Land supply and demand intelligent matching and benefit simulation method

Abstract

This invention provides a method for intelligent matching and benefit simulation of land supply and demand, comprising the following steps: transmitting data to a resource server via encrypted transmission technology; the resource server integrating all land parcel feature data and all enterprise feature data to obtain a supply and demand feature matrix; filtering based on the supply hard condition dataset and demand hard condition dataset to obtain multiple matched sets that correspond one-to-one with each vacant land plot or one-to-one with each enterprise; calculating the matching degree, revenue impact, and environmental impact; and integrating all matching degree values, simulated revenue values, and simulated environmental values ​​to obtain a matching benefit matrix. Using this method, the revenue and environmental impacts that enterprises can generate after landing on different vacant land plots can be intuitively determined, thereby guiding the enterprise's landing planning.

Description

Technical Field

[0001] This invention relates to intelligent simulation of land supply and demand, specifically a method for intelligent matching and benefit simulation of land supply and demand. Background Technology

[0002] Land resources, as a fundamental element for human survival and development, are directly related to the quality of regional economic development, the optimization of industrial structure, and the sustainability of the ecological environment through their rational allocation and efficient utilization. With the accelerating urbanization process in my country, the demand for construction land continues to grow, while the total land supply is limited and its spatial distribution is uneven, leading to increasingly prominent structural contradictions between supply and demand. Traditional land resource allocation methods mainly rely on manual experience and static indicator approvals, making it difficult to accurately quantify the complex matching relationship between land parcel characteristics and enterprise needs, resulting in the dilemma of both land idling and difficulties in project implementation.

[0003] At the data infrastructure level, existing land enterprise information management suffers from severe fragmentation. Plot characteristic data and enterprise characteristic data are typically scattered across different departments or offline archives, lacking a unified integration platform. Some enterprises, concerned about the leakage of trade secrets or data security risks, refuse to provide critical information, further exacerbating information asymmetry. Data transmission largely relies on manual collection and carrying, which is inefficient and time-consuming. For example, some land change surveys take several days, while the service radius at the grassroots level exceeds 15 kilometers, requiring grassroots staff or enterprise employees to make multiple trips to complete information registration. This inefficient process not only prolongs the decision-making cycle but also causes matching results to deviate from actual needs due to data lag.

[0004] At the technical application level, although existing matching methods have incorporated computer-aided models, they still have significant limitations. On the one hand, traditional models mostly use rule base filtering, which can only handle a small number of hard conditions, while ignoring comprehensive factors such as industrial environment requirements, energy supply, and labor resource distribution. On the other hand, existing algorithms lack the ability to collaboratively evaluate multiple objectives and cannot simultaneously quantify matching degree, economic benefits, and environmental impact.

[0005] Furthermore, current technologies lack mechanisms for parallel simulation and dynamic optimization across multiple land parcels. In enterprise site selection scenarios involving multiple parcels, traditional methods cannot simultaneously simulate the benefits of multiple candidate parcels, making it difficult to support globally optimal decision-making. Simultaneously, model parameter adjustments rely on manual intervention, failing to establish a self-learning mechanism based on feedback data, leading to a disconnect between configuration schemes and real-time requirements. Therefore, a new method integrating secure data transmission, intelligent matching, and multi-benefit simulation is urgently needed to achieve a shift from "passive approval" to "proactive optimization." Summary of the Invention

[0006] To overcome existing technical problems, this invention provides a method for intelligent matching and benefit simulation of land supply and demand that integrates secure data transmission.

[0007] The present invention adopts the following technical solution.

[0008] This application provides a method for intelligent matching of land supply and demand and benefit simulation, including the following steps: S1. Collect land parcel feature data from multiple vacant land plots and enterprise feature data from multiple enterprises, and transmit them to the resource server through double-layer encryption transmission technology. The resource server integrates all the land parcel feature data and all the enterprise feature data to obtain a supply and demand feature matrix. S2. By processing the matching dual model, identify and extract the supply hard condition dataset and demand hard condition dataset from the supply and demand feature matrix. Filter according to the supply hard condition dataset and demand hard condition dataset to obtain multiple filter matching sets that correspond one-to-one with each vacant land or one-to-one with each enterprise. S3. Calculate the matching degree, revenue impact, and environmental impact for each pair of land parcel enterprises in the screening and matching set according to the supply and demand feature matrix, and obtain the corresponding matching degree value, simulated revenue value, and simulated environmental value. Integrate all the matching degree values, simulated revenue values, and simulated environmental values ​​to obtain the matching benefit matrix.

[0009] As a further improvement of the present invention, step S3 is followed by step S4: when an enterprise is established, it is marked as an established enterprise; the actual land plot feature data and the actual enterprise feature data of the land plot corresponding to the established enterprise are obtained and updated; a corresponding dynamic weight set is selected according to the enterprise type of the actual enterprise feature data; the revenue impact and environmental impact are calculated according to the actual land plot feature data, the actual enterprise feature data and the dynamic weight set to obtain the corresponding simulated revenue value and simulated environmental value; and the land plot feature data of the vacant land is updated according to the actual land plot feature data and the actual enterprise feature data. Collect revenue and environmental data of the enterprises that have settled in the area, and denote them as corrected revenue data and corrected environmental data. Calculate the simulation difference based on the corrected revenue data, corrected environmental data, simulated revenue data, and simulated environmental data, and optimize the values ​​of the revenue weight coefficient and environmental weight coefficient in the dynamic weight set based on the simulation difference.

[0010] As a further improvement of the present invention, the resource server is equipped with a cloud decoder, and the enterprise feature data includes multiple sub-enterprise feature data and enterprise type; The specific steps for transmitting data to the resource server using double-layer encryption technology include: selecting a corresponding dynamic weight set based on the enterprise type; if there is a matching weight coefficient, revenue weight coefficient, or environmental weight coefficient in the dynamic weight set that corresponds to the sub-enterprise feature data and is greater than a preset importance encryption threshold, then the sub-enterprise feature data is recorded as the core feature for calculation. The feature names of all the land parcel feature data and enterprise feature data are classified according to the preset local language discriminator to obtain multiple privacy core features and multiple non-core features. The privacy core features and computation core features are marked as high-encryption features, and the non-core features are marked as low-encryption features. The local encoder corresponding to the cloud decoder generates multiple latent vectors from all land parcel feature data and enterprise feature data. The latent vectors are then reconstructed using the local decoder configured with the same settings as the cloud decoder to obtain reconstructed data. The mean square error between the reconstructed data and the corresponding land parcel feature data or enterprise feature data is calculated, and a reconstruction confidence coefficient is generated. A key is generated by using a key derivation function to construct a trust coefficient corresponding to the highly encrypted feature. The latent vector of the highly encrypted feature is then encrypted using the key to obtain an encrypted latent vector. Multiple masquerading reconstruction loss data corresponding one-to-one with the encrypted latent vector are randomly generated. The encrypted latent vector and its corresponding masquerading reconstruction loss data, along with the remaining latent vectors and their corresponding reconstruction trust coefficients, are then transmitted in batches to the resource server. The key is transmitted to the resource server through a second encrypted communication channel. The resource server then obtains the land parcel feature data and enterprise feature data through symmetric decryption and decoding.

