Multi-source information fusion data space analysis method and system based on knowledge graph

Abstract

The invention provides a multi-source information fusion data spatial analysis method and system based on a knowledge graph, and relates to the technical field of data analysis, and the method comprises the steps: carrying out the quality calibration and fusion of multi-source heterogeneous data, and obtaining standardized fusion data; a city system knowledge graph is constructed, the city system knowledge graph comprises city element nodes with category and hierarchical relations and association strength of the city element nodes, and city function communities are constructed and optimized through a modularity optimization method; an urban element diffusion wave equation model is constructed based on an urban system knowledge graph, an urban element state propagation rule is analyzed, urban function division is performed, an urban system coupling differential equation set model is established to predict urban element state change, an urban system vulnerability index is constructed, and urban system toughness analysis is performed. According to the method, multi-source city data can be effectively fused, the city system knowledge graph is constructed, and decision support is provided for city space planning and management.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and particularly to a multi-source information fusion data space analysis method and system based on a knowledge graph. Background Art

[0002] In recent years, urban system data has shown characteristics such as multi-source heterogeneity, spatio-temporal heterogeneity, and complex associations. Traditional urban data analysis methods mainly focus on the processing and analysis of single data sources, and it is difficult to fully utilize the rich information contained in multi-source data. With the development of knowledge graph technology, data fusion and knowledge representation methods based on semantic associations provide new research ideas for urban system analysis. By modeling entity relationships, the knowledge graph can effectively describe the complex associations between urban elements and provide technical support for the holistic analysis of urban systems.

[0003] However, the existing technologies have the following deficiencies. Existing multi-source data fusion methods often use static methods such as simple weighted averaging or principal component analysis, and fail to consider the spatio-temporal heterogeneity and uncertainty of data, resulting in insufficient reliability of the fusion results; traditional urban functional zoning methods are mainly based on spatial clustering algorithms, lacking in-depth analysis of the semantic associations and evolution laws between urban elements, and it is difficult to reflect the dynamic characteristics of urban functional zoning; existing urban system resilience assessment methods mostly use static indicator systems, ignoring the coupling relationships and propagation effects between urban elements, and unable to accurately depict the vulnerability characteristics of urban systems.

[0004] In summary, the present invention aims to solve the following technical problems: propose a data quality assessment method based on multivariate statistical process control and recursive Bayesian estimation to achieve dynamic fusion of multi-source heterogeneous data; construct a method for constructing an urban knowledge graph based on formal concept analysis and random walk algorithms to reveal the semantic associations and hierarchical structures between urban elements; establish an urban system dynamics model based on diffusion wave equations and coupled differential equations to achieve dynamic identification of urban functional zoning and quantitative assessment of system resilience. Through the organic combination of the above technical solutions, it provides theoretical support and method guidance for the refined management and scientific decision-making of urban systems. Summary of the Invention

[0005] An embodiment of the present invention provides a multi-source information fusion data space analysis method and system based on a knowledge graph, which can solve the problems in the existing technologies.

[0006] In the first aspect of the embodiment of the present invention,

[0007] A multi-source information fusion data space analysis method based on a knowledge graph is provided, including:

[0008] For multi-source heterogeneous data in the urban system, the multivariate statistical process control method is used to establish data quality benchmark thresholds and perform outlier detection to obtain preliminary calibrated data. The least squares registration algorithm is used for the preliminary calibrated data to calculate the spatial transformation matrix to obtain calibrated data in a unified coordinate system; the recursive Bayesian estimation is used to calculate the credibility score of the calibrated data, and the Markov decision process is used for dynamic adjustment to obtain the data fusion weight, and the calibrated data is weighted and fused to obtain standardized fusion data;

[0009] The double-buffer mechanism is used to receive the standardized fusion data and convert it into an initial graph structure. The urban ontology model is constructed based on formal concept analysis and the nodes of the initial graph structure are mapped to obtain urban element nodes with category and hierarchical relationships; for the urban element nodes, the random walk algorithm with a time-series decay factor is used to obtain dynamic representation vectors and calculate the association strength between nodes. The urban functional communities are constructed by the modularity optimization method, and the urban system knowledge graph is optimized and reconstructed by the minimum graph structure entropy criterion;

[0010] Based on the urban system knowledge graph, an element diffusion wave equation model is constructed and solved to obtain the state propagation law. The spatial correlation is obtained through geographically weighted regression to get the urban functional zoning; for the functional zoning, a system coupling differential equation group model is established and solved to obtain the prediction results of the element state changes. The time-series dependence relationship is calculated based on conditional mutual information to obtain the causal chain of element changes. The system vulnerability index is constructed through the causal chain, and the multi-objective optimization is used to obtain the urban system resilience analysis results.

[0011] In an alternative embodiment,

[0012] For multi-source heterogeneous data in the urban system, the multivariate statistical process control method is used to establish data quality benchmark thresholds and perform outlier detection to obtain preliminary calibrated data. Using the least squares registration algorithm for the preliminary calibrated data to calculate the spatial transformation matrix to obtain calibrated data in a unified coordinate system includes:

[0013] Obtain multi-source heterogeneous data of the urban system, perform dimensionality reduction processing on the multi-source heterogeneous data of the urban system through the principal component analysis method to obtain the principal component space and the residual space, calculate the principal component statistics of the multi-source heterogeneous data of the urban system based on the principal component space, and calculate the residual statistics of the multi-source heterogeneous data of the urban system based on the residual space;

[0014] Based on the principal component statistics, the first control limit is calculated through probability distribution, and the parameters of the probability distribution are determined by the number of principal components and the number of samples; based on the residual statistics, the second control limit is calculated through weighted distribution, and the weights of the weighted distribution are determined by the eigenvalues of the residual space; the first control limit and the second control limit are combined to construct the data quality benchmark threshold;

[0015] Perform outlier detection on the multi-source heterogeneous data of the urban system using the data quality benchmark threshold, mark the data exceeding the data quality benchmark threshold as outliers and remove them to obtain preliminary calibrated data;

[0016] Construct a spatial point set for the preliminary calibrated data, calculate the covariance matrix of the spatial point set, and perform eigenvalue decomposition on the covariance matrix to obtain a rotation matrix;

[0017] Based on the rotation matrix, construct a least squares objective function, and by minimizing the least squares objective function, solve for the spatial transformation matrix and translation vector, where the least squares objective function represents the sum of the squared distances of the spatial point set in a unified coordinate system;

[0018] Apply the spatial transformation matrix and the translation vector to the preliminary calibrated data, and transform the preliminary calibrated data to a unified urban spatial coordinate system to obtain calibrated data in the unified coordinate system.

[0019] In an alternative embodiment,

[0020] Calculate the credibility score of the calibrated data using recursive Bayesian estimation, perform dynamic adjustment through a Markov decision process to obtain the data fusion weight, and perform weighted fusion on the calibrated data to obtain the standardized fusion data, including:

[0021] Construct a state space model including a state equation and an observation equation, calculate the evolution value of the credibility state of the calibrated data over time through the state equation, introduce a normal distribution system noise, calculate the mapping value between the observed value and the credibility state through the observation equation, and introduce a normal distribution observation noise;

[0022] Use the probability density propagation equation to calculate the propagation result of the posterior probability density at the previous moment to the current moment to obtain the prior probability density, calculate the integration result of the new observation information and the prior probability density based on the conditional probability formula to obtain the posterior probability density, and calculate the mathematical expectation of the posterior probability density to obtain the credibility score of the calibrated data;

[0023] Perform discretization processing on the distribution interval of the credibility score to obtain the state space, set credibility enhancement, attenuation, and maintenance as the action space, generate a state transition probability matrix, construct a Markov decision model, and set an immediate reward function including a credibility improvement reward term and a credibility fluctuation penalty term in the Markov decision model;

[0024] Based on the state transition probability matrix and the immediate reward function, perform value function iteration calculation, obtain the optimal value function through iterative update, generate the mapping relationship from state to action based on the optimal value function to obtain the optimal policy, and calculate the dynamic adjustment result of the credibility score using the optimal policy, and perform normalization processing on the adjusted credibility score to obtain the data fusion weight;

[0025] Generate a spatial kernel function with an adaptive bandwidth parameter, perform local smoothing processing on the data fusion weight using the spatial kernel function to obtain the smoothed fusion weight, and calculate the weighted average value of the calibration data at the same spatial position using the smoothed fusion weight;

[0026] Repeat the weighted fusion calculation for the calibration data at all spatial positions in the urban system, introduce spatial neighborhood constraints in the fusion calculation to determine spatial continuity, and combine the weighted average values of each spatial position to generate the standardized fusion data of the urban system.

[0027] In an alternative embodiment,

[0028] Adopt a double-buffer mechanism to receive the standardized fusion data and convert it into an initial graph structure, construct an urban ontology model based on formal concept analysis and map the nodes of the initial graph structure to obtain urban element nodes with category and hierarchical relationships, including:

[0029] Set two buffers with equal sizes, namely the first buffer and the second buffer, allocate the first buffer and the second buffer to independent physical storage spaces through memory mapping. When the first buffer receives the standardized fusion data of the urban system, the second buffer performs spatio-temporal attribute separation operation to obtain spatio-temporal separation data. The first buffer sets a data reception completion flag bit. When the data reception completion flag bit is set, exchange the physical storage space mapping relationship between the first buffer and the second buffer, and use the spatio-temporal separation data as the input data for urban element construction;

[0030] Construct a parallel pipeline architecture including a data parsing stage, a node generation stage, and an edge relationship construction stage. In a multi-processor environment, connect each processing stage through pipeline registers. The data parsing stage divides the spatio-temporal separation data according to a preset data structure template and adds semantic tags to obtain the data parsing result. The node generation stage creates graph nodes based on the data parsing result and assigns node identifiers to each graph node through a distributed identifier generator to obtain a graph node set. The edge relationship construction stage calculates the neighborhood relationship of the nodes in the graph node set using a spatial index structure, and establishes edge connections based on the neighborhood relationship and node attribute similarity to obtain the initial graph structure;

[0031] A formal context is established, which includes a set of urban element objects, a set of urban element attributes, and a set of relationships between objects and attributes. Context compatibility verification is performed on the objects and attributes in the formal context. An incremental algorithm is used to perform a closure operation on the formal context to obtain a hierarchical graph of the concept lattice, and the hierarchical graph is converted into an urban ontology model described in a knowledge representation language;

[0032] The semantic similarity between the graph node attributes in the initial graph structure and the concept classes in the urban ontology model is calculated. Based on the maximum value of the semantic similarity, the concept class corresponding to the node is determined to obtain the node category. The node hierarchical relationship including the subordinate relationship and the composition relationship is constructed through ontology inference rules to obtain a node relationship set. The transitive closure iterative calculation based on the adjacency matrix is performed on the node relationship set, and the calculation result is combined with the node category to obtain an urban element node graph with category and hierarchical relationships.

