Intelligent policy resource matching method and system

Abstract

The invention relates to the technical field of policy resource intelligent matching, in particular to a policy resource intelligent matching method and system, and the method comprises the steps: data collection, matching strategy recognition, visual hierarchical feedback and strategy adjustment and optimization. The data acquisition is used for acquiring declaration subject attribute data and policy term text data, and the matching strategy identification is used for generating optimal target matching strategy data of the declaration subject attribute data and the policy term text data; according to the invention, technologies such as a BERT embedding model, a graph neural network algorithm module, a particle swarm search algorithm module and a natural language processing interface are introduced, so that efficient and accurate matching between a declaration subject and policy resources is realized; the conflict between policy dynamics and system staticity, the conflict between semantic complexity and man-machine interaction simplicity and the conflict between resource supply accuracy and service efficiency are effectively solved, and the scientificity and practicability of policy resource intelligent matching are remarkably improved.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent matching of policy resources, and in particular to a method and system for intelligent matching of policy resources. Background Art

[0002] As policy resource management becomes increasingly complex and diverse, how to achieve intelligent matching between applicants and policy resources has become a key issue in improving government efficiency and service quality. In existing technologies, policy resource intelligent matching systems mainly rely on the following three technical solutions: The first type of solution uses a static matching method based on a rule engine, manually constructing IF-THEN logical rules to compare policy terms with the conditions of the reporting entity. This method has obvious flaws: first, policy terms are frequently updated (such as the high annual revision rate of tax preferential policies), and the lag in manually maintaining the rule base leads to poor timeliness of matching results; second, in scenarios with multi-dimensional condition combinations (such as simultaneously meeting "high-tech enterprise qualifications + additional deductions for R&D expenses"), logical conflicts between rules are likely to increase, seriously affecting matching accuracy; third, cross-regional policy differences (such as the gradient of subsidy standards between the eastern coastal areas and the western regions) require repeated development of localized rules, which has high maintenance costs.

[0003] The second type of solution is based on keyword matching and text similarity calculation, using algorithms such as TF-IDF and BM25 to achieve rapid retrieval of policy texts and demand descriptions. The core problem of this method is that the lack of semantic understanding ability leads to the misjudgment of synonyms, such as equating "technology-based SMEs" with "innovative small and micro enterprises"; at the same time, the logical connection between policy elements is ignored, affecting the substantive consistency of the matching results with the original policy text.

[0004] The third type of solution attempts to introduce a single machine learning model and use algorithms such as decision trees and random forests for classification and prediction. Although this method improves the degree of automation, it has three major technical bottlenecks: first, it is highly dependent on feature engineering and requires manual labeling of policy labels. When faced with a policy library of millions, the labeling cost surges; second, the cold start problem is prominent. After the release of a new policy, it takes 3-6 months of training data accumulation to achieve effective matching accuracy; third, the model has poor interpretability, and it is impossible to explain the matching basis to users in the government information disclosure scenario.

[0005] These deficiencies lead to three contradictions in the current system's practical application: the conflict between policy dynamism and system staticness, the conflict between semantic complexity and simplicity of human-computer interaction, and the conflict between resource supply accuracy and service efficiency. Furthermore, the current system lacks a dynamic feedback mechanism, making it unable to optimize matching strategies based on user application results, resulting in a long-term decline in matching accuracy.

[0006] Therefore, there is an urgent need to propose an intelligent matching method and system for policy resources that can effectively solve the above-mentioned defects. Summary of the Invention

[0007] To address the deficiencies in the existing intelligent matching system for policy resources in terms of policy dynamics, semantic complexity, and resource supply accuracy, the present invention provides an intelligent matching method and system for policy resources. By introducing technologies such as the BERT embedding model, graph neural network algorithm module, particle swarm search algorithm module, and natural language processing interface, it realizes the efficient and accurate matching of declaration entities and policy resources, effectively solves the contradictions between policy dynamics and system staticity, semantic complexity and simple human-computer interaction, and resource supply accuracy and service efficiency, and significantly improves the scientificity and practicality of intelligent matching of policy resources.