[0011] As a further improvement of the present invention, the dual-model processing and matching includes a natural language processing model and a rule matching model; The specific steps for identifying and extracting the supply hard condition dataset and demand hard condition dataset from the supply and demand feature matrix by processing the matching dual model include: converting the text feature descriptions in the supply and demand feature matrix into semantic vectors through the natural language processing model, converting the semantic vectors into standardized structured data, integrating the standardized structured data of the same vacant land to obtain the land parcel hard condition data, integrating all the land parcel hard condition data to obtain the supply hard condition dataset, integrating the standardized structured data of the same enterprise to obtain the enterprise hard condition data, and integrating all the enterprise hard condition data to obtain the demand hard condition dataset.

[0012] As a further improvement of the present invention, the natural language processing model includes a pre-trained domain knowledge graph, wherein multiple nodes and multiple edges in the domain knowledge graph correspond to multiple land enterprise normative concepts and multiple relationships between concepts, respectively. The specific steps for converting the semantic vector into standardized structured data include: processing the semantic vector through the named entity recognition model in the natural language processing model to obtain an original sequence with entity labels, wherein the original sequence includes multiple labeled fragmented words; By scanning all fragmented words through the domain knowledge graph and matching them with similar semantic land enterprise normative concepts, at least one fragmented word is replaced with a corresponding land enterprise normative concept to obtain an enhanced sequence. The remaining fragmented words in the enhanced sequence are then reorganized into entities to obtain at least one entity. If the context distance between multiple entities is less than the preset combination distance, and the nodes corresponding to two entities have the same side in the domain knowledge graph, then the entities are integrated to obtain a standard statement, the remaining entities are combined according to syntactic rules to obtain a standard statement, and all the standard statements are integrated to obtain standardized structured data.

[0013] As a further improvement of the present invention, the enterprise feature data includes multiple sub-enterprise feature data, enterprise type, and a reconstructed trust coefficient corresponding one-to-one with each of the sub-enterprise feature data; Between steps S2 and S3, step S21 is also included: traversing the enterprise feature data of each enterprise in the supply and demand feature matrix, selecting the corresponding dynamic weight set according to the enterprise type, determining whether each sub-enterprise feature data under the enterprise feature data is blank or invalid data, and whether the matching weight coefficient, revenue weight coefficient, or environmental weight coefficient corresponding to the sub-enterprise feature data in the dynamic weight set is greater than zero. If so, the enterprise is marked as a blank enterprise. It is then identified whether the blank enterprise has a chain enterprise with the same name and the enterprise feature data and land feature data corresponding to the chain enterprise with the same name. If so, the similarity between the enterprise feature data of the blank enterprise and the enterprise feature data of each chain enterprise with the same name is calculated to obtain several enterprise feature similarities. The similarity between each vacant land and the land feature data of the chain enterprise with the same name is calculated to obtain several enterprise environmental similarities. Using the enterprise feature similarity and enterprise environment similarity as weights, the sub-enterprise feature data of the chain enterprises with the same name are weighted and averaged to calculate the alternative feature data. The alternative feature data is used to cover the corresponding blank and invalid data. The substitution loss coefficient is calculated based on all the enterprise feature similarities and enterprise environment similarities. The substitution loss coefficient is used to cover the corresponding reconstruction trust coefficient. If there is no chain enterprise with the same name, and no enterprise characteristic data and land characteristic data corresponding to the chain enterprise with the same name, then generate dynamic average characteristic data based on the enterprise characteristic data of the blank enterprise and cover the corresponding blank invalid data, and cover the corresponding reconstructed trust coefficient with the preset non-migrated data trust weight.

[0014] As a further improvement of the present invention, the expression for calculating the reconstruction trust coefficient is as follows: , in, It is the reconstruction confidence coefficient corresponding to the k-th latent vector. It is a trust decay factor. It is the mean square error. Is it land parcel characteristic data or enterprise characteristic data? It's about reconstructing the data.

[0015] As a further improvement of the present invention, the supply and demand feature matrix includes multiple plot feature data, multiple enterprise feature data, and multiple reconstructed trust coefficients that correspond one-to-one with the plot feature data or one-to-one with the enterprise feature data. The enterprise feature data includes multiple sub-enterprise feature data and enterprise types. The calculation of matching degree, benefit impact and environmental impact includes: selecting the corresponding dynamic weight set according to the enterprise type, and selecting the matching weight coefficient, benefit weight coefficient and environmental weight coefficient with a value greater than zero in the dynamic weight set; The matching degree value is obtained by calculating and normalizing the selected matching weight coefficient, the enterprise feature data and land feature data corresponding to the matching weight coefficient, and the reconstruction trust coefficient corresponding to the enterprise feature data and land feature data; The simulated revenue value is calculated based on the selected revenue weighting coefficient, the enterprise characteristic data and land parcel characteristic data corresponding to the revenue weighting coefficient, and the reconstruction trust coefficient corresponding to the enterprise characteristic data and land parcel characteristic data. The simulated environmental value is calculated based on the selected environmental weight coefficient, the enterprise characteristic data and land parcel characteristic data corresponding to the environmental weight coefficient, and the reconstruction trust coefficient corresponding to the enterprise characteristic data and land parcel characteristic data.

