A land space layout adjustment decision-making method based on planning conflict identification

Abstract

The application discloses a land space layout adjustment decision-making method based on planning conflict identification, and belongs to the technical field of land space planning and management. The method comprises the following steps: constructing a time-space conflict feature vector library containing multi-dimensional attributes; generating a semantic knowledge graph containing causal relationships, time sequence relationships, spatial relationships and subordinate relationships; realizing accurate prediction of the influence intensity and duration of the conflict state over time; realizing trend analysis and risk level evaluation of the cross-scale conflict propagation path; automatically identifying the explicit conflicts in the current layout and the potential conflicts based on the propagation prediction, and intelligently classifying the composite conflicts by using fuzzy clustering; automatically generating a differentiated conflict intervention strategy recommendation scheme in stages; solving a land space layout adjustment scheme; and judging whether the layout adjustment scheme meets the preset conflict relief standard.

Description

Technical Field

[0001] This application relates to the field of land spatial planning and management technology, and more specifically, to a decision-making method for adjusting land spatial layout based on planning conflict identification. Background Technology

[0002] Currently, some conflicts in national spatial planning are handled through manual identification and experience-based judgment, which suffers from problems such as limited data dimensions, limited identification accuracy, and delayed analysis results, making it difficult to meet the needs of dynamic monitoring and multi-scale management. Furthermore, once conflicts occur, they often exhibit cross-regional diffusion and delayed manifestation. Failure to identify and predict them in a timely manner can lead to a chain reaction of impacts on subsequent spatial layout adjustments, even exacerbating the uncertainty of spatial governance. In addition, existing layout adjustment methods are mostly based on static optimization, failing to systematically consider the temporal evolution and spatial hierarchical transmission mechanisms of conflict, resulting in some adjustment schemes exhibiting insufficient adaptability and limited effectiveness of intervention measures in actual implementation. Therefore, there is an urgent need for a layout adjustment support tool that can integrate multi-source data, deeply analyze the causes of conflicts, and possess predictive and intervention capabilities to enhance the foresight of planning and the proactiveness of governance.

[0003] In conclusion, how to accurately identify and analyze territorial spatial conflicts in a dynamic and complex spatial environment, and formulate feasible and adaptable layout adjustment decision-making schemes, has become an urgent technical problem to be solved. Summary of the Invention

[0004] To overcome a series of shortcomings in existing technologies, the purpose of this application is to provide a decision-making method for adjusting the spatial layout of national territory based on planning conflict identification, which is applied to a national territory planning decision-making system. The national territory planning decision-making system includes a multi-source data fusion module, a spatiotemporal conflict analysis module, a conflict propagation prediction module, a national territory conflict knowledge graph construction module, a scheduling decision-making module, and a layout optimization module, and includes the following steps.

[0005] Step 1: The multi-source data fusion module performs spatiotemporal registration and semantic unification of multi-source heterogeneous spatial data, and automatically extracts the spatial distribution characteristics, temporal evolution patterns and semantic attribute features of land use conflicts, thereby constructing a spatiotemporal conflict feature vector library containing multi-dimensional attributes.

[0006] Step 2: Using the territorial conflict knowledge graph construction module, based on the fused multi-source data, automatically extract conflict event entities, conflict subject entities, spatial location entities, and attribute feature entities to construct a four-element semantic network, identify the complete causal chain of conflict causes, and generate a semantic knowledge graph containing causal relationships, temporal relationships, spatial relationships, and subordinate relationships.

[0007] Step 3: Based on the pre-trained spatiotemporal lag effect model and semantic enhancement of the land conflict knowledge graph, the spatiotemporal conflict analysis module is used to automatically capture and quantify the spatiotemporal propagation and lag effects of land use conflicts between different regions. Combined with the causal chain reasoning results in the knowledge graph, the intensity and duration of the impact of the conflict state over time can be accurately predicted.

[0008] Step 4: The conflict propagation prediction module intelligently identifies the optimal propagation path of land use conflicts at different spatial scales and quantifies the obstacles of various influencing factors to conflict propagation, thereby realizing trend analysis and risk level assessment of cross-scale conflict propagation paths.

[0009] Step 5: Based on the preset spatiotemporal lag threshold and influence intensity threshold, automatically identify explicit conflicts and potential conflicts based on propagation prediction in the current layout, and use fuzzy clustering to intelligently classify complex conflicts.

[0010] Step 6: Through the scheduling decision module, based on the identified explicit and potential conflicts, comprehensively assess the urgency of conflict mitigation, the rationality of intervention measures, and the existing resource allocation capacity, automatically generate phased differentiated conflict intervention strategy recommendations, including strategy combinations and their priority ranking for the emergency response period, short-term governance period, medium-term adjustment period, and long-term consolidation period.

[0011] Step 7: Construct a multi-objective optimization model with the goals of minimizing conflict, maximizing benefits, and optimizing costs through the layout optimization module. Integrate rigid rule constraints and spatiotemporal lag factors to solve for the land spatial layout adjustment scheme.

[0012] Furthermore, the multi-source data fusion module employs a hierarchical registration strategy during spatiotemporal registration: First, spatial data from different coordinate systems are uniformly converted to the WGS84 geographic coordinate system to achieve coordinate benchmark consistency; then, common land feature features in the spatial data are identified through feature point matching algorithms, and affine transformation matrices are used to perform geometric correction on the data; finally, timestamp alignment methods are used to perform temporal standardization on data from different time phases to ensure consistency in the time dimension; in the semantic unification process, a unified land use classification coding system is constructed, semantic conversion between different data sources is achieved through ontology mapping technology, and a confidence-weighted fusion method is combined to handle semantic conflicts, ensuring that the fused data achieves spatial accuracy down to the meter level while maintaining semantic consistency.

[0013] Furthermore, the multi-source data fusion module includes a data preprocessing unit, a coordinate transformation unit, and a semantic mapping unit. Specifically: the data preprocessing unit is responsible for quality inspection and noise filtering of remote sensing image data, GIS vector data, statistical survey data, and field measurement data from different sensor platforms; the coordinate transformation unit, based on a high-precision geodetic datum and projection parameters, completes accurate coordinate transformation between multiple spatial reference systems and uses a seven-parameter Bursa model for unified processing of three-dimensional spatial coordinates; the semantic mapping unit constructs an ontology knowledge base based on industry standards, defines a standardized land use classification system and attribute description specifications, and achieves semantic alignment between heterogeneous data sources through semantic similarity calculation and concept hierarchy matching, ensuring consistency in the semantic representation of the same geographic entity across different data sources.

[0014] Furthermore, the pre-trained spatiotemporal lag effect model employs a deep recurrent neural network architecture, integrating an encoder-decoder structure and an attention mechanism module. Specifically, the encoder uses a bidirectional long short-term memory network to capture forward and backward dependencies in the time series; the decoder uses a multi-head self-attention mechanism to dynamically learn the influence weights between different time steps. During model training, supervised learning samples are constructed based on historical conflict data, and input and target sequence pairs are generated using a sliding time window. The loss function is set as a weighted combination of mean squared error and time decay weights, and parameters are iteratively updated using the Adam optimization algorithm. In the model validation phase, time series cross-validation is used to evaluate prediction performance metrics including mean absolute error, root mean square error, and correlation coefficient. Simultaneously, sensitivity analysis is used to determine the optimal lag window length and influence decay parameters, ensuring good prediction stability across multiple spatiotemporal scales.

[0015] Furthermore, the spatiotemporal conflict analysis module includes a conflict identification unit, an intensity calculation unit, and a propagation analysis unit. Specifically: the conflict identification unit, based on the land use compatibility matrix and spatial proximity criteria, identifies combinations of land use types with potential conflicts and marks potentially conflicting areas by setting spatial buffer thresholds and time interval parameters; the intensity calculation unit quantifies conflict intensity using a multi-factor weighted evaluation method, comprehensively considering factors such as land use type differences, spatial overlap area, population density, economic value, and ecological sensitivity, and uses the analytic hierarchy process (AHP) to determine the weights of each factor and establish a conflict intensity evaluation index system; the propagation analysis unit constructs a conflict propagation network based on graph theory and network analysis methods, where nodes represent spatial units and edges represent conflict propagation paths, and analyzes the spatial patterns and impact range of conflict propagation by calculating node centrality, clustering coefficients, and path lengths.

[0016] Furthermore, the conflict propagation prediction module simulates the propagation process of conflict in a spatial network based on a multi-agent modeling method. Each agent corresponds to a spatial unit and has state transition rules and interaction behavior patterns. The agent states include four types: no conflict, potential conflict, explicit conflict, and conflict resolution. When the conflict intensity of neighboring agents exceeds a preset threshold, a state transition is triggered. The propagation probability is jointly determined by the distance decay function, terrain barrier factor, and policy intervention intensity. Multiple random experiments are conducted using the Monte Carlo simulation method to statistically analyze the conflict propagation path, arrival time, and impact range under different scenarios, generating probability distribution maps and risk level classifications. Sensitivity analysis is used to identify the key parameters that have the greatest impact on the propagation process.

[0017] Furthermore, the scheduling decision-making module includes an urgency assessment unit, a feasibility analysis unit, a resource allocation unit, a strategy generation unit, a planning scenario classification unit, an intelligent indicator adjustment unit, and a real-time dynamic adjustment triggering unit. Among them: the urgency assessment unit constructs an urgency evaluation indicator system based on conflict propagation prediction results and causal analysis of knowledge graphs; the feasibility analysis unit is used to assess the implementation conditions of intervention measures; the resource allocation unit allocates the optimal resource combination for intervention strategies based on the constraints of currently available resources; the strategy generation unit takes the urgency assessment, feasibility analysis, and resource allocation results as inputs for decision-making conditions and automatically generates a phased intervention strategy combination; the planning scenario classification unit divides decision scenarios into four types: economic development orientation, ecological protection priority, social equity consideration, and comprehensive coordination.

[0018] Furthermore, step 3 includes the following steps.

[0019] The quality of historical land use conflict data is checked, and the data of each region is mapped to a unified grid cell based on a spatial weight matrix. Then, the sequence is aligned according to the timestamp to generate a standardized spatiotemporal dataset.

[0020] The spatiotemporal conflict analysis module is used to calculate the conflict intensity index of each grid unit. Combining the neighborhood location characteristics and semantic information in the national conflict knowledge graph, a multi-dimensional temporal feature vector sequence that integrates causal relationships is constructed based on a sliding time window.

