Land site selection recommendation method and system fusing constraint conditions and optimization algorithm

Abstract

The application relates to a land use site selection recommendation method and system combining constraint conditions and optimization algorithms, which comprises the following steps: acquiring a constraint feature set, dividing the constraint feature set into strong constraint features and weak constraint features, and constructing a site selection knowledge graph; acquiring project demand information, screening a strong constraint set and a weak constraint set based on the site selection knowledge graph; obtaining a compliant land information set based on the strong constraint set, extracting fuzzy clauses in the weak constraint set, and generating auxiliary decision suggestions based on the site selection knowledge graph; solving the compliant land information set and the weak constraint set, generating a site selection recommendation solution set, and outputting the site selection recommendation solution set to a user end; the application automatically analyzes planning clauses and spatial data through a multi-channel feature extraction model, intelligently classifies constraint conditions by constructing a site selection knowledge graph, generates a recommendation solution set in combination with a multi-objective optimization algorithm, and has the advantages of improving the automatic processing degree of complex planning clauses and realizing multi-objective optimization decision.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of land site selection recommendation, and in particular to a land site selection recommendation method and system fusing constraint conditions and optimization algorithms. BACKGROUND

[0002] With the development of smart cities and territorial space planning, using computer technology to assist in land site selection for construction projects has become an important direction to improve the scientific nature and efficiency of planning.

[0003] In the prior art, the land site selection method based on geographic information systems mainly has the following technical defects: first, in terms of planning article processing, the existing method highly depends on manual interpretation and is difficult to automatically process a large number of complex planning regulations, especially cannot effectively analyze the reference relationship between articles, exceptions and a large number of descriptive requirements, resulting in low system automation and high knowledge maintenance cost; second, in terms of constraint condition processing, the existing method mainly uses rigid screening logic and lacks intelligent differentiation ability for strong and weak constraint conditions, which cannot effectively process the ambiguity clauses and flexibility requirements in planning articles; third, in terms of decision support, the existing system cannot realize deep semantic association between spatial data and planning rules, cannot automatically schedule multi-source heterogeneous data according to project requirements, and lacks the ability to solve multi-objective optimization problems, resulting in single site selection scheme and lack of scientific basis. SUMMARY

[0004] In order to solve the above-mentioned defects, the present application provides a land site selection recommendation method and system fusing constraint conditions and optimization algorithms.

[0005] The above-mentioned invention purpose of the present application is realized by the following technical scheme:

[0006] A land site selection recommendation method fusing constraint conditions and optimization algorithms, comprising the steps of:

[0007] Obtaining planning data information and spatial data information and inputting them into a pre-trained multi-channel feature extraction model to obtain a constraint feature set, the constraint feature set including spatial constraint features and attribute constraint features;

[0008] Dividing the constraint feature set into strong constraint features and weak constraint features by a pre-defined constraint classification rule and constructing a site selection knowledge graph;

[0009] Obtaining project requirement information from a user end and extracting target area information and attribute requirement information, screening a strong constraint set and a weak constraint set corresponding to the target area information and the attribute requirement information based on the site selection knowledge graph;

[0010] Based on a strong constraint set, the target area information is subjected to spatial superposition analysis to obtain a compliance land use information set, the fuzzy clauses in the weak constraint set are extracted, and an auxiliary decision suggestion is generated based on the site selection knowledge graph;

[0011] A preset multi-objective optimization algorithm is used to solve the compliance land use information set and the weak constraint set to generate a site selection recommendation solution set.

[0012] The site selection recommendation solution set and the auxiliary decision suggestion are integrated into a site selection decision report and output to the user end.

[0013] In a preferred example, the multi-channel feature extraction model includes a natural language channel, a spatial analysis channel, and an integrated output layer. The planning data information and spatial data information are obtained and input into a pre-trained multi-channel feature extraction model to obtain a constraint feature set, which includes spatial constraint features and attribute constraint features. The steps include:

[0014] The natural language channel identifies the natural language clauses in the planning data information and performs semantic analysis to extract attribute constraint features.

[0015] The spatial analysis channel performs spatial relationship reasoning on the spatial data information to extract spatial constraint features.

[0016] The integrated output layer aligns and cross-modal fuses the attribute constraint features and the spatial constraint features to generate the constraint feature set.

[0017] In a preferred example, the natural language channel identifies the natural language clauses in the planning data information and performs semantic analysis to extract attribute constraint features. The steps include:

[0018] The natural language channel performs reference relationship analysis on the identified natural language clauses and establishes traceable associations.

[0019] The natural language channel identifies the effectiveness level, context, and modal words of the natural language clauses, and determines the constraint enforcement level of the natural language clauses based on the identification results and pre-set level determination rules.

[0020] The natural language channel generates attribute constraint features based on the natural language clauses and their constraint enforcement levels.

[0021] In a preferred example, the spatial analysis channel performs spatial relationship reasoning on the spatial data information to extract spatial constraint features. The steps include:

[0022] The spatial analysis channel performs basic spatial analysis on the spatial data, and maps the geometric relationship obtained by the basic spatial analysis into preset spatial relationship semantics, the basic spatial analysis including overlay analysis, buffer analysis, network analysis and visual range analysis;

[0023] The spatial analysis channel matches the mapped spatial relationship semantics with the preset spatial constraint rule template, and generates a spatial constraint condition vector;

[0024] The spatial analysis channel standardizes and encodes the spatial constraint condition vector, and outputs a spatial constraint feature.

[0025] In a preferred example, the application can be further configured to: the step of dividing the constraint feature set into strong constraint features and weak constraint features by the pre-defined constraint classification rule, and constructing the site selection knowledge graph, comprising the steps of:

[0026] The constraint feature set is matched with the pre-constructed classification rule library by the pre-defined constraint classification rule, so as to divide the constraint feature set into strong constraint features and weak constraint features;

[0027] The strong constraint features and the weak constraint features are instantiated as feature nodes in the site selection knowledge graph, and a core concept node is created, the core concept node including a project demand concept node, a spatial data source concept node and a calculation logic concept node;

[0028] According to the pre-defined domain ontology, a semantic relationship chain between the core concept node and the feature node in the site selection knowledge graph is established.

[0029] In a preferred example, the application can be further configured to: the step of obtaining project demand information from the user end, and extracting target area information and attribute demand information, screening a strong constraint set and a weak constraint set corresponding to the target area information and the attribute demand information based on the site selection knowledge graph, comprising the steps of:

[0030] A project instance node is created in the site selection knowledge graph based on the target area information and the attribute demand information, and a query condition of the site selection knowledge graph is generated, the query condition including a spatial position parameter and a project attribute label;

[0031] Taking the project instance node as a query starting point, all feature nodes associated with the spatial position parameter and the project attribute label are retrieved along the semantic relationship chain in the site selection knowledge graph;

[0032] The strong constraint features and the weak constraint features corresponding to the retrieved feature nodes are summarized to form a strong constraint set and a weak constraint set.

[0033] The application can be further configured in a preferred example: the site selection knowledge graph includes spatial data source nodes, the step of performing spatial overlay analysis on the target region information based on the strong constraint set to obtain the compliance land use information set, extracting the ambiguity clauses in the weak constraint set, and generating the auxiliary decision-making suggestion based on the site selection knowledge graph includes the steps of:

[0034] analyzing the spatial constraint conditions in the strong constraint set and generating corresponding spatial analysis instruction sequences;

[0035] identifying the corresponding relevant spatial data source nodes based on the spatial analysis instruction sequences, and scheduling the original spatial data of the identified spatial data source nodes to generate a regular data set after corresponding data processing;

[0036] performing spatial overlay analysis on the target region information based on the spatial analysis instruction sequences and the regular data set, and outputting the spatial region intersection that meets all spatial constraint conditions as the compliance land use information set.

