Low-efficiency land intelligent identification and redevelopment potential evaluation method and system

Abstract

The invention discloses a low-efficiency land intelligent identification and redevelopment potential evaluation method and system. The method comprises the steps of obtaining land parcel attribute data; performing nonlinear correlation analysis on the attribute data based on the multi-dimensional index system to identify the low-efficiency land, including: preliminarily identifying whether the land is the low-efficiency land; determining space units belonging to low-efficiency land and optimizing and identifying low-efficiency types; fine recognition of low-efficiency sub-regions caused by mismatching of minimum unit structures in space units of low-efficiency land; constructing a potential evaluation index system to quantify the low-efficiency land evaluation redevelopment potential, and analyzing the redevelopment priority of the low-efficiency land evaluation plot through an entropy weight method and a multi-criterion decision; and integrating a geographic information system platform, realizing visual display of low-efficiency land use, performing policy simulation based on the redevelopment priority, outputting a simulation result and screening an optimal redevelopment mode. According to the invention, accurate identification, classification evaluation and redevelopment potential dynamic analysis of low-efficiency land can be realized.

Description

Technical Field

[0001] This application relates to the field of land resource management technology, specifically to a method and system for intelligent identification of inefficient land use and assessment of its redevelopment potential. Background Technology

[0002] In the fields of land resource management, urban renewal, and redevelopment of inefficient land, accurate identification and redevelopment potential assessment of inefficient land are crucial for rational land resource planning and improving the quality of urban development. Accurate identification and assessment help relevant departments better understand land use, formulate scientific and rational land policies and urban renewal plans, thereby improving land use efficiency and promoting sustainable urban development.

[0003] To address the issues of identifying inefficient land use and assessing its redevelopment potential, existing technologies typically rely on linear weighted evaluation models, such as the entropy weight method and the analytic hierarchy process (AHP). These methods assign weights to various economic and social indicators and then perform linear calculations to assess whether land is inefficiently used and its redevelopment potential. Additionally, some existing studies employ simple threshold classification methods, judging land use efficiency based on the values ​​of certain single indicators.

[0004] Existing technologies have significant shortcomings. On the one hand, current methods for identifying inefficient land use lack consideration of the nonlinear relationship between land parcel attributes and inefficiency levels, and struggle to handle the interactive effects of multi-dimensional features, thus failing to accurately reflect the true state of land use. On the other hand, existing research is still insufficient in explaining the theoretical presuppositions and logical relevance of redevelopment models, as well as summarizing and deducing multi-factor interaction mechanisms. It fails to fully consider the nonlinear relationship between indicators and inefficient use levels and neglects the identification of high-value land parcels with synergistic effects. Furthermore, current methods for identifying inefficient land use are largely driven by government land revenue and taxation, lacking consideration of comprehensive human-centered benefits, and often ignoring the impact of locational differences on land use efficiency, making it difficult to achieve refined identification. Summary of the Invention

[0005] To achieve accurate identification, classification assessment, and dynamic analysis of the redevelopment potential of inefficient land, this application provides a method and system for intelligent identification and redevelopment potential assessment of inefficient land.

[0006] Firstly, this application provides a method for intelligent identification and redevelopment potential assessment of inefficient land use, including: Acquire multi-dimensional attribute data of the land parcel and perform preprocessing; Nonlinear correlation analysis is performed on the collected data based on a multi-dimensional indicator system to identify inefficient land use. The nonlinear correlation analysis includes three levels of identification analysis. Preliminary identification includes: calculating location entropy, coupling degree, and coordination degree based on a location hierarchical model and a multi-dimensional indicator coupling and coordination model; dividing spatial units based on a Voronoi diagram; and setting threshold division and feature combinations to determine whether the divided spatial units are inefficient land use. Optimized identification includes: performing nonlinear clustering of multi-dimensional indicator features on the spatial units initially identified as inefficient land use to obtain inefficient land use feature clusters and determine the spatial units and inefficiency types belonging to inefficient land use. Refined identification includes: using graph neural networks to identify inefficient sub-regions caused by the mismatch of the smallest unit structure within the spatial units of inefficient land use. A potential assessment index system was constructed to quantify the redevelopment potential of inefficient land use, and the redevelopment priority of inefficient land use assessment plots was analyzed through entropy weight method and multi-criteria decision analysis. By integrating a geographic information system platform, we can visualize inefficient land use, conduct policy simulations based on redevelopment priorities, output simulation results, and select the optimal redevelopment model.

[0007] By adopting the above scheme, a multi-dimensional indicator system is used to conduct nonlinear correlation analysis on the multi-attribute data of land parcels. Inefficient land use is accurately identified through a three-level identification process. An evaluation indicator system is constructed to quantify redevelopment potential and assess priorities. A geographic information system platform is used to realize visualization and policy simulation, and the optimal redevelopment model is selected. This achieves multi-scale accurate identification from land parcels to the smallest unit, as well as integrated closed-loop management of inefficient land use identification, evaluation, and decision-making.

[0008] Preferably, the primary identification further includes: A dynamic location hierarchy model is set up to replace the location hierarchy model. The dynamic location hierarchy model adopts a temporal spatial syntax analysis algorithm to calculate the spatial integration degree over several consecutive years and perform location division. For each spatial unit, multi-scale location entropy is calculated, and the smallest scale location entropy is selected as the final location entropy. The multi-scale location entropy includes: a first-scale entropy covering a single spatial unit, a second-scale location entropy covering a single spatial unit and its adjacent spatial units, and a third-scale location entropy covering spatial units within the administrative region where the single spatial unit is located. Different scale entropies are set with corresponding scale judgment location entropy thresholds, and the annual change rate of the location entropy at the corresponding scale is calculated. If the annual change rate of location entropy is lower than the preset annual change rate of location entropy, the judgment location entropy threshold at the corresponding scale is adjusted.

[0009] By adopting the above scheme, multi-scale location entropy calculation is performed using a dynamic location gradation model and thresholds are set. The judgment threshold is adjusted in combination with the annual change rate of location entropy, which more accurately considers the dynamic changes in the location conditions of the plot and improves the accuracy of preliminary identification of inefficient land use.

