POI association graph is constructed, and the functional areas of POIs are merged and adjusted using a graph neural network model, which can fully consider the spatial association and interaction between different functional areas, overcome the limitations of traditional methods that ignore the mutual influence of functional areas, and improve the comprehensiveness and accuracy of urban functional analysis. By setting different distance thresholds and community division parameters, urban functions can be analyzed at multiple spatial scales. It will greatly improve the accuracy, comprehensiveness and practicality of urban function analysis, and provide important technical support for optimizing urban spatial structure and improving urban governance. BRIEF DESCRIPTION OF THE DRAWINGS
[0070] Figure 1It is a schematic diagram of the process of the distance-based urban function co-location analysis method according to an embodiment of the present invention;
[0071] Figure 2 It is a schematic diagram of the structure of the distance-based urban function co-location analysis system according to an embodiment of the present invention. DETAILED DESCRIPTION
[0072] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0073] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.
[0074] Figure 1 FIG. 1 is a flow chart of a distance-based urban function colocation analysis method according to an embodiment of the present invention. Figure 1 As shown, the method includes:
[0075] S101. Acquire POI geographic point of interest data of the city to be analyzed, wherein the POI data includes geographic coordinate information and functional attribute information of the POI, and calculate the geographic distance between each POI and other POIs based on the acquired POI data to obtain a city POI distance matrix; binarize the city POI distance matrix according to a preset distance threshold to obtain a spatiotemporal POI association matrix; calculate the connection strength matrix between POI points based on the spatiotemporal POI association matrix using an improved Lesile-Kronenfeld homonymous quotient algorithm;
[0076] In an optional embodiment,
[0077] Based on the acquired POI data, the geographical distance between each POI and other POIs is calculated to obtain the city POI distance matrix; according to the preset distance threshold, the city POI distance matrix is binarized to obtain the spatiotemporal POI association matrix, including:
[0078] According to the timestamp information in the acquired POI data, the time axis is divided into several time windows, and the POI data in each time window constitutes a time slice; for each time slice, a multi-level POI association matrix is constructed according to the preset distance threshold, and the distance matrix is binarized to obtain a multi-level POI association matrix reflecting the spatial association strength of POIs at different distance scales;
[0079] For each pair of POIs, based on their functional attribute information, the POI functional attribute similarity matrix is obtained through a similarity calculation algorithm, and the semantic association matrix is obtained by combining the POI association matrix and the POI functional attribute similarity matrix;
[0080] By integrating the multi-level POI association matrix and the semantic association matrix through the weighted average algorithm, a set of time slice POI association matrices reflecting the comprehensive association strength of POIs in different time slices are obtained;
[0081] The POI association matrices of each time slice are spliced in chronological order to obtain a spatiotemporal POI association matrix, where the rows of the matrix represent POIs, the columns represent time slices, and the matrix elements represent the comprehensive association strength of the corresponding POI with other POIs in the corresponding time slice.
[0082] In an optional embodiment,
[0083] The POI function attribute similarity matrix is obtained through the similarity calculation algorithm. The semantic association matrix is obtained by combining the POI association matrix and the POI function attribute similarity matrix, including:
[0084] The calculation formula of semantic association matrix is as follows:
[0085]
[0086] in, represents the semantic association matrix, i represents the i-th time slice, j represents the distance scale, (p, q) represents a pair of POIs, represents the POI association matrix at the i-th time slice and the j-th distance scale, S i represents the POI functional attribute similarity matrix at the i-th time slice;
[0087] The calculation formula of the time slice POI association matrix is as follows:
[0088]
[0089] Among them, R i represents the time slice POI association matrix, m represents the mth distance scale, ω j represents the weight of the POI association matrix at the jth distance scale, v jrepresents the weight of the semantic association matrix at the jth distance scale.
[0090] For example, first divide the time window, and divide the time axis into several time windows, such as hours, days, weeks, etc., according to the timestamp information of the POI data. The selection of time windows needs to be determined according to research needs and data characteristics. For example, if the spatiotemporal laws of daily activities of urban residents are analyzed, it can be divided into hours or days; if the long-term evolution of urban functional areas is analyzed, it can be divided into weeks or months. The POIs in each time window constitute a time slice, which reflects the distribution and association status of the POIs in that time period.
[0091] After dividing the time window, for each time slice, a multi-level POI association matrix and a semantic association matrix are generated for the corresponding time slice. The multi-level POI association matrix reflects the spatial association strength of POIs in the time slice at different distance scales, and the semantic association matrix considers the impact of the similarity of POI functional attributes on the association strength. Through this step, a set of time slice POI association matrices can be obtained.
[0092] After obtaining the POI association matrix of each time slice, they are spliced in chronological order to form a spatiotemporal POI association matrix. The rows of this matrix represent POIs and the columns represent time slices. Through the spatiotemporal POI association matrix, the changes in the association strength of POIs in different time periods can be intuitively presented, revealing the temporal dynamics of the POI association pattern. For example, by observing the changes in the element values of each row in the matrix, the fluctuations in the association strength of a specific POI with other POIs in different time periods can be analyzed; by comparing the differences between different rows, the similarities and differences in the association patterns of different types of POIs in the time dimension can be found.