[0037] The application can be further configured in a preferred example: the step of performing spatial overlay analysis on the target region information based on the strong constraint set to obtain the compliance land use information set, extracting the ambiguity clauses in the weak constraint set, and generating the auxiliary decision-making suggestion based on the site selection knowledge graph includes the steps of:

[0038] identifying and extracting the descriptive and relevant ambiguity clauses in the weak constraint set;

[0039] For the identified ambiguity clauses, searching for a number of historical cases in the case library of the site selection knowledge graph;

[0040] aggregating and analyzing the processing methods and results of the historical cases to generate an auxiliary decision-making suggestion containing quantitative recommended values, qualitative processing methods, and potential risk prompts.

[0041] The application can be further configured in a preferred example: the site selection knowledge graph includes spatial data source nodes, and the step of integrating the site selection recommendation solution set and the auxiliary decision-making suggestion into a site selection decision report and outputting it to the user end includes the steps of:

[0042] converting each scheme solution in the site selection recommendation solution set into a decision vector, and generating a decision matrix based on the converted decision vector;

[0043] performing semantic association between each item of metadata in the decision matrix and the corresponding data source nodes in the site selection knowledge graph to generate a semantic enhanced matrix;

[0044] generating a site selection decision report based on the semantic enhanced matrix, the compliance land use information set, and the auxiliary decision-making suggestion.

[0045] The second application object of the present application is achieved by the following technical scheme.

[0046] A land site selection recommendation system fusing constraint conditions and optimization algorithms, comprising:

[0047] A feature extraction module is configured to acquire planning data information and spatial data information, and input the information into a pre-trained multi-channel feature extraction model to obtain a constraint feature set, the constraint feature set including spatial constraint features and attribute constraint features.

[0048] A graph construction module is configured to divide the constraint feature set into strong constraint features and weak constraint features according to a pre-defined constraint classification rule, and construct a site selection knowledge graph.

[0049] A constraint screening module is configured to acquire project demand information from a user end, extract target area information and attribute demand information, and screen a strong constraint set and a weak constraint set corresponding to the target area information and the attribute demand information based on the site selection knowledge graph.

[0050] An information generation module is configured to perform spatial overlay analysis on the target area information based on the strong constraint set to obtain a compliant land information set, extract fuzzy clauses in the weak constraint set, and generate an auxiliary decision suggestion based on the site selection knowledge graph.

[0051] An optimization solving module is configured to solve the compliant land information set and the weak constraint set by using a pre-set multi-objective optimization algorithm to generate a site selection recommendation solution set.

[0052] A decision integration module is configured to integrate the site selection recommendation solution set and the auxiliary decision suggestion into a site selection decision report and output the report to the user end.

[0053] In summary, the land site selection recommendation method and system fusing constraint conditions and optimization algorithms provided by the present application automatically analyze planning clauses and spatial data by using a multi-channel feature extraction model, construct a site selection knowledge graph to realize intelligent classification of constraint conditions, and generate a recommendation solution set by using a multi-objective optimization algorithm, thereby improving the degree of automation in processing complex planning clauses and the degree of intelligent differentiation of constraint strength, and realizing the advantages of multi-objective optimization decision. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 is a flowchart of an embodiment of the land site selection recommendation method fusing constraint conditions and optimization algorithms of the present application;

[0055] Figure 2 is an implementation flowchart of step S10 in the embodiment of the land site selection recommendation method fusing constraint conditions and optimization algorithms of the present application;

[0056] Figure 3is an implementation flowchart of step S11 in an embodiment of a land site selection recommendation method fusing constraint conditions and optimization algorithms. DETAILED DESCRIPTION

[0057] The following will be described in detail below with reference to the accompanying drawings. Figures 1-3 The present application will be further described in detail.

[0058] In an embodiment, as shown in the figure, the present application discloses a land site selection recommendation method fusing constraint conditions and optimization algorithms, specifically including the following steps: Figure 1

[0059] S10: Obtain planning data information and spatial data information, and input them into a pre-trained multi-channel feature extraction model to obtain a constraint feature set, the constraint feature set including spatial constraint features and attribute constraint features;

[0060] In the embodiment, the planning data information refers to the general term of normative documents and provisions in the field of urban planning in the form of natural language text, the sources of which include but are not limited to overall planning of national space at various levels, special planning, technical regulations, and design conditions of specific plots, etc. The planning data information usually exists in the form of unstructured documents, containing various regulations and requirements required for land site selection; the spatial data information refers to data with geographic coordinate reference and capable of representing surface spatial features and phenomena, in the embodiment, specifically refers to multi-source geographic spatial data related to site selection, such as basic geographic information data, land use status data, various planning control line data, infrastructure data, and environmental sensitive area data, etc.; the multi-channel feature extraction model refers to a neural network architecture capable of processing text and spatial data in parallel, for example, using a natural language processing module to analyze planning provisions, and a geographic spatial analysis module to extract spatial relationships; the constraint feature set refers to a set of machine-readable features extracted from the original data after processing by the multi-channel feature extraction model, capable of representing planning restrictions and conditions, the constraint feature set is a highly summarized and structured expression of the original complex data, providing a unified input for subsequent classification, reasoning and optimization calculation; the spatial constraint feature refers to a restrictive condition describing the spatial distribution, geometric shape and spatial relationship of geographic entities extracted from spatial data, the essence of the spatial constraint feature is the dependence of position and spatial relationship; the attribute constraint feature refers to a restrictive condition describing the non-spatial attributes of geographic entities themselves extracted from planning text, the essence of the attribute constraint feature is the attribute regulation independent of spatial position.

[0061] ​Exemplarily, the spatial constraint features can include: location relationship, such as “whether the land plot is located in a high-risk geological disaster area” and “whether the distance between the land plot boundary and the river trunk road meets the safety setback requirement”; shape relationship, such as “whether the shape of the land plot is regular and conducive to layout” and “whether the slope of the land plot exceeds the threshold value of allowable construction”; connectivity relationship, such as “whether the accessibility of the land plot to the nearest highway exit meets the logistics demand”;

[0062] The attribute constraint features can include: strength index, such as “the plot ratio (attribute) is not greater than (operator) 2.0 (value)” and “the building density (attribute) is not less than (operator) 30% (value)”; function index, such as “the land use property (attribute) must be (operator) a type of industrial land (value)”; and style index, such as “the building height limit (attribute) shall not exceed (operator) 50 meters (value)”.

[0063] S20: classifying the constraint feature set into strong constraint features and weak constraint features through a pre-defined constraint classification rule, and constructing a site selection knowledge graph;

[0064] In the embodiment, the constraint classification rule can adopt a classification rule based on the level of effectiveness, for example, determining a planning provision with an effectiveness level higher than a local standard as a strong constraint. The strong constraint feature refers to a constraint condition in the constraint feature set with strong mandatory and non-flexible space. The strong constraint feature usually corresponds to the prohibitive and restrictive requirements in the planning provision, and is a bottom line that must be met by the site selection scheme. In the technical implementation, the scheme that violates the strong constraint feature will be directly excluded. The weak constraint feature refers to a constraint condition in the constraint feature set with guiding and optimized space. The weak constraint feature usually corresponds to the desirability and target requirements in the planning provision, and is a flexible index for measuring the pros and cons of the site selection scheme. In the technical implementation, the weak constraint feature is usually converted into an optimized objective function in a multi-objective optimization problem. The site selection knowledge graph refers to a structured semantic network for representing site selection field knowledge, which is composed of nodes (representing entities or concepts, such as constraint conditions, data sources, and project types) and edges (representing the relationship between nodes, such as “dependent on” and “applicable to”). The site selection knowledge graph constructs an entity relationship network through a graph database, for example, establishing an association between the ecological protection area node and the plot ratio node.