[0033] In an alternative embodiment,

[0034] A random walk algorithm with a time decay factor is used for the urban element nodes to obtain dynamic representation vectors. The association strength between nodes is calculated through the dynamic representation vectors. Based on the association strength, a modularity optimization method is used to construct urban functional communities, and the urban functional communities are optimized and reconstructed according to the minimum graph structure entropy criterion to obtain an urban system knowledge graph, including:

[0035] A basic transition matrix is constructed according to the spatial distance between urban element nodes; a time decay factor is introduced into the basic transition matrix to form a time adjustment matrix, where the time decay factor decreases as the time interval increases; the basic transition matrix is multiplied by the time adjustment matrix to obtain a time transition probability matrix. Based on the time transition probability matrix, random walk sampling is performed on the urban element nodes to obtain a node neighborhood sequence, and the node neighborhood sequence is mapped to a low-dimensional vector space to obtain a node dynamic representation vector;

[0036] Based on adjacent urban element nodes, node pairs are determined. The cosine similarity between the node dynamic representation vectors of the node pairs is calculated, and within a preset time window range, the interaction times of the node pairs within the time window range are counted. The cosine similarity and the interaction times are weighted and combined to obtain the node association strength, and the node association strength is filled to form an association strength matrix;

[0037] Extract the sum of the association strengths of the internal nodes of the community from the association strength matrix, statistically analyze the degree distribution of the urban element nodes and calculate the expected random association strength, subtract the expected random association strength from the sum of the association strengths to construct a modularity objective function, perform iterative optimization to maximize the modularity objective function, and divide the urban element nodes with node association strengths greater than the preset strength threshold into initial urban functional communities;

[0038] Calculate the graph structure entropy of the initial urban functional community, randomly adjust the community membership of the urban element nodes to generate multiple candidate solutions, calculate the change in the graph structure entropy generated by each candidate solution, perform the candidate solution that reduces the graph structure entropy the most, and repeat the node adjustment until the change in the graph structure entropy is less than the preset change threshold to obtain an optimized urban functional community;

[0039] Define the optimized urban functional community as a functional unit, statistically analyze the association strength between the internal nodes of the functional unit to calculate the association degree between the functional units, and organize the functional units and the corresponding association degrees into an urban system knowledge graph.

[0040] In an alternative embodiment,

[0041] Construct an element diffusion wave equation model based on the urban system knowledge graph and solve it to obtain the state propagation law. Through geographically weighted regression analysis of spatial correlation, the urban functional zoning includes:

[0042] Obtain the spatial position coordinates and influence range of the functional units in the urban system knowledge graph, and set the spatial positions of the functional units as diffusion wave source points; extract the association degrees between the functional units, convert them into wave propagation speed coefficients, and calculate the corresponding wave attenuation coefficients; establish an urban element diffusion wave equation centered on the diffusion wave source points, multiply the influence intensity corresponding to the functional units by the wave attenuation coefficients to calculate the external excitation terms, and substitute the external excitation terms into the urban element diffusion wave equation;

[0043] Divide the computational domain into irregular triangular grid cells centered on the diffusion wave source points, and increase the size of the irregular triangular grid cells correspondingly as the distance from the diffusion wave source points increases; establish a linear interpolation function within the irregular triangular grid cells, and calculate the stiffness matrix of the irregular triangular grid cells according to the linear interpolation function; establish a mass matrix representing the node mass distribution, and substitute the mass matrix and the stiffness matrix into the urban element diffusion wave equation to obtain a discretized matrix equation;

[0044] Perform time-step integration calculations, and perform predictor-corrector iterations within each time step: Substitute the mass matrix and the stiffness matrix into the calculations to predict displacement fields, velocity fields, and acceleration fields, and perform equilibrium iteration calculations to correct displacement fields, corrected velocity fields, and corrected acceleration fields; Combine the corrected displacement fields, the corrected velocity fields, and the corrected acceleration fields into a spatio-temporal evolution sequence, obtain a state quantity propagation velocity field and a state quantity gradient field based on the spatio-temporal evolution sequence, calculate the density distribution of functional units based on the state quantity propagation velocity field and the state quantity gradient field, and obtain the spatial distribution density of functional units; Interpolate and calculate the degree of association to obtain the functional association strength distribution;

[0045] Construct a spatial weighted regression model, set the regression coefficients to vary continuously with spatial positions, calculate the density distribution of observation points to determine the bandwidth of the spatial weight function, and perform spatial weighting of observation points; Input the spatial distribution density of functional units and the functional association strength distribution into the spatial weighted regression model;

[0046] Perform weighted least squares calculations to obtain the spatial distribution of the regression coefficients, calculate the spatial variance of the regression coefficients, determine regions with significant spatial heterogeneity, and perform natural break classification in the regions with significant spatial heterogeneity to divide urban functional zones with significant spatial differences.

[0047] In an alternative embodiment,

[0048] Establish a system coupling differential equation model for the functional zones and solve it to obtain the prediction results of the state changes of the elements. Calculate the temporal dependence relationship based on conditional mutual information to obtain the causal chain of element changes. Construct a system vulnerability index through the causal chain, and use multi-objective optimization to obtain the urban system resilience analysis results, including:

[0049] Establish an urban system coupling differential equation model for the urban functional zones, numerically solve the urban system coupling differential equation model using the adaptive step spectral element method, divide the time domain into multiple non-uniform sub-intervals, construct high-order basis functions using Legendre polynomials within each non-uniform sub-interval, dynamically adjust the time step through error estimation, substitute the high-order basis functions into the urban system coupling differential equation model for Galerkin projection, and obtain the prediction results of the state changes of urban elements;

[0050] Calculate the marginal information entropy of each urban element in the prediction result of the urban element state change, calculate the joint information entropy of any pair of urban elements, and the marginal information entropy of the remaining urban elements outside the pair of urban elements; combine the marginal information entropy, joint information entropy and marginal information entropy of the remaining urban elements in the pair of urban elements to calculate the conditional mutual information; determine whether there is a temporal dependence relationship between the pair of urban elements according to the magnitude relationship between the conditional mutual information and the preset significance threshold; traverse all pairs of urban elements in the prediction result of the urban element state change to establish a causal chain of urban element changes.

[0051] Calculate the in-degree value and out-degree value of each node corresponding to the urban element in the causal chain of urban element changes, multiply the in-degree value by the in-degree weight to obtain the in-degree weighted value, multiply the out-degree value by the out-degree weight to obtain the out-degree weighted value, and add the in-degree weighted value and the out-degree weighted value to obtain the urban system vulnerability index.

[0052] Establish a multi-objective optimization model aiming to minimize the urban system vulnerability index and the regulation cost, set the urban resource constraint as the constraint condition of the multi-objective optimization model, and solve the multi-objective optimization model through the non-dominated sorting genetic algorithm to obtain a set of urban system resilience analysis results that meet the constraint conditions.

[0053] In the second aspect of the embodiments of the present invention,

[0054] Provide a multi-source information fusion data space analysis system based on a knowledge graph, including:

[0055] The first unit is used to establish a data quality benchmark threshold for the multi-source heterogeneous data in the urban system by using the multivariate statistical process control method and perform outlier detection to obtain preliminary calibrated data, calculate the spatial transformation matrix for the preliminary calibrated data by using the least squares registration algorithm to obtain calibrated data in a unified coordinate system; calculate the credibility score of the calibrated data by using recursive Bayesian estimation, perform dynamic adjustment through the Markov decision process to obtain the data fusion weight, and perform weighted fusion on the calibrated data to obtain standardized fusion data.

[0056] The second unit is used to receive the standardized fusion data by using a double buffer mechanism and convert it into an initial graph structure, construct an urban ontology model based on formal concept analysis and map the nodes of the initial graph structure to obtain urban element nodes with category and hierarchical relationships; use the random walk algorithm with a temporal decay factor for the urban element nodes to obtain dynamic representation vectors and calculate the association strength between nodes, construct urban functional communities by using the modularity optimization method, and optimize and reconstruct by using the graph structure entropy minimization criterion to obtain the urban system knowledge graph.

[0057] The third unit is used to construct a factor diffusion wave equation model based on the urban system knowledge graph, solve it to obtain the state propagation law, and obtain the urban functional zoning by analyzing the spatial correlation through geographically weighted regression; establish a system coupling differential equation group model for the functional zoning, solve it to obtain the prediction result of the factor state change, calculate the time series dependence relationship based on conditional mutual information to obtain the causal chain of factor changes, construct a system vulnerability index through the causal chain, and adopt multi-objective optimization to obtain the urban system resilience analysis result.

[0058] In the third aspect of the embodiments of the present invention,

[0059] a kind of electronic device is provided, including:

[0060] a processor;

[0061] a memory for storing instructions executable by the processor;

[0062] wherein, the processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0063] In the fourth aspect of the embodiments of the present invention,

[0064] a kind of computer-readable storage medium is provided, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the aforementioned method is realized.