[0016] The beneficial effects of this invention are as follows: By standardizing scattered land parcel characteristic data and enterprise characteristic data into a computable matrix, the time-consuming process of manual collection and multiple trips is avoided, reducing the data update cycle from "days" to "near real-time." Double-layer encryption technology significantly enhances the willingness to share data while protecting enterprise and land confidentiality. The calculation of the matching benefit matrix breaks through the limitations of traditional rule bases that only focus on hard conditions. It not only considers hard conditions such as area and location but also integrates soft indicators such as industrial relevance and environmental compatibility. By quantifying the comprehensive impact on regional benefits and the environment, it can intuitively determine the revenue and environmental impact that an enterprise can generate after establishing itself on different vacant land plots. This supports simultaneous benefit simulation for enterprises selecting multiple land parcels, thereby guiding their site planning. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the present invention. Detailed Implementation

[0019] The accompanying drawings are for illustrative purposes only and should not be construed as limiting the scope of this patent. To better illustrate this embodiment, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual dimensions of the product.

[0020] It will be understood by those skilled in the art that certain well-known structures and their descriptions may be omitted in the accompanying drawings. The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0021] Reference Figure 1 A method for intelligent matching of land supply and demand and efficiency simulation includes the following steps: S1. Collect land parcel feature data from multiple vacant land plots and enterprise feature data from multiple enterprises, and transmit them to the resource server through double-layer encryption transmission technology. The resource server integrates all the land parcel feature data and all the enterprise feature data to obtain a supply and demand feature matrix. The land parcel characteristic data includes a structured set of information describing the physical, locational, planning (planning includes not only government land use planning but also existing planning of land occupied by nearby enterprises, so that it can be correlated with the enterprises when calculating the revenue and environmental impact to determine whether it is a positive or negative benefit. In addition, it can also include the impact of the enterprises, such as the frequent loading and unloading needs of nearby enterprises, which leads to frequent road occupation) and ecological attributes. It can be used as a supply-side input parameter to characterize the factors that make land resources available and restrictive.

[0022] Enterprise characteristic data includes a structured set of information reflecting demand characteristics such as industry type, specific product type, investment scale, production mode, and environmental behavior, which can be used as input parameters on the demand side. It is important to note that specific product type is a strong indicator of the positive or negative returns between established and non-established enterprises. For example, if an area is full of factories in a certain industry, but their products have upstream and downstream supporting relationships, it can be judged as having positive returns. Conversely, if the specific product types are the same and highly homogenized, they can only compete with neighboring enterprises through price wars, which can be judged as having negative returns. In a specific embodiment, enterprise characteristic data can be extracted from enterprise registration information, environmental impact assessment reports, industry access catalogs, and historical land use behavior of chain enterprises. If these are not available, enterprise characteristic data from established enterprises with the same industry type, similar investment scale, and similar production mode can be obtained for manual evaluation, or the data can be left blank for subsequent execution step S21.

[0023] Furthermore, enterprise characteristic data may include, but is not limited to, one or more of the following: industry category data, investment intensity per unit area data, and energy consumption per unit output value data. However, its use is not recommended in practice because enterprise investment behavior is highly unpredictable.

[0024] The resource server is specifically an information processing platform that centrally stores, integrates, and manages land and enterprise data. It can be used to aggregate and uniformly manage multi-source heterogeneous data, providing data support for subsequent analysis. In this embodiment, the resource server is deployed in the cloud or a local data center, and has data receiving interfaces, storage modules, and computing scheduling capabilities, supporting concurrent access and task distribution.

[0025] S2. By processing the matching dual model, identify and extract the supply hard condition dataset and demand hard condition dataset from the supply and demand feature matrix. Filter according to the supply hard condition dataset and demand hard condition dataset to obtain multiple filter matching sets that correspond one-to-one with each vacant land or one-to-one with each enterprise. Traditional models often use rule base filtering, which can only handle a small number of hard conditions. This invention, however, first uses a unified processing approach with a matching dual-model approach, thus standardizing the data.

[0026] For each vacant land plot, enterprises meeting its supply requirements are selected, or for each enterprise, plots meeting its demand requirements are selected. In one specific embodiment, this operation can be achieved through conditional filtering and matching using SQL queries, or by constructing a plot-enterprise relationship network using a graph database and traversing valid paths, thereby generating a preliminary set of feasible matching schemes. This narrows down the scope of subsequent evaluations and significantly reduces the computational load. Specifically, for example, a plot near a water source may be unsuitable for enterprises engaged in the chemical industry, or near residential buildings may be unsuitable for noisy enterprises such as factories, etc.

[0027] S3. Calculate the matching degree, revenue impact, and environmental impact for each pair of land parcel enterprises in the screening and matching set according to the supply and demand feature matrix, and obtain the corresponding matching degree value, simulated revenue value, and simulated environmental value. Integrate all the matching degree values, simulated revenue values, and simulated environmental values ​​to obtain the matching benefit matrix.

[0028] Matching degree is an indicator that quantifies the degree of fit between a company and a land plot under non-rigid conditions, reflecting the level of soft fit between them. Simulated revenue is a predicted positive contribution of a company's establishment on a land plot to the regional economy, assessing the potential impact of different matching schemes on local finances, employment, and output. Simulated environmental value is a quantitative indicator that predicts the pressure or damage to the ecological environment caused by a company's establishment on a land plot, measuring the potential negative ecological consequences of site selection decisions and supporting sustainable development assessments.

[0029] As a further improvement of the present invention, step S3 is followed by step S4: when an enterprise is established, it is marked as an established enterprise; the actual land plot feature data and the actual enterprise feature data of the land plot corresponding to the established enterprise are obtained and updated; a corresponding dynamic weight set is selected according to the enterprise type of the actual enterprise feature data; the revenue impact and environmental impact are calculated according to the actual land plot feature data, the actual enterprise feature data and the dynamic weight set to obtain the corresponding simulated revenue value and simulated environmental value; and the land plot feature data of the vacant land is updated according to the actual land plot feature data and the actual enterprise feature data. Collect revenue and environmental data of the enterprises that have settled in the area, and denote them as corrected revenue data and corrected environmental data. Calculate the simulation difference based on the corrected revenue data, corrected environmental data, simulated revenue data, and simulated environmental data, and optimize the values ​​of the revenue weight coefficient and environmental weight coefficient in the dynamic weight set based on the simulation difference.

[0030] The purpose of this step is to improve the accuracy of revenue and environmental impact calculations. When calculating revenue and environmental impacts before a company is established, predictions are made based on the land and company characteristic data at that time. However, the land and company characteristic data change significantly, so it is not possible to make predictions and optimize calculation parameters based on the land and company characteristic data at that time.