[0021] The multidimensional temporal feature vector sequence is input into the encoder part of the pre-trained spatiotemporal lag effect model. A bidirectional long short-term memory network is used to capture the forward and backward associations in the sequence. Combined with the semantic enhancement information of the knowledge graph, the hidden state vector containing lag information and causal chain reasoning is output.

[0022] The decoder uses an attention mechanism to weight and summarize the hidden state vectors, integrates the causal chain reasoning results in the knowledge graph, and maps the weight distribution at different time steps to the prediction parameters of the conflict impact intensity and duration at the current moment, thereby achieving accurate prediction of the conflict state over time.

[0023] Based on the lag coefficient and spatial propagation probability in the decoder output, combined with semantic enhancement analysis of the knowledge graph, the prediction results are mapped back to the geographic grid, automatically capturing and quantifying the spatiotemporal propagation paths and corresponding time delays of land use conflicts between regions, thereby generating a spatiotemporal propagation trajectory map that includes propagation and lag effects.

[0024] Error analysis is performed between the predicted conflict intensity and duration and the measured data. If the prediction error exceeds the set threshold, the lag window length or spatial weight parameters in the model are dynamically adjusted based on the sensitivity analysis results and knowledge graph feedback, and iterative optimization is performed until the prediction accuracy meets the set standard.

[0025] Furthermore, step 4 includes the following steps.

[0026] At the three spatial levels of region, county, and plot, conflict propagation networks are constructed based on grid units or administrative boundary units, respectively. Nodes represent corresponding spatial units, and edge weights are determined by the conflict propagation probability and distance decay function between adjacent nodes to ensure accurate mapping of network structures at different scales.

[0027] A multi-agent model is introduced into each network layer to simulate the conflict propagation process. Agents migrate between nodes according to preset state transition rules. At the same time, the shortest path algorithm is used to automatically extract the optimal propagation path of the conflict from the simulation results.

[0028] The resistance value corresponding to the comprehensive barrier factor is calculated for each edge on the path. After normalization, the path cost weight is assigned to achieve accurate quantification of multiple influencing factors.

[0029] The optimal propagation paths and their resistance costs identified in each level of the network are summarized. By comparing path length, cumulative resistance value, and propagation time delay, cross-scale propagation trend characteristics are extracted, providing a quantitative basis for the trend differences between macro and micro levels.

[0030] Based on the total path resistance and propagation probability, and combined with the threshold quantile method, a risk score is applied to each propagation path, dividing it into three levels of risk: low, medium, and high, thereby identifying potential key conflict diffusion corridors and high-risk sections.

[0031] The optimal propagation path, resistance distribution, and risk level are displayed in the form of map layers, and trend analysis and risk assessment reports are automatically generated, including cross-scale path comparison charts, resistance factor contribution bar charts, and risk area distribution heat maps.

[0032] Furthermore, the decision-making method for adjusting the spatial layout of the national territory also includes the following steps.

[0033] Step 8: During the generation of layout adjustment schemes, the impact of the schemes on future conflict evolution is evaluated in real time through the spatiotemporal fitness evaluation function. The layout adjustment schemes are judged to meet the preset conflict mitigation standards by combining the conflict propagation prediction results and the spatiotemporal lag effect analysis results.

[0034] Step 9: If the layout adjustment scheme does not meet the preset conflict mitigation criteria, return to the multi-objective optimization model to solve again until the conflict mitigation criteria are met.

[0035] Step 10: If the layout adjustment scheme has met the preset conflict mitigation criteria, then generate the Pareto optimal solution set that takes into account both current conflict mitigation and future risk prevention, and output the final land space layout adjustment decision scheme.

[0036] Furthermore, step 8 includes the following steps.

[0037] The spatial structure parameters, land use type distribution, and adjustment range of the current layout adjustment plan are input into the spatiotemporal fitness evaluation function. Combined with the future conflict situation characteristics output by the conflict propagation prediction module, the key impact indicators of the layout plan on the conflict evolution process are extracted.

[0038] The evaluation indicators of the layout scheme in terms of conflict mitigation effect, future development adaptability and implementation rationality are calculated respectively, and a comprehensive evaluation vector is constructed accordingly.

[0039] Based on a predetermined expert weighting system, each indicator is weighted, and the overall fitness score of the scheme is generated by weighted summation.

[0040] The obtained fitness score is compared with the set conflict mitigation criteria to quantitatively determine whether the current layout scheme meets the minimum threshold requirements for conflict intensity reduction, conflict area reduction, and propagation path blocking indicators.

[0041] Based on the analysis results of the spatiotemporal lag effect, the time delay and persistence changes in the conflict propagation process after the layout adjustment are examined to further verify whether the control effect of the scheme on the conflict evolution trend at different time points in the future meets expectations.

[0042] Compared with existing technologies, this application has the following beneficial effects: By constructing a knowledge graph of land conflict that integrates multi-source heterogeneous data, and combining a pre-trained spatiotemporal lag effect model with multi-agent propagation simulation, this application realizes the causal chain analysis, cross-scale propagation path prediction and intelligent layout optimization of land use conflict, thereby providing refined and evolvable decision support for conflict intervention and adjustment in land spatial planning. Attached Figure Description

[0043] Figure 1 This is a flowchart illustrating a land spatial layout adjustment decision-making method based on planning conflict identification, as disclosed in an embodiment of this application. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of the embodiments of this invention will be described in more detail below with reference to the accompanying drawings. In the drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The described embodiments are some embodiments of this invention, but not all embodiments.

[0045] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] The embodiments and directional terms described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0047] like Figure 1 As shown, a decision-making method for adjusting the spatial layout of land based on planning conflict identification is applied to a land spatial planning decision-making system. The land spatial planning decision-making system includes a multi-source data fusion module, a spatiotemporal conflict analysis module, a conflict propagation prediction module, a land conflict knowledge graph construction module, a scheduling decision-making module, and a layout optimization module. The method includes the following steps.

[0048] Step 1: The multi-source data fusion module performs spatiotemporal registration and semantic unification of multi-source heterogeneous spatial data, and automatically extracts the spatial distribution characteristics, temporal evolution patterns and semantic attribute features of land use conflicts, thereby constructing a spatiotemporal conflict feature vector library containing multi-dimensional attributes.

[0049] Step 2: Using the territorial conflict knowledge graph construction module, based on the fused multi-source data, automatically extract conflict event entities, conflict subject entities, spatial location entities, and attribute feature entities to construct a four-element semantic network, identify the complete causal chain of conflict causes, and generate a semantic knowledge graph containing causal relationships, temporal relationships, spatial relationships, and subordinate relationships.

[0050] Step 3: Based on the pre-trained spatiotemporal lag effect model and semantic enhancement of the land conflict knowledge graph, the spatiotemporal conflict analysis module is used to automatically capture and quantify the spatiotemporal propagation and lag effects of land use conflicts between different regions. Combined with the causal chain reasoning results in the knowledge graph, the intensity and duration of the impact of the conflict state over time can be accurately predicted.

[0051] Step 4: The conflict propagation prediction module intelligently identifies the optimal propagation path of land use conflicts at different spatial scales and quantifies the obstacles of various influencing factors to conflict propagation, thereby realizing trend analysis and risk level assessment of cross-scale conflict propagation paths.

[0052] Step 5: Based on the preset spatiotemporal lag threshold and influence intensity threshold, automatically identify explicit conflicts and potential conflicts based on propagation prediction in the current layout, and use fuzzy clustering to intelligently classify complex conflicts.

[0053] Step 6: Through the scheduling decision module, based on the identified explicit and potential conflicts, comprehensively assess the urgency of conflict mitigation, the rationality of intervention measures, and the existing resource allocation capacity, automatically generate phased differentiated conflict intervention strategy recommendations, including strategy combinations and their priority ranking for the emergency response period, short-term governance period, medium-term adjustment period, and long-term consolidation period.

[0054] Step 7: Construct a multi-objective optimization model with the goals of minimizing conflict, maximizing benefits, and optimizing costs through the layout optimization module. Integrate rigid rule constraints and spatiotemporal lag factors to solve for the land spatial layout adjustment scheme.

[0055] Step 8: During the generation of layout adjustment schemes, the impact of the schemes on future conflict evolution is evaluated in real time through the spatiotemporal fitness evaluation function. The layout adjustment schemes are judged to meet the preset conflict mitigation standards by combining the conflict propagation prediction results and the spatiotemporal lag effect analysis results.

[0056] Step 9: If the layout adjustment scheme does not meet the preset conflict mitigation criteria, return to the multi-objective optimization model to solve again until the conflict mitigation criteria are met.

[0057] Step 10: If the layout adjustment scheme has met the preset conflict mitigation criteria, then generate the Pareto optimal solution set that takes into account both current conflict mitigation and future risk prevention, and output the final land space layout adjustment decision scheme.

[0058] This method, by introducing a multi-module collaborative mechanism, solves the decision-making challenges arising from overlapping conflicts, spatiotemporal mismatches, and resource scarcity in land use planning, significantly improving the scientific rigor and foresight of planning adjustments. Specifically, firstly, a multi-source data fusion module is used to perform spatiotemporal unification and semantic standardization processing on heterogeneous spatial data, effectively extracting the spatial distribution and evolution characteristics of land use conflicts and constructing a high-dimensional feature library to enhance the accuracy and comprehensiveness of conflict identification. Secondly, by constructing a land conflict knowledge graph, conflict subjects, events, spaces, and attribute entities are organically linked to form a multi-dimensional causal relationship network, enabling structured extraction and visual representation of conflict causes. Based on this, a spatiotemporal lag effect model, combined with a semantically enhanced knowledge graph, is used to quantitatively analyze conflict propagation paths and intensity, achieving accurate prediction of dynamic conflict evolution trends and providing a scientific basis for subsequent intervention measures. Furthermore, different types of conflicts are intelligently classified to improve the efficiency of analyzing and responding to complex conflicts. Subsequently, based on the conflict level and urgency, multi-stage intervention strategies are matched, balancing emergency response and long-term governance to form a targeted strategy combination, enhancing the flexibility and adaptability of spatial governance. At the layout optimization level, a multi-objective optimization model is constructed, comprehensively considering benefits, costs, and rule constraints to deduce the layout adjustment effects under different intervention strategies. The fitness function is used to dynamically evaluate the inhibitory effect of the scheme on future conflict evolution, achieving iterative optimization of the planning scheme. Finally, under the condition of meeting the conflict mitigation threshold, a Pareto optimal solution set that considers both current and future risk prevention is output, providing a comprehensive decision-making scheme with scientific support and operability for actual national spatial layout. Overall, this method achieves a complete closed loop from conflict identification and propagation prediction to scheduling decision-making and optimal execution, overcoming the problems of traditional national spatial adjustment methods being static, isolated, and dependent on experience.