[0065] In the embodiments of the present invention, through multi-source heterogeneous data quality control, coordinate system unification, credibility evaluation and weighted fusion, problems such as multi-source heterogeneity, inconsistency and incompleteness of urban data are effectively solved, the data quality and fusion accuracy are significantly improved, and a reliable data basis is provided for urban system analysis; by using formal concept analysis and graph calculation methods, an urban system knowledge graph including categories, hierarchies and dynamic relationships is constructed, realizing the refined expression and correlation analysis of urban elements, and being able to deeply explore the internal structure and operation law of the urban system; through constructing a diffusion wave equation model, a coupling differential equation group model and causal chain analysis, the prediction of the state propagation law, future change trend and vulnerability of urban elements is realized, providing scientific decision-making support for urban planning and management, and helping to improve the resilience and sustainable development ability of the urban system. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 It is a schematic flow chart of the multi-source information fusion data spatial analysis method based on the knowledge graph in the embodiments of the present invention;

[0067] Figure 2 It is a schematic structural diagram of the multi-source information fusion data spatial analysis system based on the knowledge graph in the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0068] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0069] The following uses specific embodiments to elaborate on the technical solutions of the present invention in detail. These several specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0070] Figure 1 It is a schematic flowchart of a multi-source information fusion data space analysis method based on a knowledge graph according to an embodiment of the present invention, as Figure 1 shown. The method includes:

[0071] S101. For the multi-source heterogeneous data in the urban system, use the multivariate statistical process control method to establish the data quality benchmark threshold and perform outlier detection to obtain the preliminary calibrated data. Use the least squares registration algorithm to calculate the spatial transformation matrix for the preliminary calibrated data to obtain the calibrated data in the unified coordinate system; use recursive Bayesian estimation to calculate the credibility score of the calibrated data, and perform dynamic adjustment through the Markov decision process to obtain the data fusion weight, and perform weighted fusion on the calibrated data to obtain the standardized fusion data;

[0072] S102. Use the double buffer mechanism to receive the standardized fusion data and convert it into the initial graph structure. Based on formal concept analysis, construct the urban ontology model and map the nodes of the initial graph structure to obtain the urban element nodes with category and hierarchical relationships; use the random walk algorithm with a time series decay factor for the urban element nodes to obtain the dynamic representation vectors and calculate the association strength between nodes, construct the urban functional community through the modularity optimization method, and optimize and reconstruct according to the minimum graph structure entropy criterion to obtain the urban system knowledge graph;

[0073] S103. Based on the urban system knowledge graph, construct an element diffusion wave equation model and solve it to obtain the state propagation law. Analyze the spatial correlation through geographically weighted regression to obtain the urban functional zoning; establish a system coupling differential equation model for the functional zoning and solve it to obtain the prediction results of the element state changes. Calculate the time series dependence relationship based on conditional mutual information to obtain the causal chain of element changes, construct the system vulnerability index through the causal chain, and use multi-objective optimization to obtain the urban system resilience analysis results.

[0074] In an optional implementation, for multi-source heterogeneous data in an urban system, a multivariate statistical process control method is used to establish a data quality benchmark threshold and perform outlier detection to obtain preliminary calibration data, and a least squares registration algorithm is used to calculate a spatial transformation matrix for the preliminary calibration data to obtain calibration data in a unified coordinate system, including:

[0075] Acquire multi-source heterogeneous data of the urban system, perform dimensionality reduction processing on the multi-source heterogeneous data of the urban system by a principal component analysis method to obtain a principal component space and a residual space, calculate the principal component statistics of the multi-source heterogeneous data of the urban system based on the principal component space, and calculate the residual statistics of the multi-source heterogeneous data of the urban system based on the residual space;

[0076] Based on the principal component statistic, a first control limit is obtained by probability distribution calculation, and the parameters of the probability distribution are determined by the number of principal components and the number of samples; based on the residual statistic, a second control limit is obtained by weighted distribution calculation, and the weight of the weighted distribution is determined by the eigenvalue of the residual space; the first control limit and the second control limit are combined to construct a data quality benchmark threshold;

[0077] Using the data quality benchmark threshold, outlier detection is performed on the multi-source heterogeneous data of the urban system, and data exceeding the data quality benchmark threshold is marked as an outlier and removed to obtain preliminary calibration data;

[0078] constructing a spatial point set for the preliminary calibration data, calculating a covariance matrix of the spatial point set, and performing eigenvalue decomposition on the covariance matrix to obtain a rotation matrix;

[0079] A least squares objective function is constructed based on the rotation matrix, and a spatial transformation matrix and a translation vector are solved by minimizing the least squares objective function, wherein the least squares objective function represents the sum of squares of distances of the spatial point set in a unified coordinate system;

[0080] The spatial transformation matrix and the translation vector are applied to the preliminary calibration data, and the preliminary calibration data is converted into a unified urban space coordinate system to obtain calibration data in a unified coordinate system.

[0081] In a specific implementation, first, multi-source heterogeneous data in the urban system is obtained. For example, data on urban traffic flow, air quality, meteorological information, population density, etc. are obtained from different sensors, databases, and public data platforms. These data may have different data formats, different time resolutions, and different spatial coverage.

[0082] Then, perform dimensionality reduction on the obtained multi-source heterogeneous data. Using the principal component analysis method, project the high-dimensional data onto the low-dimensional principal component space and residual space. The principal component space contains the main variation information of the data, while the residual space contains the noise and abnormal information of the data. For example, perform principal component analysis on data such as traffic flow, air quality, and meteorological information to extract the main influencing factors. Suppose there are 1000 samples, and each sample contains 5 variables. Through principal component analysis, these 5 variables can be reduced to 2 principal components, and these two principal components can explain 80% of the variance of the original data.

[0083] Next, calculate the data quality benchmark threshold. Based on the data distribution in the principal component space, calculate the first control limit. This control limit is determined by the number of principal components and the number of samples. For example, by calculating the mean and standard deviation of the principal component score values, set the first control limit as the mean plus or minus three times the standard deviation. Based on the data distribution in the residual space, calculate the second control limit. This control limit is obtained by weighting the eigenvalues of the residual space. For example, multiply the square value of each residual by the reciprocal of the corresponding eigenvalue, and then sum to obtain a statistic. Set the second control limit as a quantile of this statistic, such as the 95% quantile. Combine the first control limit and the second control limit to construct the data quality benchmark threshold.

[0084] Subsequently, perform outlier detection. Compare the multi-source heterogeneous data with the data quality benchmark threshold, and the data exceeding the threshold is marked as an outlier and removed. For example, if the principal component score value or residual statistic of a certain sample exceeds the set control limit, then this sample is marked as an outlier. Suppose 10 samples are marked as outliers, and after removal, the remaining 990 samples are used as the preliminary calibration data.

[0085] Next, unify the coordinate system for the preliminary calibration data. Construct a spatial point set and calculate the covariance matrix of the point set. Perform eigenvalue decomposition on the covariance matrix to obtain the rotation matrix. For example, extract the spatial coordinate information from the preliminary calibration data to construct a spatial point set. Calculate the covariance matrix of the point set and perform eigenvalue decomposition on it to obtain the rotation matrix.

[0086] Then, calculate the spatial transformation matrix. Based on the rotation matrix, construct a least squares objective function. This objective function characterizes the sum of the squared distances of the spatial point set in the unified coordinate system. By minimizing this objective function, solve for the spatial transformation matrix and the translation vector. For example, through an iterative optimization algorithm, solve for the spatial transformation matrix and the translation vector that minimize the sum of the squared distances of the spatial point set in the unified coordinate system.

[0087] Finally, perform coordinate transformation. Apply the calculated spatial transformation matrix and translation vector to the preliminary calibration data to convert the data into a unified urban spatial coordinate system, obtaining the calibration data in the unified coordinate system. For example, apply the spatial transformation matrix and translation vector to the spatial coordinates of the remaining 990 samples to convert them into the unified coordinate system.

[0088] In this embodiment, outlier detection and elimination are performed through the data quality benchmark threshold, effectively improving the accuracy and reliability of the data, laying a solid foundation for subsequent data analysis and applications; through the least squares registration algorithm, multi-source heterogeneous data is converted into a unified urban spatial coordinate system, eliminating the spatial differences in the data fusion process, facilitating cross-data-source analysis and comparison; through data calibration, the consistency and comparability of the data are improved, enabling the data to better serve urban planning, management, and decision-making. For example, the calibrated data can be used for urban traffic flow prediction, air quality assessment, and optimal allocation of urban resources.

[0089] In an alternative embodiment, the credibility score of the calibration data is calculated using recursive Bayesian estimation and dynamically adjusted through a Markov decision process to obtain the data fusion weight. The weighted fusion of the calibration data to obtain the standardized fusion data includes:

[0090] Construct a state space model including a state equation and an observation equation. Calculate the evolution value of the credibility state of the calibration data over time through the state equation, and introduce a normal distribution system noise. Calculate the mapping value between the observation value and the credibility state through the observation equation, and introduce a normal distribution observation noise.

[0091] Use the probability density propagation equation to calculate the propagation result of the posterior probability density at the previous moment to the current moment to obtain the prior probability density. Calculate the integration result of the new observation information and the prior probability density based on the conditional probability formula to obtain the posterior probability density. Calculate the mathematical expectation of the posterior probability density to obtain the credibility score of the calibration data.

[0092] Perform discretization processing on the distribution interval of the credibility score to obtain the state space. Set credibility enhancement, attenuation, and maintenance as the action space, generate the state transition probability matrix, construct a Markov decision model, and set an immediate reward function including a credibility improvement reward term and a credibility fluctuation penalty term in the Markov decision model.

[0093] Based on the state transition probability matrix and the immediate reward function, perform value function iteration calculation. Obtain the optimal value function through iterative update. Generate the mapping relationship from the state to the action based on the optimal value function to obtain the optimal policy. Use the optimal policy to calculate the dynamic adjustment result of the credibility score, and perform normalization processing on the adjusted credibility score to obtain the data fusion weight.