[0010] Preferably, the primary identification further includes: Based on the multi-dimensional indicator coupling and collaboration model, time-series fusion is performed on each dimension indicator to obtain static indicators and dynamic trend indicators for each dimension. Based on the static indicators and dynamic trend indicators for each dimension, the static nonlinear coupling degree and dynamic nonlinear coupling degree are calculated and obtained. The time-series nonlinear coupling degree obtained by weighted calculation is used as the final coupling degree.

[0011] By adopting the above scheme and combining a multi-dimensional indicator coupling and synergy model for time series fusion, static indicators and dynamic trend indicators of each dimension are obtained, and then the time series nonlinear coupling degree is calculated as the final coupling degree. This allows for a more accurate capture of the nonlinear correlation of multi-dimensional indicators in the time series, thereby improving the accuracy and reliability of the initial identification of inefficient land use.

[0012] Preferably, the optimized identification further includes: For spatial units initially identified as inefficient land use, a multi-dimensional index feature nonlinear clustering algorithm, DBSCAN, is used. The clustering process includes: using the spatial units initially identified as inefficient land use as candidate plots, standardizing the multi-dimensional index features of each spatial unit; calculating the attribute distance of non-spatial attributes between spatial units based on the Mahalanobis distance algorithm, calculating the spatial distance between spatial units using the Huffman distance algorithm, obtaining the mixed distance between spatial units through weighted calculation, and constructing a mixed distance matrix; setting DBSCAN clustering parameters; traversing the spatial units to identify core points, boundary points, and noise points; performing spatial constraint verification to filter spatially discrete clusters; and outputting spatially continuous inefficient land use and its inefficiency type.

[0013] By adopting the above scheme, the DBSCAN clustering algorithm is used to perform nonlinear clustering on the initially identified inefficient land spatial units. A hybrid distance matrix is ​​constructed by combining attribute distance and spatial distance to filter spatial discrete clusters, thereby more accurately identifying spatially continuous inefficient land and its inefficiency types.

[0014] Preferably, the optimized identification further includes: constructing a heterogeneous temporal graph attention network model; the node definition in the heterogeneous temporal graph attention network model includes: spatial units as core nodes, POIs and roads as auxiliary nodes; the edge definition includes: spatially adjacent edges and functionally related edges; the heterogeneous temporal graph attention network model is equipped with spatial attention and feature attention mechanisms, and embeds an LSTM module to capture the temporal trend of input dynamic indicators; and is trained and generated by multi-dimensional indicator features corresponding to spatial units that are historically labeled as inefficient; the multi-dimensional indicator features corresponding to spatial units initially identified as inefficient land use are input into the heterogeneous temporal graph attention network model to obtain the output judgment result of whether the spatial unit is inefficient land use; The refined identification also includes: using the inefficiency score, inefficiency type and feature weight in the output spatial unit in the judgment result of whether the spatial unit is an inefficient land use to correct the preset minimum unit inefficiency probability threshold and minimum unit structure mismatch feature weight threshold in the process of identifying inefficient sub-regions caused by minimum unit structure mismatch in the spatial unit of inefficient land use, so as to optimize the identification result of inefficient sub-regions caused by minimum unit structure mismatch.

[0015] By adopting the above scheme, in addition to using clustering algorithms to identify inefficient land use, a heterogeneous time-series graph attention network model is constructed and trained in combination with multi-dimensional index features to verify whether the spatial units initially identified as inefficient land use are indeed inefficient land use. At the same time, relevant information in the judgment results is used to correct the judgment threshold in the fine recognition, thereby improving the accuracy of inefficient sub-region identification.

[0016] Preferably, the construction of the potential assessment index system quantifies the redevelopment potential of inefficient land use, and the redevelopment priority of inefficient land use assessment plots is analyzed through entropy weight method and multi-criteria decision analysis, including: Combining policy adaptation theory, market benefit model, and ecological constraints, a potential assessment index system is constructed, which includes economic power, social power, and ecological power. Each potential assessment index is standardized. The information entropy and difference coefficient of each potential assessment index are calculated using the entropy weight method to obtain the entropy weight of the corresponding potential assessment index. The obtained entropy weights are adjusted and corrected using the AHP method to obtain the corresponding weight of each potential assessment index. Using a multi-criteria decision analysis algorithm, a weighted standardized matrix is ​​constructed, and the weights of the potential assessment indicators related to the identified inefficiency types are further adjusted to determine the ideal solution. The Euclidean distance between the spatial unit and the ideal solution is calculated, and the proximity is calculated. The proximity is compared with a preset proximity threshold to determine the proximity range corresponding to the proximity, and the redevelopment priority is determined based on the proximity range. Based on the dimensions of each potential assessment indicator, the proximity is decomposed into sub-dimensions, and the magnitude of the proximity of each sub-dimension is compared. The redevelopment direction is determined based on the dimension corresponding to the maximum sub-dimension proximity.

[0017] By adopting the above scheme, a multi-dimensional indicator system for evaluation is constructed to comprehensively consider the redevelopment potential of inefficient land. The entropy weight method and AHP method are used to determine the indicator weights, and the indicator weights are adjusted according to the type of inefficiency to improve the accuracy of weight calculation. The multi-criteria decision analysis algorithm is used to more accurately determine the redevelopment priority, and the proximity decomposition of the dimensions is used to determine the redevelopment direction, guiding the specific direction of inefficient land redevelopment.

[0018] Preferably, the step of performing policy simulation based on redevelopment priorities, outputting simulation results, and selecting the optimal redevelopment model includes: A redevelopment policy simulation engine is set up; the redevelopment policy simulation engine has several built-in redevelopment modes, and each redevelopment mode can simulate the effect of at least one of the following policies: floor area ratio incentive policy, mixed land use policy, ecological restoration policy, and stock renewal policy; simulation resources and depth are divided according to the redevelopment priority of each spatial unit. The higher the redevelopment priority, the higher the proportion of simulation resources and the deeper the simulation level; scenario simulation is performed based on each redevelopment mode, the simulation results of each redevelopment mode are output, the closeness is recalculated, and the policy corresponding to the optimal redevelopment mode is selected.

[0019] By adopting the above scheme, simulation resources and depth are reasonably allocated according to redevelopment priorities, and the policy effects of various redevelopment models are simulated in depth. The optimal redevelopment model corresponding to the most suitable and accurate policy is selected by recalculating the closeness, thereby providing a more targeted and effective decision-making basis for the redevelopment of inefficient land.