[0093] Based on the constructed spatiotemporal POI association matrix, data mining and machine learning methods can be used to deeply analyze the spatiotemporal evolution of POI association patterns;
[0094] Specifically, the association strength of each POI at different time slices is regarded as a time series, and time series analysis methods such as Fourier transform and wavelet analysis are used to extract the periodic characteristics of POI association strength and find the time period regularity of POI association pattern. For example, by analyzing the time series of association strength of POI in commercial areas, it may be found that it presents obvious daily and weekly cycles, reflecting the daily and weekly rhythms of commercial activities.
[0095] Using methods such as frequent pattern mining and sequential pattern mining, we can find high-frequency co-occurrence POI association patterns from the spatiotemporal POI association matrix and reveal the association rules of different types of POIs in the spatiotemporal dimension. For example, through frequent pattern mining, we may find that residential POIs and commercial POIs frequently co-occur during peak hours in the morning and evening, reflecting the consumption needs of residents on their way to and from get off work.
[0096] Cluster analysis is performed on the spatiotemporal POI association matrix to aggregate POIs with similar spatiotemporal association patterns into one category, revealing the differences and evolution patterns of urban functional areas in different time periods. For example, by performing spectral clustering on the spatiotemporal POI association matrix, it can be found that there are multiple dynamic functional areas in the city, such as commercial office areas active during the day and entertainment and leisure areas active at night, reflecting the spatiotemporal differentiation characteristics of urban functions.
[0097] By using anomaly detection algorithms, such as local anomaly factor (LOF) and isolated forest, we can identify abnormal POI association patterns from the spatiotemporal POI association matrix and discover abnormal events and emergencies in the spatiotemporal evolution of urban functions. For example, through anomaly detection, we can find abnormal situations such as large-scale gathering or evacuation of people in the city, providing decision support for urban emergency management.
[0098] In this embodiment, by setting different distance thresholds, a multi-level POI association matrix can be constructed at different distance scales. This method can reveal the spatial association strength between POIs from different spatial granularity angles, and provide a more comprehensive and three-dimensional POI association relationship representation. By integrating the semantic association matrix with the spatial association matrix, a comprehensive association matrix that comprehensively considers the spatial proximity and functional similarity of POIs can be obtained. This method can more comprehensively characterize the association relationship between POIs, not only considering the geographic spatial distribution of POIs, but also considering the functional attribute characteristics of POIs, and providing a more semantically rich POI association representation. By analyzing the spatiotemporal POI association matrix, the spatiotemporal evolution law of urban functional areas can be revealed, and the differences in POI association patterns in different periods can be found, providing an important basis for understanding the dynamic development of cities. The spatiotemporal POI association matrix provides a quantitative measure of the comprehensive association strength between POIs, and contains rich prior information in spatial, semantic and temporal dimensions. This prior information can be used as an important input for community discovery algorithms and functional area division algorithms, guiding the algorithms to optimize and adjust in spatial, semantic and temporal dimensions, and improving the accuracy and rationality of community division and functional area identification.
[0099] S102. Apply a community discovery algorithm to divide the city POIs into multiple communities based on the connection strength matrix between POI points, and the POIs in each community are geographically clustered; for each community obtained, the functional attribute information of the POIs within it is counted to obtain the functional attribute distribution of the community;
[0100] Through the community discovery algorithm, a community division result can be obtained, in which each POI is assigned to a specific community.
[0101] In an optional embodiment,
[0102] The community discovery algorithm is applied to divide the city POIs into multiple communities based on the connection strength matrix between POI points. The POIs in each community are clustered geographically. For each community obtained, the functional attribute information of the POIs within it is counted to obtain the functional attribute distribution of the community, including:
[0103] Based on the connection strength matrix between POI points, a multi-level POI association network is constructed according to different distance thresholds. A multi-level community discovery algorithm is applied to the multi-level POI association network to identify cross-level community structures. A geospatial adjacency matrix is constructed and introduced as a constraint in the community discovery algorithm. The boundaries of each community are optimized and smoothed, and the city POIs are divided into multiple communities.
[0104] Use the spatial clustering algorithm to perform spatial clustering and boundary optimization on the POIs in the community, obtain the community optimized boundary, calculate the mean and standard deviation of the geographic coordinates of the POIs in the community, and determine the degree of spatial aggregation of the POIs;
[0105] The kernel density estimation algorithm is used to identify high-density POI clustering areas, count the number and proportion of POIs of different functional types in the community, identify the dominant functions and functional mixed characteristics of the community, and obtain the functional attribute distribution of each community.