[0094] Generate a spatial kernel function with adaptive bandwidth parameters, locally smooth the data fusion weights using the spatial kernel function to obtain smoothed fusion weights, and calculate the weighted average of the calibrated data at the same spatial position using the smoothed fusion weights;

[0095] Repeat the weighted fusion calculation for the calibrated data at all spatial positions in the urban system. Introduce spatial neighborhood constraints in the fusion calculation to determine spatial continuity, and combine the weighted averages at each spatial position to generate the standardized fusion data of the urban system.

[0096] In a specific implementation, first, construct a state space model. This model consists of two parts: a state equation and an observation equation. The state equation describes the variation law of the data credibility state over time and takes into account the influence of system noise, assuming that the noise follows a normal distribution. The observation equation describes the relationship between the observed value and the credibility state, also considering the influence of observation noise and assuming that the noise follows a normal distribution. For example, assume that the credibility state value at a certain moment is 0.8. After calculation by the state equation, the credibility state value at the next moment is 0.82. At the same time, the observation equation calculates the observed value as 0.78 based on the current state.

[0097] Next, calculate the credibility score using the probability density propagation equation. First, calculate the prior probability density at the current moment based on the posterior probability density at the previous moment. This process describes the propagation of the probability density over time. Then, combine the new observation information and update the prior probability density using the conditional probability formula to obtain the posterior probability density. Finally, calculate the mathematical expectation of the posterior probability density as the credibility score of the calibrated data at the current moment. For example, assume that the posterior probability density at the previous moment is a normal distribution with a mean of 0.8 and a variance of 0.01. After probability density propagation and observation update, the posterior probability density at the current moment becomes a normal distribution with a mean of 0.81 and a variance of 0.008. Then, the credibility score at the current moment is 0.81.

[0098] Then, construct a Markov decision model to dynamically adjust the credibility score. Discretize the distribution interval of the credibility score to form a state space. For example, divide the interval from 0 to 1 into 10 states. Set three actions: credibility enhancement, attenuation, and maintenance, to form an action space. Generate a state transition probability matrix according to the state transition rule. For example, when the current state is 0.8 and the "enhancement" action is executed, there is a certain probability of transitioning to the state of 0.85, and there is also a certain probability of remaining in the state of 0.8 or decaying to the state of 0.75. Set an immediate reward function, which includes a credibility improvement reward term and a credibility fluctuation penalty term. For example, credibility improvement will receive a positive reward, while credibility fluctuation will be penalized.

[0099] Based on the state transition probability matrix and the immediate reward function, perform value function iteration calculation. Through continuous iterative updates, finally obtain the optimal value function. According to the optimal value function, generate the mapping relationship from state to action, that is, the optimal policy. Apply the optimal policy to calculate the dynamic adjustment result of the credibility score and perform normalization processing to obtain the data fusion weight. For example, assume the current state is 0.8 and the optimal policy is "enhance", then adjust the credibility score to a higher value, such as 0.85.

[0100] Finally, perform data fusion. Generate a spatial kernel function with an adaptive bandwidth parameter to perform local smoothing on the data fusion weight to obtain the smoothed fusion weight. For example, use a Gaussian kernel function to smooth the weight so that the weights at spatially adjacent positions are closer. Use the smoothed fusion weight to calculate the weighted average of the calibrated data at the same spatial position. In the fusion calculation, introduce spatial neighborhood constraints to ensure spatial continuity. For example, only consider the calibrated data within a certain range of the target position for fusion. Combine the weighted averages at each spatial position to generate the final standardized fusion data of the urban system. Suppose there are two spatially adjacent calibrated data with values of 10 and 12 respectively, and the corresponding smoothed fusion weights are 0.4 and 0.6 respectively, then the fused value is 10×0.4 + 12×0.6 = 11.2.

[0101] In this embodiment, through recursive Bayesian estimation and Markov decision process, the credibility of the calibrated data is dynamically evaluated and adjusted, effectively identifying and reducing the impact of low-credibility data, and improving the reliability of the final fusion data; spatial neighborhood constraints and smoothing processing are introduced during the fusion process to ensure the spatial continuity of the fusion data and more accurately reflect the spatial characteristics of the urban system; through the weighted fusion method, multi-source calibrated data are effectively integrated, making full use of the information from different data sources and improving the accuracy and integrity of the final fusion data.

[0102] In an alternative embodiment, a double-buffer mechanism is used to receive the standardized fusion data and convert it into an initial graph structure. Based on formal concept analysis, a city ontology model is constructed and the nodes of the initial graph structure are mapped to obtain city element nodes with category and hierarchical relationships, including:

[0103] Set two buffers of equal size, namely the first buffer and the second buffer. Allocate the first buffer and the second buffer to independent physical storage spaces through memory mapping. When the first buffer receives the standardized fusion data of the urban system, the second buffer performs a spatio-temporal attribute separation operation to obtain spatio-temporally separated data. A data reception completion flag bit is set in the first buffer. When the data reception completion flag bit is set, swap the physical storage space mapping relationships of the first buffer and the second buffer, and use the spatio-temporally separated data as the input data for urban feature construction;

[0104] Construct a parallel pipeline architecture including a data parsing stage, a node generation stage, and an edge relationship construction stage. In a multi-processor environment, connect each processing stage through pipeline registers. The data parsing stage splits the spatio-temporally separated data according to a preset data structure template and adds semantic tags to obtain a data parsing result. The node generation stage creates graph nodes based on the data parsing result and assigns node identifiers to each graph node through a distributed identifier generator to obtain a set of graph nodes. The edge relationship construction stage uses a spatial index structure to calculate the neighborhood relationships of the nodes in the set of graph nodes, and establishes edge connections based on the neighborhood relationships and node attribute similarities to obtain an initial graph structure;

[0105] Establish a formal context, which includes a set of urban feature objects, a set of urban feature attributes, and a set of relationships between objects and attributes. Perform context compatibility verification on the objects and attributes in the formal context, and use an incremental algorithm to perform a closure operation on the formal context to obtain a hierarchical graph of the concept lattice, and convert the hierarchical graph into an urban ontology model described in a knowledge representation language;

[0106] Calculate the semantic similarity between the graph node attributes in the initial graph structure and the concept classes in the urban ontology model, determine the concept class corresponding to the node based on the maximum value of the semantic similarity to obtain the node category, construct a node hierarchical relationship including subordination relationships and composition relationships through ontology inference rules to obtain a set of node relationships, perform a transitive closure iterative calculation based on the adjacency matrix on the set of node relationships, and combine the calculation results with the node category to obtain an urban feature node graph with category and hierarchical relationships.

[0107] In a specific implementation manner, first, the preparatory work is to set up a double-buffer mechanism. Allocate two memory areas of equal size as the first buffer and the second buffer, and map them to independent physical storage spaces through memory mapping technology to avoid mutual interference during the data processing process. For example, two 1GB memory spaces can be allocated and named Buffer1 and Buffer2 respectively.

[0108] Next, start receiving the standardized integration data of the urban system. The standardized integration data of the urban system continuously flows into the first buffer. Assume that the data currently received by the first buffer is information about roads, buildings, and parks, including their names, locations, areas, construction times, and other attributes. While the first buffer is receiving data, the second buffer performs spatio-temporal attribute separation operations on the previously received data. For example, separate the geometric shape and construction time of the road.

[0109] When the data reception in the first buffer is completed, a data reception completion flag is set. Once this flag is set, the system exchanges the physical storage space mapping relationship between the first buffer and the second buffer. This means that the pointer originally pointing to Buffer1 now points to Buffer2, and vice versa. Now, the data after spatio-temporal attribute separation in the second buffer will be used as the input data for urban feature construction.

[0110] Then, construct a parallel pipeline architecture to process the data. This architecture includes a data parsing stage, a node generation stage, and an edge relationship construction stage. These three stages are connected by pipeline registers and executed in parallel in a multi-processor environment to improve data processing efficiency.

[0111] The data parsing stage splits the spatio-temporally separated data according to a preset data structure template and adds semantic tags. For example, tag the "name" field of road data as "road_name" and the "location" field as "road_location". The node generation stage creates graph nodes based on the data parsing results and assigns unique node identifiers to each graph node using a distributed identifier generator. For example, create a node for a road named "Central Street" and assign the ID "road_001". The edge relationship construction stage uses a spatial index structure (such as an R-tree) to calculate the neighborhood relationships of the nodes in the graph node set. For example, determine whether there is an intersection relationship between two roads. Then, establish edge connections based on the neighborhood relationships and node attribute similarities (such as the similarity of road types) to finally obtain an initial graph structure.

[0112] Next, construct the urban ontology model. First, establish a formal context. The formal context includes a set of urban element objects (such as roads, buildings, parks), a set of urban element attributes (such as name, location, area, function), and a set of relationships between objects and attributes. For example, the road "Central Street" has attributes such as "named Central Street" and "located in the city center". Then, perform context compatibility verification on the objects and attributes in the formal context to ensure the logical consistency of the data. Next, perform a closure operation on the formal context using an incremental algorithm to obtain the hierarchical graph of the concept lattice. Finally, convert the hierarchical graph into an urban ontology model described based on a knowledge representation language (such as OWL).

[0113] Finally, calculate the semantic similarity between the graph node attributes in the initial graph structure and the concept classes in the urban ontology model. For example, calculate the similarity between the "Central Street" node and the "transportation facility" concept class. Determine the concept class corresponding to the node based on the maximum value of the semantic similarity to obtain the node category. For example, determine that the "Central Street" node belongs to the "road" category. Construct a node hierarchical relationship including subordination relationships (such as "Central Street" belongs to "road") and composition relationships (such as "road" is composed of "road surface", "street lamp", etc.) through ontology inference rules to obtain a set of node relationships. Perform iterative calculation of the transitive closure based on the adjacency matrix on the set of node relationships, and combine the calculation results with the node categories to finally obtain an urban element node graph with categories and hierarchical relationships.