[0065] S30: obtaining project demand information from the user end, extracting target area information and attribute demand information, and screening a strong constraint set and a weak constraint set corresponding to the target area information and the attribute demand information based on the site selection knowledge graph;

[0066] In the present embodiment, the project demand information refers to the initial and comprehensive description of the to-be-sited project proposed by the user end, the project demand information is the total input and starting point of starting the site selection analysis task, and is usually in the form of natural language, form or file, etc., the project demand information defines the basic range and direction of the site selection task; the target area information refers to the defined data about the search range of the site selection space extracted from the project demand information, which clearly indicates the geographic range in which the site selection analysis task needs to search and evaluate the land use, the target area information provides the geographic boundary and calculation range for all subsequent spatial analysis operations (such as overlay analysis); the attribute demand information refers to the structured data about the non-spatial characteristics and requirements of the project extracted from the project demand information, which describes the inherent attribute requirements of the project for the land, and the attribute demand information is the core matching basis for intelligently screening the related constraint rules (strong / weak constraint set) from the site selection knowledge graph; the strong constraint set refers to the set of all mandatory rule conditions that must be met for the current specific site selection project, which is dynamically screened from the complete site selection knowledge graph, and the strong constraint set defines the legality and feasibility boundary of the site selection scheme; the weak constraint set refers to the set of all guiding rule conditions representing the optimization direction and desired target for the current specific site selection project, which is dynamically screened from the complete site selection knowledge graph, and the weak constraint set defines the optimization degree and potential value of the site selection scheme, which is the target that needs to be weighed and optimized in the feasible solution space.

[0067] S40: performing spatial overlay analysis on the target area information based on the strong constraint set to obtain a compliant land information set, extracting the ambiguity clauses in the weak constraint set, and generating an auxiliary decision suggestion based on the site selection knowledge graph;

[0068] In the present embodiment, the spatial overlay analysis refers to the spatial analysis technology in geographic information system, which finds the regions that meet multiple spatial conditions by performing logical operations such as intersection, merging and erasing on multiple spatial data layers based on geometric position, in the present embodiment, the spatial overlay analysis is used to execute the strong constraint feature, and quickly filters out the compliant land information set that meets all the hard spatial conditions from the target area; the compliant land information set refers to the set of land or spatial units that meet all the mandatory rules in the strong constraint set within the range of the target area information specified by the user, which is verified by spatial analysis, and represents the feasible solution space of the site selection problem; the ambiguity clause refers to the descriptive and qualitative requirement that cannot be directly quantified as an accurate mathematical model or a simple logical rule extracted from the planning provisions or project demand, the ambiguity clause usually contains subjective judgment and value trade-off; the auxiliary decision suggestion refers to the reference opinion with data support and logical basis provided for the user end by calling the internal knowledge base and intelligent algorithm (such as case-based reasoning) for the ambiguity clauses in the weak constraint set, so as to enhance the intelligent level of decision-making.

[0069] Exemplarily, the auxiliary decision suggestion content usually contains:

[0070] Quantitative reference: give a recommended value or range based on case statistics (e.g. “the average green ratio of similar projects is 35%, the recommended reference”);

[0071] Qualitative scheme: recommend feasible treatment or best practice (e.g. “retreat-style building layout can be used to echo the mountain shape”);

[0072] Risk warning: point out possible risks or conflicts (e.g. “the east side of the mountain is visually sensitive, and over-excavation may pose a risk”).

[0073] S50: adopt a preset multi-objective optimization algorithm to solve the compliance land information set and the weak constraint set, and generate a site selection recommendation solution set;

[0074] In this embodiment, the multi-objective optimization algorithm refers to a mathematical method for solving problems with multiple conflicting optimization objectives. In this embodiment, the multi-objective optimization algorithm is used to weigh different objectives represented by multiple weak constraint features within the compliance land information set, so as to find a solution that achieves the best balance in all aspects, i.e. the site selection recommendation solution set. The multi-objective optimization algorithm can use genetic algorithm or particle swarm algorithm, for example, setting land cost and traffic accessibility as the optimization objective function. The site selection recommendation solution set refers to a set or an optimal (or approximately optimal) land use site selection scheme generated by the multi-objective optimization algorithm after comprehensive weighing and calculation of each plot in the compliance land information set on multiple objectives defined by the weak constraint set. The site selection recommendation solution set is a scheme set that reflects different weighting preferences.

[0075] S60: integrate the site selection recommendation solution set and the auxiliary decision suggestion into a site selection decision report and output to the user end.

[0076] In this embodiment, the site selection decision report refers to the final output structured result. The site selection decision report integrates the quantitative comparison results of the site selection recommendation solution set and the qualitative analysis of the auxiliary decision suggestion, aiming to provide a clear and well-founded decision reference for the user end, rather than just a simple scheme list.

[0077] Specifically, the natural language clauses in the planning data are converted into attribute constraints through semantic analysis, and the spatial data are extracted through superimposition analysis to obtain spatial constraints; the constraint features are divided into mandatory and non-mandatory according to the level of effectiveness, and a knowledge graph containing spatial relationships and clause references is constructed; after the user inputs the project requirements, the system searches for the matching constraint conditions in the knowledge graph, and filters out the mandatory constraints and the adjustable weak constraints; the spatial superimposition analysis excludes the areas that do not meet the mandatory constraints, and the remaining land is used as the input of the optimization algorithm; the fuzzy clauses in the weak constraints are matched to generate quantitative suggestions through historical cases, and the multi-objective optimization algorithm outputs multiple candidate schemes, and finally a decision report containing scheme comparison and risk prompt is formed.

[0078] Through the above technical solutions, the application realizes the automatic processing of planning clauses and spatial data, reduces manual intervention; the semantic association under the support of the knowledge graph can dynamically schedule multi-source data, and improve the accuracy of spatial analysis; the comparable and optional schemes generated by the multi-objective optimization algorithm provide quantitative basis for the decision user end, and the decision-making suggestions help to solve the execution problems of the fuzzy clauses. Through the above steps and features, the application significantly improves the rationality and decision efficiency of the land site selection scheme on the basis of ensuring compliance.

[0079] In an embodiment, the multi-channel feature extraction model includes a natural language channel, a spatial analysis channel, and an integrated output layer, as shown in Figure 2 The step S10 includes:

[0080] S11: The natural language channel identifies the natural language clauses in the planning data information and performs semantic analysis thereon to extract attribute constraint features;

[0081] In the embodiment, the natural language channel refers to a module for processing text type planning data, which can be implemented by using a deep learning based semantic analysis model. By identifying the effectiveness level, context and modal words in the natural language clauses, attribute constraint conditions with regulatory effectiveness are automatically extracted. Semantic analysis refers to the core processing action of the natural language channel. Semantic analysis usually does not refer to simple keyword extraction, but refers to syntax and semantic analysis of the clauses to identify the core predicates (operators), entities (planning indicators), numerical values (standard values) and their logical relationships, and finally form a calculable logical expression, for example, "the volume rate shall not be higher than 2.0" is analyzed as FAR (volume rate) ≤ 2.0. The attribute constraint feature refers to the output result of the natural language channel, which refers to the attribute regulation extracted from the planning clause text, which is independent of the spatial position and describes the attributes that the land or project itself should satisfy. The attribute constraint feature is usually represented in the form of structured data (such as key-value pair, triple or feature vector), for example, {“constraint type”:“upper limit”,“indicator”:“volume rate”,“value”:“2.0”}.

[0082] S12: The spatial analysis channel performs spatial relationship reasoning on the spatial data information to extract spatial constraint features;

[0083] In this embodiment, the spatial analysis channel refers to a module for processing geographic spatial data, which can be implemented by using a GIS spatial analysis tool chain. The geometric relationship is converted into computable spatial constraint semantics through overlay analysis and buffer analysis. The spatial relationship reasoning refers to the core processing action of the spatial analysis channel, which refers to deriving higher-level semantic relationships from the geometric properties of the original spatial data through computational geometry, topology, etc. For example, from the spatial position of the land plot and the ecological protection layer, the semantic relationships with planning significance such as “adjacent”, “intrusion”, and “away” are inferred. The spatial constraint features refer to the output results of the spatial analysis channel, which specifically refer to the restrictive conditions directly related to geographic location and spatial interaction extracted from the spatial data. They represent the constraints formed by the distribution, combination, and interaction of geographic entities in space, such as {“constraint type”: “prohibition”, “spatial relationship”: “intersection”, “target feature”: “ecological control boundary”}.