[0059] Furthermore, the multi-source data fusion module employs a hierarchical registration strategy during spatiotemporal registration: First, spatial data from different coordinate systems are uniformly converted to the WGS84 geographic coordinate system to achieve coordinate benchmark consistency; then, common land feature features in the spatial data are identified through feature point matching algorithms, and affine transformation matrices are used to perform geometric correction on the data; finally, timestamp alignment methods are used to perform temporal standardization on data from different time phases to ensure consistency in the time dimension; in the semantic unification process, a unified land use classification coding system is constructed, semantic conversion between different data sources is achieved through ontology mapping technology, and a confidence-weighted fusion method is combined to handle semantic conflicts, ensuring that the fused data achieves spatial accuracy down to the meter level while maintaining semantic consistency.

[0060] In this embodiment, a hierarchical registration strategy is employed to achieve high-precision fusion of heterogeneous spatial data. Based on a unified coordinate system, time reference, and semantic encoding, it effectively addresses challenges such as inconsistent coordinate references, large timestamp differences, and semantic classification conflicts among multi-source data. Affine transformation matrix and feature point matching techniques improve spatial alignment accuracy. Combined with semantic ontology mapping and a confidence weighting mechanism, the accuracy and robustness of semantic fusion are enhanced. Ultimately, comprehensive unification of the fused data across time, space, and semantic dimensions is achieved, providing a high-quality data foundation for subsequent conflict analysis and decision optimization.

[0061] Furthermore, the multi-source data fusion module includes a data preprocessing unit, a coordinate transformation unit, and a semantic mapping unit. Specifically: the data preprocessing unit is responsible for quality inspection and noise filtering of remote sensing image data, GIS vector data, statistical survey data, and field measurement data from different sensor platforms; the coordinate transformation unit, based on a high-precision geodetic datum and projection parameters, completes accurate coordinate transformation between multiple spatial reference systems and uses a seven-parameter Bursa model for unified processing of three-dimensional spatial coordinates; the semantic mapping unit constructs an ontology knowledge base based on industry standards, defines a standardized land use classification system and attribute description specifications, and achieves semantic alignment between heterogeneous data sources through semantic similarity calculation and concept hierarchy matching, ensuring consistency in the semantic representation of the same geographic entity across different data sources.

[0062] This embodiment effectively improves the coordination and consistency between different data sources through a multi-source data fusion module. The preprocessing unit ensures the reliability and usability of the raw data and reduces noise interference. The coordinate transformation unit uses a seven-parameter model to unify spatial references and enhance 3D positioning accuracy. The semantic mapping unit establishes an ontology knowledge base to achieve normalization of semantics for various data types, resolving the discrepancies in the semantic representation of geographic entities and providing unified, high-quality data support for subsequent conflict analysis and spatial reasoning.

[0063] Furthermore, the knowledge graph construction module for land conflict includes an entity extraction unit, a relationship identification unit, a knowledge fusion unit, and a causal reasoning unit. Specifically: the entity extraction unit automatically identifies four types of core entities from the fused multi-source data. Conflict event entities include attributes such as conflict type, occurrence time, duration, and scope of impact; conflict subject entities include stakeholders and their roles; spatial location entities include geographic identifiers such as administrative division codes, land parcel codes, latitude and longitude coordinates, and spatial levels; and attribute feature entities include quantitative attributes such as economic indicators, environmental indicators, social indicators, and policy indicators. The relationship identification unit automatically identifies four types of semantic relationships between entities: causal relationships are represented by semantic tags such as "cause," "influence," "promote," and "hinder"; temporal relationships are represented by time tags such as "prior to," "simultaneous," "after," and "continuous"; spatial relationships are represented by spatial tags such as "adjacent," "containing," "overlapping," and "separated"; and subordinate relationships are represented by... The knowledge graph is represented by organizational labels such as "affiliation," "jurisdiction," "responsibility," and "participation," and each relationship is assigned a confidence weight in the range of 0-1. The knowledge fusion unit identifies the same entity from different data sources by calculating the semantic and structural similarity between entities, merges entities using clustering algorithms and expert rules, and ensures the integrity and consistency of the knowledge graph through relationship consistency checks and conflict resolution mechanisms. The causal reasoning unit adopts a hybrid reasoning method, combining rule-based symbolic reasoning and statistical probabilistic reasoning. Symbolic reasoning constructs a causal rule base based on domain expert knowledge, stating that "if condition A and condition B, then the probability of result C is P." It discovers potential causal relationships through forward chain reasoning and backward chain reasoning. Probabilistic reasoning uses a Bayesian network model to model causal relationships as conditional probability distributions, calculates the joint probability of complex causal chains through the assumption of conditional independence between variables, and introduces fuzzy logic and evidence theory to handle incomplete and uncertain causal evidence.

[0064] This embodiment constructs a well-structured and semantically complete knowledge graph of land conflict, providing systematic and reasonable knowledge support for planning decisions. The entity extraction unit accurately identifies various types of information, including conflict events, subjects, locations, and attributes, comprehensively reconstructing conflict scenarios. The relationship identification unit reveals causal, temporal, spatial, and subordinate relationships between entities, enhancing the logical connectivity between data. The knowledge fusion unit effectively resolves redundancy and conflicts in cross-source data, ensuring the accuracy and consistency of the graph structure. The causal reasoning unit combines symbolic logic and probabilistic models to automatically construct and evolve causal chains under uncertain conditions, significantly improving the intelligence level of conflict cause analysis and risk prediction.

[0065] Furthermore, the construction process of the spatiotemporal conflict feature vector database includes the following steps: First, the spatial autocorrelation index, landscape fragmentation index, and boundary complexity index of land use conflict are calculated, and the periodic patterns, trend characteristics, and anomalous changes of conflict evolution are analyzed by setting different time windows; then, based on expert knowledge and statistical analysis, a subset of key features with high influence on conflict prediction is selected; next, continuous features are standardized, discrete features are encoded using one-hot encoding, and time-series features are subjected to differential transformation, finally constructing a standardized feature vector representation; simultaneously, a mapping index between feature vectors and corresponding spatial locations and timestamps is established to achieve efficient spatiotemporal feature query and management.

[0066] This embodiment constructs a standardized spatiotemporal conflict feature vector library, enabling multidimensional feature extraction and structured representation of land use conflicts. By introducing indicators such as spatial autocorrelation, fragmentation, and boundary complexity, the spatial distribution and morphological characteristics of conflicts are quantified. Combining time window analysis with evolution trends and anomalous fluctuations enhances the ability to perceive the dynamic evolution of conflicts. Feature selection and standardization ensure the validity and comparability of vector data, while differential transformation enhances the expressive power of temporal changes. Finally, the established spatiotemporal mapping index enables rapid location and management of feature data, providing efficient and accurate foundational support for conflict prediction and decision-making.

[0067] Furthermore, the pre-trained spatiotemporal lag effect model employs a deep recurrent neural network architecture, integrating an encoder-decoder structure and an attention mechanism module. Specifically, the encoder uses a bidirectional long short-term memory network to capture forward and backward dependencies in the time series; the decoder uses a multi-head self-attention mechanism to dynamically learn the influence weights between different time steps. During model training, supervised learning samples are constructed based on historical conflict data, and input and target sequence pairs are generated using a sliding time window. The loss function is set as a weighted combination of mean squared error and time decay weights, and parameters are iteratively updated using the Adam optimization algorithm. In the model validation phase, time series cross-validation is used to evaluate prediction performance metrics including mean absolute error, root mean square error, and correlation coefficient. Simultaneously, sensitivity analysis is used to determine the optimal lag window length and influence decay parameters, ensuring good prediction stability across multiple spatiotemporal scales.

[0068] This embodiment utilizes a deep recurrent neural network architecture, integrating bidirectional memory and attention mechanisms, to accurately model the lag effects of land use conflicts in their temporal evolution. The encoder effectively extracts temporal dependencies, while the decoder dynamically allocates weights for each time step, enhancing the model's ability to identify the impact of conflict delays. During training, historical data and sliding window-generated samples are combined, and a time decay mechanism is incorporated into the loss function design, better reflecting the actual evolutionary process. The validation phase employs multiple error metrics and sensitivity analysis to ensure robust predictive capabilities across different spatiotemporal scales, providing reliable support for subsequent propagation analysis and decision-making interventions.

[0069] Furthermore, the spatiotemporal conflict analysis module includes a conflict identification unit, an intensity calculation unit, and a propagation analysis unit. Specifically: the conflict identification unit, based on the land use compatibility matrix and spatial proximity criteria, identifies combinations of land use types with potential conflicts and marks potentially conflicting areas by setting spatial buffer thresholds and time interval parameters; the intensity calculation unit quantifies conflict intensity using a multi-factor weighted evaluation method, comprehensively considering factors such as land use type differences, spatial overlap area, population density, economic value, and ecological sensitivity, and uses the analytic hierarchy process (AHP) to determine the weights of each factor and establish a conflict intensity evaluation index system; the propagation analysis unit constructs a conflict propagation network based on graph theory and network analysis methods, where nodes represent spatial units and edges represent conflict propagation paths, and analyzes the spatial patterns and impact range of conflict propagation by calculating node centrality, clustering coefficients, and path lengths.

[0070] This embodiment achieves accurate identification, quantitative assessment, and trend analysis of land use conflicts by constructing a systematic conflict analysis process. The conflict identification unit utilizes a compatibility matrix and spatial proximity principles, combined with buffer zones and time parameters, to effectively screen potential conflict areas. The intensity calculation unit quantifies the severity of conflicts using a multi-factor weighted method, taking into account spatial, social, economic, and ecological dimensions to improve the objectivity and applicability of the evaluation results. The propagation analysis unit introduces graph theory methods to construct a propagation network, revealing the spatial diffusion paths and clustering characteristics of conflicts, providing a clear basis for subsequent risk assessment and intervention strategy formulation.