[0106] In an optional embodiment,
[0107] The geospatial adjacency matrix is introduced as a constraint in the community discovery algorithm to optimize and smooth the boundaries of each community, including:
[0108] The objective function of the multi-level community discovery algorithm after introducing geographic space constraints is as follows:
[0109]
[0110] Where L represents the number of layers of the POI association network, s represents the sth POI, t represents the tth POI, l represents the lth layer POI association network, N represents the number of POIs, represents the adjacency matrix of the POI association network at the lth layer, represents the probability that the s-th POI and the t-th POI belong to the same community in the l-th layer network, represents the marginal probability distribution of the s-th POI in the l-th layer network, represents the marginal probability distribution of the t-th POI in the l-th layer network, λ represents the weight coefficient of the geographic space constraint term, S st represents the geospatial adjacency matrix, represents the community belonging vector of the s-th POI in the l-th layer network, represents the community belonging vector of the t-th POI in the l-th layer network;
[0111] Use the spatial clustering algorithm to perform spatial clustering and boundary optimization on the POIs in the community to obtain regular and compact community boundaries, calculate the mean and standard deviation of the geographic coordinates of the POIs in the community, and determine the degree of spatial aggregation of the POIs, including:
[0112] For each community, the mean and standard deviation of the geographic coordinates of the POIs within it are calculated, and the spatial clustering function is used to determine the degree of spatial aggregation of the POIs. The calculation formula is as follows:
[0113]
[0114] Among them, K k (r) represents the spatial aggregation degree of POIs in the community, C k Represents community, |C k | indicates community C k The number of POIs in the system, I(·) represents the exponential function, μ k Indicates community C k The density of POIs within the network.
[0115] Exemplarily, firstly, a city interest point association matrix is constructed as the input of the community discovery algorithm, and a multi-level community discovery algorithm is applied to simultaneously identify the community structure of interest points at multiple scales to obtain a comprehensive community division result;
[0116] Specifically, a multi-level interest point association network is constructed. According to different distance thresholds or association strength thresholds, multiple network levels reflecting the association relationship of interest points at different scales are constructed. A multi-level community discovery algorithm is applied to the multi-level interest point association network. The network structure of multiple levels is considered at the same time to identify the cross-level community structure.
[0117] According to the geographic coordinate information of POI, the spatial adjacency relationship between POI is calculated and the geographic spatial adjacency matrix is constructed. The adjacency matrix represents the spatial proximity between POIs, and the adjacency relationship can be determined using the distance threshold or K nearest neighbor method.
[0118] Adding the geospatial adjacency matrix as a constraint encourages the algorithm to group spatially adjacent POIs into the same community, improving the spatial continuity and compactness of the community. Indicators such as spatial modularity can be used to measure the spatial quality of community division results.
[0119] After obtaining the community division results, the boundaries of each community are optimized and smoothed to make the community more continuous and compact in geographic space. Spatial clustering algorithms can be used to perform spatial clustering and boundary optimization on the POIs of the community to obtain more regular and beautiful community boundaries.
[0120] Calculate the mean and standard deviation of the geographic coordinates of POIs in the community to determine the degree of spatial aggregation of POIs; use spatial statistical methods such as kernel density estimation to further characterize the spatial distribution pattern of POIs in the community and identify high-density POI aggregation areas;
[0121] Count the number and proportion of POIs of different functional types in the community, identify the dominant functions and mixed functional characteristics of the community; analyze other attribute information such as the business composition and brand distribution of points of interest in the community, and comprehensively characterize the functional attribute characteristics of the community.
[0122] Optionally, construct a POI spatiotemporal dataset, divide the POI data into time windows to represent the urban functional status in different periods, conduct community discovery and functional attribute analysis in different time periods, and obtain community division results and functional attribute distribution characteristics in different periods; analyze the spatiotemporal dynamic evolution pattern of community functions, identify the changing trends, stability and mutation points of community functions, and use time series analysis methods to characterize the regularity and abnormality of community function evolution; combine urban development background information to explain the driving factors and mechanisms of community function evolution, analyze the impact of different factors on community function evolution, and provide an explanation and prediction basis for community function evolution.
[0123] In this embodiment, by constructing a multi-level POI association network and applying a multi-level community discovery algorithm, the cross-level community structure of urban POI can be identified under different distance thresholds. This method can divide urban POI into communities from different spatial scales, reveal the hierarchical organizational characteristics of urban functional areas, and provide more comprehensive and systematic POI community division results. Introducing the geospatial adjacency matrix as a constraint in the community discovery algorithm, optimizing and smoothing the boundaries of each community, can effectively improve the geospatial continuity and compactness of the community division results. Through the spatial aggregation degree index, the compactness of POI distribution in the community can be judged, and the highly concentrated core functional areas can be identified, providing a basis for analyzing the spatial organization mode of urban functions. By calculating the spatial aggregation degree, functional type proportion, functional mixing degree and other indicators of POI in the community, quantitative indicators of multiple dimensions are provided for quantitative evaluation of community characteristics. The community division results obtained based on multi-level community discovery and spatial clustering optimization have clear boundaries and regular shapes, which are easy to explain and understand. Combined with the functional attribute distribution information of the community, intuitive visualization charts and maps can be generated to show the spatial distribution, dominant functions, functional mixture and other characteristics of urban functional areas, providing intuitive and explainable decision support for urban planning and management departments.
[0124] S103. Use a deep learning model to obtain the functional attribute representation of the target POI, build a POI association graph based on the functional attribute representation of the target POI, merge and adjust the functional areas of the POI according to the preset graph neural network model, obtain the merged urban community division results, and the target functional attributes of each community, and complete the distance-based urban functional co-location analysis.
[0125] Among them, POI functional attribute representation refers to the vector representation obtained through deep learning model learning, which is used to characterize the functional attribute characteristics of each POI. It is the encoding and abstraction of POI functional attribute information, mapping multi-source heterogeneous data such as POI text description and image information into a compact vector space.