[0084] S13: The integration output layer performs feature alignment and cross-modal fusion on the attribute constraint features and the spatial constraint features to generate a constraint feature set.

[0085] In this embodiment, the integration output layer refers to a module for coordinating different modal data, which can be implemented by using a cross-modal attention mechanism. The semantic gap between natural language and spatial data is eliminated through feature alignment to form a unified expression of the constraint feature set. Feature alignment refers to the first operation performed by the integration output layer. Since the dimensions, distributions, and semantic levels of the feature vectors from the text and spatial channels may be different, feature alignment aims to establish a correspondence between the two feature spaces. For example, through an attention mechanism or a projection network, the model learns that the ecological protection area concept in the text features and the ecological control boundary region in the spatial features refer to each other. Cross-modal fusion is the next operation after feature alignment, which refers to the deep integration of the aligned bimodal features to produce information complementary and enhanced joint feature representation. Simple cross-modal fusion methods include concatenation or weighted averaging, while advanced cross-modal fusion methods such as tensor-based fusion or cross-modal attention mechanism can capture more complex inter-modal interactions to generate a 1+1>2 fusion effect. The feature constraint set refers to the final product of the integration output layer, which contains all the constraint information extracted and fused from multiple sources. The feature constraint set is the input for subsequent knowledge graph construction and classification operations.

[0086] Specifically, the natural language channel first performs reference relationship analysis on the provisions in the planning data, establishes traceable associations between provisions, such as identifying the supplementary explanation relationship of a provision to another provision; then generates attribute constraint features with constraint labels according to the effectiveness level determination rules, such as classifying rigid planning provisions into a high coercive level and local rules into a low coercive level; the spatial analysis channel performs basic spatial analysis on the input vector or raster data, such as generating a 500-meter buffer range for road data, and mapping the results of such geometric operations to semantic labels such as "construction prohibited area"; the integration output layer maps attribute constraint features and spatial constraint features to the same dimension through vector space projection, and generates a constraint feature set containing multi-modal information in a weighted splicing manner.

[0087] Through the above technical solutions, the application realizes automatic feature extraction of planning provisions and spatial data, reduces the degree of manual participation and improves processing efficiency; reduces rule bias caused by subjective interpretation through structured constraint feature sets; enhances semantic consistency between different data sources through cross-modal fusion, and lays a data foundation for subsequent constraint classification and knowledge graph construction.

[0088] In an embodiment, as shown in FIG. 11, step S11 includes: Figure 3

[0089] S111: The natural language channel performs reference relationship analysis on the identified natural language provisions and establishes traceable associations;

[0090] In this embodiment, reference relationship analysis refers to establishing logical associations by analyzing the mutual reference relationship between provisions, which can be implemented by using a reference pattern matching algorithm based on regular expressions to solve the problem of rule conflicts caused by isolated interpretation of provisions; traceable association refers to the link established in the knowledge graph or database that can be bidirectionally queried, which ensures that when needed, all provisions cited by a provision and all provisions citing it can be quickly located from any provision node, forming a complete traceability chain, greatly enhancing the transparency and interpretability of the knowledge system.

[0091] S112: The natural language channel identifies the effectiveness level, context, and modal words of the natural language provisions, and determines the constraint coercive level of the natural language provisions based on the identification results and the pre-set level determination rules;

[0092] ​In this embodiment, the hierarchy of effectiveness refers to the different levels of validity of planning provisions due to the varying authority of their source regulations; the context refers to the surrounding textual environment upon which the provisions are understood, including the chapter in which the provision is located, applicable preconditions, and exceptions; modal words refer to linguistic concepts, specifically those words in the provisions that express the intensity of obligation, permission, or desire, such as "must" and "prohibited" indicating mandatory obligations, "shall" and "should" indicating general requirements, "appropriate" and "may" indicating recommendations or permission, and "encourage" indicating expectations. Modal words directly reflect the degree of binding force of the constraints; the identification of the hierarchy of effectiveness can be specifically implemented using a regulation hierarchy reasoning model based on knowledge graphs to eliminate conflicts of effectiveness between provisions at different levels; modal Word recognition can be implemented using a natural language processing model based on an attention mechanism to quantify the strength of the binding force of a provision. The level determination rule refers to the calculation logic that quantifies multi-dimensional information into a comprehensive force score. It is usually a rule base or calculation model. Its input is information such as the identified effectiveness level, context, and modal words. The output is a comparable numerical value or level label, such as mandatory, requirement, recommendation, etc. The constraint force level refers to the output result of the level determination rule. Usually, the constraint force level is a quantitative or graded representation of the strength of the binding force of the provision. It will become an important metadata attribute of the attribute constraint features corresponding to the natural language provision. It is used to distinguish between rigid constraints and flexible objectives in subsequent processes (such as multi-objective optimization) and to perform differentiated processing.

[0093] S113: Natural language channel generates attribute constraint features based on natural language provisions and their constraint enforcement level.

[0094] Specifically, when natural language provisions are input into the natural language channel, the system first scans the citation markers in the provisions, such as "according to Article X" or "referring to XX standard," to automatically establish a traceable link between provisions. Then, based on factors such as the level of the provision's source and the time of promulgation, and in conjunction with a pre-set ranking table of regulatory effectiveness, the system assigns corresponding constraint priority values ​​to the provisions. Simultaneously, it extracts restrictive words from the provisions and transforms them into constraint strength coefficients through semantic analysis. Finally, it fuses the provisions with the parsed citation relationships, effectiveness levels, contextual information, and modal words to generate a structured feature vector containing multi-dimensional constraint attributes.

[0095] Through the above technical solution, this application realizes multi-dimensional feature analysis of planning clauses, and can automatically construct a constraint system with complete reference relationships and effectiveness levels, providing accurate semantic input for subsequent strong constraint screening and knowledge graph construction, and can automatically identify core constraint conditions with high-level enforcement force.

[0096] In one embodiment, step S12 includes:

[0097] S121: The spatial analysis channel performs basic spatial analysis on the spatial data, and maps the geometric relationship obtained by the basic spatial analysis into a preset spatial relationship semantic, the basic spatial analysis including overlay analysis, buffer analysis, network analysis, and visual range analysis;

[0098] In the embodiment, the basic spatial analysis refers to an operation of geometric relationship calculation on the spatial data through overlay analysis, buffer analysis, network analysis, and visual range analysis, and can be specifically implemented by using a spatial analysis tool chain of a GIS platform, and is used for extracting topological relationship and spatial distribution characteristics from original spatial data; the geometric relationship refers to a spatial relationship description directly obtained by the basic spatial analysis, for example, two polygons "intersect", a point is "inside" a polygon, a road "crosses" a region, at this time, the geometric relationship has not been given a specific domain semantic; the spatial relationship semantic refers to converting the geometric relationship calculation result into a semantic label with a domain meaning, and can be specifically implemented by using a pre-defined semantic mapping table, for example, mapping the buffer analysis result into a "adjacent protected area" semantic label, to solve the semantic gap problem between the spatial data and the rule provisions; the overlay analysis refers to a method similar to analyzing by stacking multiple transparent papers together, and specifically refers to stacking two or more spatial data layers (usually polygon layers) in geometry, and generating a new layer and new attributes according to the overlapping relationship; the buffer analysis refers to automatically establishing a certain width of a surrounding influence area, i.e., a buffer zone, around one or more point, line, and surface geographic elements; the network analysis refers to analyzing the path, flow, and resource allocation based on a network system composed of interconnected linear elements (such as roads and pipelines), and considering connectivity, impedance such as time, distance, and cost; the visual range analysis, also known as the visibility analysis, is used to determine which regions are visible and which regions are not visible from one or more observation points in a three-dimensional terrain scene.

[0099] Exemplarily, typical application scenarios of the overlay analysis include:

[0100] Land suitability evaluation: multiple restrictive layers such as "ecological control boundary", "basic farmland", and "slope" are overlaid to find out the constructible area not belonging to any restricted area;

[0101] Planning compliance analysis: the "project site selection range" is overlaid with the "land use overall planning layer" to check whether the land use property is consistent;

[0102] Typical application scenarios of the buffer analysis include:

[0103] Infrastructure protection: a 100-meter buffer zone is established on both sides of a river as an ecological buffer belt, and construction is prohibited in the area;

[0104] Service range analysis: Take a school as the center point, and establish a 1-kilometer walking buffer zone to analyze the residential area covered by the buffer zone.