[0071] Furthermore, the intensity calculation unit also includes an ecological value change rate calculation subunit and a socio-cultural value assessment subunit. The ecological value change rate calculation subunit comprises four assessment functional blocks: supply services, regulation services, cultural services, and support services. Specifically, the supply services functional block calculates the value gains and losses of food production, freshwater supply, forest products, and genetic resources, employing a pricing method combining the substitution cost approach and the market price approach, and establishing a service value coefficient table covering different ecosystem types such as farmland, forests, wetlands, and grasslands. The regulation services functional block assesses value changes in climate regulation, hydrological regulation, soil conservation, and pollution remediation, using the shadow price approach and the cost-benefit approach to calculate carbon sequestration, runoff regulation, and soil remediation. The value of soil erosion control is assessed by introducing a spatiotemporal differential adjustment coefficient to consider the marginal value differences of adjustment services in different locations; the cultural service functional blocks are evaluated using the conditional value method and the selection experiment method to assess the value of landscape aesthetics, cultural heritage, and leisure and entertainment, and a cultural service value assessment model based on the public's willingness to pay is established, constructing a cultural service importance classification system at the national (weight 1.0), provincial (0.8), municipal (0.6), and county (0.4) levels; the value of support service functional blocks is assessed by evaluating the value of basic ecological processes such as biodiversity maintenance, soil formation, and nutrient cycling, and the indirect economic value of support services is calculated using the production function method and the protection cost method, and an ecosystem health evaluation model is established to quantify the stability changes of support services. The socio-cultural value assessment subunit comprises three assessment functional blocks: impact on social equity, cultural landscape value, and changes in social cohesion. The social equity impact functional block establishes a Gini coefficient change rate index to assess the impact of layout adjustments on regional development equity, calculates changes in the distribution of benefits among different income groups, introduces a spatial justice index to assess the equity changes in accessibility of public service facilities, and sets a weighting coefficient for vulnerable groups, giving higher weight to the impact on low-income groups and the elderly. The cultural landscape value functional block constructs a landscape value evaluation system based on historical and cultural elements, using the analytic hierarchy process (AHP) to determine the weights of historical buildings, traditional villages, cultural routes, and intangible cultural heritage carriers. The social cohesion change functional block assesses the impact of changes in community network structure on social capital and calculates a community stability index.

[0072] This embodiment quantifies the impact of land space adjustments on regional comprehensive functions through a dual value assessment approach encompassing both ecological and socio-cultural aspects. On one hand, it constructs a value calculation system based on multiple assessment methods and ecosystem types, focusing on four categories of ecosystem services: supply, regulation, culture, and support, thereby improving the quantitative accuracy of ecological value changes. On the other hand, it introduces indicators such as social equity, cultural landscape, and social cohesion to comprehensively assess the potential impact of planning schemes on cultural heritage and social stability. Overall, it achieves a systematic quantification of multi-dimensional values, contributing to optimized spatial layout and enhanced ecological security and social harmony.

[0073] Furthermore, the spatiotemporal conflict analysis module and the land conflict knowledge graph construction module establish a semantic enhancement mechanism. Through the causal relationships and semantic attributes provided by the knowledge graph, the conflict identification results are semantically labeled and the causes are explained. In the conflict identification process, entity matching and relational reasoning of the knowledge graph are introduced to improve the accuracy of conflict type judgment. In the intensity calculation process, the causal weights and influence paths of the knowledge graph are integrated to enhance the scientific nature of the conflict impact assessment. In the propagation analysis process, the spatial and temporal relationships of the knowledge graph are combined to optimize the construction of the propagation network and path analysis, realizing the transformation from simple data statistical analysis to semantic understanding and knowledge reasoning.

[0074] This embodiment enhances the explanatory power and intelligence of conflict identification and assessment by linking the spatiotemporal conflict analysis module with a knowledge graph. In the identification stage, causal relationships and semantic tags from the graph are introduced to assign semantic meaning to conflict areas, achieving more accurate type judgment and automated causal tracing. In intensity calculation, the causal chains and influence weights in the graph are combined to improve the logical consistency and scientific basis of conflict intensity assessment. In propagation analysis, spatial and temporal relationships in the graph are used to assist in network structure construction, making the conflict propagation path more closely aligned with actual evolutionary characteristics. This achieves a shift from data-driven to knowledge-driven analysis, enhancing intelligent reasoning capabilities and analytical depth.

[0075] Furthermore, the conflict propagation prediction module simulates the propagation process of conflict in a spatial network based on a multi-agent modeling method. Each agent corresponds to a spatial unit and has state transition rules and interaction behavior patterns. The agent states include four types: no conflict, potential conflict, explicit conflict, and conflict resolution. When the conflict intensity of neighboring agents exceeds a preset threshold, a state transition is triggered. The propagation probability is jointly determined by the distance decay function, terrain barrier factor, and policy intervention intensity. Multiple random experiments are conducted using the Monte Carlo simulation method to statistically analyze the conflict propagation path, arrival time, and impact range under different scenarios, generating probability distribution maps and risk level classifications. Sensitivity analysis is used to identify the key parameters that have the greatest impact on the propagation process.

[0076] This embodiment utilizes a multi-agent modeling approach to simulate the spatial propagation mechanism of conflict, showcasing the dynamic characteristics of conflict evolution. Each agent represents a geographical unit, possessing a clear state definition and response rules, and can transition states based on conflict pressure in neighboring areas, reflecting the chain reaction of conflict escalation in reality. The propagation probability is jointly regulated by factors such as spatial distance, terrain conditions, and policy intervention, enhancing the model's fit to real-world scenarios. Through Monte Carlo simulation and multi-scenario statistics, the uncertainty of conflict propagation is quantified, and a risk level map is generated, providing data support for risk prediction and intervention strategies.

[0077] Furthermore, the scheduling decision-making module adopts a multi-criteria decision analysis framework, including an urgency assessment unit, a feasibility analysis unit, a resource allocation unit, and a strategy generation unit. Specifically, the urgency assessment unit, based on conflict propagation prediction results and causal analysis using a knowledge graph, constructs an urgency evaluation index system encompassing four dimensions: the speed of conflict impact expansion, the scale of potential economic losses, the level of social stability risk, and the degree of ecological environment degradation. The analytic hierarchy process (AHP) is used to determine the weights of each index, and a comprehensive urgency score is calculated using a fuzzy comprehensive evaluation method, categorized into four levels according to score range: extremely urgent (≥0.8), urgent (0.6-0.8), moderate (0.4-0.6), and delayed (<0.4). The feasibility analysis unit assesses the implementation conditions of intervention measures from four perspectives: technical feasibility, economic feasibility, policy feasibility, and social feasibility. Technical feasibility considers factors such as the maturity of existing technology, the availability of professional personnel, and equipment support. Economic feasibility assesses factors such as investment costs, operating costs, expected returns, and the difficulty of fundraising. Policy feasibility analyzes the completeness of laws and regulations, the convenience of administrative approvals, and policy... Factors such as support levels, public acceptance of social feasibility assessments, difficulty in coordinating with stakeholders, and scope of social impact are considered in the comprehensive evaluation results to generate a feasibility level. The resource allocation unit, based on current available resource constraints, including the total amount and allocation ratio of fiscal funds, the number and professional structure of technical personnel, the type and usage status of technical equipment, and the implementation time window and schedule, employs a resource allocation optimization algorithm combining linear programming and integer programming to allocate optimal resource combinations for intervention strategies with different urgency and feasibility levels, ensuring maximum intervention effectiveness under resource constraints. The strategy generation unit, based on decision tree algorithms and expert system methods, uses urgency assessment, feasibility analysis, and resource allocation results as inputs for decision-making. Through a rule-matching mechanism of "if urgency is X, feasibility is Y, and resource conditions are Z, then the recommended strategy is W," it automatically generates phased intervention strategy combinations. The strategy content includes specific intervention measures, implementation schedules, division of responsibilities among stakeholders, expected effect indicators, and risk control measures. A heuristic search algorithm is used to optimize the overall effect of the strategy combinations.

[0078] This embodiment achieves the scientific formulation of conflict intervention strategies and the rational allocation of resources through multi-criteria analysis. The urgency assessment unit comprehensively considers four indicators: conflict spread rate, economic loss, social risk, and environmental degradation, quantifying the urgency of conflict resolution and ensuring that high-risk issues are addressed first. Feasibility analysis comprehensively evaluates the implementation conditions of intervention measures from technical, economic, policy, and social perspectives, ensuring the practical operability of the plan. The resource allocation unit combines financial resources, personnel structure, equipment status, and time schedule, using optimization algorithms to achieve optimal allocation of limited resources and improve intervention efficiency. The strategy generation unit relies on decision trees and expert systems to automatically formulate phased intervention plans through rule matching, covering measures, timelines, division of responsibilities, and risk control. Heuristic search is used to optimize the overall effect, ensuring that the strategy is both scientifically sound and practically feasible.

[0079] Furthermore, the scheduling decision-making module also includes a planning scenario classification unit, an intelligent indicator adjustment unit, and a real-time dynamic adjustment trigger unit. The planning scenario classification unit divides decision-making scenarios into four types: economic development orientation, ecological protection priority, social equity consideration, and comprehensive coordination. The indicator weights for the economic development orientation scenario are configured as follows: economic benefits (0.5) > social benefits (0.3) > ecological benefits (0.2); the weights for the ecological protection priority scenario are configured as follows: ecological benefits (0.5) > social benefits (0.3) > economic benefits (0.2); the weights for the social equity consideration scenario are configured as follows: social benefits (0.4) > ecological benefits (0.35) > economic benefits (0.25); and the weights for the comprehensive coordination scenario are configured as follows: economic benefits (0.35) = ecological benefits (0.35) = social benefits. Benefit (0.3); The intelligent indicator adjustment unit adopts a dual mechanism of weight optimization based on machine learning and weight adjustment based on feedback learning. Machine learning weight optimization takes historical conflict data, planning implementation effect and expert evaluation results as input, extracts key influencing factors and effect indicators through feature engineering, automatically updates indicator weights based on prediction accuracy and outputs a scenario-based dynamic weight matrix; Feedback learning weight adjustment adjusts weights according to the difference between actual implementation effect and expected goal, and uses particle swarm optimization algorithm to find the optimal weight combination; When there is a significant conflict between different goals, the weight coordination procedure is started, the TOPSIS method is used to evaluate the comprehensive effect of different weight combinations, Pareto front analysis is introduced to find the optimal balance point among multiple goals, and a consistency verification mechanism for weight adjustment is established to ensure logical rationality. The system establishes a three-tiered mechanism of threshold triggering, periodic evaluation, and learning evolution, which dynamically adjusts trigger units in real time. The threshold triggering mechanism is configured to automatically adjust weights when a key indicator deviates from the expected value by more than 20%, initiate policy-adaptive weight adjustments when new major policies are released, and activate an emergency weight adjustment mode when unexpected environmental or economic events occur. The periodic evaluation mechanism is configured to conduct an applicability assessment of the indicator system quarterly, a weight effectiveness check every six months, and a comprehensive optimization and upgrade of the indicator system annually. The learning evolution mechanism is configured to accumulate best practice cases and lessons learned in different scenarios, forming an adaptive learning intelligent evaluation mechanism to continuously improve evaluation accuracy and practicality.