[0008] The objectives of the present invention can be achieved through the following technical solutions: In the first aspect, the present invention provides an intelligent matching method for policy resources, including the following steps: S1. Data collection: Collect declaration entity attribute data and policy clause text data; S2. Matching strategy identification: Use the particle swarm search algorithm to perform specific matching strategy identification processing based on the adaptability score and matching probability, and generate target matching strategy data; Use the BERT embedding model to calculate the semantic similarity between the declaration entity attribute data and the policy constraint condition keywords to generate an adaptability score; use the graph neural network algorithm to match the declaration entity attribute data and the policy clause text data to obtain a matching probability; S3. Visual hierarchical feedback: Execute the matching result based on the declaration entity attribute data, policy clause text data, and target matching strategy data, and display it through a visual interface in a hierarchical manner; S4. Strategy adjustment and optimization: Collect user feedback opinions, and generate strategy adjustment judgment data according to the matching strategy adjustment instructions. When the user feedback opinion is that no adjustment is required, directly end the current policy matching task; when the user feedback opinion is that adjustment is required, perform strategy optimization operations on the target matching strategy data and generate optimized matching strategy data.

[0009] As a preferred technical solution of the present invention, the data collection includes the steps of: Online input enterprise attribute information through the government service platform to generate declaration entity attribute data; the declaration entity attribute data includes enterprise type, industry field, registered place, registered capital, employee scale, qualification certificate, R & D investment ratio, and credit rating; Input the text information of policy terms online through the government service platform and generate policy term text data; the policy term text information includes policy name, scope of application, and support direction.

[0010] As a preferred technical solution of the present invention, the generating of the adaptability score includes the steps of: Construct a set of policy constraint condition keywords based on the declared entity attribute data; the policy constraint condition keywords include blacklisted enterprises, persons subject to enforcement for dishonesty, parties to major tax violation cases, environmental protection violation records, work safety accident records, and major legal lawsuits; Use the BERT embedding model to calculate the semantic similarity between the declared entity attribute data and the policy constraint condition keywords, and generate an adaptability score.

[0011] As a preferred technical solution of the present invention, the matching probability includes the steps of: Construct a set of policy classification label data; the policy classification label data includes scientific and technological innovation category, energy conservation and emission reduction category, rural revitalization category, employment and entrepreneurship category, and financial service category; Perform attribute feature matching according to the declared entity attribute data and the set of policy classification label data, use the graph neural network algorithm to model the multi-dimensional features of the declared entity attribute data, and combine the policy term text data to obtain the matching probability between the declared entity attribute data and the policy classification label data.

[0012] As a preferred technical solution of the present invention, the particle swarm search algorithm includes: Parameter setting: Set the number of particles, the number of iterations, and initialize the inertia weight , acceleration constants and ; Particle swarm initialization: Each particle represents a complete target matching strategy data, set the feasible domain interval of the target matching strategy data, the initial position of the particle is randomly generated in the feasible domain, the initial velocity is 0 or a small-range random value, the individual best is the initial position of the particle, and the global best is the position of the particle with the highest fitness; Fitness function and iteration function design: Define the fitness function, particle velocity iteration function, and position iteration function in combination with the adaptability score and the matching probability. If the fitness of the current particle position is better than the historical value, then update , if the current particle position is better than all other particle positions, then update ; The fitness function and iteration function include: Fitness function calculation formula: ; Calculation formula of particle velocity iteration function: ; Calculation formula of particle position iteration function: ; Among them, and are weight coefficients, and = 1, is the target matching strategy, is the adaptability score. If the semantic similarity between the declared subject attribute data and the keywords of the policy constraint conditions exceeds the preset threshold, then is 0. If it does not exceed the preset threshold, then is 1, is the matching probability between the declared subject attribute data output by the graph neural network and the policy classification label data, and are random numbers between, increasing the search randomness, and are respectively the velocities of the th particle at the th and th iterations, and are respectively the positions of the th particle at the th and th iterations; Termination condition definition: reaching the maximum number of iterations or the global optimum being less than the threshold for N consecutive times; Output the target matching strategy data; The particle swarm search algorithm is divided into three parts: a global exploration stage, a local optimization stage, and a convergence stage; In the global exploration stage, it is used to quickly cover the search space and avoid falling into local optima; in the local optimization stage, it is used to focus on high-potential regions and improve the accuracy of the target matching strategy data; in the convergence stage, it is used to stabilize the optimal target matching strategy data and prevent oscillations.

[0013] As a preferred technical solution of the present invention, the visual hierarchical feedback includes the steps of: Transmit the declared subject attribute data, policy clause text data, and target matching strategy data to the terminal device through the cloud computing network, and orderly and hierarchically display the declared subject attribute data, policy clause text data, and target matching strategy data through the visual interface; Synchronously generate a matching strategy adjustment selection button module on the visual interface; the matching strategy adjustment selection button module includes a matching strategy needs adjustment selection button module and a matching strategy does not need adjustment selection button module.