[0114] In this embodiment, the double-buffer mechanism and the parallel pipeline architecture can effectively utilize the multi-core processor resources, accelerate the data processing speed, and thus shorten the construction time of the urban element atlas; by constructing the urban ontology model, the attributes and relationships of urban elements are formally described, making the urban element atlas have richer semantic information and facilitating knowledge reasoning and query; the formal concept analysis method can effectively identify the potential relationships between urban elements and construct a complete concept hierarchy, thereby improving the accuracy and integrity of the urban element atlas.

[0115] In an alternative implementation, a random walk algorithm with a time-series decay factor is used for the urban element nodes to obtain dynamic representation vectors. The association strength between nodes is calculated through the dynamic representation vectors. Based on the association strength, an urban function community is constructed by a modularity optimization method, and the urban function community is optimized and reconstructed according to the graph structure entropy minimization criterion to obtain an urban system knowledge graph including:

[0116] Construct a basic transfer matrix based on the spatial distances between urban element nodes; introduce a temporal decay factor into the basic transfer matrix to form a temporal adjustment matrix, where the temporal decay factor decreases as the time interval increases; multiply the basic transfer matrix by the temporal adjustment matrix to obtain a temporal transfer probability matrix, and based on the temporal transfer probability matrix, perform random walk sampling on the urban element nodes to obtain a node neighborhood sequence, and map the node neighborhood sequence to a low-dimensional vector space to obtain a node dynamic representation vector;

[0117] Based on adjacent urban element nodes, determine node pairs, calculate the cosine similarity between the node dynamic representation vectors of the node pairs, and within a preset time window range, count the interaction times of the node pairs within the time window range, and weightedly combine the cosine similarity and the interaction times to obtain a node association strength, and fill the node association strength to form an association strength matrix;

[0118] Extract the sum of the association strengths of the nodes within the community from the association strength matrix, count the degree distribution of the urban element nodes and calculate the expected random association strength, subtract the expected random association strength from the sum of the association strengths to construct a modularity objective function, perform iterative optimization to maximize the modularity objective function, and divide the urban element nodes with node association strengths greater than a preset strength threshold into initial urban functional communities;

[0119] Calculate the graph structure entropy of the initial urban functional community, randomly adjust the community membership of the urban element nodes to generate multiple candidate solutions, calculate the change in the graph structure entropy generated by each candidate solution, perform the candidate solution that reduces the graph structure entropy the most, repeat the node adjustment until the change in the graph structure entropy is less than a preset change threshold to obtain an optimized urban functional community;

[0120] Define the optimized urban functional community as a functional unit, count the association strength between the nodes within the functional unit to calculate the association degree between functional units, and organize the functional units and the corresponding association degrees into an urban system knowledge graph.

[0121] In a specific implementation manner, first, construct the spatial relationship between urban element nodes. Use various Points of Interest (POIs) in the city as urban element nodes, such as restaurants, shopping malls, hospitals, schools, etc. Collect the geographical location information (latitude and longitude) of these POIs, calculate the straight-line distance between any two POIs, and construct a basic transfer matrix. Each row and each column of this matrix represents a POI, and the elements in the matrix represent the spatial distance between the corresponding two POIs. For example, the element value in the i-th row and j-th column of the matrix represents the distance between POI i and POI j. The closer the distance, the smaller the value, and vice versa.

[0122] Next, introduce a temporal decay factor to adjust the spatial relationship in the time dimension. Considering that people's activities in the city have temporal regularity, for example, the travel patterns on weekdays and weekends may be different, so it is necessary to introduce a temporal decay factor. The temporal decay factor is a function that decreases as the time interval increases, such as an exponential decay function or an inverse function. Multiply the temporal decay factor by the distance value in the basic transition matrix to obtain a temporal adjustment matrix. For example, if the interaction frequency between POIi and POI j is high within a certain time period, the corresponding temporal decay factor will be large, thus enhancing the association strength between them.

[0123] Then, multiply the basic transition matrix by the temporal adjustment matrix to obtain a temporal transition probability matrix. This matrix reflects the probability of transferring from one POI to another within a specific time period. Based on the temporal transition probability matrix, perform random walk sampling on the urban element nodes. The random walk simulates people's movement trajectories in the city. Starting from a POI, randomly jump to the next POI according to the transition probability, and repeat multiple times to form a node neighborhood sequence. For example, starting from the POI "mall", a possible neighborhood sequence generated may be "mall - restaurant - cinema - mall - supermarket". Map the node neighborhood sequence of each POI to a low-dimensional vector space to obtain a node dynamic representation vector. Word embedding techniques, such as Word2Vec or GloVe, can be used to convert the node neighborhood sequence into a dense vector.

[0124] Next, calculate the association strength between nodes. Select adjacent POI node pairs and calculate the cosine similarity between their dynamic representation vectors. At the same time, count the number of interactions between each POI node pair within a preset time window range, such as one week or one month. The number of interactions can be obtained by counting the number of times two POIs are visited within the same time period, such as the check-in data posted by users on social media. Combine the cosine similarity and the number of interactions with weights to obtain the node association strength. For example, if the cosine similarity between two POIs is high and the number of interactions within the time window is also large, the association strength between them will be high. Fill the association strength of all POI node pairs into an association strength matrix.

[0125] Based on the association strength matrix, construct urban functional communities. First, calculate the sum of the association strengths of the nodes within each community. Then, count the degree distribution of all POI nodes and calculate the expected random association strength. Subtract the expected random association strength from the sum of the within-community association strengths to construct a modularity objective function. Through an iterative optimization algorithm, such as the Louvain algorithm, maximize the modularity objective function to divide the POI nodes into initial urban functional communities. Divide the POI nodes with an association strength greater than a preset strength threshold into the same community.

[0126] In order to optimize the community structure, the concept of graph structure entropy is introduced. The graph structure entropy of the initial urban functional community is calculated. Then, the community affiliation of some POI nodes is randomly adjusted to generate multiple candidate solutions. The change in graph structure entropy of each candidate solution is calculated. The candidate solution that reduces the graph structure entropy the most is selected and the community structure is updated. The process of node adjustment and solution selection is repeated until the change in graph structure entropy is less than the preset change threshold, and the optimized urban functional community is obtained.

[0127] Finally, construct the urban system knowledge graph. Define the optimized urban functional communities as functional units, such as "catering and entertainment area", "commercial shopping area", "residential area", etc. Count the association strength between the internal nodes of each functional unit, and calculate the degree of association between functional units. Organize the functional units and their corresponding association degrees into an urban system knowledge graph, such as using the graph database Neo4j for storage and visualization.

[0128] Data example: Assume there are three POIs: A (restaurant), B (cinema), and C (shopping mall). The distances between them are: AB: 1km, AC: 2km, and BC: 1km. Assume the time window is one week, the number of interactions between AB is 50 times, AC is 20 times, and BC is 30 times. The final urban system knowledge graph is calculated through the above steps. For example, the catering and entertainment areas (A, B) are highly correlated with the commercial shopping area (C).

[0129] In this embodiment, the dynamic relationship between urban element nodes can be captured, so as to have a more detailed understanding of the urban spatial structure and functional organization; through dynamic representation and community optimization, different functional areas in the city can be more accurately identified, such as commercial areas, residential areas, entertainment areas, etc.; the constructed urban system knowledge graph can provide a more comprehensive understanding of the urban system, which is helpful for applications such as urban planning, traffic management, and commercial site selection.

[0130] In an optional implementation, a factor diffusion wave equation model is constructed based on the urban system knowledge graph and the state propagation law is solved. The urban functional zoning is obtained by analyzing the spatial correlation through geographical weighted regression, including:

[0131] The spatial position coordinates and influence range of the functional unit in the urban system knowledge graph are obtained, and the spatial position of the functional unit is set as the diffusion wave source point; the degree of association between the functional units is extracted, converted into a wave propagation velocity coefficient, and the corresponding wave attenuation coefficient is calculated; an urban element diffusion wave equation centered on the diffusion wave source point is established, and the influence intensity corresponding to the functional unit is multiplied by the wave attenuation coefficient to calculate the external excitation term, and the external excitation term is substituted into the urban element diffusion wave equation;

[0132] The computational domain is divided into irregular triangular grid cells centered on the diffusion wave source point. As the distance from the diffusion wave source point increases, the size of the irregular triangular grid cells is correspondingly increased. A linear interpolation function is established within the irregular triangular grid cells, and the stiffness matrix of the irregular triangular grid cells is calculated according to the linear interpolation function. A mass matrix representing the node mass distribution is established, and the mass matrix and the stiffness matrix are substituted into the urban element diffusion wave equation to obtain a discretized matrix equation.

[0133] Time-step integration calculation is performed, and a predictor-corrector iteration is performed within each time step: the mass matrix and the stiffness matrix are substituted into the calculation to predict the displacement field, velocity field, and acceleration field, and equilibrium iteration calculation is performed to correct the displacement field, corrected velocity field, and corrected acceleration field; the corrected displacement field, the corrected velocity field, and the corrected acceleration field are combined into a spatio-temporal evolution sequence, and the state quantity propagation velocity field and the state quantity gradient field are obtained according to the spatio-temporal evolution sequence. Based on the state quantity propagation velocity field and the state quantity gradient field, the density distribution of the functional unit is calculated to obtain the spatial distribution density of the functional unit. Interpolation calculation is performed on the correlation degree to obtain the functional correlation intensity distribution.

[0134] A spatial weighted regression model is constructed, the regression coefficient is set to vary continuously with the spatial position, the density distribution of the observation points is calculated to determine the bandwidth of the spatial weight function, and spatial weighting of the observation points is performed; the spatial distribution density of the functional unit and the functional correlation intensity distribution are input into the spatial weighted regression model.

[0135] Weighted least squares calculation is performed to obtain the spatial distribution of the regression coefficient, the spatial variance of the regression coefficient is calculated, the region with significant spatial heterogeneity is determined, and natural segmentation classification is performed in the region with significant spatial heterogeneity to divide the urban functional zones with significant spatial differences.