[0080] This embodiment achieves scenario-adaptive and dynamic weight optimization in planning and scheduling decisions. By configuring indicator weights according to different scenarios and combining machine learning and feedback algorithms to dynamically adjust decision preferences, it can flexibly switch between different development goals and automatically trigger adjustment mechanisms based on real-time data changes. This triple triggering mechanism ensures the timeliness and adaptability of planning responses, supporting emergency response and policy adaptation. Overall, it enhances the intelligence, flexibility, and practicality of land and space scheduling strategies, and improves the scientific nature of decision-making under multi-objective collaboration.

[0081] Furthermore, the trend analysis of cross-scale conflict propagation paths employs a hierarchical analysis method, dividing the study area into three spatial levels: regional, county, and plot. Corresponding grid resolutions and analysis units are configured for each level. At the regional level, macroscopic analysis identifies inter-regional conflict propagation trends and major corridors, with a grid size of 5 km × 5 km. At the county level, mesoscopic analysis identifies local conflict hotspots and propagation nodes, with a grid size of 1 km × 1 km. At the plot level, detailed analysis identifies specific conflict locations and their propagation paths, with a grid precision refined to 100 m × 100 m. Information transmission and feedback mechanisms ensure vertical coordination between scales, guaranteeing the consistency and accuracy of conflict propagation path analysis across multiple spatial scales.

[0082] This embodiment achieves the assessment of cross-scale conflict propagation paths through hierarchical analysis. Appropriate grid resolutions are used at different spatial levels, progressively refining the analysis from regional (5 km × 5 km) to county (1 km × 1 km), and then to plot (100 m × 100 m), ensuring both macro-level trends and micro-level details are considered. Each level coordinates with the others through information sharing and feedback mechanisms, guaranteeing the continuity and accuracy of propagation path identification. This comprehensively reveals the diffusion patterns of conflicts at different spatial scales, providing a reliable basis for the scientific formulation of hierarchical intervention strategies.

[0083] Furthermore, the preset spatiotemporal lag thresholds are determined based on the statistical analysis results of historical conflict data: by calculating the spatiotemporal lag distribution characteristics of historical conflict events, the thresholds are set using the percentile method, where the time lag threshold is taken as the 75th percentile of historical data and the spatial lag threshold is taken as the 80th percentile of historical propagation distance; the impact intensity thresholds are divided into four intensity levels—minor, moderate, severe, and extremely severe—using the quartile method according to the conflict impact severity grading standard, and the optimal cut-off point is determined through ROC curve analysis to ensure that the thresholds can effectively identify key conflicts while avoiding misjudgments due to oversensitivity.

[0084] This embodiment scientifically determines the threshold settings for spatiotemporal lag and impact intensity through systematic statistical analysis of historical conflict data. The time lag threshold uses the 75th percentile, reflecting the typical time range of most conflict propagation; the spatial lag threshold is selected from the 80th percentile, covering the propagation distance of most conflicts. The impact intensity threshold is based on the classification standard of conflict severity, combined with quartile method and ROC curve optimization, ensuring accurate classification and practical application value. This threshold setting method effectively balances sensitivity and accuracy, providing a robust basis for conflict identification and risk assessment.

[0085] Furthermore, the phased intervention strategy is dynamically planned based on a time priority matrix and resource constraints. Intervention measures are divided into four implementation phases according to their urgency and resource demand intensity: Emergency Response Phase (0-3 months): Primarily targeting conflicts rated as extremely urgent or urgent, this phase employs temporary rapid response measures, including establishing spatial control buffer zones, implementing temporary land use restrictions, activating emergency response plans, and coordinating relevant departmental actions. The key objective is to quickly control the spread of conflict and stabilize the situation. Short-Term Governance Phase (3-12 months): This phase implements direct intervention measures for identified overt conflicts, including localized land use adjustments, relocation and renovation of conflict facilities, restoration of damaged ecological environments, and stakeholder coordination and negotiation. The key objective is to eliminate current prominent conflicts and restore normal order. Medium-Term Governance Phase… The entire period (1-3 years) involves implementing systematic structural adjustment measures based on the results of layout optimization, including repositioning regional functions, optimizing and upgrading industrial structures, reconstructing infrastructure networks, and improving institutional mechanisms. The key objectives are to establish a long-term governance mechanism and enhance system stability. The long-term consolidation period (3 years or more) involves institutional construction and capacity building measures, including improving relevant laws and regulations, establishing an intelligent monitoring and early warning system, enhancing multi-departmental collaborative governance capabilities, and cultivating sustainable development models. The key objective is to achieve the organic unity of conflict prevention and sustainable development. A dynamic adjustment mechanism is established during the strategy formulation process at each stage. By monitoring changes in key indicators and evaluating the effectiveness of interventions in real time, a strategy correction procedure is automatically triggered when the actual implementation results deviate from the expected goals, ensuring the scientific nature and adaptability of the phased strategies.

[0086] This implementation plan, taking into account time priorities and resource constraints, scientifically divides the process into four phases to ensure the targeted and continuous nature of intervention measures. The emergency response phase focuses on high-urgency conflicts, rapidly implementing temporary control and emergency measures to prevent the conflict from escalating. The short-term governance phase addresses overt conflicts, implementing specific adjustments and repairs to restore regional order. The medium-term adjustment phase promotes structural optimization and institutional improvement, strengthening the governance system. The long-term consolidation phase focuses on institutional development and capacity building to achieve sustainable development. A dynamic adjustment mechanism runs throughout the entire process, flexibly revising strategies based on real-time monitoring and evaluation to ensure effective connection and continuous optimization of intervention measures at each stage.

[0087] Furthermore, the identification of explicit conflicts employs a multi-criteria decision analysis method, comprehensively considering factors such as spatial overlap, functional incompatibility, environmental impact, and socioeconomic losses. A weighted comprehensive evaluation model is used to calculate the conflict severity index. The identification of potential conflicts is based on propagation prediction results. A probability threshold method is used to determine areas where the propagation probability exceeds a set threshold as potential conflict zones. At the same time, a spatiotemporal buffer zone is set for early warning, taking into account the propagation time window and the scope of impact. The identification results include conflict location, conflict type, impact level, probability of occurrence, and expected time information, which are visualized through a geographic information system.

[0088] In this embodiment, explicit conflict identification employs a multi-criteria comprehensive evaluation, combining factors such as spatial overlap, functional compatibility, environmental impact, and socioeconomic losses to construct a weighted model that quantifies conflict severity, ensuring comprehensive and scientific judgment results. Potential conflicts, on the other hand, rely on probability thresholds for propagation prediction to identify areas with a high probability of propagation. Furthermore, the inclusion of spatiotemporal buffer zones enables early warning. The final identification results encompass the specific location, type, impact intensity, probability of occurrence, and estimated time of the conflict, presented intuitively through a GIS platform. This facilitates decision-makers' timely understanding of the conflict situation and the development of targeted measures.

[0089] Furthermore, in the design of the objective function of the multi-objective optimization model, the conflict minimization objective comprehensively considers the severity and processing cost of various conflicts through a weighted summation method; the benefit maximization objective covers three sub-objectives: economic benefits, ecological benefits, and social benefits; the cost optimization objective covers land consolidation costs, infrastructure construction costs, and ecological restoration costs; the weights of each objective function are determined by a combination of the analytic hierarchy process (AHP) and the Delphi method, with domain experts invited to evaluate the weights and consistency checks to ensure the rationality of the weight settings; the model solution adopts the classic genetic algorithm, introducing an elite retention strategy to balance the convergence and diversity of solutions, with the parameters set as follows: population size 50, crossover probability 0.8, mutation probability 0.1, and maximum number of generations 200.

[0090] This embodiment integrates three major objectives—conflict minimization, benefit maximization, and cost optimization—through weighted aggregation, comprehensively addressing the key needs in land use adjustment. Conflict minimization focuses on measuring conflict severity and response costs; benefit objectives cover multiple levels including economic, ecological, and social aspects; and cost objectives are refined to include land consolidation, infrastructure, and ecological restoration. Weight determination employs a combination of analytic hierarchy process (AHP) and Delphi method, with expert review and consistency checks to ensure a scientifically sound weight allocation. A genetic algorithm, coupled with an elite retention strategy, effectively balances search efficiency and solution diversity. Reasonable parameter configuration ensures stable and high-quality optimization results within 200 generations.

[0091] Furthermore, the multi-objective optimization model dynamically adjusts the objective function structure and constraint settings according to different planning scenarios. In the economic development-oriented scenario, the weight of the conflict minimization objective is reduced to 0.3, the weight of the benefit maximization objective is increased to 0.5, and the weight of the cost optimization objective is set to 0.2. At the same time, some ecological constraints are relaxed, allowing moderate development under the premise of strict environmental impact assessment. In the ecological protection priority scenario, the weight of the conflict minimization objective is increased to 0.5, the weight of the benefit maximization objective is reduced to 0.2, and the weight of the cost optimization objective is set to 0.3, strengthening the ecological protection red line, environmentally sensitive areas, etc. Rigid constraints are added, including a bottom-line constraint that the ecosystem service function will not be reduced. In scenarios that balance social equity, a social equity objective sub-function is added to the objective function, with the weights adjusted to conflict minimization (0.35), benefit maximization (0.3), cost optimization (0.2), and social equity (0.15). Constraints are added to include the balanced allocation of public service facilities and the protection of the basic rights of vulnerable groups. In comprehensive coordination scenarios, a dynamic balance mechanism for the three objectives is established, fine-tuning the weights of each objective within ±10% based on real-time evaluation results to ensure the coordinated development of economic, ecological, and social benefits. When the planning scenario changes, the objective function is gradually adjusted using a weight interpolation method to avoid sudden jumps in the solution space caused by abrupt changes. The interpolation formula is: W new (t)=W old ×(1-t)+W target ×t, where: W old This represents the currently used optimization weight configuration, i.e., the set of objective function weight values ​​before the scene switch occurred; W target This represents the new weight configuration target to be achieved after the planning scenario changes, i.e., the set of weight values ​​that different objective functions should have under the new planning scenario; W new (t) represents the current iteration weight vector in the transition phase, used to realize the transition from W old To W target The system ensures a smooth transition; t is the transition progress parameter, which increases linearly from 0 to 1, and the transition period is set to 5-10 optimization iteration cycles; when the optimization result deviates significantly from the target orientation of the selected scenario, the system automatically adjusts the constraints or reconfigures the target weights to ensure the consistency between the final solution and the scenario expectation; at the same time, it generates corresponding optimization solutions for different scenarios, provides decision-makers with multiple choices through comparative analysis, and establishes a scenario sensitivity analysis function to assess the impact of changes in key parameters on the optimization results under different scenarios.