[0031] Actual land parcel characteristic data is a set of information reflecting the true use status of a land parcel after a company has established itself, obtained through on-site inspections or administrative registration updates. This data can be used to calibrate the land parcel assumptions in the initial model, supporting subsequent simulations and parameter optimization. In one specific embodiment of this invention, actual land parcel characteristic data can be acquired through land use verification by natural resources authorities, dynamic monitoring of remote sensing images, or collection by IoT sensors. Actual land parcel characteristic data may include, but is not limited to, one or more of the following: land use status change data, infrastructure access status data, and measured environmental carrying capacity data.

[0032] Actual enterprise characteristic data includes operational indicators such as investment intensity, employment scale, energy consumption, and emissions actually generated after the enterprise commences production and operation. This data can be used to provide real demand-side behavioral data, serving as a basis for verifying the accuracy of model predictions. In one specific embodiment of the invention, actual enterprise characteristic data can be obtained through enterprise reports from statistical departments, invoicing data from the tax system, online environmental monitoring platforms, and energy audit reports.

[0033] The dynamic weight set is a combination of influencing factor weights set according to different enterprise types. It is used to adjust the relative importance of revenue and environmental indicators in the comprehensive evaluation, and can be used to achieve differentiated evaluation standards, making the benefit simulation more in line with industry characteristics. In this embodiment, the dynamic weight set is generated based on historical matching cases. Each enterprise type corresponds to an independent set of weight configurations. Specific enterprise types can be classified according to industry classification codes or specific business products in the business scope.

[0034] Acquiring and updating the actual land parcel feature data and the actual enterprise feature data can be achieved by periodically synchronizing land use change information through the government big data platform API, or by manually supplementing the data through the enterprise-side reporting system by submitting quarterly operation reports. This ensures that the model input information is consistent with the real situation and enhances the realism of subsequent simulations.

[0035] The characteristic data of vacant land parcels is an initial set of information describing unoccupied land resources in terms of physical attributes, location conditions, planning, and policy constraints. The planning includes dynamic competitive environment information, which can be used as a basis for subsequent enterprise site selection matching, reflecting the current state of regional land supply and changes in its competitive environment. Dynamic competitive environment information can be used to identify implicit patterns such as industrial agglomeration effects, resource competition pressures, or increasingly stringent regulatory trends through correlation analysis of multiple existing enterprise-located parcels and enterprise operation data.

[0036] Collect the company's revenue and environmental data, denoted as adjusted revenue and adjusted environmental values. This data can be extracted from verified data entries in authoritative statistical annual reports, tax collection systems, and pollution source monitoring platforms. Furthermore, this operation can be achieved by using ETL tools to extract tax details from the integrated fiscal system, or by leveraging an environmental IoT platform to capture real-time monitoring averages of the company's discharge outlets. This allows for the establishment of objective and accurate performance benchmarks, supporting the validation of model effectiveness.

[0037] Specifically, the simulation difference is calculated by subtracting terms one by one or by using the relative error formula, forming a quantifiable prediction error matrix. Optimizing the parameters for calculating the benefit and environmental impacts can be achieved by applying least squares regression to correct the linear model parameters, or by using a neural network fine-tuning mechanism to update the nonlinear impact function, thereby improving the model's accuracy in predicting similar future projects.

[0038] As a specific example of this invention, taking the post-landing assessment and model optimization of a company mainly engaged in PCB products as an example, the system acquires actual enterprise characteristic data such as actual power load, number of employees, annual output value, and environmental impact assessment compliance status, while updating actual land plot characteristic data such as road connectivity and municipal pipeline access status. Based on the company's classification as "circuit board processing," the system calls the corresponding dynamic weight set to recalculate its simulated revenue and environmental values ​​under the current conditions. Simultaneously, the system collects its actual tax contribution and wastewater discharge concentration from tax and environmental protection platforms as corrections to the revenue and environmental values. Comparison reveals that the original model underestimated the long-term tax growth potential of this type of enterprise, resulting in a positive simulation error. Based on this, the system increases the proportion of the growth factor in the revenue weight coefficient of the dynamic weight set and fine-tunes the environmental weight coefficient to better reflect the characteristics of low-emission, high-value-added projects. This optimization result will be applied to subsequent site selection recommendations for similar enterprises to improve matching accuracy.

[0039] As a further improvement of the present invention, the resource server is equipped with a cloud decoder, and the enterprise feature data includes multiple sub-enterprise feature data and enterprise type; The specific steps for transmitting data to the resource server using double-layer encryption technology include: selecting a corresponding dynamic weight set based on the enterprise type; if there is a matching weight coefficient, revenue weight coefficient, or environmental weight coefficient in the dynamic weight set that corresponds to the sub-enterprise feature data and is greater than a preset importance encryption threshold, then the sub-enterprise feature data is recorded as the core feature for calculation. The feature names of all the land parcel feature data and enterprise feature data are classified according to the preset local language discriminator to obtain multiple privacy core features and multiple non-core features. The privacy core features and computation core features are marked as high-encryption features, and the non-core features are marked as low-encryption features. The preset importance encryption threshold is set according to the security policy and serves as a baseline for comparing weight coefficients. This prevents unauthorized individuals from using data with low privacy but high weights for matching degree, simulated profit, and simulated environment values ​​to deduce the best land plot for a company, thus affecting the company's subsequent development plans. Specifically, it can be set to 0.4. It's important to note that the feature name is a fixed "subheading" when acquiring land plot and company feature data. The subsequent conversion to standardized structured data involves content entered by staff under the corresponding "subheading" (because automatically entered content is generally filled directly according to the nodes and edges of the domain knowledge graph). The preset local language discriminator integrates a mapping table between feature names and classification levels, classifying and grading by directly recognizing the fixed feature names.

[0040] Select the corresponding dynamic weight set based on the enterprise type. Specifically, read the industry classification code, look up the pre-stored weight set mapping table, and load the corresponding parameter group.