[0020] Secondly, this application provides a system for intelligent identification and redevelopment potential assessment of inefficient land use, comprising: The data acquisition and preprocessing module is used to acquire multi-dimensional attribute data of the land parcel and perform preprocessing. The inefficient land use intelligent identification module is used to identify inefficient land use by performing nonlinear correlation analysis on the collected data based on a multi-dimensional indicator system. The nonlinear correlation analysis includes three levels of identification analysis. Preliminary identification includes: calculating the location entropy, coupling degree, and coordination degree based on a location hierarchical model and a multi-dimensional indicator coupling and coordination model; dividing spatial units based on a Voronoi diagram; and setting threshold division and feature combinations to determine whether the divided spatial units are inefficient land use. Optimized identification includes: performing nonlinear clustering of multi-dimensional indicator features on the spatial units initially identified as inefficient land use to obtain inefficient land use feature clusters and determine the spatial units and inefficiency types belonging to inefficient land use. Refined identification includes: using graph neural networks to identify inefficient sub-regions caused by the mismatch of the smallest unit structure within the spatial units of inefficient land use. The redevelopment potential dynamic assessment module is used to construct a potential assessment index system to quantify the redevelopment potential of inefficient land use, and to analyze the redevelopment priority of inefficient land use assessment plots through entropy weight method and multi-criteria decision analysis. The redevelopment decision simulation and screening module is used to integrate a geographic information system platform to visualize inefficient land use, simulate policies based on redevelopment priorities, output simulation results, and screen the optimal redevelopment model.

[0021] By adopting the above scheme, we can collect multi-dimensional attribute data of land parcels, perform nonlinear correlation analysis to accurately identify inefficient land use, construct an indicator system to quantify redevelopment potential and assess priorities, and use a geographic information system platform for visualization and policy simulation to select the optimal mode, thereby improving the accuracy of inefficient land use identification and the scientific nature of redevelopment decisions.

[0022] Thirdly, this application provides a computer-readable storage medium including a stored computer program, wherein the computer program, when running, controls the device where the computer-readable storage medium is located to perform the method described above.

[0023] Fourthly, this application provides a computer device, the computer device including a memory, a processor and a program stored in the memory and executable thereon, the program being executed by the processor to implement the steps of the method described above.

[0024] In summary, this application has the following beneficial effects: 1. Acquire multi-attribute data and combine it with a multi-dimensional indicator system for nonlinear correlation analysis to overcome the shortcomings of existing linear weighted models that ignore nonlinear relationships; achieve accurate multi-scale identification from land parcels to the smallest unit through a three-level identification process; and use graph neural networks to analyze the inherent structural mismatch mechanism of inefficient land use to achieve accurate and dynamic identification of inefficient land use; visualize inefficient land use through a geographic information system platform and select the optimal redevelopment model to achieve integrated closed-loop management of inefficient land use identification, assessment, and decision-making. 2. A dynamic location grading model is set up to replace the traditional location grading model. It takes into account the dynamic changes of location conditions over time and improves the accuracy of judging the inefficiency of land parcels. It comprehensively considers static indicators and dynamic trend indicators of various dimensions, more accurately reflects the coupling degree between multi-dimensional indicators, and improves the accuracy and reliability of the initial identification of inefficient land use. 3. The DBSCAN clustering algorithm is used to perform nonlinear clustering on the initially identified inefficient land spatial units, which can more accurately identify spatially continuous inefficient land and its inefficiency types; a heterogeneous temporal graph attention network model is constructed and used for training and prediction to effectively capture the temporal trend of dynamic indicators and provide a basis for the judgment of inefficient land; at the same time, the output results of this model are used to correct the threshold in fine identification, further improving the identification accuracy of inefficient sub-regions caused by the mismatch of the smallest unit structure within the inefficient land spatial units. Attached Figure Description

[0025] Figure 1 This is a flowchart of the intelligent identification and redevelopment potential assessment method for inefficient land use described in a specific embodiment; Figure 2 This is a schematic diagram of the structure of the intelligent identification and redevelopment potential assessment system for inefficient land use described in a specific embodiment. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0027] like Figure 1 As shown in the illustration, this application discloses a method for intelligent identification and redevelopment potential assessment of inefficient land use, including steps such as data acquisition, inefficient land use identification, potential assessment, and policy simulation screening. Specifically, it identifies inefficient land use by acquiring multi-attribute data of land parcels, performs nonlinear correlation analysis, constructs an assessment index system to quantify potential and determine priorities, and finally integrates a geographic information system platform for visualization and policy simulation. Each step will be described in detail below.

[0028] S1. Obtain multi-source data of the land parcel and perform preprocessing.

[0029] Specifically, multi-source big data such as satellite remote sensing imagery, GIS data, POI data, and mobile signaling data are used to obtain multi-source data of land parcels, including various dimensions of attribute data such as spatial attributes, economic attributes, social attributes, and ecological attributes.

[0030] Spatial attribute data can be obtained through satellite remote sensing imagery and geographic information systems, such as the area, shape, location, building area, plot ratio, and three-dimensional spatial complexity of a land parcel. Economic attribute data can be obtained from government statistical departments and corporate financial statements, such as tax revenue per mu (unit of land area), energy output per unit, industrial suitability, and land appreciation potential. Social attribute data can be obtained through population censuses and mobile signaling data, such as population vitality index and public service gap rate. Ecological attribute data can be obtained through environmental monitoring equipment and remote sensing image analysis, such as carbon sink gain rate, urban heat island effect index, and ecological sensitivity.

[0031] The acquired data underwent preprocessing, including standardization, denoising, and spatial registration, to construct a multi-dimensional feature database at the plot level. This database includes spatial, economic, social, and ecological feature data, as shown in Figure 1 below. Table 1

[0032] S2. Based on a multi-dimensional indicator system, nonlinear correlation analysis is performed on the collected data to identify inefficient land use.