[0092] This embodiment achieves adaptive optimization and adjustment of planning objectives under different scenarios. By dynamically changing the objective function structure and constraints, a balance between economic, ecological, and social goals is ensured. Weighted interpolation ensures the continuity of the scenario switching process and avoids fluctuations in the solution space. The system can automatically correct the optimization path according to the degree of deviation and generate multiple optimization schemes for comparison and selection, enhancing the flexibility, robustness, and interpretability of planning decisions, while improving the adaptability to complex scenario changes.

[0093] Furthermore, a collaborative optimization mechanism is established between the layout optimization module and the scheduling decision module. The phased intervention strategies output by the scheduling decision module are used as constraints and guiding principles for layout optimization. Strategy consistency constraints are added to the multi-objective optimization model to ensure the coordination and unity between the optimization results and the phased intervention strategies. At the same time, strategy implementation cost and time-series benefit terms are introduced into the objective function to comprehensively consider the cost-benefit ratio and time value of strategies at different stages. A hierarchical solution strategy is adopted in the optimization process. First, key constraint variables are determined according to the requirements of emergency and short-term strategies. Then, the medium- and long-term layout scheme is optimized on this basis to achieve an organic combination of short-term emergency response and long-term system optimization, thereby improving the overall scientificity and operability of the layout adjustment scheme.

[0094] This embodiment incorporates a phased intervention strategy into the constraints of the optimization model, ensuring consistency between the layout adjustment plan and the intervention measures. By introducing strategy implementation cost and time-efficiency indicators, the economic efficiency and timeliness of strategies at different stages are comprehensively evaluated. The optimization adopts a hierarchical solution, first satisfying the key constraints of emergency and short-term strategies, and then advancing the design of medium- and long-term solutions, achieving an organic integration of short-term response and long-term planning.

[0095] Furthermore, rigid rules and restrictions include legal and regulatory constraints, ecological protection constraints, infrastructure constraints, and safety protection constraints. Among them: legal and regulatory constraints include restrictions on land use and development intensity determined by statutory plans such as overall land use planning, urban master planning, and environmental protection planning; ecological protection constraints include strict protection requirements for ecological protection red lines, environmentally sensitive areas, water source protection areas, and nature reserves; infrastructure constraints include safety protection distances and functional guarantee requirements for important transportation facilities, energy facilities, water conservancy facilities, communication facilities, and other infrastructure; safety protection constraints include development and construction restrictions in areas prone to geological disasters, flood-prone areas, and hazardous chemical production and storage areas; and spatiotemporal lag factors are embedded in the optimization model through dynamic weight adjustment: when the layout scheme violates soft constraints, punitive weights are assigned to relevant terms in the objective function, and the weight adjustment range is calculated in stages according to the degree of deviation and the duration of the impact, thereby achieving a coordinated balance between the strict enforcement of hard constraints and the flexible guidance of soft constraints.

[0096] In this embodiment, rigid rules and restrictions cover four major aspects: laws and regulations, ecological protection, infrastructure, and safety protection. They clearly define land use, development intensity, and protection requirements to ensure planning compliance and environmental safety. Legal and regulatory constraints ensure land use complies with statutory plans at all levels; ecological protection constraints strictly limit development activities in ecologically sensitive areas; infrastructure constraints maintain safe distances and normal functioning of key facilities; and safety protection constraints prevent geological disasters, flood risks, and safety hazards from hazardous chemicals. Spatiotemporal lag factors are incorporated into the optimization model with dynamic weights, imposing tiered penalties on violations of soft constraints. Weight adjustments, combined with deviation magnitude and duration, achieve a flexible balance between strict adherence to rigid rules and flexible guidance.

[0097] Furthermore, the spatiotemporal fitness evaluation function is constructed based on a multi-criteria decision analysis method, covering three dimensions: conflict mitigation effect, future development adaptability, and scheme implementation rationality. Specifically: in evaluating the conflict mitigation effect, the direct effectiveness of the scheme in reducing conflict is quantified by calculating the rate of change in conflict intensity, the rate of reduction in conflict area, and the rate of blocking conflict propagation before and after the layout adjustment; in evaluating future development adaptability, multiple development scenarios such as rapid economic development, stable population growth, and priority for environmental protection are set up using scenario analysis methods, and the adaptability score and robustness index of the scheme under each scenario are evaluated to reflect the scheme's ability to cope with future uncertainties; in evaluating the scheme implementation rationality, factors such as implementation cost, technical difficulty, policy support, and public acceptance are comprehensively considered, and the feasibility is quantitatively assessed using a fuzzy comprehensive evaluation method; by weighting and summarizing the evaluation results of the three dimensions, the overall spatiotemporal fitness score is obtained, and sensitivity analysis and uncertainty analysis mechanisms are introduced to improve the robustness and explanatory power of the evaluation results.

[0098] In this embodiment, the spatiotemporal fitness evaluation function integrates three dimensions: conflict mitigation effect, future development adaptability, and the rationality of scheme implementation, enabling a comprehensive assessment of the layout adjustment scheme. Regarding conflict mitigation, the direct governance effectiveness of the scheme is quantified by analyzing changes in conflict intensity and area, as well as the effect of propagation blocking. Future adaptability is assessed through scenario analysis, examining the stability and resilience of the scheme under different economic, demographic, and environmental development backgrounds. The rationality of implementation is evaluated by combining cost, technical difficulty, policy support, and public acceptance, using fuzzy comprehensive evaluation for quantification. The results from these three parts are weighted and integrated to generate an overall fitness score, supplemented by sensitivity and uncertainty analysis to ensure the scientific validity and practical guiding value of the evaluation results.

[0099] Furthermore, the conflict mitigation criteria include quantitative and qualitative standards. Quantitative standards determine whether a solution meets basic requirements by setting minimum thresholds for conflict intensity reduction rate, conflict area reduction rate, and conflict propagation blocking rate. The conflict intensity reduction rate must be no less than 10%, the conflict area reduction rate no less than 8%, and the conflict propagation blocking rate no less than 15%. Qualitative standards, through expert evaluation and stakeholder participation, comprehensively assess the rationality and acceptability of the solution. Evaluation criteria include whether it aligns with regional development strategies, meets public needs, fully considers ecological and environmental protection, and is feasible for implementation. Fuzzy reasoning is introduced during the standard judgment process to transform quantitative indicators and qualitative evaluation results into fuzzy membership degrees, and a comprehensive judgment result is derived through a fuzzy rule base and reasoning engine. A solution is deemed acceptable when both quantitative and qualitative standards are met; otherwise, it needs to return to the optimization process for re-solution.

[0100] In this embodiment, the conflict mitigation criteria encompass both quantitative and qualitative aspects. Quantitative criteria set thresholds for conflict intensity reduction rate (≥10%), conflict area reduction rate (≥8%), and conflict propagation blocking rate (≥15%) to measure the effectiveness of the proposed solutions in actual conflict governance. Qualitative criteria, through expert review and stakeholder participation, comprehensively assess whether the solutions align with regional development strategies, meet public needs, balance ecological protection, and are feasible. The evaluation process employs fuzzy reasoning, transforming quantitative indicators and subjective evaluations into fuzzy membership degrees. A rule base and reasoning engine are used to comprehensively determine the suitability of the solutions. A solution is only approved when both quantitative and qualitative criteria are met; otherwise, it needs to be regressed and optimized to ensure the final solution is both scientifically effective and practically feasible.

[0101] Furthermore, the re-solution process of the multi-objective optimization model adopts a parameter fine-tuning strategy, appropriately adjusting parameters according to the quality of the current solution and the search state, including adjusting the population size between 40 and 60, the crossover probability between 0.7 and 0.9, and the mutation probability between 0.05 and 0.15. During the re-solution process, a multi-population parallel evolution mechanism is adopted, dividing the main population into 3-5 subpopulations for parallel search, and performing optimal individual migration between populations every 10 generations. A convergence judgment criterion is set during the solution process: convergence is determined when the objective function improvement is less than 0.001 for 20 consecutive generations.

[0102] In this embodiment, when resolving the multi-objective optimization model, a parameter fine-tuning strategy is adopted to dynamically adjust the population size (40-60), crossover probability (0.7-0.9), and mutation probability (0.05-0.15) to adapt to the search progress and improve the quality of the solution. A multi-population parallel evolution mechanism is introduced, dividing the main population into 3 to 5 subpopulations for simultaneous searching, enhancing the algorithm's diversity and global exploration capabilities. Furthermore, optimal individual migration is performed every 10 generations to promote information sharing. Convergence is determined based on the improvement in the objective function; if the improvement is less than 0.001 for 20 consecutive generations, the algorithm is considered converged, ensuring an efficient and stable solution process.

[0103] Furthermore, the strategy for balancing current conflict mitigation and future risk prevention adopts a time-weighted approach, assigning different weights to the impact of conflicts at different times through an exponentially decaying time decay function, with recent conflicts receiving higher weights and long-term conflicts receiving lower weights. The risk prevention strategy is based on scenario analysis, setting three categories of future development scenarios: routine scenarios, challenge scenarios, and opportunity scenarios, and formulating corresponding countermeasures for each scenario. The dynamic adjustment mechanism of the strategy relies on feedback information during implementation for real-time optimization, and by monitoring changes in key node indicators, the strategy adjustment procedure is triggered when the indicators deviate from the expected range. The strategy effectiveness evaluation adopts a cost-benefit analysis method, comprehensively comparing input costs and expected benefits, while also considering social and environmental benefits to ensure the economic feasibility and sustainability of the strategy.

[0104] This embodiment employs a time-weighted approach, utilizing an exponential decay function to assign differentiated weights to the impact of conflicts at different time stages, prioritizing the handling of near-term conflicts while also considering the prevention of long-term risks. Risk control incorporates scenario analysis, setting three future scenarios: routine, challenging, and opportunistic, and developing targeted countermeasures to enhance the flexibility and relevance of the strategy. A dynamic adjustment mechanism optimizes the strategy in real-time based on implementation feedback, monitoring changes in key indicators and promptly correcting deviations to ensure the continued effectiveness of intervention measures. Strategy evaluation utilizes cost-benefit analysis, comprehensively considering economic input, social impact, and environmental benefits to ensure the economic rationality and long-term sustainable development of the plan.