[0105] Typical application scenarios of network analysis include:

[0106] Optimal path planning: Calculate the shortest or fastest driving route from point A to point B.

[0107] Facility service range: Analyze which areas a fire station can cover within a 5-minute drive.

[0108] Municipal pipeline connectivity: Analyze which areas will be affected by the closure of a water pipe valve.

[0109] Typical application scenarios of viewshed analysis include:

[0110] Landscape view corridor protection: Analyze whether a newly planned building at a certain location will destroy the important urban landscape view corridor.

[0111] Infrastructure site selection: Select locations with the best view or the largest signal coverage for viewing platforms, signal towers, wind turbines, etc.

[0112] Security deployment: Determine the best installation location for surveillance cameras to maximize their monitoring view.

[0113] S122: The spatial analysis channel matches the mapped spatial relationship semantics with the preset spatial constraint rule template and instantiates it to generate a spatial constraint condition vector.

[0114] In this embodiment, the spatial constraint rule template refers to a structured rule expression containing spatial constraint logic, which can be stored in XML or JSON format, and is used to instantiate semantic tags into computable constraint conditions. For example, a spatial constraint rule template can be: IF [geographical entity A] [spatial relationship] [geographical entity B] THEN [constraint action]. Matching and instantiation refer to the process of binding spatial relationship semantics with spatial constraint rule templates. Matching is to find the appropriate rule template corresponding to the current semantic relationship, for example, identifying that the "invasion" relationship should match the "prohibit development" type of rule template. Instantiation is to fill in specific entities (such as "current plot" and "core ecological patch") and relationships (such as "invasion") into the slots of the template to generate a complete and executable rule, for example, instantiated as: IF current plot invades core ecological patch THEN prohibit development. The spatial constraint condition vector refers to the conversion of the instantiated spatial constraint rule into a numerical feature representation that is easier for machines to process. Each spatial constraint rule can be represented as a point or a vector in a multi-dimensional vector space, and different dimensions of the spatial constraint condition vector can represent different constraint types, strengths, or parameters.

[0115] S123: The spatial analysis channel normalizes and encodes the spatial constraint condition vector, and outputs a spatial constraint feature.

[0116] In this embodiment, normalization and encoding refers to converting the spatial constraint condition vector into a unified format feature representation, which can be implemented by binary encoding or vector normalization method, to ensure the comparability and computability of spatial constraint conditions from different sources.

[0117] Specifically, the spatial analysis channel first performs overlay analysis on the input vector data, raster data and network data to identify spatial intersection areas, performs buffer analysis to define the influence range of a certain distance, calculates traffic accessibility indicators through network analysis, and evaluates visual sensitive areas through visual range analysis; then, the analysis results are matched with a preset semantic mapping table, for example, the overlapping areas identified by overlay analysis are mapped to the "prohibited construction area" semantic, and the buffer analysis results are mapped to the "ecological sensitive area" semantic; then, the corresponding spatial constraint rule template is called according to the project type, for example, the rule template of "distance from residential area not less than 500 meters" is matched for industrial projects, the semantic label is combined with the rule parameter to generate a vectorized constraint condition; finally, the numerical parameters and Boolean conditions in the vector are converted into a unified dimensional feature vector through normalization and encoding, eliminating the dimensional difference and realizing the integration processing of multi-source constraint conditions.

[0118] Through the above technical solutions, the application realizes the automatic extraction and structured expression of spatial constraint features, avoids subjective errors caused by manual interpretation of spatial data and planning provisions, and improves the processing efficiency of spatial constraint conditions; through the matching mechanism of semantic mapping and rule template, the corresponding constraint conditions of complex spatial relationships can be accurately identified, and the processing capacity of the system for fuzzy spatial clauses is enhanced; the normalization and encoding method ensures the compatibility of spatial constraint features generated from different data sources, and provides a standardized input data format for subsequent multi-objective optimization.

[0119] In an embodiment, step S20 comprises:

[0120] S21: Match the constraint feature set with the pre-constructed classification rule library through the pre-defined constraint classification rule, so as to divide the constraint feature set into strong constraint features and weak constraint features;

[0121] In the embodiment, the constraint classification rule refers to an automatic matching rule established based on the specified effectiveness level, article compulsory level and spatial constraint non-negotiability, which can be specifically implemented by using a clause effectiveness evaluation model based on a rule engine, and is used to realize objective classification of constraint features without manual intervention; the pre-constructed classification rule library refers to a set or database for storing and managing all pre-defined constraint classification rules, and the classification rule library is an extensible knowledge component, allowing quick and consistent classification of a large number of input constraint features automatically by querying and matching the rule library.

[0122] S22: Instantiating the strong constraint features and the weak constraint features into feature nodes in the site selection knowledge graph, and creating core concept nodes including a project demand concept node, a spatial data source concept node and a calculation logic concept node;

[0123] In the embodiment, the strong constraint feature refers to a restrictive condition with high compulsory effectiveness or spatial non-adjustability, which can be specifically implemented by analyzing the prohibitive clauses and spatial rigid boundaries in planning articles, and is used to ensure compliance of the site selection scheme; the weak constraint feature refers to a guiding condition with negotiation allowance, which can be specifically implemented by identifying the suggestive expressions and flexible indicators in planning articles, and is used to support multi-objective optimization decision; instantiation refers to a process of creating a specific individual with actual attributes from an abstract concept or class, which in the embodiment is a process of converting abstract constraint feature data into specific feature node individuals in the knowledge graph; the site selection knowledge graph refers to a knowledge representation system using a graph structure to express the association relationship between constraint conditions and planning elements, which can be specifically implemented by using a Neo4j graph database, and is used to realize structured storage and semantic query of constraint conditions; the feature node refers to an entity representing a specific constraint condition in the site selection knowledge graph, and the feature node is a specific embodiment of the constraint feature in the graph, for example, a node named “volume rate upper limit 2.0” is a feature node, and the node attributes of the feature node can include detailed parameters, source and compulsory level of the constraint; the core concept node refers to an entity node representing a basic category of planning elements, which can be specifically defined by using an OWL ontology modeling method, and is used to establish a domain framework of the knowledge graph; wherein, the project demand concept node refers to an abstract template or category for defining and representing the “project demand” entity in the knowledge graph; the spatial data source concept node refers to an abstract template or category for defining and representing the source and identity of the “spatial data” in the knowledge graph; and the calculation logic concept node refers to an abstract template or category for defining and representing an algorithm or spatial analysis tool required for verifying a certain constraint condition in the knowledge graph.

[0124] S23: Establishing a semantic relationship chain between the core concept nodes and the feature nodes in the site selection knowledge graph according to the pre-defined domain ontology.

[0125] In the embodiment, the predefined domain ontology refers to a formal specification describing concepts, concept attributes and relationships between concepts in the site selection field, which is equivalent to a blueprint of a knowledge graph, and it is predefined which core concept nodes can exist and which types of semantic relationship chains are allowed to be established between them. The domain ontology ensures that the constructed knowledge graph is logically consistent and understandable; the semantic relationship chain refers to a logical association path connecting different nodes, which can be encoded using RDF triples, and is used to support reasoning and tracing of constraint conditions. The semantic relationship chain defines the specific relationship between nodes and is the core means of expressing knowledge in a knowledge graph.

[0126] Specifically, in the constraint feature classification stage, the pre-constructed classification rule library stores rule matching patterns of different levels of effectiveness, such as automatically classifying clauses containing modal words such as “must” and “prohibit” as strong constraints; during the instantiation of feature nodes, each constraint feature is converted into a graph node containing attribute labels, such as instantiating “ecological management boundary” as a strong constraint node with spatial coordinate attributes; the creation of core concept nodes uses ontology modeling methods, and three types of basic entities such as project requirements, data sources and calculation logic are predefined, such as defining “traffic accessibility analysis model” as a calculation logic concept node; the establishment of semantic relationship chains is achieved by analyzing the logical dependencies between planning elements, such as establishing a “restricted by” relationship between the “industrial land” project requirement node and the “soil pollution risk” constraint node.