[0136] The diffusion wave source point specifically refers to the spatial position corresponding to the urban functional unit in the knowledge graph, which serves as the starting point for influence diffusion. Similar to the "wave source" in physics, it is used to describe how the influence of this functional unit spreads outward. For example, a newly built large commercial center will attract surrounding residents to shop and consume, and drive changes in the surrounding areas. This commercial center is a "diffusion wave source point" because it is the starting point for influence diffusion, and these influences spread around like "waves".

[0137] The natural segmentation classification specifically refers to a data classification method. It automatically finds appropriate demarcation points according to the distribution of the data itself, divides the data into different categories, so that the data within the same category is similar, while the differences between categories are relatively large. For example, when dividing urban functional areas, if the commercial density in some areas is much higher than that in other areas, the algorithm will automatically classify these areas as "commercial areas" instead of forcing them to be mixed with ordinary residential areas; it is more scientific than setting thresholds artificially and can classify reasonably according to the actual distribution of the data, rather than making a rigid division according to fixed rules.

[0138] In a specific implementation, first, construct a knowledge graph of the urban system. This graph covers various functional units in the city, such as residential areas, commercial areas, industrial areas, parks, etc., and records their spatial location coordinates, influence ranges, and the degree of association with each other. For example, the influence range of a shopping mall may be 1 kilometer around it, and its degree of association with nearby residential areas and restaurants is relatively high, while its degree of association with industrial areas farther away is relatively low. This information can be obtained through various means such as open data, on-site research, and sensor data, and stored in a database to form a knowledge graph containing nodes (functional units) and edges (association relationships).

[0139] Next, set the spatial location of the functional unit as the diffusion wave source point, and calculate the wave propagation speed coefficient and wave attenuation coefficient based on the degree of association between functional units. The higher the degree of association, the larger the propagation speed coefficient and the smaller the attenuation coefficient, indicating that this functional unit has a greater influence on the surrounding area and a wider propagation range. For example, the influence of a large hospital is usually greater than that of a small clinic and has a wider propagation range. Suppose the degree of association between a shopping mall and a residential area is 0.8, then it can be converted into a propagation speed coefficient and an attenuation coefficient according to an empirical formula. For example, the speed coefficient is 0.8×10, and the attenuation coefficient is 1 / (0.8×5).

[0140] Then, establish the urban element diffusion wave equation. This equation describes the propagation law of urban elements (such as population, commercial activities, traffic flow, etc.) in space. The equation contains an external excitation term, which is calculated by multiplying the influence intensity of the functional unit by the wave attenuation coefficient. For example, the influence intensity of a shopping mall can be evaluated according to indicators such as its scale and turnover.

[0141] To solve this equation, divide the computational domain into irregular triangular grid cells and establish a linear interpolation function. The size of the grid cells increases with the increase of the distance from the wave source point to improve the calculation efficiency. For example, the grid cells near the shopping mall can be set smaller to more finely simulate its influence range, while the grid cells far from the shopping mall can be set larger.

[0142] Within each grid cell, the stiffness matrix is calculated according to the linear interpolation function, and the mass matrix representing the node mass distribution is established. Substitute the mass matrix and the stiffness matrix into the wave equation to obtain the discretized matrix equation.

[0143] The time-step integration method is used to solve the discretized matrix equation. Within each time step, a predictor-corrector iteration is performed to calculate the displacement field, velocity field, and acceleration field. These field quantities are combined to form a spatio-temporal evolution sequence, and based on this sequence, the propagation velocity field of the state quantity and the gradient field of the state quantity are calculated. For example, the change in population density at different time steps and the gradient change of population density in space can be calculated.

[0144] Based on the propagation velocity field of the state quantity and the gradient field of the state quantity, the density distribution of the functional units is calculated to obtain the spatial distribution density of the functional units. For example, the density distribution of residential areas, commercial areas, and industrial areas within a certain region can be calculated. At the same time, the correlation degree between functional units is interpolated to obtain the distribution of functional association strength.

[0145] Finally, a spatial weighted regression model is constructed, and the spatial distribution density of functional units and the distribution of functional association strength are input into this model. The spatial distribution of the regression coefficients is obtained through weighted least squares calculation, and the spatial variance of the regression coefficients is calculated to determine the regions with significant spatial heterogeneity. Natural break classification is performed in these regions to divide the urban functional zones with significant spatial differences. For example, based on the spatial distribution of the regression coefficients, the city can be divided into different functional zones, such as the core commercial area, residential area, industrial area, etc.

[0146] In this embodiment, by establishing the diffusion wave equation of urban elements and combining with the spatial weighted regression model, the spatial distribution characteristics of urban functional zones can be more precisely characterized, and the regions with significant spatial differences can be identified; by calculating the spatial variance of the regression coefficients, the regions with significant spatial heterogeneity can be identified, and the interaction relationship between different functional zones can be deeply understood; it can provide a scientific basis for urban planning, traffic management, resource allocation, etc., which helps to optimize the urban spatial structure and improve the urban operation efficiency.

[0147] In an alternative embodiment, a system coupling differential equation group model is established for the functional zones and solved to obtain the prediction results of the element state changes. The time-series dependence relationship is calculated based on conditional mutual information to obtain the causal chain of element changes. The system vulnerability index is constructed through the causal chain, and multi-objective optimization is used to obtain the urban system resilience analysis results, including:

[0148] A coupled differential equation model of the urban system is established for the urban functional zoning. The adaptive step - size spectral element method is used to numerically solve the coupled differential equation model of the urban system. The time domain is divided into multiple non - uniform sub - intervals. In each non - uniform sub - interval, Legendre polynomials are used to construct high - order basis functions. The time step is dynamically adjusted through error estimation. The high - order basis functions are substituted into the coupled differential equation model of the urban system for Galerkin projection to obtain the prediction results of the changes in the states of urban elements.

[0149] Calculate the marginal information entropy of each urban element in the prediction results of the changes in the states of urban elements, calculate the joint information entropy of any combination of urban element pairs, and the marginal information entropy of the remaining urban elements other than the urban element pairs. Combine the marginal information entropy, joint information entropy, and marginal information entropy of the remaining urban elements in the urban element pairs to calculate the conditional mutual information. Determine whether there is a temporal dependence relationship between the urban element pairs according to the magnitude relationship between the conditional mutual information and the preset significance threshold. Traverse all urban element pairs in the prediction results of the changes in the states of urban elements to establish a causal chain of urban element changes.

[0150] Calculate the in - degree value and out - degree value of each node corresponding to an urban element in the causal chain of urban element changes. Multiply the in - degree value by the in - degree weight to obtain the in - degree weighted value, multiply the out - degree value by the out - degree weight to obtain the out - degree weighted value, and add the in - degree weighted value and the out - degree weighted value to obtain the urban system vulnerability index.

[0151] Establish a multi - objective optimization model with the goal of minimizing the urban system vulnerability index and the regulation cost. Set the urban resource constraints as the constraints of the multi - objective optimization model. Solve the multi - objective optimization model through the non - dominated sorting genetic algorithm to obtain a set of urban system resilience analysis results that meet the constraints.

[0152] The Legendre polynomials specifically refer to a set of special polynomials, which are commonly used in numerical calculations and physical modeling. They are independent of each other within a specific interval and do not affect each other. Therefore, they are suitable as the basis functions of mathematical models. Using Legendre polynomials, complex functions can be expressed as a combination of multiple simple terms, making the calculation more accurate.

[0153] The Galerkin projection specifically refers to a numerical method for solving equations. Its core idea is to use a set of basis functions (such as Legendre polynomials) to approximate the solution of the equation, and then require the error to be as small as possible in the directions of these basis functions. In this way, the original complex equation can be converted into an algebraic equation system, which is convenient for computer solution.

[0154] In a specific implementation, first, the city is divided into different functional zones, such as residential areas, commercial areas, industrial areas, etc. For each functional zone, key urban elements are identified, such as population, economy, environment, infrastructure, etc. Then, according to the interaction relationships between the elements, a coupled differential equation system model is established to describe the variation laws of each element over time. The mutual promotion or inhibition relationships between the elements, as well as the influence of external factors, are considered in the model. For example, population growth promotes economic development but also increases environmental pressure.

[0155] Taking a simple urban system containing three elements of population, economy, and environment as an example. Population growth is affected by birth rate, death rate, and migration rate; economic development is affected by population size, investment, and technological level; environmental quality is affected by population density, economic activities, and pollution control. These influencing factors are interrelated and constitute a complex dynamic system.

[0156] Next, the adaptive step - size spectral element method is used to solve the coupled differential equation system model. The time domain is divided into multiple non - uniform sub - intervals, and high - order basis functions are constructed using Legendre polynomials within each sub - interval. The time step is dynamically adjusted through error estimation to ensure the calculation accuracy. The high - order basis functions are substituted into the coupled differential equation system model for Galerkin projection to obtain the prediction results of the changes in the urban element states. For example, predicting the trends of population, economy, and environment in the next decade.

[0157] Then, based on the prediction results, the causal chains of the changes in urban elements are calculated. First, the marginal information entropy of each urban element, the joint information entropy of any two combinations of urban elements, and the marginal information entropy of the remaining urban elements are calculated. Then, these information entropies are combined to calculate the conditional mutual information. According to the magnitude relationship between the conditional mutual information and a preset significance threshold, it is judged whether there is a time - series dependence relationship between two urban elements. For example, if the conditional mutual information between population change and economic change is greater than the significance threshold, it is considered that population change has a significant impact on economic change.

[0158] Suppose the conditional mutual information between population and economy is 0.8, which is greater than the preset significance threshold of 0.5; the conditional mutual information between population and environment is 0.6, which is also greater than the significance threshold; while the conditional mutual information between economy and environment is 0.3, which is less than the significance threshold. Then a causal chain can be established: population change affects economic change, which in turn affects environmental change.