[0126] The POI association graph is a data representation based on a graph structure, which is used to describe the association relationship between POIs in a city.
[0127] In an optional embodiment,
[0128] A deep learning model is used to obtain the functional attribute representation of the target POI. A POI association graph is constructed based on the functional attribute representation of the target POI. The functional areas of POIs are merged and adjusted according to the preset graph neural network model. The merged urban community division results and the target functional attributes of each community are obtained to complete the distance-based urban functional co-location analysis, including:
[0129] Convolutional neural network and long short-term memory network are used to extract features from text and image data respectively. The text and image features are fused through the attention mechanism, and the importance weights of different modal features are automatically learned to obtain the functional attribute representation of the target POI.
[0130] Based on the functional attribute representation of the target POI and the geographic coordinate information of the POI, a POI association graph is constructed. Each POI is regarded as a node in the graph, and edges are constructed according to the geographic distance and functional attribute similarity between each POI. The attribute information of the POI is used as the initial feature of the POI node.
[0131] According to the preset graph neural network model, a clustering algorithm is used to divide the initial features of POI nodes into communities, and the functional areas of POIs are merged and adjusted according to the division results to obtain the results of urban functional area division and the target functional attributes of each community, completing the distance-based urban functional co-location analysis.
[0132] For example, for each POI, its related text description and image data are collected, and a pre-trained convolutional neural network (such as ResNet, VGGNet, etc.) is used to extract features from the image data to obtain an image feature vector, and a pre-trained long short-term memory network is used to extract features from the text description to obtain a text feature vector. The image features and text features are fused through the attention mechanism, and the importance weights of different modal features are automatically learned to obtain the functional attribute representation vector of the POI.
[0133] Each POI is regarded as a node in the graph, and the initial feature of the node is the functional attribute representation vector of the POI. The geographical distance between each pair of POIs is calculated, and the edges between the POIs are constructed according to the preset distance threshold. The functional attribute similarity between each pair of POIs is calculated, and the edges between the POIs are constructed according to the preset similarity threshold. The geographical coordinate information and functional attribute representation of the POI are used as the attribute information of the node to complete the POI association graph.
[0134] Select a suitable graph neural network model, such as graph convolutional network (GCN), graph attention network (GAT), etc., take the POI association graph as input, use the graph neural network model to update and aggregate the node features, and learn the high-order feature representation of the nodes by stacking multiple layers of graph neural networks. Use supervised learning to take known urban functional area labels as training targets to optimize the parameters of the graph neural network model.
[0135] Clustering algorithms (such as k-means, DBSCAN, etc.) are used to divide the node features output by the graph neural network into communities. Based on the community division results, the functional areas of POIs are merged and adjusted to obtain preliminary urban functional area division results. The spatial adjacency and functional attribute similarity of POIs in the community are considered, the functional area boundaries are optimized and smoothed, and the functional attribute distribution of POIs in each community is statistically analyzed to determine the dominant functions and target functional attributes of the community.
[0136] Optionally, visualize the merged urban functional zoning results, generate an urban functional zoning map, analyze the functional attribute differences and similarities between different communities, identify co-location areas with similar functional attributes, calculate the functional mixture index of each community, evaluate the diversity and complexity of community functions, and generate an urban functional co-location analysis report, including community functional attribute descriptions, co-location area identification results, and functional mixture evaluation.
[0137] In an optional embodiment,
[0138] According to the preset graph neural network model, a clustering algorithm is used to divide the initial features of POI nodes into communities, and the functional areas of POIs are merged and adjusted according to the division results, including:
[0139] According to the preset graph neural network model, the POI association graph is encoded, and the embedded representation of the node is learned and updated;
[0140] Based on the updated embedding representation of the nodes, a clustering algorithm is used to divide the nodes into communities. According to the division results, the functional areas of POIs are merged and adjusted to obtain the final urban functional area division results.
[0141] Exemplarily, the trained graph neural network model is applied to the POI association graph, and forward propagation is performed on each POI node to obtain the node embedding representation output by the graph neural network model as the updated POI node feature. The updated node embedding representation integrates the original functional attribute information and graph structure information of the POI node, and can better characterize the role and importance of the POI node in the association graph.
[0142] Select a suitable clustering algorithm, such as k-means, DBSCAN, spectral clustering, etc., to cluster the updated POI node embedding representation. According to the clustering results, the POI nodes are divided into different communities, and the POI nodes in each community are closer in the embedding space.
[0143] Analyze the POI nodes in each community, count the distribution of the functional attributes of the POIs in the community, and identify the dominant functions of the community. According to the dominant functions and spatial adjacency of the community, merge adjacent communities with similar functions to form a larger functional area, and optimize and smooth the boundaries of the merged functional areas to make the boundaries of the functional areas more regular and compact.