[0127] Through the above technical solutions, the application realizes the automatic classification and structured storage of planning constraint conditions, solves the processing bottleneck problem caused by low efficiency of manual interpretation; the construction of the knowledge graph makes the scattered constraint conditions form an organic whole, supports quick retrieval of associated constraints through graph traversal, and overcomes the performance limitations of traditional relational databases in complex association queries; the explicit expression of semantic relationship chains enhances the explainability of constraint conditions and provides a clear reasoning path for subsequent optimization algorithms.

[0128] In an embodiment, step S30 comprises:

[0129] S31: creating a project instance node in the site selection knowledge graph based on the target area information and the attribute requirement information, and generating a query condition of the site selection knowledge graph, the query condition comprising a spatial position parameter and a project attribute label;

[0130] In the embodiment, the project instance node refers to an entity node in the site selection knowledge graph for representing a specific project requirement, which can be dynamically generated by using an entity creation interface in the knowledge graph, and is realized by binding the geographic range coordinates of the target area and the keywords of the attribute requirements to the node attributes; the query condition refers to the retrieval keywords or filtering parameters extracted from the attributes of the project instance node for searching the related constraints in the knowledge graph; wherein the spatial location parameter in the query condition can be a geographic coordinate range or an administrative division code, which defines the geographic range of the search and ensures that only the spatial constraints related to the area are retrieved; the project attribute label can be a standardized classification code of the project type, construction scale or functional positioning, which defines the business context of the search and ensures that only the rules applicable to the project are retrieved; the spatial location parameter and the project attribute label can be specifically extracted from the requirement text input by the user by using the natural language processing technology and mapped to the preset label system.

[0131] S32: Taking the project instance node as the query starting point, traversing along the semantic relationship chain in the site selection knowledge graph, and retrieving all feature nodes associated with the spatial location parameter and the project attribute label;

[0132] In the embodiment, the query starting point refers to the initial node specified in the graph traversal operation, and all searches and reasoning will start from this query starting point and expand outward; the traversal of the semantic relationship chain refers to the graph traversal query based on the ontology relationship path defined in advance in the knowledge graph, which can be specifically implemented by using the traversal algorithm of the graph database, such as breadth-first search combined with relationship type filtering.

[0133] S33: The strong constraint features and weak constraint features corresponding to the retrieved feature nodes are summarized to form a strong constraint set and a weak constraint set.

[0134] Specifically, when the user end submits the project requirement, an automatically created project instance node with a unique identifier is created in the site selection knowledge graph, and the spatial boundary coordinates of the target area and the parsed attribute requirement label are attached to the node; the spatial location parameter in the query condition is converted into polygon data in the standard geographic reference system, and the attribute label can be mapped to the preset industry standard classification system; in the knowledge graph traversal process, the project instance node is taken as the starting point, and the search is expanded along multiple semantic paths such as land type association, spatial adjacency relationship, specified reference relationship, etc., and the feature nodes directly or indirectly associated with the current spatial range and attribute label are dynamically identified; all visited feature nodes automatically trigger the constraint feature extraction mechanism, and the associated strong constraint conditions and weak constraint conditions are collected into the corresponding sets to form a complete constraint condition system.

[0135] By the technical solution, the application realizes intelligent matching of project requirements and planning constraint conditions, significantly improves the efficiency of pre-analysis of land site selection; meanwhile, the knowledge graph-based associated retrieval can find cross-domain constraint clauses that are easily ignored by humans, such as composite constraint conditions triggered in environmental protection regulations and traffic planning, to ensure compliance of recommended land; and finally, the strong constraint set and the weak constraint set provide accurate input conditions for subsequent multi-objective optimization, reducing the deviation of optimization results caused by rule omission in traditional methods.

[0136] In an embodiment, the site selection knowledge graph includes a spatial data source node, and step S40 includes:

[0137] S41: Analyzing the spatial constraint conditions in the strong constraint set and generating a corresponding spatial analysis instruction sequence;

[0138] In this embodiment, the spatial constraint condition refers to a constraint condition in the strong constraint set that needs to be verified by spatial analysis technology, and the spatial constraint condition usually describes a spatial relationship that a geographic entity must satisfy or avoid, for example, "the plot boundary keeps a distance of not less than 50 meters from a river" is a typical spatial constraint condition; the spatial analysis instruction sequence refers to a set of executable spatial analysis operation steps converted from the spatial constraint condition, which can be implemented by using a rule engine to analyze the constraint condition and generate an instruction sequence including operation steps such as overlay analysis and buffer analysis, to guide the subsequent spatial data processing flow.

[0139] S42: Identifying corresponding related spatial data source nodes based on the spatial analysis instruction sequence, and scheduling original spatial data of the identified spatial data source nodes to generate a regular data set after corresponding data processing;

[0140] In this embodiment, the spatial data source node refers to a node in the site selection knowledge graph that stores spatial data entities and metadata, which can be implemented by identifying the data source node related to the spatial analysis instruction through a preset node attribute tag in the knowledge graph, to dynamically schedule multi-source heterogeneous spatial data; the original spatial data refers to multi-source heterogeneous spatial data that has not been preprocessed and is scheduled from the data source, which may have different formats, different coordinate systems, redundancies or errors, and is usually difficult to be directly used for accurate spatial calculation; the corresponding data processing refers to a series of standardization and cleaning operations on the original spatial data, aiming to eliminate data differences and ensure that it can be used for subsequent analysis, and the core processing includes: coordinate system unification, format conversion and data cleaning, etc.; the regular data set refers to a spatial data set processed by coordinate system conversion and format standardization, which can be implemented by using a spatial data conversion tool to perform projection transformation and format unification on the original data, to ensure the compatibility of data from different sources in spatial overlay analysis.

[0141] S43: Based on the spatial analysis instruction sequence and the regularized data set, the target area information is subjected to spatial overlay analysis, and the spatial region intersection satisfying all spatial constraint conditions is output as the compliance land information set.

[0142] In this embodiment, the spatial region intersection refers to the common region satisfying two or more spatial conditions (layers), and logically, the spatial region intersection is equivalent to the "and" in Boolean operation. The final compliance land must be the region satisfying all spatial constraint conditions, i.e., the spatial intersection of the layers corresponding to these constraint conditions.

[0143] Specifically, the spatial constraint conditions in the strong constraint set are parsed into spatial analysis instruction sequences containing overlay order and parameters, such as converting "avoid ecological protection zone within 500 meters" into buffer analysis instruction; the spatial data source nodes in the site knowledge graph are automatically identified according to the instruction requirements, such as identifying the node storing the ecological protection zone boundary data; the original spatial data is converted and formatted to form a unified data set, such as converting the vector data in different coordinate systems into WGS84 coordinate system; the spatial overlay analysis performs multi-level spatial operation on the regularized data according to the spatial analysis instruction sequence, such as first performing buffer analysis and then performing overlay intersection operation, and finally outputting the spatial region set satisfying all strong constraint conditions.

[0144] Through the above technical solutions, the application realizes the automatic analysis and efficient spatial analysis of strong constraint conditions, provides accurate compliance land input data for subsequent optimization algorithms, and reduces the intensity of manual intervention and the risk of operation errors.

[0145] In an embodiment, step S50 includes:

[0146] S51: Identify and extract the descriptive related ambiguity clauses in the weak constraint set;

[0147] In this embodiment, the ambiguity clause refers to the descriptive constraint condition that cannot be directly quantified in the planning clause or requirement, which can be realized by using the semantic pattern recognition method in natural language processing technology, such as identifying the clause containing "coordination", "suitable", "reasonable" and other ambiguous expressions through keyword matching and context dependency analysis.