[0033] Specifically, considering the nonlinear correlation between land parcel attribute data and inefficiency, a hierarchical intelligent identification strategy is designed by combining a location hierarchy model, multidimensional feature clustering, and machine learning algorithms. This strategy performs nonlinear correlation analysis on the collected data to accurately identify inefficient land use. This embodiment includes a three-level identification process. The first level, during preliminary identification, calculates the location entropy, coupling degree, and coordination degree based on the location hierarchy model and the multidimensional indicator coupling and coordination model.

[0034] Specifically, location stratification models (such as spatial syntax analysis) are used to calculate location condition indicators (such as integration degree) for each plot, and the plots are divided into different areas such as core area (top 25% of integration degree), secondary core area (integration degree between 25% and 50%), general area (integration degree between 50% and 70%), and edge area (bottom 25% of integration degree) based on these indicators.

[0035] A multi-dimensional index coupling and coordination model is used to calculate the coupling and coordination degree between each land parcel to reflect the overall condition of the land parcels. This involves constructing multiple evaluation index subsystems such as "people," "land," and "industry," or "people," "land," "industry," and "ecology," and calculating the coupling and coordination degree of each land parcel's comprehensive score in each system. For example, the formulas for calculating the coupling and coordination degree of people (population density, per capita income, etc.), land (development intensity, land price, etc.), and industry (industry density, industry diversity) are as follows: D= Where U1, U2, and U3 represent the comprehensive indices of the three subsystems, respectively. This involves standardizing the indicators in each subsystem and determining the weights using entropy weighting or principal component analysis to calculate the comprehensive index. T represents the overall development level of the three subsystems, determined by… The land parcels are divided into different coordination levels based on the weighted calculation and the degree of coordination.

[0036] Spatial units are divided based on Voronoi diagrams; this includes generating a Voronoi diagram using the centroid of each plot as the seed point. If larger spatial units are needed, Voronoi units are merged according to the following rules: Rule 1: If two adjacent Voronoi units belong to the same locational hierarchy (e.g., both are core areas) and have the same coupling coordination level, then they are merged; Rule 2: If one of two adjacent Voronoi units has a lower entropy value, and merging them would increase the entropy value of the new unit, then they are merged; Rule 3: A minimum area threshold is set; if the unit area is too small, it is merged with the most similar adjacent unit.

[0037] Calculate the location entropy for each spatial unit. For each spatial unit, statistically analyze the land use type (or industry type) composition of the plots within it. This differs from traditional location entropy, which measures the degree of specialization of a particular industry or land use type in a region. The calculation formula is... , e represents the scale (e.g., number of employees, output value, etc.) of a specific industry (or land use type) within the research area, while e represents the total scale of all industries within the research area. E represents the scale of the industry in the background region (such as an entire city or country), and E represents the total scale of all industries in the background region. In this embodiment, the location entropy uses Shannon information entropy to measure land use mixing and is used to assess whether the land use structure of a single spatial unit is "inefficient" (single-function may imply low vitality or low coordination). Assuming there are n land use types, the area proportion of each land use type in the unit is calculated. The location entropy of this unit is calculated using the following formula: The system identifies inefficient land use by setting thresholds and feature combinations. These thresholds and feature combinations determine whether the divided spatial units are inefficient land use. Feature combination rules can be set in advance (e.g., ecological constraint exclusion: Voronoi units with an overlap rate greater than 50% with the ecological red line are excluded and identified as non-human-caused inefficiency) or obtained by training a neural network. The location entropy threshold can be obtained by adjusting the preset original entropy value based on the location level mapping to the corresponding location level adjustment coefficient (e.g., core area adjustment coefficient is 1.2, secondary core area is 1.1) and the coordination degree influence coefficient mapping to the coordination degree value (e.g., coordination degree coefficient = 1 + (coupling coordination degree - 0.5) * 0.4).

[0038] The second level: During the optimization identification process, multi-dimensional index feature nonlinear clustering is performed on the spatial units initially identified as inefficient land use to obtain inefficient land use feature clusters and determine the spatial units and inefficiency types belonging to inefficient land use.

[0039] Specifically, the DBSCAN clustering algorithm is used to identify spatially discrete inefficient patches. The process involves clustering spatial units based on multiple dimensions of characteristics to construct a digital feature spectrum. For example, clustering the "people, land, industry, and livelihood" characteristics of land parcels to determine the spatial units and inefficiency types belonging to inefficient land use.

[0040] Level 3: During fine-grained identification, graph neural networks are used to identify inefficient sub-regions caused by structural mismatches in the smallest unit within the spatial units of inefficient land use, based on mixed-use land parcels.

[0041] Specifically, for mixed-use land parcels (different industry types), considering that the inefficiency of such parcels often stems from "spatial structural mismatch of functional units" rather than inefficiency of a single attribute, such as: a spatial unit may meet the overall floor area ratio and GDP per unit area standards, but its internal commercial units are surrounded by industrial units (circulation obstruction), resulting in a high vacancy rate for commercial units (local inefficiency). To further achieve precise identification of inefficient land use, graph neural networks are used to identify inefficient sub-regions caused by structural mismatch of the smallest unit within the spatial units of inefficient land use.

[0042] The process of identifying inefficient sub-regions using graph neural networks involves dividing the spatial units of inefficient land into 100m×100m grids as nodes of a graph neural network (GNN). For each node, features such as functional type, utilization intensity, accessibility, and supporting facilities are extracted to establish a graph structure. This includes defining the spatial adjacency and functional association weights between nodes. A small number of mismatched inefficient units are manually labeled as samples. A graph convolutional network or graph attention network is then used to classify the nodes, outputting the labels for the smallest units, including: efficient units, inefficient units, structurally mismatched units, and their mismatch types.

[0043] S3. Construct a potential assessment index system to quantify the redevelopment potential of inefficient land use, and use the entropy weight method and multi-criteria decision analysis to prioritize the redevelopment of inefficient land use assessment plots.

[0044] Specifically, combining policy adaptation theory, market benefit models, and ecological constraints, a potential assessment index system is constructed, encompassing economic, social, and ecological forces. In this embodiment, the construction formula is as follows: , , In the formula, E represents economic power. To enhance the value of the renovation, The policy coefficient is (0.8-1.2). These are the weighting coefficients; For social force, To create new jobs, To improve the coverage of public service facilities, These are the weighting coefficients. Among them, economic power can be calculated through factors such as value-added from renovation and policy coefficients; social power can consider factors such as job growth and public service coverage; and ecological power can involve factors such as carbon sink gains and heat island mitigation rates.