[0144] In this embodiment, the semantic information and visual information of POI are learned through the deep learning model to automatically fuse text and image features, and a more comprehensive and accurate representation of POI functional attributes is obtained. By constructing a POI association graph, the geographic spatial relationship and functional attribute similarity between POIs are fully considered to provide more abundant and effective information for functional area division. The high-order feature representation of POI nodes is learned by using a graph neural network model to fully explore the complex association relationship between POIs and improve the accuracy and reliability of functional area division. Through clustering algorithms and functional area merging optimization, continuous, compact, and functionally similar urban functional areas are obtained to achieve high-quality division of urban spatial structure. Generate urban functional co-location analysis results, provide community functional attribute description, co-location area identification, functional mixing degree evaluation and other information, provide decision support for urban planning and management departments, promote the optimal configuration and coordinated development of urban functions, achieve high-precision, explainable urban functional area division and co-location analysis, and provide important decision support for urban spatial structure optimization and functional area coordinated development.
[0145] In an optional embodiment,
[0146] The method further comprises:
[0147] Based on the business attributes and metadata of POIs, they are initially classified into basic functional categories;
[0148] Using machine learning and expert systems, we further subdivide the basic functional categories into subcategories, and further locate the functions by analyzing the names and frequency characteristics of POIs;
[0149] Based on the determined basic functional categories, isotope quotient values are calculated by a multi-scale isotope quotient calculation engine;
[0150] According to the parity quotient value, a city function parity network diagram is constructed, in which nodes represent POIs of different functional types, and edge weights correspond to parity quotient values, reflecting the mutual influence and dependence between functions;
[0151] According to the community detection algorithm, functional clusters are automatically identified and divided to reveal the modular characteristics of the urban functional spatial structure;
[0152] Provides an interactive visualization interface to display the distribution pattern of functional clusters.
[0153] In another embodiment provided by the present invention,
[0154] Based on the business attributes and metadata of POIs, they are initially classified into basic functional categories, such as residential, commercial, office, leisure, etc.;
[0155] By using machine learning and expert systems, we further subdivide functional subcategories based on basic classification, and achieve more accurate functional positioning by analyzing the spatial distribution, frequency characteristics and interaction relationships of POIs.
[0156] In the embodiment of the present invention, in order to ensure the accuracy, richness and completeness of urban function identification, this article divides the POI data more accurately and meticulously. This operation also has a great impact on the calculation results of functional colocation due to the different specific subdivision types of POIs in the same category. For example, Internet companies and steel mills belong to the "company enterprise" category, but their actual surrounding functions are completely different; at the same time, the difference in facility levels will also affect the identification of functional clusters. For example, comprehensive tertiary hospitals and community health centers are both medical facilities, but due to the huge difference in their energy levels, the impact on surrounding functions is very different. Therefore, it is difficult to ignore the differences in specific types, levels and even scales between POIs, and it is necessary to carefully classify POIs based on the above differences.
[0157] The classification aims to achieve the following goals: to make the existing POI classification correspond to industry and national standards such as urban functional land, 15-minute living circle related guidelines, urban residential area planning and design standards (GB50180-2018) and national economic industry classification (GB / T 4754-2017) as much as possible; to describe the functional type more clearly than the original classification while trying not to lose the rich information of the original POI; each type of POI after classification is not too many or too few, so that it can be used for isotope quotient calculation; POI classifications in different years are often inconsistent, and they will be unified by establishing a unified standard; and non-functional data in POIs are deleted.
[0158] To this end, the present invention proposes a "two-step mapping method" to reclassify POIs.
[0159] The first step is mapping between the original sub-category and the middle category and the new functional type. The present invention preliminarily sorts and classifies POIs according to the sub-category and the middle category, and establishes a mapping table between the original sub-category and the middle category and the new functional type. Since the sub-category classifications of different cities are slightly different, after establishing a detailed sub-category mapping table, there will always be some POIs in different cities that are difficult to classify, so the middle category is used to perform a secondary mapping of these POIs;
[0160] The second step is to map the keywords in the POI name to the new function type. This paper extracts an ordered keyword table from the POI name attributes to form a classification judgment. When a keyword in the keyword table appears in the POI name, it is classified into the corresponding class. When multiple words in the keyword table appear, the class corresponding to the keyword ranked first in the table is selected as the type of this POI. Using the keyword table, the remaining POIs are further split.
[0161] Multi-scale homotopy quotient calculation engine: an algorithm component used to quantify the homotopy relationship strength between different functional areas within a city at multiple spatial scales. The engine can:
[0162] Adaptively select appropriate spatial analysis units, such as grids, blocks, or custom geographic boundaries;
[0163] Applying advanced statistical models, based on the improved Leslie-Kronenfeld isotope quotient formula, the isotope quotient value of any two functional types within a certain distance interval is calculated to reflect the closeness of their coexistence and dependence.
[0164] In an embodiment of the present invention, the present invention incorporates the distance between functions into consideration based on the CLQ (colocation quotient) method proposed by Leslie and Kronenfeld, and proposes a distance-based co-location quotient (DBCLQ) calculation method.