[0041] The local encoder corresponding to the cloud decoder generates multiple latent vectors from all land parcel feature data and enterprise feature data. The latent vectors are then reconstructed using the local decoder configured with the same settings as the cloud decoder to obtain reconstructed data. The mean square error between the reconstructed data and the corresponding land parcel feature data or enterprise feature data is calculated, and a reconstruction confidence coefficient is generated. The local encoder and cloud encoder can be one or more of the following: a fully connected neural network encoder, a convolutional neural network-based encoder, or a variational autoencoder-based encoder. The local decoder is a reconstruction model symmetrical to the local encoder structure, used to recover the original feature data from latent vectors. Latent vectors are low-dimensional continuous numerical vectors generated by the local encoder, carrying the core semantic information of the original land parcel or enterprise feature data. The reconstruction trust coefficient is a quantitative indicator that measures the degree of deviation between the data reconstructed by the local decoder and the original data. It provides quality feedback signals during the data compression process and, in the case of encryption, assists the resource server in determining whether the received latent vectors are trustworthy.

[0042] As a further improvement of the present invention, the expression for calculating the reconstruction trust coefficient is as follows: , in, It is the reconstruction confidence coefficient corresponding to the k-th latent vector. It is a trust decay factor. It is the mean square error. Is it land parcel characteristic data or enterprise characteristic data? It's about reconstructing the data.

[0043] A key is generated by using a key derivation function to construct a trust coefficient corresponding to the highly encrypted feature. The latent vector of the highly encrypted feature is then encrypted using the key to obtain an encrypted latent vector. Multiple masquerading reconstruction loss data corresponding one-to-one with the encrypted latent vector are randomly generated. The encrypted latent vector and its corresponding masquerading reconstruction loss data, along with the remaining latent vectors and their corresponding reconstruction trust coefficients, are then transmitted in batches to the resource server. The key is transmitted to the resource server through a second encrypted communication channel. The resource server then obtains the land parcel feature data and enterprise feature data through symmetric decryption and decoding.

[0044] Specifically, the key derivation function takes a reconstructed trust coefficient with high encryption features as input and outputs a dedicated key for encrypting the latent vector. The key derivation function can be one or more of, but not limited to, PBKDF2, HKDF, and SCRYPT, and can be randomly selected and combined for encryption. The key is generated by the key derivation function based on the reconstructed trust coefficient and changes dynamically with each transmission, exhibiting strong unpredictability. It is transmitted to the resource server through a second encrypted communication channel. Specifically, a temporary channel based on asymmetric encryption or a channel based on a pre-shared key (PSK) can be established, and the key can be encapsulated and transmitted through this channel. Alternatively, random selection can be used for encrypted transmission. The symmetric decryption process is the reverse of the aforementioned encryption and encoding steps.

[0045] Of course, as another embodiment of the present invention, a key can be generated by a secure random number generator and transmitted through a second channel, and the reconstructed loss system can be encrypted and sent together with the latent vector.

[0046] Furthermore, if the masquerading reconstruction loss data is more easily exposed due to significant statistical differences, then generating masquerading reconstruction loss data is prohibited, and the latent vector with low encryption features can be sent together.

[0047] As a further improvement of the present invention, the dual-model processing and matching includes a natural language processing model and a rule matching model; The specific steps for identifying and extracting the supply hard condition dataset and demand hard condition dataset from the supply and demand feature matrix by processing the matching dual model include: converting the text feature descriptions in the supply and demand feature matrix into semantic vectors through the natural language processing model, converting the semantic vectors into standardized structured data, integrating the standardized structured data of the same vacant land to obtain the land parcel hard condition data, integrating all the land parcel hard condition data to obtain the supply hard condition dataset, integrating the standardized structured data of the same enterprise to obtain the enterprise hard condition data, and integrating all the enterprise hard condition data to obtain the demand hard condition dataset.

[0048] Specifically, natural language processing models can be fine-tuned based on pre-trained language models, such as BERT and RoBERTa, and extract deep semantic features of text through encoder architecture.

[0049] As a further improvement of the present invention, the natural language processing model includes a pre-trained domain knowledge graph, wherein multiple nodes and multiple edges in the domain knowledge graph correspond to multiple land enterprise normative concepts and multiple relationships between concepts, respectively. The pre-trained domain knowledge graph is based on policy documents, industry standards, local regulations, and historical approval cases for ontology modeling and triple extraction, and is pre-trained using graph neural networks for embedding. Nodes are basic semantic units representing specific land or enterprise-related normative concepts in the domain knowledge graph. They can be used as semantic anchors of the knowledge graph, carrying standardized terms and their unique identifiers. Specifically, nodes are one or more specific contents categorized according to different semantic categories, such as land use nature nodes, industry classification nodes, and environmental control nodes. Edges are directed connections connecting two nodes, indicating a certain deterministic semantic relationship between them. They can be used to characterize logical dependencies or constraints between concepts, supporting semantic combination and rule derivation. In specific embodiments of this invention, edges can be one or more of compatible, exclusive, or subordinate relationships, categorized according to different relationship types. Land enterprise normative concepts are standard terms clearly defined under the framework of natural resource management and industry access policies, possessing legal or administrative effect. They can be used as target points for unstructured text mapping, ensuring that the parsing results meet the policy semantic consistency requirements. The relationship between concepts is a verifiable semantic connection expressed by edges between two land enterprise normative concepts. It is an important auxiliary means to determine whether entities should be integrated into standard statements.

[0050] The specific steps for converting the semantic vector into standardized structured data include: processing the semantic vector through the named entity recognition model in the natural language processing model to obtain an original sequence with entity labels, wherein the original sequence includes multiple labeled fragmented words; By scanning all fragmented words through the domain knowledge graph and matching them with similar semantic land enterprise normative concepts, at least one fragmented word is replaced with a corresponding land enterprise normative concept to obtain an enhanced sequence. The remaining fragmented words in the enhanced sequence are then reorganized into entities to obtain at least one entity. If the context distance between multiple entities is less than the preset combination distance, and the nodes corresponding to two entities have the same side in the domain knowledge graph, then the entities are integrated to obtain a standard statement, the remaining entities are combined according to syntactic rules to obtain a standard statement, and all the standard statements are integrated to obtain standardized structured data.

[0051] The named entity recognition model is based on BiLSTM-CRF or Span-based Transformer architecture and fine-tuned on land and enterprise domain corpora. The original sequence is a linear text sequence of words with attached type labels, which can be used to preserve the original semantic fragments and their preliminary semantic categories, serving as the initial input for semantic alignment in the knowledge graph. For the combination distance, the criterion is greater than the distance based on positional adjacency. It can be dynamically generated based on whether the context forms a standard statement. For example, if the adjacent preceding text forms a standard statement, the combination distance is the distance to itself within the preceding text; otherwise, the combination distance can be extended along the context direction.