[0045] For each potential assessment indicator, standardization is performed. Specifically, data preprocessing is conducted beforehand, such as using a truncation method to correct indicator values ​​that exceed reasonable ranges: for example, if the development cost-return ratio is greater than 1.5, it is counted as 1.5; if it is less than 0, it is counted as 0. Because the attributes of each indicator are different, the standardization formulas are divided into two categories, such as: Standardization formula for positive indicators: In the formula, Let be the standardized value of spatial unit i in index j; This represents the original value of spatial unit i in index j; / The maximum / minimum value of indicator j; the standardized formula for the negative indicator is: The information entropy and difference coefficient of each potential assessment indicator are calculated using the entropy weight method to obtain the entropy weight of the corresponding potential assessment indicator. Specifically, this includes calculating the information entropy of indicator j: ;in, ; Calculate the coefficient of difference for index j =1- ; Calculate entropy weight .

[0046] The entropy weights obtained are adjusted and corrected using the Analytic Hierarchy Process (AHP) to obtain the corresponding weights for each potential assessment indicator. The AHP method can be combined with expert experience and judgment to optimize the entropy weights. This includes pairwise comparisons of the importance of primary indicators E, S, and G by experts, establishing a judgment matrix using a 1-9 scaling method, conducting consistency checks, and calculating the consistency index. ;in, To determine the largest eigenvalue of a matrix, if If the value is less than a preset threshold (e.g., 0.1), the matrix is ​​considered valid; otherwise, it is corrected. The calculation of the primary indicator weights E / S / G does not involve decomposing the adjustment coefficients of the primary indicator E / S / G. For example, if the weight of primary indicator E is 0.4, the weight of secondary indicator... Upgrade and increase value The policy coefficient weight is Final indicator weights: .

[0047] Using the Multi-Criterion Decision Analysis (TOPSIS) algorithm, a weighted normalization matrix is ​​constructed to determine the ideal solution. The Euclidean distance between the spatial cell and the ideal solution is calculated, and the proximity is then determined. Specifically, a weighted normalization matrix R is constructed. The ideal solution is The negative ideal solution is ; Calculate the Euclidean distance between spatial element i and the ideal solution. ; Calculate proximity: .

[0048] The approximation level is compared with a preset immediacy threshold to determine the immediacy range, and the redevelopment priority is determined based on the immediacy range. The immediacy ranges are set as follows: less than 0.5, 0.5-0.7, and greater than 0.7, corresponding to high, medium, and low priorities, respectively.

[0049] Furthermore, to further improve the accuracy of redevelopment potential priority assessment and determine redevelopment directions to assist in subsequent policy dynamic simulations, the weights of the corresponding potential assessment indicators for each identified inefficiency type are adjusted. Inefficiency types generally include: industrial inefficiency, spatial utilization inefficiency, lack of social services, ecological constraint inefficiency, and mixed inefficiency. Correspondingly, the weights of the primary indicators E / S / G are adjusted according to different inefficiency types. For example, for industrial inefficiency plots, the weight of economic power (E) is increased from 0.3 to 0.4; for ecological constraint plots, the weight of ecological power (G) is increased from 0.25 to 0.35. Furthermore, a dimensional proximity decomposition is performed based on each potential assessment indicator dimension. The magnitude of the proximity of each dimensional is compared, and the redevelopment direction is determined based on the dimension corresponding to the largest dimensional proximity or a dimension corresponding to a greater than the preset dimensional proximity. The dimensional proximity is as follows: , , ;in: The formula is: Similarly, calculate , .like and It has been identified as a high priority and a direction for industrial upgrading.

[0050] S4. Integrate a geographic information system platform to visualize inefficient land use, conduct policy simulations based on redevelopment priorities, output simulation results, and select the optimal redevelopment model.

[0051] Specifically, a redevelopment policy simulation engine is set up; the redevelopment policy simulation engine has several built-in redevelopment modes, and each redevelopment mode can simulate the effect of at least one of the following policies: floor area ratio incentive policy, mixed land use policy, ecological restoration policy, and stock renewal policy. In this embodiment, twelve redevelopment modes are set up, such as: floor area ratio incentive mode, industrial tax incentive mode, mixed land use mode, ecological compensation mode, old building renovation mode, and land consolidation mode.

[0052] Simulation resources and depth are allocated according to the redevelopment priority of each spatial unit. The higher the redevelopment priority, the higher the proportion of simulation resources and the deeper the simulation level. For example, the simulation resource proportion of high-priority spatial units is 60%-70%, and the simulation level is refined to the building level; the simulation resource proportion of medium-priority spatial units is 20%-30%, and the simulation level is refined to the spatial unit level; the simulation resource proportion of low-priority spatial units is 10%, and the simulation level is refined to the region level.

[0053] Scenario simulations are performed based on each redevelopment model, and the simulation results for each model are output. In other words, the redevelopment policy simulation engine outputs quantitative performance indicators for the corresponding scenarios, recalculates the fit, and selects the policy corresponding to the optimal redevelopment model. For example, using a spatial unit with a high priority number of P001, a scenario simulation is performed using the floor area ratio reward model. The economic, social, and ecological forces of the corresponding spatial unit under the simulated scenario are obtained, and the fit is recalculated. The scenario with the highest fit is the optimal solution.

[0054] In addition, the core development directions already obtained in the aforementioned steps, such as industrial upgrading, mixed-use development, ecological restoration, and stock renewal, can be combined with the determined redevelopment directions. The redevelopment mode can be selected from the corresponding redevelopment mode that conforms to the determined redevelopment direction, while the simulation of the mode that does not conform to the determined redevelopment direction can be ignored.

[0055] A specific embodiment, which more accurately reflects the location characteristics, coupling degree, and inefficiency of a land parcel, improves the accuracy of inefficient land use identification, and makes the identification results more consistent with the actual situation, the method further includes: Specifically, the difference from the above embodiments is that the primary identification also includes setting a dynamic location classification model to replace the location classification model.

[0056] The dynamic location classification model uses a temporal spatial parsing algorithm to calculate the spatial integration degree over several consecutive years (e.g., 3 years) and divide the locations accordingly. Based on whether the average spatial integration degree over N years is less than a preset integration degree threshold range, the model is divided into core area, secondary core area, general area and edge area.