[0165] The present invention incorporates the distance between functions based on the CLQ (colocation quotient) method proposed by Leslie and Kronenfeld, and proposes a distance-based co-location quotient (DBCLQ) calculation method. The basic structure of the classic CLQ is a conditional probability formula, and the calculation result reflects the ratio of the actual number of A with B as the co-location function to the expected value when the null hypothesis is randomly distributed. In order to further discuss the relationship between functions A and B at different scales, this method introduces the distance parameter d into the original equation. The formula for calculating the global CLQ within the distance range of d1 to d2 including the distance parameter is as follows:
[0166]
[0167] in To show the isotopic index of A to B in the distance range from d1 to d2, N A Indicates the total number of type A points; N' BWhen A and B are not of the same class, it is the total number of points of class B. When A and B are of the same class, considering that points cannot be related to themselves, the value is (N B -1); N is the total number of POIs, also considering that a point cannot be related to itself (N-1); It indicates the percentage of A points with B as the co-location function of A within the distance range (d1, d2);
[0168] in, The calculation formula is as follows:
[0169]
[0170] B in the formula ij (1, 0) is a judgment function. If the jth point around the i-th point of type A is of type B, then B ij (1, 0) takes 1, otherwise it takes 0. In order to ensure the continuity of distance calculation at different scales, B is not introduced. ij The coefficient function of (1,0). Represents the total number of points within the distance from d1 to d2.
[0171] The above formula calculates The actual meaning of the value is a probability ratio, which can be understood as follows: when A appears, the ratio of the average probability of point B appearing within the distance range of d1 to d2 to the average probability of point B appearing randomly when A does not exist. Then when A appears, the average probability of B appearing within the distance range of 0-50m from the A-type function is twice that when there is no A.
[0172] In terms of distance range selection, considering the radiation scale and calculation cost of general functional associations, this paper chooses to use the distance interval of 50-500m, and forms six spatial weight matrices of 0-50m, 50-100m, 100-200m, 200-300m, 300-400m, and 400-500m according to the Euclidean distance between two points in the city.
[0173] Based on the calculated homology quotient value, a city function homology network diagram is constructed, in which the nodes represent POIs of different functional types, and the weights of the edges correspond to the homology quotient value, reflecting the mutual influence and dependence between functions;
[0174] Implement complex network analysis techniques, such as community detection algorithms, to automatically identify and divide functional clusters and reveal the modular nature of the spatial structure of urban functions;
[0175] Provides an interactive visualization interface to display the distribution pattern of functional clusters, helping decision makers understand and optimize urban planning layout.
[0176] In order to visualize the spatial relationship of functional co-location, this method first constructs a complex network based on functional connections and uses a community detection algorithm to extract functional clusters. The specific method is as follows:
[0177] Each POI point in the space is combined with other POI points within a 500m Euclidean distance to form a point pair set, and the distance between the two is calculated. For example, if the distance between the mth A point and the mth B point is d, then a point pair containing the element {A m ,B n ,d} dataset;
[0178] After the above calculation, each category and a distance interval can be mapped to the corresponding effective CLQ and p value, forming a m ,B n ,CLQ ABd ,p} data set, get point A m and B n The connection strength F between them, CLQ i is the critical value of CLQ. Due to the limitation of calculation amount and presentation effect, CLQ is often i The value of is related to the number of point sets that need to be presented and the type of connection that the research focuses on. In actual operation, it should be compared and adjusted as needed;
[0179]
[0180] Arrange point elements at the real spatial position of POI points, connect each point to form a line with the connection strength F between points, and build a connection network;
[0181] Since the number of points to be classified is extremely large and the network is relatively complex, in order to obtain the global optimal solution while ensuring efficiency, the present invention selects the Leiden algorithm based on CPM (constant potts model) to cluster the above complex network.
[0182] The present invention integrates the "two-step mapping method", multi-scale homology quotient calculation and functional homology network construction technology to achieve in-depth analysis of urban functional spatial organization, showing the following significant beneficial effects:
[0183] Refined urban function identification: The two-step mapping method not only improves the accuracy of POI classification, but also captures subtle differences in urban functions through the refined classification in the second step, providing more accurate basic data for subsequent analysis.
[0184] Enhanced understanding of spatial associations: Multi-scale isotope quotient calculation enables analysis beyond a single spatial level, revealing the interactions and dependencies between different functional areas at multiple scales, and providing a quantitative basis for understanding the complex functional network of cities.
[0185] Promoting the discovery of functional clusters: The construction of functional co-location networks and the ability to identify clusters help to directly gain insight into the aggregation patterns of urban functional areas, and provide intuitive visual support and data support for identifying urban functional centers, marginal areas, and potential development hotspots.
[0186] Auxiliary decision support: Through the application of this system, urban planners and managers can formulate more reasonable and effective urban planning strategies based on scientific data analysis results, optimize resource allocation, and promote the sustainable development of the city and the balanced layout of functional areas.
[0187] Improve research and planning efficiency: Automated and integrated analysis processes reduce manual intervention and improve work efficiency, while ensuring the consistency and repeatability of analysis results, which is conducive to interdisciplinary research and long-term urban monitoring projects.
[0188] In summary, the present invention provides an efficient, accurate and forward-looking analysis tool for urban planning, regional development research and urban management, which effectively promotes the progress of smart city construction.