[0052] By scanning all fragmented words in the domain knowledge graph and matching them with similar semantic concepts of land enterprise norms, this operation can be achieved by calculating the cosine similarity between the semantic vector of each fragmented word and the embedding vector of each node in the knowledge graph, and selecting the node with the highest score as the matching target. For example, words like "edge," "close to," "near," "convenient," and "easy" in "subway edge" can be unified to obtain a unique land enterprise norm concept. Entity reorganization can be achieved by using a dependency parser to identify subject-verb-object structures or by guiding the reorganization path based on the co-occurrence frequency of nodes in the knowledge graph. This can achieve the technical effect of restoring over-segmented policy semantic units, such as reorganizing "high-end" + "manufacturing" + "industry" into "high-end manufacturing industry."

[0053] By integrating all standard statements to obtain standardized structured data, each standard statement can be parsed into key-value pairs or triples, categorized by field name, and deduplicated and merged to form structured records. Specifically, this operation can be performed by storing standard statements using RDF triples, thereby achieving the technical effect of outputting machine-readable structured data that directly supports subsequent matching calculations.

[0054] As a specific example of this invention, a company describes in its land acquisition materials that "a data center is planned to be built, requiring dual power supply, close to the fiber optic backbone network, and not located in a water source protection area." The domain knowledge graph maps this to the land company's normative concepts of "new infrastructure land," "dual power supply guarantee facilities," "communication infrastructure corridor," and "primary drinking water source protection zone." Specifically, there is a "belongs to" edge between the nodes corresponding to "data center" ("new infrastructure land") and "dual power supply guarantee facilities"), and an "excludes" edge between the nodes corresponding to "data center" and "primary drinking water source protection zone." Furthermore, the contextual distance between these two nodes in the original text is three land company normative concepts, which is less than the preset combination distance. Based on this, the system generates the standard statements "new infrastructure land shall not be located within the primary drinking water source protection zone" and "new infrastructure land is located within the power supply range of dual power supply guarantee facilities." The remaining entities generate the standard statement "should be located in the vicinity of the fiber optic backbone network" according to syntactic rules. Finally, this is integrated into standardized structured data for the rule matching model to verify whether it violates the ecological protection red line policy. This mechanism makes the spatial constraints that were originally implicit in free texts explicit, structured, and computable.

[0055] As a further improvement of the present invention, the enterprise feature data includes multiple sub-enterprise feature data, enterprise type, and a reconstructed trust coefficient corresponding one-to-one with each of the sub-enterprise feature data; Between steps S2 and S3, step S21 is also included: traversing the enterprise feature data of each enterprise in the supply and demand feature matrix, selecting the corresponding dynamic weight set according to the enterprise type, determining whether each sub-enterprise feature data under the enterprise feature data is blank or invalid data, and whether the matching weight coefficient, revenue weight coefficient, or environmental weight coefficient corresponding to the sub-enterprise feature data in the dynamic weight set is greater than zero. If so, the enterprise is marked as a blank enterprise. It is then identified whether the blank enterprise has a chain enterprise with the same name and the enterprise feature data and land feature data corresponding to the chain enterprise with the same name. If so, the similarity between the enterprise feature data of the blank enterprise and the enterprise feature data of each chain enterprise with the same name is calculated to obtain several enterprise feature similarities. The similarity between each vacant land and the land feature data of the chain enterprise with the same name is calculated to obtain several enterprise environmental similarities. Using the enterprise feature similarity and enterprise environment similarity as weights, the sub-enterprise feature data of the chain enterprises with the same name are weighted and averaged to calculate the alternative feature data. The alternative feature data is used to cover the corresponding blank and invalid data. The substitution loss coefficient is calculated based on all the enterprise feature similarities and enterprise environment similarities. The substitution loss coefficient is used to cover the corresponding reconstruction trust coefficient. If there is no chain enterprise with the same name, and no enterprise characteristic data and land characteristic data corresponding to the chain enterprise with the same name, then generate dynamic average characteristic data based on the enterprise characteristic data of the blank enterprise and cover the corresponding blank invalid data, and cover the corresponding reconstructed trust coefficient with the preset non-migrated data trust weight.

[0056] As mentioned above, the dynamic weight set is a set of variable parameters configured according to the differences in enterprise type. It includes the weight allocation of sub-features under the three evaluation dimensions of matching, revenue, and environment. It can be used to guide the system to determine which sub-enterprise feature data is meaningful in the current evaluation task and avoid misjudging irrelevant fields.

[0057] Blank or invalid data refers to missing or non-compliant field values ​​in enterprise feature data, manifested as empty data, incorrect formatting, or values ​​exceeding the reasonable range. The process involves iterating through the enterprise feature data of each enterprise in the supply and demand feature matrix. Specifically, this is achieved by using regular expression rules to match invalid formats or calling Pandas' `isnull()` function to identify missing fields, thereby locating the objects requiring repair. Chain enterprises with the same name can be enterprise entities with the same registered brand, group affiliation, or operational model replication relationship as the current enterprise. They can be used as reference sources for data completion, providing transferable operational and location characteristics.

[0058] Enterprise feature similarity is a numerical indicator that quantifies the degree of similarity between a blank enterprise and a chain enterprise with the same name in terms of industry attributes, investment behavior, and production models. It can be used to reflect the consistency level of internal enterprise attributes and is used in the weighted completion process. Specifically, the cosine similarity method is used for calculation.

[0059] Corporate environment similarity measures the degree of proximity between a target vacant plot of land and a chain of businesses with the same name in terms of location, infrastructure, and policy constraints. It can be used to assess the portability of site selection scenarios and support cross-regional experience transfer. Specifically, corporate environment similarity is calculated by constructing a weighted comparison using an AHP (Analog-Hybrid Hierarchical Analysis) model or by using Euclidean distance to measure standardized differences in geographical features.

[0060] The substitution loss coefficient is a quantitative indicator reflecting the uncertainty of information about surrogate feature data relative to the true observations. It can be used to characterize the quality level of the data completion process and replace the original reconstruction confidence coefficient in subsequent modeling confidence assessments. Calculating the substitution loss coefficient using enterprise feature similarity and enterprise environment similarity aims to further reduce the value of the subsequent reconstruction confidence coefficient, significantly reducing the weight of the transferred value in subsequent calculations and preventing erroneous data from contaminating benefit and environmental impact calculations.