[0057] The initial identification also includes: performing multi-scale location entropy calculation for each spatial unit, and selecting the minimum scale location entropy or the average scale location entropy from the multi-scale location entropy as the final location entropy.

[0058] The multi-scale location entropy includes: a first-scale entropy covering a single spatial unit, a second-scale location entropy covering a single spatial unit and its adjacent spatial units, and a third-scale location entropy covering spatial units within the administrative region (administrative street / district) where the single spatial unit is located. Each scale entropy has a corresponding scale-based determination location entropy threshold (a preset original location entropy threshold). The annual change rate of the location entropy at the corresponding scale is calculated. If the annual change rate of the location entropy is lower than the preset annual change rate of the location entropy, the determination location entropy threshold at the corresponding scale is adjusted (i.e., the determination location entropy threshold at the corresponding scale is reduced), which is the preset original location entropy threshold at the corresponding scale.

[0059] In addition to the above-mentioned optimization of inefficient identification for location entropy, the method can also optimize coupling coordination to improve inefficient identification. The method also includes: considering that the original calculation of coupling degree only considers the synergy of multi-dimensional indicators and does not consider the temporal dynamic changes of indicators, in order to further enhance the amplification of synergy by indicator differences, a design is made to perform temporal fusion (capturing the dual characteristics of static level and dynamic trend) and difference enhancement (amplifying the negative impact of the imbalance between indicators on coupling degree through the difference of temporal variation coefficient and trend slope), so as to realize the dual coupling design of level and trend.

[0060] Specifically, the indicators across various dimensions are standardized in advance, including: standardizing static indicators using positive and negative indicators from a single year; calculating dynamic trend indicators, including calculating the trend slope and completing trend standardization; calculating the time-series coefficient of variation to reflect fluctuation differences; and constructing a comprehensive indicator, weighting the static indicators and the dynamic indicators adjusted based on the time-series coefficient of variation, as shown in the following formula: In the formula, This represents the static average level of the k-th dimension indicator. This represents the standardized value of the trend of the k-th dimension indicator. is the time series variation coefficient of the k-th dimension indicator, and N is the number of years; A comprehensive index constructed based on static and dynamic trend indicators across various dimensions. The static and dynamic nonlinear coupling degrees are calculated accordingly, and the weighted temporal nonlinear coupling degree is used as the final coupling degree; the formula is: In the formula, For static nonlinear coupling degree, the comprehensive index is linked to the time series fluctuation difference. The greater the time series difference, the higher the weight of the coupling degree. ; In the formula, The dynamic nonlinear coupling degree is represented by p and q, which are dimension indices. represents the cosine similarity of the trend slopes of dimensions p and q; the closer the value is to 1, the stronger the trend coherence.

[0061] In one specific embodiment, considering that the existing DBSCAN distance calculation only uses Euclidean distance to calculate attribute similarity and does not independently consider the spatial correlation of land parcels (e.g., two land parcels with similar attributes but spatially separated by 10km should not be classified into the same category), in order to further achieve more accurate identification and classification of inefficient land use, spatial constraints need to be introduced and the distance algorithm optimized. The optimized identification also includes: performing multi-dimensional index feature nonlinear clustering on spatial units initially identified as inefficient land use using the DBSCAN clustering algorithm, and designing a hybrid distance as the DBSCAN distance. The distance criterion and specific clustering process include: First, spatial units initially identified as inefficient land use are used as candidate inefficient land use plots, and the multi-dimensional indicator features of each spatial unit are standardized; second, the attribute distance of non-spatial attributes between spatial units is calculated based on the Mahalanobis distance algorithm; third, the spatial distance between spatial units is calculated using the Huffman distance algorithm; finally, the mixed distance between spatial units is obtained by weighted calculation, and a mixed distance matrix is ​​constructed; DBSCAN clustering parameters are set, including: mixed distance threshold, minimum number of neighborhoods of core points; traversing spatial units, identifying core points, boundary points, and noise points; performing spatial constraint verification (such as setting spatial constraint rules, if the maximum spatial distance in the cluster is less than the mixed distance threshold, if there is a spatial distance greater than the mixed distance threshold, then directly judging the mixed distance as 1, considering it as no similarity), filtering spatially discrete clusters; and outputting spatially continuous inefficient land use and its inefficiency type.

[0062] In one specific embodiment, a heterogeneous temporal graph attention network model is used to judge the initially identified inefficient land use, and the threshold in the fine-grained identification is adjusted based on the judgment result. This can further improve the accuracy of inefficient land use identification, more accurately uncover hidden inefficient land use patterns, and analyze the underlying mechanisms of inefficient land use. The method also includes: The optimized identification also includes: constructing a heterogeneous time series graph attention network model, inputting multi-dimensional indicator features corresponding to spatial units initially identified as inefficient land use into the heterogeneous time series graph attention network model, and obtaining the output judgment result of whether the spatial unit is inefficient land use, so as to verify the accuracy of the initial identification result.

[0063] The heterogeneous temporal graph attention network model defines nodes as follows: spatial units as core nodes, and POIs and roads as auxiliary nodes; edge definitions include: spatially adjacent edges and functionally related edges (such as the association between industrial plots and logistics POIs); the heterogeneous temporal graph attention network model is equipped with spatial attention and feature attention mechanisms, and embeds an LSTM module to capture the temporal trend of input dynamic indicators; and is generated through training using multi-dimensional indicator features corresponding to spatial units that have been historically labeled as inefficient.

[0064] The refined identification also includes: using the inefficiency score, inefficiency type and feature weight in the output spatial unit in the judgment result of whether the spatial unit is an inefficient land use to correct the preset minimum unit inefficiency probability threshold and minimum unit structure mismatch feature weight threshold in the process of identifying inefficient sub-regions caused by minimum unit structure mismatch in the spatial unit of inefficient land use, so as to optimize the identification result of inefficient sub-regions caused by minimum unit structure mismatch.