[0189] Figure 2 FIG. 1 is a schematic diagram of the structure of the distance-based city function co-location analysis system according to an embodiment of the present invention. Figure 2 As shown, the system comprises:
[0190] The first unit is used to obtain POI geographic interest point data of the city to be analyzed, wherein the POI data includes geographic coordinate information and functional attribute information of the POI, and based on the obtained POI data, calculate the geographic distance between each POI and other POIs to obtain a city POI distance matrix; binarize the city POI distance matrix according to a preset distance threshold to obtain a spatiotemporal POI association matrix; and calculate the connection strength matrix between POI points based on the spatiotemporal POI association matrix using an improved Lesile-Kronenfeld homonymous quotient algorithm;
[0191] The second unit is used to apply the community discovery algorithm, based on the connection strength matrix between POI points, to divide the city POIs into multiple communities, and the POIs in each community are clustered geographically; for each community obtained, the functional attribute information of the internal POIs is counted to obtain the functional attribute distribution of the community;
[0192] The third unit is used to use a deep learning model to obtain the functional attribute representation of the target POI, build a POI association graph based on the functional attribute representation of the target POI, merge and adjust the functional areas of POI according to the preset graph neural network model, obtain the merged urban community division results, and the target functional attributes of each community, and complete the distance-based urban functional co-location analysis.
[0193] According to a third aspect of the embodiments of the present invention,
[0194] An electronic device is provided, comprising:
[0195] processor;
[0196] a memory for storing processor-executable instructions;
[0197] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.
[0198] A fourth aspect of the embodiments of the present invention is:
[0199] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.
[0200] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.
[0201] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A distance-based urban functional colocation analysis method, characterized in that: include: The POI geographic point of interest data of the city to be analyzed is obtained, wherein the POI data includes the geographic coordinate information and functional attribute information of the POI; based on the obtained POI data, the geographic distance between each POI and other POIs is calculated to obtain a city POI distance matrix; according to a preset distance threshold, the city POI distance matrix is binarized to obtain a spatiotemporal POI association matrix; based on the spatiotemporal POI association matrix, an improved Lesile-Kronenfeld homonymous quotient algorithm is used to calculate the connection strength matrix between POI points; Using the community discovery algorithm, based on the connection strength matrix between POI points, the city POIs are divided into multiple communities, and the POIs in each community are clustered geographically; for each community obtained, the functional attribute information of the internal POIs is counted to obtain the functional attribute distribution of the community; A deep learning model is used to obtain the functional attribute representation of the target POI, and a POI association graph is constructed based on the functional attribute representation of the target POI. The functional areas of POIs are merged and adjusted according to the preset graph neural network model to obtain the merged urban community division results and the target functional attributes of each community, completing the distance-based urban functional co-location analysis. Based on the acquired POI data, the geographical distance between each POI and other POIs is calculated to obtain the city POI distance matrix; According to the preset distance threshold, the city POI distance matrix is binarized to obtain the spatiotemporal POI association matrix, including: According to the timestamp information in the acquired POI data, the time axis is divided into several time windows, and the POI data in each time window constitutes a time slice; for each time slice, a multi-level POI association matrix is constructed according to the preset distance threshold, and the distance matrix is binarized to obtain a multi-level POI association matrix reflecting the spatial association strength of POIs at different distance scales; For each pair of POIs, based on their functional attribute information, the POI functional attribute similarity matrix is obtained through a similarity calculation algorithm, and the semantic association matrix is obtained by combining the POI association matrix and the POI functional attribute similarity matrix; By integrating the multi-level POI association matrix and the semantic association matrix through the weighted average algorithm, a set of time slice POI association matrices reflecting the comprehensive association strength of POIs in different time slices are obtained; The POI association matrices of each time slice are spliced in chronological order to obtain a spatiotemporal POI association matrix, where the rows of the matrix represent POIs, the columns represent time slices, and the matrix elements represent the comprehensive association strength between the corresponding POI and other POIs in the corresponding time slice; The community discovery algorithm is applied to divide the city POIs into multiple communities based on the connection strength matrix between POI points. The POIs in each community are clustered geographically. For each community obtained, the functional attribute information of the POIs within it is counted to obtain the functional attribute distribution of the community, including: Based on the connection strength matrix between POI points, a multi-level POI association network is constructed according to different distance thresholds. A multi-level community discovery algorithm is applied to the multi-level POI association network to identify cross-level community structures. A geospatial adjacency matrix is constructed and introduced as a constraint in the community discovery algorithm. The boundaries of each community are optimized and smoothed, and the city POIs are divided into multiple communities. Use the spatial clustering algorithm to perform spatial clustering and boundary optimization on the POIs in the community, obtain the community optimized boundary, calculate the mean and standard deviation of the geographic coordinates of the POIs in the community, and determine the degree of spatial aggregation of the POIs; Using the kernel density estimation algorithm, we identify high-density POI clusters, count the number and proportion of POIs of different functional types in the community, identify the dominant functions and mixed functional characteristics of the community, and obtain the functional attribute distribution of each community. A deep learning model is used to obtain the functional attribute representation of the target POI. A POI association graph is constructed based on the functional attribute representation of the target POI. The functional areas of POIs are merged and adjusted according to the preset graph neural network model. The merged urban community division results and the target functional attributes of each community are obtained to complete the distance-based urban functional co-location analysis, including: Convolutional neural network and long short-term memory network are used to extract features from text and image data respectively. The text and image features are fused through the attention mechanism, and the importance weights of different modal features are automatically learned to obtain the functional attribute representation of the target POI. Based on the functional attribute representation of the target POI and the geographic coordinate information of the POI, a POI association graph is constructed. Each POI is regarded as a node in the graph, and edges are constructed according to the geographic distance and functional attribute similarity between each POI. The attribute information of the POI is used as the initial feature of the POI node. According to the preset graph neural network model, a clustering algorithm is used to divide the initial features of POI nodes into communities, and the functional areas of POIs are merged and adjusted according to the division results to obtain the results of urban functional area division and the target functional attributes of each community, completing the distance-based urban functional co-location analysis.