[0159] Traverse all pairs of urban elements to establish a complete causal chain of urban element changes. According to the causal chain, calculate the in-degree value and out-degree value of each node corresponding to the urban element. Multiply the in-degree value by the in-degree weight to obtain the in-degree weighted value, and multiply the out-degree value by the out-degree weight to obtain the out-degree weighted value. Add the in-degree weighted value and the out-degree weighted value to obtain the urban system vulnerability index. For example, the in-degree of the population node is 0, and the out-degree is 2. Assuming that both the in-degree weight and the out-degree weight are 1, the vulnerability index of the population node is 2. The in-degree of the economic node is 1, the out-degree is 1, and the vulnerability index is 2. The in-degree of the environmental node is 2, the out-degree is 0, and the vulnerability index is 2.

[0160] Finally, establish a multi-objective optimization model aiming to minimize the urban system vulnerability index and the regulation cost. Set the urban resource constraints as the constraints of the multi-objective optimization model. For example, the available financial budget, land resources, etc. Solve the multi-objective optimization model through the non-dominated sorting genetic algorithm to obtain a set of urban system resilience analysis results that meet the constraints, including the optimal resource regulation strategy. For example, increasing the investment in environmental protection can reduce the environmental vulnerability and enhance the urban system resilience.

[0161] In this embodiment, the adaptive step-size spectral element method is used to solve the coupled differential equation group model, which can more accurately predict the change trend of the urban element state and provide a more reliable basis for urban planning and management; through conditional mutual information analysis, the complex temporal dependence relationship between urban elements can be identified, the internal operation mechanism of the urban system can be revealed, and the law of urban system evolution can be helped to understand; based on the multi-objective optimization model, a more effective resource regulation strategy can be formulated to reduce the urban system vulnerability, enhance the urban system resilience, and strengthen the urban's ability to cope with various disturbances.

[0162] Figure 2 This is a schematic structural diagram of the multi-source information fusion data space analysis system based on the knowledge graph according to the embodiment of the present invention, as Figure 2 shown, the system includes:

[0163] The first unit is used to establish a data quality benchmark threshold for the multi-source heterogeneous data in the urban system by using the multivariate statistical process control method and perform outlier detection to obtain preliminary calibrated data, calculate the spatial transformation matrix for the preliminary calibrated data by using the least squares registration algorithm to obtain the calibrated data in the unified coordinate system; calculate the credibility score of the calibrated data by using recursive Bayesian estimation, and perform dynamic adjustment through the Markov decision process to obtain the data fusion weight, and perform weighted fusion on the calibrated data to obtain the standardized fusion data;

[0164] The second unit is used to receive the standardized fusion data by adopting a double-buffer mechanism and convert it into an initial graph structure, construct an urban ontology model based on formal concept analysis and map the nodes of the initial graph structure to obtain urban element nodes with category and hierarchical relationships; perform a random walk algorithm with a time-series decay factor on the urban element nodes to obtain dynamic representation vectors and calculate the association strength between nodes, construct urban functional communities by means of modularity optimization method, and optimize and reconstruct according to the minimum criterion of graph structure entropy to obtain an urban system knowledge graph;

[0165] The third unit is used to construct an element diffusion wave equation model based on the urban system knowledge graph and solve it to obtain the state propagation law, and obtain the urban functional zoning by analyzing the spatial correlation through geographically weighted regression; establish a system coupling differential equation group model for the functional zoning and solve it to obtain the prediction result of the element state change, calculate the time-series dependence relationship based on conditional mutual information to obtain the causal chain of element changes, construct a system vulnerability index through the causal chain, and obtain the urban system resilience analysis result by means of multi-objective optimization.

[0166] In the third aspect of the embodiments of the present invention,

[0167] A kind of electronic device is provided, including:

[0168] A processor;

[0169] A memory for storing instructions executable by the processor;

[0170] Wherein, the processor is configured to call the instructions stored in the memory to execute the method described above.

[0171] In the fourth aspect of the embodiments of the present invention,

[0172] A computer-readable storage medium is provided, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method described above is implemented.

[0173] The present invention can be a method, a device, a system and / or a computer program product. The computer program product can include a computer-readable storage medium, on which computer-readable program instructions for executing various aspects of the present invention are uploaded.

[0174] 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 foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-source information fusion data space analysis method based on knowledge graph, characterized by: include: For multi-source heterogeneous data in urban systems, a multivariate statistical process control method is used to establish a data quality benchmark threshold and perform outlier detection to obtain preliminary calibration data. The least squares registration algorithm is used to calculate the spatial transformation matrix of the preliminary calibration data to obtain calibration data in a unified coordinate system. Recursive Bayesian estimation is used to calculate the credibility score of the calibration data, and the Markov decision process is used to dynamically adjust it to obtain the data fusion weight. The calibration data is weighted and fused to obtain standardized fused data. A double buffer mechanism is used to receive standardized fusion data and convert it into an initial graph structure. Based on formal concept analysis, an urban ontology model is constructed and the initial graph structure nodes are mapped to obtain urban element nodes with category and hierarchical relationships. A random walk algorithm with a time-series decay factor is used to obtain dynamic representation vectors for urban element nodes and calculate the strength of association between nodes. An urban functional community is constructed through a modularity optimization method, and the graph structure entropy minimization criterion is used to optimize and reconstruct the urban system knowledge graph. Based on the urban system knowledge graph, a factor diffusion wave equation model is constructed and the state propagation law is obtained by solving it. The spatial correlation is analyzed by geographical weighted regression to obtain the urban functional zoning. For the functional zoning, a system coupling differential equation group model is established and solved to obtain the factor state change prediction results. The time series dependency is calculated based on the conditional mutual information to obtain the causal chain of factor changes. The system vulnerability index is constructed through the causal chain, and the urban system resilience analysis results are obtained by multi-objective optimization.

2. The method according to claim 1, characterized in that: For multi-source heterogeneous data in urban systems, a multivariate statistical process control method is used to establish a data quality benchmark threshold and perform outlier detection to obtain preliminary calibration data. The least squares registration algorithm is used to calculate the spatial transformation matrix of the preliminary calibration data to obtain calibration data in a unified coordinate system, including: Acquire multi-source heterogeneous data of the urban system, perform dimensionality reduction processing on the multi-source heterogeneous data of the urban system by a principal component analysis method to obtain a principal component space and a residual space, calculate the principal component statistics of the multi-source heterogeneous data of the urban system based on the principal component space, and calculate the residual statistics of the multi-source heterogeneous data of the urban system based on the residual space; Based on the principal component statistic, a first control limit is obtained by probability distribution calculation, and the parameters of the probability distribution are determined by the number of principal components and the number of samples; based on the residual statistic, a second control limit is obtained by weighted distribution calculation, and the weight of the weighted distribution is determined by the eigenvalue of the residual space; the first control limit and the second control limit are combined to construct a data quality benchmark threshold; Using the data quality benchmark threshold, outlier detection is performed on the multi-source heterogeneous data of the urban system, and data exceeding the data quality benchmark threshold is marked as an outlier and removed to obtain preliminary calibration data; constructing a spatial point set for the preliminary calibration data, calculating a covariance matrix of the spatial point set, and performing eigenvalue decomposition on the covariance matrix to obtain a rotation matrix; A least squares objective function is constructed based on the rotation matrix, and a spatial transformation matrix and a translation vector are solved by minimizing the least squares objective function, wherein the least squares objective function represents the sum of squares of distances of the spatial point set in a unified coordinate system; The spatial transformation matrix and the translation vector are applied to the preliminary calibration data, and the preliminary calibration data is converted into a unified urban space coordinate system to obtain calibration data in a unified coordinate system.

3. The method according to claim 1, characterized in that Recursive Bayesian estimation is used to calculate the credibility score of the calibration data, and the Markov decision process is used to dynamically adjust the data fusion weights. The calibration data is weighted and fused to obtain standardized fusion data, including: Construct a state space model including state equations and observation equations, calculate the time evolution value of the credibility state of the calibration data through the state equation, and introduce the normally distributed system noise, calculate the mapping value between the observation value and the credibility state through the observation equation, and introduce the normally distributed observation noise; The probability density propagation equation is used to calculate the propagation result of the posterior probability density from the previous moment to the current moment to obtain the prior probability density, and the integration result of the new observation information and the prior probability density is calculated based on the conditional probability formula to obtain the posterior probability density, and the mathematical expectation of the posterior probability density is calculated to obtain the credibility score of the calibration data; Discretize the distribution interval of the credibility score to obtain the state space, set credibility enhancement, attenuation, and maintenance as the action space, generate the state transition probability matrix, build a Markov decision model, and set an immediate reward function including a credibility enhancement reward item and a credibility fluctuation penalty item in the Markov decision model; Based on the state transition probability matrix and the instant reward function, iterative calculation of the value function is performed, an optimal value function is obtained through iterative updating, a mapping relationship from state to action is generated based on the optimal value function to obtain an optimal strategy, a dynamic adjustment result of the credibility score is calculated using the optimal strategy, and the adjusted credibility score is normalized to obtain a data fusion weight; Generate a spatial kernel function with an adaptive bandwidth parameter, use the spatial kernel function to perform local smoothing on the data fusion weight to obtain a smooth fusion weight, and use the smooth fusion weight to calculate the weighted average of the calibration data at the same spatial position; The weighted fusion calculation is repeatedly performed on the calibration data of all spatial locations in the urban system. Spatial neighborhood constraints are introduced in the fusion calculation to determine the spatial continuity. The weighted average values ​​of each spatial location are combined to generate standardized fusion data of the urban system.