[0065] The heterogeneous temporal graph attention network model uses spatial units corresponding to the entire inefficient land use as nodes. The inputs are multi-dimensional indicators of the spatial units, the temporal dynamic features corresponding to each indicator, and spatial correlation features. The output is an inefficiency score for the corresponding spatial unit. and inefficient confidence For example, a value between 0 and 1 indicates a higher overall inefficiency, and the corresponding inefficiency type label is... and simultaneously output feature weights For example, the per capita GDP weight of spatial units with inefficient industries is 0.38, and the employment density weight is 0.29.

[0066] Correspondingly, the neural network constructed to identify inefficient sub-regions caused by structural mismatches in the smallest unit within spatial units of inefficient land use, as a micro-level GNN, has key judgment thresholds during graph neural network inference, including: node inefficiency probability judgment threshold. (If the posterior probability of each unit being an inefficient unit is greater than the node inefficiency probability judgment threshold, it is judged as an inefficient unit), and the unit mismatch feature weight threshold (i.e., the feature weights corresponding to the units judged as structural mismatch units and their mismatch types). ).

[0067] By using the overall inefficiency score of the heterogeneous temporal graph attention network model as the core independent variable, and combining it with the inefficiency confidence score for reliability correction, the final inefficiency probability threshold of the minimum corrected unit is obtained, as shown in the formula: In the formula, For a fixed threshold, This represents the temporal decay coefficient in a heterogeneous temporal graph attention network model.

[0068] The original feature weights are the average weights from the neural network model training and are undirected. However, in heterogeneous temporal graph attention network models, it is necessary to determine which core features are causing inefficiency, and the threshold for unit mismatched feature weights is corrected as follows: .

[0069] like Figure 2 As shown in the embodiment of this application, a system for intelligent identification and redevelopment potential assessment of inefficient land use is disclosed, comprising: The data acquisition and preprocessing module 100 is used to acquire multi-dimensional attribute data of the land parcel and perform preprocessing. The inefficient land use intelligent identification module 200 is used to identify inefficient land use by performing nonlinear correlation analysis on the collected data based on a multi-dimensional indicator system. The nonlinear correlation analysis includes three levels of identification analysis. Preliminary identification includes: calculating the location entropy, coupling degree, and coordination degree based on a location hierarchical model and a multi-dimensional indicator coupling and coordination model; dividing spatial units based on a Voronoi diagram; and setting threshold division and feature combinations to determine whether the divided spatial units are inefficient land use. Optimized identification includes: performing nonlinear clustering of multi-dimensional indicator features on the spatial units initially identified as inefficient land use to obtain inefficient land use feature clusters and determine the spatial units and inefficiency types belonging to inefficient land use. Refined identification includes: using graph neural networks to identify inefficient sub-regions caused by the mismatch of the smallest unit structure within the spatial units of inefficient land use. The redevelopment potential dynamic assessment module 300 is used to construct a potential assessment index system to quantify the redevelopment potential of inefficient land use and to analyze the redevelopment priority of inefficient land use assessment plots through entropy weight method and multi-criteria decision analysis. The redevelopment decision simulation and screening module 400 is used to integrate a geographic information system platform to realize the visualization of inefficient land use, conduct policy simulation based on redevelopment priorities, output simulation results, and screen the optimal redevelopment model.

[0070] This application also discloses a computer-readable storage medium.

[0071] Specifically, the computer-readable storage medium stores a computer program that can be loaded by a processor and executed, such as the aforementioned method for intelligent identification and redevelopment potential assessment of inefficient land use. The computer-readable storage medium includes, for example, various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0072] This application also discloses a computer device.

[0073] Specifically, the computer device includes a memory and a processor, with the memory storing a computer program that can be loaded by the processor and executed to perform the aforementioned intelligent identification and redevelopment potential assessment method for inefficient land use.

[0074] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. A method for intelligent identification and redevelopment potential assessment of inefficient land use, characterized in that, include: Acquire multi-dimensional attribute data of the land parcel and perform preprocessing; Nonlinear correlation analysis is performed on the collected data based on a multi-dimensional indicator system to identify inefficient land use. The nonlinear correlation analysis includes three levels of identification analysis. Preliminary identification includes: calculating location entropy, coupling degree, and coordination degree based on a location hierarchy model and a multi-dimensional index coupling and coordination model; dividing spatial units based on a Voronoi diagram; and setting threshold division and feature combinations to determine whether the divided spatial units are inefficient land use. Optimized identification includes: performing multi-dimensional index feature nonlinear clustering on the spatial units initially identified as inefficient land use to obtain inefficient land use feature clusters and determine the spatial units and inefficiency types belonging to inefficient land use. Refined identification includes: using graph neural networks to identify inefficient sub-regions caused by the mismatch of the smallest unit structure within the spatial units of inefficient land use. A potential assessment index system was constructed to quantify the redevelopment potential of inefficient land use, and the redevelopment priority of inefficient land use assessment plots was analyzed through entropy weight method and multi-criteria decision analysis. By integrating a geographic information system platform, we can visualize inefficient land use, conduct policy simulations based on redevelopment priorities, output simulation results, and select the optimal redevelopment model.

2. The method for intelligent identification and redevelopment potential assessment of inefficient land use according to claim 1, characterized in that, The primary identification also includes: A dynamic location hierarchy model is set up to replace the location hierarchy model. The dynamic location hierarchy model adopts a temporal spatial syntax analysis algorithm to calculate the spatial integration degree over several consecutive years and perform location division. For each spatial unit, multi-scale location entropy is calculated, and the smallest scale location entropy is selected as the final location entropy. The multi-scale location entropy includes: a first-scale entropy covering a single spatial unit, a second-scale location entropy covering a single spatial unit and its adjacent spatial units, and a third-scale location entropy covering spatial units within the administrative region where the single spatial unit is located. Different scale entropies are set with corresponding scale judgment location entropy thresholds, and the annual change rate of the location entropy at the corresponding scale is calculated. If the annual change rate of location entropy is lower than the preset annual change rate of location entropy, the judgment location entropy threshold at the corresponding scale is adjusted.

3. The method for intelligent identification and redevelopment potential assessment of inefficient land use according to claim 1, characterized in that, The primary identification also includes: Based on the multi-dimensional indicator coupling and collaboration model, time-series fusion is performed on each dimension indicator to obtain static indicators and dynamic trend indicators for each dimension. Based on the static indicators and dynamic trend indicators for each dimension, the static nonlinear coupling degree and dynamic nonlinear coupling degree are calculated and obtained. The time-series nonlinear coupling degree obtained by weighted calculation is used as the final coupling degree.