[0105] Furthermore, the final decision-making plan for adjusting the national land spatial layout includes four parts: a spatial layout adjustment map, an implementation schedule, policy recommendations, and an effect evaluation report. The spatial layout adjustment map is drawn using standard cartographic specifications and includes a current layout map, a planned layout map, an adjustment comparison map, and detailed maps of key areas. The map scale is set according to different levels of needs: 1:25,000 to 1:50,000 for city and county levels, 1:5,000 to 1:10,000 for township levels, and 1:1,000 to 1:2,000 for key areas. The implementation schedule is arranged according to urgency, implementation difficulty, and... The funding needs are considered to formulate a phased implementation plan, including short-term, medium-term, and long-term projects. Each project clearly defines the implementing entity, implementation time, funding source, and expected results. The policy recommendations are based on the policy needs of the plan's implementation, including suggestions on improving laws and regulations, funding guarantees, technical support conditions, and organizational management systems. The effectiveness evaluation report uses a combination of quantitative and qualitative evaluation methods, sets up an evaluation indicator system and evaluation methods, and establishes a tracking and monitoring mechanism for implementation effectiveness, providing a scientific basis for the dynamic adjustment and continuous improvement of the plan.

[0106] In this embodiment, the final decision-making scheme for adjusting the national land spatial layout consists of four parts: a spatial layout adjustment map, an implementation schedule, policy recommendations, and an effectiveness evaluation report. The spatial layout adjustment map is drawn according to standard cartographic specifications, covering the current situation, planning, adjustment comparisons, and detailed maps of key areas. The scale is set according to different levels: 1:25,000–1:50,000 for city / county level, 1:5,000–1:10,000 for township level, and 1:1,000–1:2,000 for key areas. The implementation schedule is divided into short-term, medium-term, and long-term projects based on urgency, implementation difficulty, and funding needs, clearly defining the implementing entity, timeline, funding sources, and expected results. The policy recommendations propose specific suggestions regarding legal improvement, funding guarantees, technical support, and management system construction to address implementation needs. The effectiveness evaluation report combines quantitative and qualitative indicators to establish a tracking and monitoring mechanism, ensuring dynamic evaluation and continuous optimization of the scheme's implementation effects.

[0107] Furthermore, step 3 includes the following steps.

[0108] The quality of historical land use conflict data is checked, and the data of each region is mapped to a unified grid cell based on a spatial weight matrix. Then, the sequence is aligned according to the timestamp to generate a standardized spatiotemporal dataset.

[0109] The spatiotemporal conflict analysis module is used to calculate the conflict intensity index of each grid unit. Combining the neighborhood location characteristics and semantic information in the national conflict knowledge graph, a multi-dimensional temporal feature vector sequence that integrates causal relationships is constructed based on a sliding time window.

[0110] The multidimensional temporal feature vector sequence is input into the encoder part of the pre-trained spatiotemporal lag effect model. A bidirectional long short-term memory network is used to capture the forward and backward associations in the sequence. Combined with the semantic enhancement information of the knowledge graph, the hidden state vector containing lag information and causal chain reasoning is output.

[0111] The decoder uses an attention mechanism to weight and summarize the hidden state vectors, integrates the causal chain reasoning results in the knowledge graph, and maps the weight distribution at different time steps to the prediction parameters of the conflict impact intensity and duration at the current moment, thereby achieving accurate prediction of the conflict state over time.

[0112] Based on the lag coefficient and spatial propagation probability in the decoder output, combined with semantic enhancement analysis of the knowledge graph, the prediction results are mapped back to the geographic grid, automatically capturing and quantifying the spatiotemporal propagation paths and corresponding time delays of land use conflicts between regions, thereby generating a spatiotemporal propagation trajectory map that includes propagation and lag effects.

[0113] Error analysis is performed between the predicted conflict intensity and duration and the measured data. If the prediction error exceeds the set threshold, the lag window length or spatial weight parameters in the model are dynamically adjusted based on the sensitivity analysis results and knowledge graph feedback, and iterative optimization is performed until the prediction accuracy meets the set standard.

[0114] This embodiment generates a unified and standardized spatiotemporal dataset by rigorously quality-checking and spatially mapping historical land use conflict data, ensuring data consistency and accuracy. Utilizing a spatiotemporal conflict analysis module, combined with neighboring region features and semantic information from a knowledge graph, a temporal feature vector integrating causal relationships is constructed to achieve a deep characterization of conflict evolution. This feature sequence is input into a pre-trained model, which, through a bidirectional long short-term memory network and attention mechanism, accurately captures the temporal correlations and causal relationships of conflicts, predicting the intensity and duration of conflict impacts. The prediction results are mapped back to a geographic grid, automatically quantifying the conflict propagation path and time delay, forming a detailed spatiotemporal propagation trajectory. By comparing predicted and measured data, model parameters are dynamically adjusted to continuously optimize prediction accuracy, enhancing the understanding and early warning capabilities for the spatiotemporal evolution of land use conflicts.

[0115] Furthermore, step 4 includes the following steps.

[0116] At the three spatial levels of region, county, and plot, conflict propagation networks are constructed based on grid units or administrative boundary units, respectively. Nodes represent corresponding spatial units, and edge weights are determined by the conflict propagation probability and distance decay function between adjacent nodes to ensure accurate mapping of network structures at different scales.

[0117] A multi-agent model is introduced into each network layer to simulate the conflict propagation process. Agents migrate between nodes according to preset state transition rules. At the same time, the shortest path algorithm is used to automatically extract the optimal propagation path of the conflict from the simulation results.

[0118] The resistance value corresponding to the comprehensive barrier factor is calculated for each edge on the path. After normalization, the path cost weight is assigned to achieve accurate quantification of multiple influencing factors.

[0119] The optimal propagation paths and their resistance costs identified in each level of the network are summarized. By comparing path length, cumulative resistance value, and propagation time delay, cross-scale propagation trend characteristics are extracted, providing a quantitative basis for the trend differences between macro and micro levels.

[0120] Based on the total path resistance and propagation probability, and combined with the threshold quantile method, a risk score is applied to each propagation path, dividing it into three levels of risk: low, medium, and high, thereby identifying potential key conflict diffusion corridors and high-risk sections.

[0121] The optimal propagation path, resistance distribution, and risk level are displayed in the form of map layers, and trend analysis and risk assessment reports are automatically generated, including cross-scale path comparison charts, resistance factor contribution bar charts, and risk area distribution heat maps.

[0122] In this embodiment, step 4 covers three spatial levels: region, county, and plot, constructing corresponding conflict propagation networks. Nodes represent spatial units, and edge weights are determined by conflict propagation probability and distance attenuation. Conflict diffusion is simulated using a multi-agent model, and the optimal propagation path is extracted using a shortest path algorithm. The barrier factors at the path edges are quantified and assigned cost weights. A comprehensive analysis of cross-scale path length, resistance, and propagation delay reveals differences in propagation trends. Based on path resistance and propagation probability, a threshold method is used for risk scoring, dividing the region into low, medium, and high-risk areas and identifying key diffusion corridors. Finally, the propagation paths, resistance distribution, and risk levels are displayed in map layer form, and a trend and risk assessment report is generated.

[0123] Furthermore, step 8 includes the following steps.

[0124] The spatial structure parameters, land use type distribution, and adjustment range of the current layout adjustment plan are input into the spatiotemporal fitness evaluation function. Combined with the future conflict situation characteristics output by the conflict propagation prediction module, the key impact indicators of the layout plan on the conflict evolution process are extracted.

[0125] The evaluation indicators of the layout scheme in terms of conflict mitigation effect, future development adaptability and implementation rationality are calculated respectively, and a comprehensive evaluation vector is constructed accordingly.

[0126] Based on a predetermined expert weighting system, each indicator is weighted, and the overall fitness score of the scheme is generated by weighted summation.

[0127] The obtained fitness score is compared with the set conflict mitigation criteria to quantitatively determine whether the current layout scheme meets the minimum threshold requirements for conflict intensity reduction, conflict area reduction, and propagation path blocking indicators.

[0128] Based on the analysis results of the spatiotemporal lag effect, the time delay and persistence changes in the conflict propagation process after the layout adjustment are examined to further verify whether the control effect of the scheme on the conflict evolution trend at different time points in the future meets expectations.

[0129] This embodiment inputs the spatial structure parameters of the layout scheme and land use adjustments into a spatiotemporal adaptability evaluation function, and extracts key impact indicators based on future conflict trends. Evaluation values ​​are calculated for three dimensions: conflict mitigation effect, future adaptability, and implementation rationality. A comprehensive adaptability score is generated based on expert weights. This score is compared with conflict mitigation standards to determine whether the scheme meets the minimum requirements for reducing conflict intensity, shrinking conflict area, and blocking conflict propagation. Simultaneously, considering the spatiotemporal lag effect, the scheme's control effect on delaying and sustaining conflict propagation time is examined to ensure it meets the expected objectives.