4. The method according to claim 1, characterized in that: A double buffer mechanism is used to receive standardized fusion data and convert it into an initial graph structure. Based on formal concept analysis, an urban ontology model is constructed and the initial graph structure nodes are mapped. The urban element nodes with category and hierarchical relationships are obtained, including: Two buffers of equal size are set, namely a first buffer and a second buffer, and the first buffer and the second buffer are allocated to independent physical storage spaces by memory mapping. When the first buffer receives standardized fusion data of the urban system, the second buffer performs a spatiotemporal attribute separation operation to obtain spatiotemporal separation data. The first buffer sets a data reception completion flag. When the data reception completion flag is set, the physical storage space mapping relationship between the first buffer and the second buffer is exchanged, and the spatiotemporal separation data is used as input data for urban element construction. Construct a parallel pipeline architecture including a data parsing stage, a node generation stage, and an edge relationship construction stage. In a multi-processor environment, each processing stage is connected through a pipeline register. The data parsing stage performs field segmentation on the time-space separation data according to a preset data structure template, and adds semantic tags to obtain a data parsing result. The node generation stage creates a graph node based on the data parsing result, and assigns a node identifier to each of the graph nodes through a distributed identifier generator to obtain a graph node set. The edge relationship construction stage uses a spatial index structure to calculate the neighborhood relationship of the nodes in the graph node set, and establishes edge connections based on the neighborhood relationship and the similarity of node attributes to obtain an initial graph structure. Establishing a formal background, the formal background includes a set of urban element objects, a set of urban element attributes, and a set of relationships between objects and attributes, performing context compatibility verification on the objects and attributes in the formal background, using an incremental algorithm to perform a closure operation on the formal background to obtain a hierarchical graph of a concept lattice, and converting the hierarchical graph into an urban ontology model described based on a knowledge representation language; The semantic similarity of the graph node attributes in the initial graph structure and the concept class in the city ontology model is calculated, and the concept class corresponding to the node is determined based on the maximum value of the semantic similarity to obtain the node category. The node hierarchical relationship including subordinate and composition relationships is constructed through ontology reasoning rules to obtain a node relationship set. The node relationship set is iteratively calculated by transitive closure based on the adjacency matrix, and the calculation result is combined with the node category to obtain an urban element node graph with category and hierarchical relationship.

5. The method according to claim 1, characterized in that A random walk algorithm with a time-series attenuation factor is used to obtain a dynamic representation vector for the city element nodes. The association strength between nodes is calculated by the dynamic representation vector. Based on the association strength, an urban functional community is constructed by a modularity optimization method. The urban functional community is optimized and reconstructed by a graph structure entropy minimization criterion, and the urban system knowledge graph is obtained, including: A basic transfer matrix is ​​constructed according to the spatial distance between the city element nodes; a time series decay factor is introduced into the basic transfer matrix to form a time series adjustment matrix, wherein the time series decay factor decreases as the time interval increases; the basic transfer matrix is ​​multiplied by the time series adjustment matrix to obtain a time series transfer probability matrix, random walk sampling is performed on the city element nodes based on the time series transfer probability matrix to obtain a node neighborhood sequence, and the node neighborhood sequence is mapped to a low-dimensional vector space to obtain a node dynamic representation vector; Based on adjacent city element nodes, node pairs are determined, the cosine similarity between the node dynamic representation vectors of the node pairs is calculated, and according to a preset time window range, the number of interactions of the node pairs within the time window range is counted, the cosine similarity and the number of interactions are weightedly combined to obtain the node association strength, and the node association strength is filled to form an association strength matrix; Extract the sum of the association strengths of the nodes within the community from the association strength matrix, count the degree distribution of the city element nodes and calculate the expected random association strength, subtract the expected random association strength from the sum of the association strengths to construct a modularity objective function, perform iterative optimization to maximize the modularity objective function, and divide the city element nodes whose node association strength is greater than a preset strength threshold into initial urban function communities; Calculate the graph structure entropy of the initial urban functional community, randomly adjust the community affiliation of the urban element nodes to generate multiple candidate solutions, calculate the graph structure entropy change generated by each candidate solution, execute the candidate solution that reduces the graph structure entropy the most, repeat the node adjustment until the graph structure entropy change is less than a preset change threshold, and obtain the optimized urban functional community; The optimized urban functional community is defined as a functional unit, the correlation strength between the internal nodes of the functional unit is counted to calculate the correlation degree between the functional units, and the functional units and the corresponding correlation degrees are organized into an urban system knowledge graph.

6. The method according to claim 1, characterized in that Based on the urban system knowledge graph, the factor diffusion wave equation model is constructed and the state propagation law is solved. The spatial correlation is analyzed through geographical weighted regression to obtain the urban functional zoning, including: The spatial position coordinates and influence range of the functional unit in the urban system knowledge graph are obtained, and the spatial position of the functional unit is set as the diffusion wave source point; the degree of association between the functional units is extracted, converted into a wave propagation velocity coefficient, and the corresponding wave attenuation coefficient is calculated; an urban element diffusion wave equation centered on the diffusion wave source point is established, and the influence intensity corresponding to the functional unit is multiplied by the wave attenuation coefficient to calculate the external excitation term, and the external excitation term is substituted into the urban element diffusion wave equation; The calculation domain is divided into irregular triangular grid units with the diffusion wave source point as the center, and the size of the irregular triangular grid unit is increased accordingly as the distance from the diffusion wave source point increases; a linear interpolation function is established in the irregular triangular grid unit, and the stiffness matrix of the irregular triangular grid unit is calculated according to the linear interpolation function; a mass matrix representing the node mass distribution is established, and the mass matrix and the stiffness matrix are substituted into the urban element diffusion wave equation to obtain a discretized matrix equation; Execute time step integral calculation, and perform estimation-correction iteration in each time step: substitute the mass matrix and the stiffness matrix into the calculation of the estimated displacement field, velocity field and acceleration field, and perform equilibrium iteration to calculate the corrected displacement field, the corrected velocity field and the corrected acceleration field; compose the corrected displacement field, the corrected velocity field and the corrected acceleration field into a spatiotemporal evolution sequence, obtain the state quantity propagation velocity field and the state quantity gradient field according to the spatiotemporal evolution sequence, calculate the density distribution of the functional unit based on the state quantity propagation velocity field and the state quantity gradient field, and obtain the spatial distribution density of the functional unit; interpolate and calculate the degree of association to obtain the functional association strength distribution; Constructing a spatial weighted regression model, setting the regression coefficient to change continuously with the spatial position, calculating the density distribution of the observation points to determine the bandwidth of the spatial weight function, and performing spatial weighting of the observation points; inputting the spatial distribution density of the functional units and the functional association intensity distribution into the spatial weighted regression model; The weighted least squares calculation is performed to obtain the spatial distribution of the regression coefficient, the spatial variance of the regression coefficient is calculated, the area of ​​significant spatial heterogeneity is determined, and natural segmentation classification is performed in the area of ​​significant spatial heterogeneity to divide the urban functional zones with significant spatial differences.

7. The method according to claim 1, characterized in that A system coupling differential equation group model is established for functional zoning and solved to obtain the prediction results of element state changes. The temporal dependency relationship is calculated based on conditional mutual information to obtain the causal chain of element changes. The system vulnerability index is constructed through the causal chain. The results of urban system resilience analysis obtained by multi-objective optimization include: A model of coupled differential equations of an urban system is established for the urban functional zoning, and an adaptive step-size spectral element method is used to numerically solve the model of coupled differential equations of an urban system, and the time domain is divided into a plurality of non-uniform sub-intervals. A high-order basis function is constructed in each non-uniform sub-interval using Legendre polynomials, and the time step is dynamically adjusted through error estimation. The high-order basis function is substituted into the model of coupled differential equations of an urban system for Galerkin projection, and a prediction result of the state change of urban elements is obtained; Calculate the edge information entropy of each urban element in the urban element state change prediction result, and calculate the joint information entropy of any urban element pair combination, as well as the edge information entropy of the remaining urban elements outside the urban element pair; combine the edge information entropy, joint information entropy and edge information entropy of the remaining urban elements in the urban element pair to calculate the conditional mutual information; determine whether there is a temporal dependency relationship between the urban element pairs according to the relationship between the conditional mutual information and a preset significance threshold; traverse all urban element pairs in the urban element state change prediction result to establish a causal chain of urban element changes; Calculate the in-degree value and out-degree value of the node corresponding to each urban element in the causal chain of urban element change, multiply the in-degree value by the in-degree weight to obtain the in-degree weighted value, multiply the out-degree value by the out-degree weight to obtain the out-degree weighted value, and add the in-degree weighted value and the out-degree weighted value to obtain the urban system vulnerability index; A multi-objective optimization model is established with the goal of minimizing the vulnerability index and regulation cost of the urban system. The urban resource constraints are set as the constraints of the multi-objective optimization model. The multi-objective optimization model is solved by a non-dominated sorting genetic algorithm to obtain a set of urban system resilience analysis results that meet the constraints.

8. A multi-source information fusion data space analysis system based on knowledge graph, used to implement the method described in any one of claims 1 to 7, characterized in that: include: The first unit is used to establish data quality benchmark thresholds and perform outlier detection for multi-source heterogeneous data in urban systems using multivariate statistical process control methods to obtain preliminary calibration data, and to calculate the spatial transformation matrix of the preliminary calibration data using the least squares registration algorithm to obtain calibration data in a unified coordinate system; to calculate the credibility score of the calibration data using recursive Bayesian estimation, to dynamically adjust it through the Markov decision process to obtain data fusion weights, and to perform weighted fusion on the calibration data to obtain standardized fused data; The second unit is used to receive standardized fusion data using a double buffer mechanism and convert it into an initial graph structure. Based on formal concept analysis, the city ontology model is constructed and the initial graph structure nodes are mapped to obtain city element nodes with category and hierarchical relationships. The random walk algorithm with a time-series decay factor is used to obtain dynamic representation vectors for the city element nodes and calculate the correlation strength between nodes. The urban functional community is constructed through the modularity optimization method, and the graph structure entropy minimization criterion is used to optimize and reconstruct the urban system knowledge graph. The third unit is used to construct an element diffusion wave equation model based on the urban system knowledge graph and solve it to obtain the state propagation law, and obtain the urban functional zoning through geographical weighted regression analysis of spatial correlation; establish a system coupling differential equation group model for functional zoning and solve it to obtain the element state change prediction results, calculate the time series dependency based on conditional mutual information to obtain the element change causal chain, construct the system vulnerability index through the causal chain, and use multi-objective optimization to obtain the urban system resilience analysis results.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method described in any one of claims 1 to 7.

10. 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 7 is implemented.

Images