[0148] S52: For the identified ambiguity clause, search for a number of historical cases in the case library of the site knowledge graph;

[0149] In this embodiment, the case library refers to a relational database or graph database storing historical project decision data, which can be realized by using the spatiotemporal index and semantic tag joint retrieval technology to quickly locate the historical cases with similar constraint characteristics as the current clause.

[0150] S53: aggregate and analyze the processing mode and result of the historical cases, and generate an auxiliary decision suggestion containing a quantitative recommendation value, a qualitative processing mode, and a potential risk prompt.

[0151] In this embodiment, the aggregate analysis refers to the structured integration of multi-source heterogeneous case data, which can be implemented by using a multi-dimensional data fusion algorithm and a decision tree induction method. The common processing mode and the differentiated result in the cases are extracted to form a reference basis for decision-making. The quantitative recommendation value refers to a suggestion expressed in a specific numerical value or numerical range obtained through aggregate analysis. The quantitative recommendation value quantifies the fuzzy requirements into an operable index. The qualitative processing mode refers to a suggestion described in a strategic or methodological form obtained through aggregate analysis. It provides a solution idea and best practice, rather than a specific number. The potential risk prompt refers to the problems or challenges that the current scheme may face, which is based on the failure or controversial situation in the historical cases and is forward-looking. It reflects the early warning capability of the scheme.

[0152] Specifically, in the process of generating an auxiliary decision suggestion, first, the text information in the weak constraint set is scanned to identify the fuzzy clauses containing non-quantitative descriptions. Then, similarity retrieval is performed in the case library of the site knowledge graph based on the semantic features of the clauses to filter out historical case data sets with the same or similar constraint conditions. Further, multi-dimensional analysis is performed on the case data set, such as statistical quantization of the execution threshold range of similar clauses in different geographical environments, induction of common processing strategies of qualitative constraints, and extraction of typical risk events caused by clause interpretation bias. Finally, the analysis results are converted into a composite decision suggestion containing numerical interval suggestions, operation guidelines, and risk warnings.

[0153] Through the above technical solution, the present application can convert the fuzzy constraints that are difficult to directly quantify into operable decision suggestions, effectively solving the problems of low decision-making efficiency and subjective bias caused by relying on artificial experience in the prior art. Through the aggregate analysis of case data, a composite suggestion that takes into account quantitative reference and qualitative guidance can be automatically generated, providing multi-dimensional decision support basis for planning personnel users.

[0154] In an embodiment, step S60 includes:

[0155] S61: convert each scheme solution in the site recommendation solution set into a decision vector, and generate a decision matrix based on the converted decision vector;

[0156] In this embodiment, the decision vector refers to converting each site selection scheme solution into a structured data representation containing multi-dimensional evaluation indicators, which can be implemented using a normalized numerical vector, and the vector dimension can include quantitative indicators such as land area, development cost, and traffic accessibility; the decision matrix refers to a mathematical matrix formed by arranging the decision vectors of all alternative schemes in rows, and each row of the decision matrix represents a scheme, and each column represents an evaluation criterion.

[0157] S62: Semantically associate each item of metadata in the decision matrix with the corresponding data source node in the site knowledge graph to generate a semantic enhanced matrix;

[0158] In this embodiment, the metadata refers to extended metadata containing data provenance information, specifically the attached description information of each data value in the decision matrix, which is used to explain the background information such as the source, calculation method, calculation time, and confidence of the data value; semantic association refers to establishing a traceable link between the metadata of the data value in the decision matrix and the related data source node in the site knowledge graph, for example, associating the data "traffic score = 0.72" with the "road network data layer" and "network analysis algorithm" nodes used to calculate it; the semantic enhanced matrix refers to the enhanced matrix obtained by establishing traceable links between the metadata in the decision matrix and the original data source through knowledge graph node association, which can be implemented using node embedding technology of a graph database, for example, binding each metadata item with the corresponding spatial data source node in the knowledge graph through a unique identifier.

[0159] S63: Based on the semantic enhanced matrix, the compliance land information set, and the auxiliary decision suggestion, generate a site selection decision report.

[0160] In this embodiment, the compliance land information set refers to spatial region data that meets all strong constraint conditions filtered through spatial overlay analysis, which can be obtained using a spatial query interface of a geographic information system, for example, performing a polygon intersection operation through PostGIS to obtain a region set that meets the planning constraint boundary; the site selection decision report refers to the final output result, which is not a simple text summary, but a comprehensive decision support information file generated based on structured data such as the semantic enhanced matrix and rich in interaction and traceability capabilities.

[0161] Specifically, each solution in the site recommendation solution set is first parsed into a decision vector containing key parameters, for example, a certain solution can be converted into a vector containing a land area of 5000 square meters, a development cost of 3000 million yuan, and a distance of 800 meters from a nearby subway station; then, the vectors of all solution sets are organized into a decision matrix, with rows representing different solutions and columns representing evaluation indexes; when generating a semantic enhanced matrix, each data item in the decision matrix is associated with the corresponding node in the knowledge graph, for example, the development cost data is associated with the economic attribute node, and the land area is associated with the spatial data source node; finally, by integrating the spatial coordinate data of the compliant land, the risk prompts in the auxiliary decision suggestions, and the association relationship in the semantic enhanced matrix, a site decision report containing solution priority ranking, data traceability link and expert suggestions is generated.

[0162] In some embodiments, the decision vector can include a dynamic weight allocation mechanism, for example, adjusting the weight coefficients of different indexes according to user needs; the semantic association process can use an automated labeling tool, for example, using natural language processing technology to identify the mapping relationship between metadata and knowledge graph nodes; the presentation form of the site decision report can support interactive visualization, for example, through a WebGIS platform to realize the linkage display of the spatial distribution heat map of the solution and the attribute index.

[0163] Through the above technical solutions, the present application solves the problems of disconnection between data and rules and opaque decision logic in the existing site report generation process, and realizes the dual enhancement of multi-dimensional quantitative evaluation and data traceability of solution sets; the structured expression of decision vectors supports quantitative comparison between different solutions, and the semantic association mechanism ensures the logical consistency of the report conclusion and the original constraint conditions, which helps users quickly understand the compliance basis and optimization direction of each solution.

[0164] It should be understood that the size of the serial number of each step in the above embodiments does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0165] In an embodiment, a land site selection recommendation system integrating constraint conditions and optimization algorithms is provided, which corresponds one-to-one to the land site selection recommendation method integrating constraint conditions and optimization algorithms in the above embodiment. The land site selection recommendation system integrating constraint conditions and optimization algorithms comprises:

[0166] The feature extraction module is configured to obtain planning data information and spatial data information, and input the planning data information and the spatial data information into a pre-trained multi-channel feature extraction model to obtain a constraint feature set, the constraint feature set comprising spatial constraint features and attribute constraint features.

[0167] a graph construction module, configured to classify the constraint feature set into strong constraint features and weak constraint features by a predefined constraint classification rule, and construct a site selection knowledge graph;

[0168] a constraint screening module, configured to acquire project demand information from a user end, extract target region information and attribute demand information, and screen a strong constraint set and a weak constraint set corresponding to the target region information and the attribute demand information based on the site selection knowledge graph;

[0169] an information generation module, configured to perform spatial overlay analysis on the target region information based on the strong constraint set to obtain a compliance land use information set, extract fuzzy clauses in the weak constraint set, and generate an auxiliary decision suggestion based on the site selection knowledge graph;

[0170] an optimization solving module, configured to solve the compliance land use information set and the weak constraint set by using a preset multi-objective optimization algorithm to generate a site selection recommendation solution set;

[0171] a decision integration module, configured to integrate the site selection recommendation solution set and the auxiliary decision suggestion into a site selection decision report and output the site selection decision report to the user end.

[0172] Specific limitations of the land use site selection recommendation system integrating constraint conditions and optimization algorithms can be seen in the limitations of the land use site selection recommendation method integrating constraint conditions and optimization algorithms in the foregoing, and will not be repeated here. Each module in the land use site selection recommendation system integrating constraint conditions and optimization algorithms can be realized by software, hardware, and combinations thereof, in whole or in part. Each module can be embedded in or independent of a processor in a computer device in hardware form, or can be stored in a memory in a computer device in software form, so as to be called and executed by a processor to perform operations corresponding to each module.