4. The method for intelligent identification and redevelopment potential assessment of inefficient land use according to claim 1, characterized in that, The optimized identification also includes: For spatial units initially identified as inefficient land use, a multi-dimensional index feature nonlinear clustering algorithm, DBSCAN, is used. The clustering process includes: using the spatial units initially identified as inefficient land use as candidate plots, standardizing the multi-dimensional index features of each spatial unit; calculating the attribute distance of non-spatial attributes between spatial units based on the Mahalanobis distance algorithm, calculating the spatial distance between spatial units using the Huffman distance algorithm, obtaining the mixed distance between spatial units through weighted calculation, and constructing a mixed distance matrix; setting DBSCAN clustering parameters; traversing the spatial units to identify core points, boundary points, and noise points; performing spatial constraint verification to filter spatially discrete clusters; and outputting spatially continuous inefficient land use and its inefficiency type.

5. The method for intelligent identification and redevelopment potential assessment of inefficient land use according to claim 1, characterized in that, The optimized identification also includes: constructing a heterogeneous temporal graph attention network model; the node definition in the heterogeneous temporal graph attention network model includes: spatial units as core nodes, POIs and roads as auxiliary nodes; the edge definition includes: spatially adjacent edges and functionally related edges; the heterogeneous temporal graph attention network model is equipped with spatial attention and feature attention mechanisms, and embeds an LSTM module to capture the temporal trend of input dynamic indicators; and is trained and generated by multi-dimensional indicator features corresponding to spatial units that are historically labeled as inefficient; the multi-dimensional indicator features corresponding to spatial units initially identified as inefficient land use are input into the heterogeneous temporal graph attention network model to obtain the output judgment result of whether the spatial unit is inefficient land use; The refined identification also includes: using the inefficiency score, inefficiency type and feature weight in the output spatial unit in the judgment result of whether the spatial unit is an inefficient land use to correct the preset minimum unit inefficiency probability threshold and minimum unit structure mismatch feature weight threshold in the process of identifying inefficient sub-regions caused by minimum unit structure mismatch in the spatial unit of inefficient land use, so as to optimize the identification result of inefficient sub-regions caused by minimum unit structure mismatch.

6. The method for intelligent identification and redevelopment potential assessment of inefficient land use according to claim 1, characterized in that, The proposed potential assessment index system quantifies the redevelopment potential of inefficient land use, and the redevelopment priorities of inefficient land use assessment plots are analyzed using the entropy weight method and multi-criteria decision analysis, including: Combining policy adaptation theory, market benefit model, and ecological constraints, a potential assessment index system is constructed, which includes economic power, social power, and ecological power. Each potential assessment index is standardized. The information entropy and difference coefficient of each potential assessment index are calculated using the entropy weight method to obtain the entropy weight of the corresponding potential assessment index. The obtained entropy weights are adjusted and corrected using the AHP method to obtain the corresponding weight of each potential assessment index. Using a multi-criteria decision analysis algorithm, a weighted standardized matrix is ​​constructed, and the weights of the potential assessment indicators related to the identified inefficiency types are further adjusted to determine the ideal solution. The Euclidean distance between the spatial unit and the ideal solution is calculated, and the proximity is calculated. The proximity is compared with a preset proximity threshold to determine the proximity range corresponding to the proximity, and the redevelopment priority is determined based on the proximity range. Based on the dimensions of each potential assessment indicator, the proximity is decomposed into sub-dimensions, and the magnitude of the proximity of each sub-dimension is compared. The redevelopment direction is determined based on the dimension corresponding to the maximum sub-dimension proximity.

7. The method for intelligent identification and redevelopment potential assessment of inefficient land use according to claim 1, characterized in that, The process of conducting policy simulations based on redevelopment priorities, outputting simulation results, and selecting the optimal redevelopment model includes: A redevelopment policy simulation engine is set up; the redevelopment policy simulation engine has several built-in redevelopment modes, and each redevelopment mode can simulate the effect of at least one of the following policies: floor area ratio incentive policy, mixed land use policy, ecological restoration policy, and stock renewal policy; simulation resources and depth are divided according to the redevelopment priority of each spatial unit. The higher the redevelopment priority, the higher the proportion of simulation resources and the deeper the simulation level; scenario simulation is performed based on each redevelopment mode, the simulation results of each redevelopment mode are output, the closeness is recalculated, and the policy corresponding to the optimal redevelopment mode is selected.

8. A smart system for identifying inefficient land use and assessing its redevelopment potential, characterized in that, include: The data acquisition and preprocessing module is used to acquire multi-dimensional attribute data of the land parcel and perform preprocessing. The inefficient land use intelligent identification module is used to perform nonlinear correlation analysis on the collected data based on a multi-dimensional indicator system to identify inefficient land use. The nonlinear correlation analysis includes three levels of identification analysis. Preliminary identification includes: calculating location entropy, coupling degree, and coordination degree based on a location hierarchy model and a multi-dimensional index coupling and coordination model; dividing spatial units based on a Voronoi diagram; and setting threshold division and feature combinations to determine whether the divided spatial units are inefficient land use. Optimized identification includes: performing multi-dimensional index feature nonlinear clustering on the spatial units initially identified as inefficient land use to obtain inefficient land use feature clusters and determine the spatial units and inefficiency types belonging to inefficient land use. Refined identification includes: using graph neural networks to identify inefficient sub-regions caused by the mismatch of the smallest unit structure within the spatial units of inefficient land use. The redevelopment potential dynamic assessment module is used to construct a potential assessment index system to quantify the redevelopment potential of inefficient land use, and to analyze the redevelopment priority of inefficient land use assessment plots through entropy weight method and multi-criteria decision analysis. The redevelopment decision simulation and screening module is used to integrate a geographic information system platform to visualize inefficient land use, simulate policies based on redevelopment priorities, output simulation results, and screen the optimal redevelopment model.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the method as described in any one of claims 1 to 7.

10. A computer device, characterized in that, The computer device includes a memory, a processor, and a program stored in and executable on the memory, the program being executed by the processor to implement the steps of the method as described in any one of claims 1 to 7.

Images