[0130] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A land space layout adjustment decision-making method based on planning conflict identification, applied to a land space planning decision-making system, characterized in that, The land space planning decision system includes a multi-source data fusion module, a time-space conflict analysis module, a conflict propagation prediction module, a land conflict knowledge graph construction module, a scheduling decision module, and a layout optimization module, including the following steps: Step 1, through the multi-source data fusion module, the multi-source heterogeneous spatial data is time-space registered and semantically unified, and the spatial distribution characteristics, time evolution mode and semantic attribute characteristics of land use conflict are automatically extracted, and then a time-space conflict feature vector library containing multi-dimensional attributes is constructed; Step 2, through the land conflict knowledge graph construction module, based on the fused multi-source data, the conflict event entity, conflict subject entity, spatial location entity and attribute feature entity are automatically extracted, a four-semantic network is constructed, and the complete cause-effect chain of conflict causes is identified, and a semantic knowledge graph containing cause-effect relationship, time sequence relationship, spatial relationship and subordination relationship is generated; Step 3, based on the pre-trained time-space lag effect model and the semantic enhancement of the land conflict knowledge graph, the time-space conflict analysis module is used to automatically capture and quantify the time-space propagation and lag effect of land use conflicts between different regions, combined with the reasoning results of the cause-effect chain in the knowledge graph, the influence intensity and duration of the conflict state over time are accurately predicted; Step 4, through the conflict propagation prediction module, the optimal propagation path of land use conflict between different spatial scales is intelligently identified, and the hindering of various influencing factors on conflict propagation is quantified, and then the trend analysis and risk level assessment of cross-scale conflict propagation path are realized; Step 5, according to the preset time-space lag threshold and influence intensity threshold, the explicit conflict in the current layout and the potential conflict based on the propagation prediction are automatically identified, and the composite conflict is intelligently classified by fuzzy clustering; Step 6, through the scheduling decision module, based on the identified explicit conflict and potential conflict, the conflict relief urgency, the rationality of intervention measures implementation and the existing resource allocation ability are comprehensively evaluated, and a phased differentiated conflict intervention strategy recommendation scheme is automatically generated, including the strategy combination and priority order of the emergency response period, the short-term governance period, the medium-term adjustment period and the long-term consolidation period; Step 7, through the layout optimization module, a multi-objective optimization model with the goal of conflict minimization, benefit maximization and cost optimization is constructed, and the rigid rule restriction and time-space lag factor are comprehensively considered, and the land space layout adjustment scheme is solved. The pre-trained spatio-temporal lag effect model adopts a deep recurrent neural network architecture, integrating an encoder-decoder structure and an attention mechanism module. The encoder part uses a bidirectional long short-term memory network to capture forward and backward dependency features in time series. The decoder part uses a multi-head self-attention mechanism to dynamically learn the influence weights between different time steps. During model training, supervised learning samples are constructed based on historical conflict data, input sequences and target sequence pairs are generated using a sliding time window, the loss function is set as the weighted combination of mean square error and time decay weight, and the Adam optimization algorithm is used for parameter iterative update. In the model validation stage, time series cross-validation method is used to evaluate the prediction performance indicators including mean absolute error, root mean square error and correlation coefficient. At the same time, sensitivity analysis is conducted to determine the optimal lag time window length and influence strength decay parameters, ensuring the model has good prediction stability in multiple spatio-temporal scales. The spatio-temporal conflict analysis module includes a conflict identification unit, a strength calculation unit and a propagation analysis unit. The conflict identification unit identifies potential conflicting land use type combinations based on land use compatibility matrix and spatial proximity criteria, and labels the areas with potential conflicts by setting spatial buffer threshold and time interval parameters. The strength calculation unit quantifies the conflict strength using a multi-factor weighted evaluation method, considering factors such as land use type difference, spatial overlap area, population density, economic value and ecological sensitivity. The analytic hierarchy process is used to determine the weight of each factor and establish a conflict strength evaluation index system. The propagation analysis unit constructs a conflict propagation network based on graph theory and network analysis method. Nodes represent spatial units, and edges represent conflict propagation paths. The spatial pattern and influence range of conflict propagation are analyzed by calculating node centrality, clustering coefficient and path length. Step 3 includes the following steps: Quality inspection is performed on historical land use conflict event data, and regional data is mapped to a unified grid cell based on a spatial weight matrix. Then, sequence alignment is performed according to the time stamp, generating a standardized spatio-temporal data set. The conflict strength index of each grid cell is calculated using the spatio-temporal conflict analysis module. The neighborhood location characteristics and semantic information in the land conflict knowledge graph are combined to construct a multi-dimensional time series feature vector sequence based on a sliding time window, which integrates causal relationships. The multi-dimensional time series feature vector sequence is input into the pre-trained spatio-temporal lag effect model encoder. A bidirectional long short-term memory network is used to capture the forward and backward associations in the sequence, and the semantic enhancement information from the knowledge graph is combined to output a hidden state vector that includes lag information and causal chain reasoning. The decoder uses an attention mechanism to weight and aggregate the hidden state vector, integrates the causal chain reasoning results from the knowledge graph, and maps the weight distribution of different time steps to conflict influence strength and duration prediction parameters at the current time, achieving accurate prediction of conflict state over time. Based on the lag coefficient in the decoder output and the spatial propagation probability, combined with the semantic enhancement analysis of the knowledge graph, the prediction results are mapped back to the geographic grid to automatically capture and quantify the spatio-temporal propagation path and corresponding time delay of land use conflicts between regions, thereby generating a spatio-temporal propagation trajectory graph containing propagation and lag effects; The predicted conflict intensity and duration are compared with the measured data for error analysis. If the prediction error exceeds the set threshold, the lag window length or spatial weight parameters in the model are dynamically adjusted based on the sensitivity analysis results and knowledge graph feedback, and iterative optimization is performed until the prediction accuracy meets the set standard.

2. The land space layout adjustment decision-making method based on planning conflict identification according to claim 1, characterized in that, During the spatio-temporal registration process, the multi-source data fusion module adopts a hierarchical registration strategy: first, spatial data in different coordinate systems are unified to the WGS84 geographic coordinate system to ensure consistent coordinate reference; Then, common feature points in the spatial data are identified using a feature point matching algorithm, and an affine transformation matrix is used for geometric correction of the data. Finally, a timestamp alignment method is used to standardize the time sequence of different temporal data to ensure consistency in the time dimension. In the semantic unification process, a unified land use classification coding system is constructed, semantic conversion between different data sources is achieved through ontology mapping technology, and semantic conflicts are handled using a confidence weighted fusion method to ensure that the fused data is consistent in semantics and has a spatial accuracy of meters.

3. The land space layout adjustment decision-making method based on planning conflict identification according to claim 1, characterized in that, The multi-source data fusion module includes a data preprocessing unit, a coordinate conversion unit, and a semantic mapping unit. The data preprocessing unit is responsible for quality inspection and noise removal of remote sensing image data, GIS vector data, statistical survey data, and field measurement data from different sensor platforms. The coordinate conversion unit performs accurate coordinate transformation between multiple spatial reference systems based on high-precision geodetic reference and projection parameters, and uses a seven-parameter Bursa model for unified processing of three-dimensional spatial coordinates. The semantic mapping unit constructs an ontology knowledge base based on industry standards, defines a standardized land use classification system and attribute description specification, and achieves semantic alignment between heterogeneous data sources through semantic similarity calculation and concept level matching to ensure consistency in the semantic representation of the same geographic entity in each data source.

4. The land space layout adjustment decision-making method based on planning conflict identification according to claim 1, characterized in that, The dispatching decision module includes an urgency assessment unit, a feasibility analysis unit, a resource allocation unit, a strategy generation unit, a planning scenario classification unit, an intelligent index adjustment unit, and a real-time dynamic adjustment trigger unit. The urgency assessment unit constructs an urgency evaluation index system based on conflict propagation prediction results and causal analysis of the knowledge graph. The feasibility analysis unit is used to evaluate the implementation conditions of intervention measures. The resource allocation unit allocates the optimal resource combination for intervention strategies based on current available resource constraints. The strategy generation unit automatically generates phased intervention strategy combinations based on the urgency assessment, feasibility analysis, and resource allocation results as decision conditions. The planning scenario classification unit divides decision scenarios into four types: economic development-oriented, ecological protection-prioritized, social fairness-considered, and comprehensive coordination.

5. The land space layout adjustment decision-making method based on planning conflict identification according to claim 1, characterized in that, The conflict propagation prediction module simulates the propagation process of conflicts in the spatial network based on a multi-agent modeling method. Each agent corresponds to a spatial unit and has state transition rules and interaction behavior patterns. The agent state includes four types: no conflict, potential conflict, explicit conflict, and conflict resolution. When the conflict intensity of adjacent agents exceeds the preset threshold, state transition is triggered, and the propagation probability is determined by the distance decay function, terrain barrier factor, and policy intervention strength. Through Monte Carlo simulation method, multiple random experiments are conducted to statistically analyze the conflict propagation path, arrival time, and influence range under different scenarios, generate probability distribution graph and risk level classification, and identify the key parameters that have the greatest impact on the propagation process through sensitivity analysis.

6. The method of claim 1, wherein, Step 4 includes the following steps: At the regional, county, and plot spatial levels, conflict propagation networks are constructed based on grid cells or administrative boundary units, respectively. Nodes represent corresponding spatial units, and edge weights are determined by the conflict transmission probability between adjacent nodes and the distance decay function to ensure accurate mapping of network structures at different scales. A multi-agent model is introduced in each network level to simulate the diffusion process of conflicts. Agents migrate between nodes according to pre-set state transition rules. Meanwhile, the shortest path algorithm is used to automatically extract the optimal propagation path of conflicts from the simulation results. The resistance value corresponding to the comprehensive barrier factor is calculated for each edge in the path, and after normalization, the path cost weight is assigned to realize accurate quantification of various influencing factors. The optimal propagation paths and their resistance costs in each level network are summarized, and by comparing the path length, resistance cumulative value, and propagation time delay, the cross-scale propagation trend characteristics are extracted to provide quantitative analysis basis for the trend differences between macro and micro levels. Based on the total value of path resistance and propagation probability, combined with the threshold quantile method, the risk score of each propagation path is implemented, and it is divided into low, medium, and high risk areas, so as to clarify the potential key conflict diffusion corridor and high-risk section. The optimal propagation path, resistance distribution, and risk level are displayed in the form of map layers, and the trend analysis and risk assessment report is automatically generated, including cross-scale path comparison chart, resistance factor contribution column chart, and risk area distribution heat map.

7. The method of claim 1, wherein, The land space layout adjustment decision method further includes the following steps: Step 8. In the layout adjustment scheme generation process, the influence of the scheme on future conflict evolution is evaluated in real time by the spatio-temporal fitness evaluation function. The conflict propagation prediction results and the spatio-temporal hysteresis effect analysis results are combined to determine whether the layout adjustment scheme meets the preset conflict mitigation standards. Specifically, the spatial structure parameters, land use type distribution, and adjustment amplitude of the current layout adjustment scheme are input into the spatio-temporal fitness evaluation function. The future conflict situation characteristics output by the conflict propagation prediction module are combined to extract key influence indicators of the layout scheme on the conflict evolution process. The evaluation indicators of the layout scheme in three dimensions of conflict mitigation effect, future development adaptability, and implementation rationality are calculated respectively, and a comprehensive evaluation vector is constructed accordingly. Based on the pre-determined expert weight system, the indicators are weighted and processed, and the comprehensive fitness score is generated by weighted summation. The obtained fitness score is compared with the set conflict mitigation standards to quantitatively determine whether the current layout scheme meets the minimum threshold requirements of conflict intensity reduction, conflict area reduction, and transmission path blocking indicators. Combined with the spatio-temporal hysteresis effect analysis results, the time delay and persistence changes in the conflict propagation process after layout adjustment are tested to further verify whether the control effect of the scheme on the conflict evolution trend at different time nodes in the future meets the expectations. Step 9. If the layout adjustment scheme does not meet the preset conflict mitigation standards, return to the multi-objective optimization model for re-solution until the conflict mitigation standards are met. Step 10. If the layout adjustment scheme has met the preset conflict mitigation standards, generate a Pareto optimal solution set that considers both current conflict mitigation and future risk prevention, and output the final land space layout adjustment decision scheme.

Images