[0173] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the foregoing embodiments of the present application have been described in detail, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A land use recommendation method that integrates constraints and optimization algorithms, characterized in that, Including the following steps: The planning data and spatial data are acquired and input into a pre-trained multi-channel feature extraction model to obtain a constraint feature set, which includes spatial constraint features and attribute constraint features. The constraint feature set is divided into strong constraint features and weak constraint features by predefined constraint classification rules, and a location knowledge graph is constructed. Obtain project requirement information from the user end, and extract target area information and attribute requirement information. Based on the site selection knowledge graph, filter the strong constraint set and weak constraint set corresponding to the target area information and attribute requirement information. Spatial overlay analysis of target area information is performed based on strong constraint set to obtain compliant land use information set, fuzzy clauses in weak constraint set are extracted, and auxiliary decision-making suggestions are generated based on site selection knowledge graph; A pre-defined multi-objective optimization algorithm is used to solve the set of compliant land use information and the set of weak constraints to generate a set of recommended site selection solutions; The site selection recommendation solution set and auxiliary decision-making suggestions are integrated into a site selection decision report and output to the user terminal.

2. The land use recommendation method integrating constraints and optimization algorithms according to claim 1, characterized in that: The multi-channel feature extraction model includes a natural language channel, a spatial analysis channel, and an integrated output layer. The step of acquiring planning data information and spatial data information and inputting them into the pre-trained multi-channel feature extraction model to obtain a constraint feature set, which includes spatial constraint features and attribute constraint features, includes the following steps: The natural language channel identifies natural language clauses in planning data and performs semantic parsing to extract attribute constraint features; The spatial analysis channel performs spatial relationship reasoning on spatial data information and extracts spatial constraint features; The integrated output layer performs feature alignment and cross-modal fusion of attribute constraint features and spatial constraint features to generate a constraint feature set.

3. The land use recommendation method integrating constraints and optimization algorithms according to claim 2, characterized in that: The steps of identifying natural language clauses in planning data and performing semantic parsing to extract attribute constraint features include the following steps: The natural language channel parses the citation relationships of the identified natural language texts and establishes traceable associations; The natural language channel identifies the effectiveness level, context, and modal words of natural language provisions, and determines the binding force level of the natural language provisions based on the identification results and pre-set level determination rules. The natural language channel generates attribute constraint features based on natural language provisions and their level of constraint enforcement.

4. The land use recommendation method integrating constraints and optimization algorithms according to claim 2, characterized in that: The spatial analysis channel performs spatial relationship reasoning on spatial data information and extracts spatial constraint features, including the following steps: The spatial analysis channel performs basic spatial analysis on spatial data and maps the geometric relationships obtained from the basic spatial analysis to preset spatial relationship semantics. The basic spatial analysis includes overlay analysis, buffer analysis, network analysis, and view area analysis. The spatial analysis channel matches and instantiates the mapped spatial relationship semantics with the preset spatial constraint rule template to generate a spatial constraint condition vector; The spatial analysis channel standardizes and encodes the spatial constraint vector, outputting spatial constraint features.

5. The land use recommendation method integrating constraints and optimization algorithms according to claim 1, characterized in that: The step of classifying the constraint feature set into strong constraint features and weak constraint features according to predefined constraint classification rules, and constructing a location selection knowledge graph, includes the following steps: By matching the constraint feature set with a pre-built classification rule library through predefined constraint classification rules, the constraint feature set is divided into strong constraint features and weak constraint features. Strong and weak constraint features are instantiated as feature nodes in the location knowledge graph, and core concept nodes are created, including project requirement concept nodes, spatial data source concept nodes, and computational logic concept nodes. Based on the predefined domain ontology, semantic relationship chains are established between core concept nodes and feature nodes in the location knowledge graph.

6. The land use recommendation method integrating constraints and optimization algorithms according to claim 5, characterized in that: The steps of obtaining project requirement information from the user terminal, extracting target area information and attribute requirement information, and filtering the strong constraint set and weak constraint set corresponding to the target area information and attribute requirement information based on the site selection knowledge graph include the following steps: Based on target area information and attribute requirement information, project instance nodes are created in the site selection knowledge graph, and query conditions for the site selection knowledge graph are generated. The query conditions include spatial location parameters and project attribute tags. Starting from the project instance node, the system traverses along the semantic relationship chain in the location knowledge graph to retrieve all feature nodes associated with spatial location parameters and project attribute tags. The strong and weak constraint features corresponding to the retrieved feature nodes are summarized to form strong constraint sets and weak constraint sets.

7. The land use recommendation method integrating constraints and optimization algorithms according to claim 1, characterized in that: The site selection knowledge graph includes spatial data source nodes. The steps of performing spatial overlay analysis on target area information based on a strong constraint set to obtain a compliant land use information set, extracting fuzzy clauses from the weak constraint set, and generating auxiliary decision-making suggestions based on the site selection knowledge graph include the following steps: Parse the spatial constraints in the set of strong constraints and generate the corresponding spatial analysis instruction sequence; Based on the spatial analysis instruction sequence, the corresponding related spatial data source nodes are identified, and the original spatial data of the identified spatial data source nodes is scheduled and processed to generate a regularized data set. Based on the spatial analysis command sequence and the regularized data set, spatial overlay analysis is performed on the target area information, and the intersection of spatial areas that meet all spatial constraints is output as a set of compliant land use information.

8. The land use recommendation method integrating constraints and optimization algorithms according to claim 1, characterized in that: The steps of performing spatial overlay analysis on target area information based on a strongly constrained set to obtain a compliant land use information set, extracting fuzzy clauses from the weakly constrained set, and generating auxiliary decision-making suggestions based on a site selection knowledge graph include the following steps: Identify and extract descriptively relevant fuzzy clauses from the set of weak constraints; For the identified ambiguous clauses, several historical cases are retrieved from the case library of the location knowledge graph; By aggregating and analyzing the processing methods and results of historical cases, auxiliary decision-making suggestions are generated, including quantitative recommendation values, qualitative processing methods, and potential risk warnings.

9. The land use recommendation method integrating constraints and optimization algorithms according to claim 1, characterized in that: The site selection knowledge graph includes spatial data source nodes. The step of integrating the site selection recommendation solution set and auxiliary decision-making suggestions into a site selection decision report and outputting it to the user terminal includes the following steps: Each solution in the location recommendation solution set is transformed into a decision vector, and a decision matrix is ​​generated based on the transformed decision vectors; Semantically associate each metadata item in the decision matrix with the corresponding data source node in the location knowledge graph to generate a semantically enhanced matrix. A site selection decision report is generated based on a semantic enhancement matrix, a set of compliant land use information, and auxiliary decision-making suggestions.

10. A land use recommendation system that integrates constraints and optimization algorithms, characterized in that, include: The feature extraction module is used to acquire planning data information and spatial data information, and input them into a pre-trained multi-channel feature extraction model to obtain a constraint feature set, which includes spatial constraint features and attribute constraint features. The graph construction module is used to divide the constraint feature set into strong constraint features and weak constraint features according to predefined constraint classification rules, and to construct a location knowledge graph. The constraint filtering module is used to obtain project requirement information from the user terminal, extract target area information and attribute requirement information, and filter strong constraint sets and weak constraint sets corresponding to the target area information and attribute requirement information based on the site selection knowledge graph. The information generation module is used to perform spatial overlay analysis on the target area information based on the strong constraint set to obtain the compliant land use information set, extract the fuzzy clauses in the weak constraint set, and generate auxiliary decision-making suggestions based on the site selection knowledge graph. The optimization and solution module is used to solve the set of compliant land use information and the set of weak constraints using a preset multi-objective optimization algorithm to generate a set of recommended site selection solutions; The decision integration module is used to integrate the site selection recommendation solution set and auxiliary decision-making suggestions into a site selection decision report and output it to the user terminal.

Images