[0061] Dynamic average feature data, generated from the statistical characteristics of existing samples of similar enterprises, serves as representative data. It fills in missing values ​​when there are no chain references, ensuring data integrity and providing reasonable estimates when reference sources are lacking. Specifically, dynamic average feature data can be generated by extracting the mean or median of relevant fields from actual enterprise feature data of enterprises in the same industry and of similar size. The preset non-migrated data trust weight can be, specifically, one-tenth of the maximum reconstruction trust coefficient in the existing data. This can be adjusted according to actual circumstances.

[0062] As a specific example of this invention, a startup AI company failed to fill in approximate energy consumption indicators and production capacity plans when applying for land use, resulting in blank and invalid data in the company's feature data. After identifying this missing data in the supply and demand feature matrix, the system found that its parent company already had three R&D bases with the same name across the country. The system calculated the similarity of the company's features with each base in terms of R&D investment and personnel structure, and combined this with the similarity of the target park with the regional environments of the bases in terms of transportation, electricity, and talent supply to generate a composite weight. Based on this, a weighted average of the actual operating data of the three bases was calculated to obtain reasonable alternative feature data to fill the gaps, and a substitution loss coefficient of 0.28 was calculated simultaneously. Ultimately, the company was included in the normal matching process and successfully recommended to a suitable land plot, with its simulated benefits and environmental impact assessments both accompanied by quality annotations, improving decision-making transparency.

[0063] As a further improvement of the present invention, the supply and demand feature matrix includes multiple plot feature data, multiple enterprise feature data, and multiple reconstructed trust coefficients that correspond one-to-one with the plot feature data or one-to-one with the enterprise feature data. The enterprise feature data includes multiple sub-enterprise feature data and enterprise types. The calculation of matching degree, benefit impact and environmental impact includes: selecting the corresponding dynamic weight set according to the enterprise type, and selecting the matching weight coefficient, benefit weight coefficient and environmental weight coefficient with a value greater than zero in the dynamic weight set; The matching degree value is obtained by calculating and normalizing the selected matching weight coefficient, the enterprise feature data and land feature data corresponding to the matching weight coefficient, and the reconstruction trust coefficient corresponding to the enterprise feature data and land feature data; The simulated revenue value is calculated based on the selected revenue weighting coefficient, the enterprise characteristic data and land parcel characteristic data corresponding to the revenue weighting coefficient, and the reconstruction trust coefficient corresponding to the enterprise characteristic data and land parcel characteristic data. The simulated environmental value is calculated based on the selected environmental weight coefficient, the enterprise characteristic data and land parcel characteristic data corresponding to the environmental weight coefficient, and the reconstruction trust coefficient corresponding to the enterprise characteristic data and land parcel characteristic data.

[0064] Similarly, selecting matching weight coefficients, benefit weight coefficients, and environmental weight coefficients with values ​​greater than zero in the dynamic weight set is to filter out dimensions with weight values ​​of zero or invalid, and only retain the coefficients that are valid for calculation.

[0065] Furthermore, it's important to note that the required content for enterprise characteristic data and land parcel characteristic data differs depending on the type of enterprise. Specifically, the specific content of the enterprise characteristic data must be determined based on the enterprise type and implementation procedures; not every business uses the same formulas for matching degree calculation, revenue impact calculation, and environmental impact calculation.

[0066] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method for intelligent matching of land supply and demand and efficiency simulation, characterized in that, Includes the following steps: S1. Collect land parcel feature data from multiple vacant land plots and enterprise feature data from multiple enterprises, and transmit them to the resource server through double-layer encryption transmission technology. The resource server integrates all the land parcel feature data and all the enterprise feature data to obtain a supply and demand feature matrix. S2. By processing the matching dual model, identify and extract the supply hard condition dataset and demand hard condition dataset from the supply and demand feature matrix. Filter according to the supply hard condition dataset and demand hard condition dataset to obtain multiple filter matching sets that correspond one-to-one with each vacant land or one-to-one with each enterprise. S3. Calculate the matching degree, revenue impact, and environmental impact for each pair of land parcel enterprises in the screening and matching set according to the supply and demand feature matrix, and obtain the corresponding matching degree value, simulated revenue value, and simulated environmental value. Integrate all the matching degree values, simulated revenue values, and simulated environmental values ​​to obtain the matching benefit matrix.

2. The method for intelligent matching of land supply and demand and efficiency simulation according to claim 1, characterized in that, Step S3 is followed by step S4: When an enterprise is established, it is marked as an established enterprise. The actual land plot feature data and the actual enterprise feature data of the land plot corresponding to the established enterprise are obtained and updated. The corresponding dynamic weight set is selected according to the enterprise type of the actual enterprise feature data. The revenue impact and environmental impact are calculated according to the actual land plot feature data, the actual enterprise feature data and the dynamic weight set to obtain the corresponding simulated revenue value and simulated environmental value. The land plot feature data of the vacant land is updated according to the actual land plot feature data and the actual enterprise feature data. Collect revenue and environmental data of the enterprises that have settled in the area, and denote them as corrected revenue data and corrected environmental data. Calculate the simulation difference based on the corrected revenue data, corrected environmental data, simulated revenue data, and simulated environmental data, and optimize the values ​​of the revenue weight coefficient and environmental weight coefficient in the dynamic weight set based on the simulation difference.

3. The method for intelligent matching of land supply and demand and benefit simulation according to claim 1, characterized in that, The resource server is equipped with a cloud decoder, and the enterprise characteristic data includes multiple sub-enterprise characteristic data and enterprise type; The specific steps for transmitting data to the resource server using double-layer encryption technology include: selecting a corresponding dynamic weight set based on the enterprise type; if there is a matching weight coefficient, revenue weight coefficient, or environmental weight coefficient in the dynamic weight set that corresponds to the sub-enterprise feature data and is greater than a preset importance encryption threshold, then the sub-enterprise feature data is recorded as the core feature for calculation. The feature names of all the land parcel feature data and enterprise feature data are classified according to the preset local language discriminator to obtain multiple privacy core features and multiple non-core features. The privacy core features and computation core features are marked as high-encryption features, and the non-core features are marked as low-encryption features. The local encoder corresponding to the cloud decoder generates multiple latent vectors from all land parcel feature data and enterprise feature data. The latent vectors are then reconstructed using the local decoder configured with the same settings as the cloud decoder to obtain reconstructed data. The mean square error between the reconstructed data and the corresponding land parcel feature data or enterprise feature data is calculated, and a reconstruction confidence coefficient is generated. A key is generated by using a key derivation function to construct a trust coefficient corresponding to the highly encrypted feature. The latent vector of the highly encrypted feature is then encrypted using the key to obtain an encrypted latent vector. Multiple masquerading reconstruction loss data corresponding one-to-one with the encrypted latent vector are randomly generated. The encrypted latent vector and its corresponding masquerading reconstruction loss data, along with the remaining latent vectors and their corresponding reconstruction trust coefficients, are then transmitted in batches to the resource server. The key is transmitted to the resource server through a second encrypted communication channel. The resource server then obtains the land parcel feature data and enterprise feature data through symmetric decryption and decoding.