2. The method according to claim 1, characterized in that The POI function attribute similarity matrix is obtained through the similarity calculation algorithm. The semantic association matrix is obtained by combining the POI association matrix and the POI function attribute similarity matrix, including: The calculation formula of semantic association matrix is as follows: in, represents the semantic association matrix, i represents the i-th time slice, j represents the distance scale, (p, q) represents a pair of POIs, represents the POI association matrix at the i-th time slice and the j-th distance scale, S i represents the POI functional attribute similarity matrix at the i-th time slice; The calculation formula of the time slice POI association matrix is as follows: Among them, R i represents the time slice POI association matrix, m represents the mth distance scale, ω j represents the weight of the POI association matrix at the jth distance scale, v j represents the weight of the semantic association matrix at the jth distance scale.
3. The method according to claim 1, characterized in that The geospatial adjacency matrix is introduced as a constraint in the community discovery algorithm to optimize and smooth the boundaries of each community, including: The objective function of the multi-level community discovery algorithm after introducing geographic space constraints is as follows: Where L represents the number of layers of the POI association network, s represents the sth POI, t represents the tth POI, l represents the lth layer POI association network, N represents the number of POIs, represents the adjacency matrix of the POI association network at the lth layer, represents the probability that the s-th POI and the t-th POI belong to the same community in the l-th layer network, represents the marginal probability distribution of the s-th POI in the l-th layer network, represents the marginal probability distribution of the t-th POI in the l-th layer network, λ represents the weight coefficient of the geographic space constraint term, S st represents the geospatial adjacency matrix, represents the community belonging vector of the s-th POI in the l-th layer network, represents the community belonging vector of the t-th POI in the l-th layer network; Use the spatial clustering algorithm to perform spatial clustering and boundary optimization on the POIs in the community to obtain regular and compact community boundaries, calculate the mean and standard deviation of the geographic coordinates of the POIs in the community, and determine the degree of spatial aggregation of the POIs, including: For each community, the mean and standard deviation of the geographic coordinates of the POIs within it are calculated, and the spatial clustering function is used to determine the degree of spatial aggregation of the POIs. The calculation formula is as follows: Among them, K k (r) represents the spatial aggregation degree of POIs in the community, C k Represents community, |C k | indicates community C k The number of POIs in the system, I(·) represents the exponential function, μ k Represents community C k The density of POIs within the network.
4. The method according to claim 1, characterized in that: According to the preset graph neural network model, a clustering algorithm is used to divide the initial features of POI nodes into communities, and the functional areas of POIs are merged and adjusted according to the division results, including: According to the preset graph neural network model, the POI association graph is encoded, and the embedded representation of the node is learned and updated; Based on the updated embedding representation of the nodes, a clustering algorithm is used to divide the nodes into communities. According to the division results, the functional areas of POIs are merged and adjusted to obtain the final urban functional area division results.
5. The method according to claim 1, further comprising: Based on the business attributes and metadata of POIs, they are initially classified into basic functional categories; Using machine learning and expert systems, we further subdivide the basic functional categories into subcategories, and further locate the functions by analyzing the names and frequency characteristics of POIs; Based on the determined basic functional categories, isotope quotient values are calculated by a multi-scale isotope quotient calculation engine; According to the parity quotient value, a city function parity network diagram is constructed, in which nodes represent POIs of different functional types, and edge weights correspond to parity quotient values, reflecting the mutual influence and dependence between functions; According to the community detection algorithm, functional clusters are automatically identified and divided to reveal the modular characteristics of the urban functional spatial structure; Provides an interactive visualization interface to display the distribution pattern of functional clusters.
6. A distance-based urban function co-location analysis system, used to implement the method described in any one of claims 1 to 5, characterized in that: include: The first unit is used to obtain POI geographic interest point data of the city to be analyzed, wherein the POI data includes geographic coordinate information and functional attribute information of the POI, and based on the obtained POI data, calculate the geographic distance between each POI and other POIs to obtain a city POI distance matrix; binarize the city POI distance matrix according to a preset distance threshold to obtain a spatiotemporal POI association matrix; and calculate the connection strength matrix between POI points based on the spatiotemporal POI association matrix using an improved Lesile-Kronenfeld homonymous quotient algorithm; The second unit is used to apply the community discovery algorithm, based on the connection strength matrix between POI points, to divide the city POIs into multiple communities, and the POIs in each community are clustered geographically; for each community obtained, the functional attribute information of the internal POIs is counted to obtain the functional attribute distribution of the community; The third unit is used to use a deep learning model to obtain the functional attribute representation of the target POI, build a POI association graph based on the functional attribute representation of the target POI, merge and adjust the functional areas of POI according to the preset graph neural network model, obtain the merged urban community division results, and the target functional attributes of each community, and complete the distance-based urban functional co-location analysis.
7. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 5 is implemented.