4. The method for intelligent matching of land supply and demand and efficiency simulation according to claim 1, characterized in that, The dual-model matching process includes a natural language processing model and a rule matching model. The specific steps for identifying and extracting the supply hard condition dataset and demand hard condition dataset from the supply and demand feature matrix by processing the matching dual model include: converting the text feature descriptions in the supply and demand feature matrix into semantic vectors through the natural language processing model, converting the semantic vectors into standardized structured data, integrating the standardized structured data of the same vacant land to obtain the land parcel hard condition data, integrating all the land parcel hard condition data to obtain the supply hard condition dataset, integrating the standardized structured data of the same enterprise to obtain the enterprise hard condition data, and integrating all the enterprise hard condition data to obtain the demand hard condition dataset.

5. The method for intelligent matching of land supply and demand and benefit simulation according to claim 4, characterized in that, The natural language processing model includes a pre-trained domain knowledge graph, in which multiple nodes and multiple edges correspond to multiple land enterprise normative concepts and multiple relationships between concepts, respectively. The specific steps for converting the semantic vector into standardized structured data include: processing the semantic vector through the named entity recognition model in the natural language processing model to obtain an original sequence with entity labels, wherein the original sequence includes multiple labeled fragmented words; By scanning all fragmented words through the domain knowledge graph and matching them with similar semantic land enterprise normative concepts, at least one fragmented word is replaced with a corresponding land enterprise normative concept to obtain an enhanced sequence. The remaining fragmented words in the enhanced sequence are then reorganized into entities to obtain at least one entity. If the context distance between multiple entities is less than the preset combination distance, and the nodes corresponding to two entities have the same side in the domain knowledge graph, then the entities are integrated to obtain a standard statement, the remaining entities are combined according to syntactic rules to obtain a standard statement, and all the standard statements are integrated to obtain standardized structured data.

6. The method for intelligent matching of land supply and demand and benefit simulation according to claim 3, characterized in that, The enterprise feature data includes multiple sub-enterprise feature data, enterprise type, and a reconstructed trust coefficient corresponding one-to-one with each of the sub-enterprise feature data. Between steps S2 and S3, step S21 is also included: traversing the enterprise feature data of each enterprise in the supply and demand feature matrix, selecting the corresponding dynamic weight set according to the enterprise type, determining whether each sub-enterprise feature data under the enterprise feature data is blank or invalid data, and whether the matching weight coefficient, revenue weight coefficient, or environmental weight coefficient corresponding to the sub-enterprise feature data in the dynamic weight set is greater than zero. If so, the enterprise is marked as a blank enterprise. It is then identified whether the blank enterprise has a chain enterprise with the same name and the enterprise feature data and land feature data corresponding to the chain enterprise with the same name. If so, the similarity between the enterprise feature data of the blank enterprise and the enterprise feature data of each chain enterprise with the same name is calculated to obtain several enterprise feature similarities. The similarity between each vacant land and the land feature data of the chain enterprise with the same name is calculated to obtain several enterprise environmental similarities. Using the enterprise feature similarity and enterprise environment similarity as weights, the sub-enterprise feature data of the chain enterprises with the same name are weighted and averaged to calculate the alternative feature data. The alternative feature data is used to cover the corresponding blank and invalid data. The substitution loss coefficient is calculated based on all the enterprise feature similarities and enterprise environment similarities. The substitution loss coefficient is used to cover the corresponding reconstruction trust coefficient. If there is no chain enterprise with the same name, and no enterprise characteristic data and land characteristic data corresponding to the chain enterprise with the same name, then generate dynamic average characteristic data based on the enterprise characteristic data of the blank enterprise and cover the corresponding blank invalid data, and cover the corresponding reconstructed trust coefficient with the preset non-migrated data trust weight.

7. The method for intelligent matching of land supply and demand and efficiency simulation according to claim 3, characterized in that, The expression for calculating the reconstructed trust coefficient is as follows: ， in, It is the reconstruction confidence coefficient corresponding to the k-th latent vector. It is a trust decay factor. It is the mean square error. Is it land parcel characteristic data or enterprise characteristic data? It's about reconstructing the data.

8. The method for intelligent matching of land supply and demand and efficiency simulation according to claim 3, characterized in that, The supply and demand feature matrix includes multiple plot feature data, multiple enterprise feature data, and multiple reconstructed trust coefficients that correspond one-to-one with the plot feature data or one-to-one with the enterprise feature data. The enterprise feature data includes multiple sub-enterprise feature data and enterprise types. The calculation of matching degree, benefit impact and environmental impact includes: selecting the corresponding dynamic weight set according to the enterprise type, and selecting the matching weight coefficient, benefit weight coefficient and environmental weight coefficient with a value greater than zero in the dynamic weight set; The matching degree value is obtained by calculating and normalizing the selected matching weight coefficient, the enterprise feature data and land feature data corresponding to the matching weight coefficient, and the reconstruction trust coefficient corresponding to the enterprise feature data and land feature data; The simulated revenue value is calculated based on the selected revenue weighting coefficient, the enterprise characteristic data and land parcel characteristic data corresponding to the revenue weighting coefficient, and the reconstruction trust coefficient corresponding to the enterprise characteristic data and land parcel characteristic data. The simulated environmental value is calculated based on the selected environmental weight coefficient, the enterprise characteristic data and land parcel characteristic data corresponding to the environmental weight coefficient, and the reconstruction trust coefficient corresponding to the enterprise characteristic data and land parcel characteristic data.

Images