[0014] As a preferred technical solution of the present invention, the strategy adjustment and optimization includes the steps of: Collecting the feedback of users on the target matching strategy data online through the matching strategy adjustment selection button module of the visualization interface, and generating user feedback data, where the user feedback data includes adjustment required instructions and adjustment not required instructions; Performing keyword judgment processing on the matching strategy adjustment instruction information of the user feedback data, and generating strategy adjustment judgment data; when the user feedback data is adjustment required instructions, it means that the user needs to adjust the matching rules and operation steps in the target matching strategy data, and the output strategy adjustment judgment data is adjustment required; when the user feedback data is adjustment not required instructions, it means that the user does not need to adjust the matching rules and operation steps in the target matching strategy data, and the output strategy adjustment judgment data is adjustment not required, and at this time, this policy matching operation is directly ended.

[0015] As a preferred technical solution of the present invention, the strategy adjustment and optimization further includes the steps of: When the strategy adjustment judgment data is adjustment required, the user inputs the matching strategy adjustment text information for the target matching strategy data online through the natural language processing interface, and generates optimized matching strategy data.

[0016] In a second aspect, the present invention also provides a policy resource intelligent matching system that executes the above-mentioned policy resource intelligent matching method. The system includes a declaration subject attribute collection module, a matching strategy generation module, a matching result feedback module, and a strategy optimization module. The declaration subject attribute collection module is used to input enterprise attribute information and policy clause text information online, and generate declaration subject attribute data; the declaration subject attribute collection module includes a declaration subject information input unit, a policy clause information input unit, a policy constraint condition storage unit, and a policy suitability judgment unit; The matching strategy generation module is used to perform specific matching strategy recognition processing according to the declaration subject attribute data and the policy clause text data, and generate target matching strategy data through data identification; The matching result feedback module executes the matching result based on the declaration subject attribute data, the policy clause text data, and the target matching strategy data, and hierarchically feedbacks and displays it through the visualization interface; the matching result feedback module includes a matching result feedback unit, a user feedback collection unit, a strategy adjustment judgment unit, and a matching strategy optimization unit; The strategy optimization module is used to perform strategy optimization operations on the target matching strategy data and generate optimized matching strategy data when the user feedback is adjustment required.

[0017] As a preferred technical solution of the present invention, the declaration subject attribute collection module includes a declaration subject information input unit, a policy clause information input unit, a policy constraint condition storage unit, and a policy suitability judgment unit; the matching result feedback module includes a matching result feedback unit, a user feedback opinion collection unit, a strategy adjustment judgment unit, and a matching strategy optimization unit.

[0018] The present invention provides a method and system for intelligent matching of policy resources, having the following beneficial effects: 1. By collecting the declaration subject attribute information and policy clause text information online through the government service platform, it provides data support for accurately matching the target policy; using the BERT embedding model to calculate the semantic similarity between the declaration subject attribute data and the policy constraint condition keywords, using the graph neural network algorithm to match the declaration subject attribute data and the policy clause text data, and obtaining the target matching strategy data through the particle swarm search algorithm, realizing the accurate matching of the declaration subject attributes and policy categories, thereby improving the applicability and efficiency of policy matching; 2. Intuitively feedback the matching results of the declaration subject attribute information, policy category information, and target matching strategy information through the visualization interface; and collect user feedback according to the visualization interface, and accurately identify the user's handling opinions on the adjustment of the matching strategy, improving the rationality and scientificity of generating the matching strategy; at the same time, the user can input the matching strategy adjustment text information for the target matching strategy data online through the natural language processing interface, thereby improving the quality and convenience of policy matching. Description of the Drawings

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0020] Figure 1 It is a schematic flowchart of a method for intelligent matching of policy resources provided by an embodiment of the present invention; Figure 2 It is a schematic flowchart of the generation and adjustment optimization of target matching strategy data provided by an embodiment of the present invention; Figure 3 It is a schematic structural diagram of a system for intelligent matching of policy resources provided by an embodiment of the present invention. Detailed Embodiments

[0021] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts also belong to the scope of protection of the present application.

[0022] The following further describes the optimal embodiments of the present invention with reference to the accompanying drawings; Embodiment 1 Please refer to Figure 1 , this embodiment provides an intelligent matching method for policy resources, including: In step S1, the declaring entity inputs enterprise attribute information online through the government service platform to generate declaring entity attribute data, and at the same time inputs policy clause text information to generate policy clause text data; The declaring entity attribute information includes specific contents such as enterprise type, industry field, place of registration, registered capital, number of employees, qualification certificates, R & D investment ratio, and credit rating; the policy clause text information includes policy name, scope of application, and support direction; the government service platform, as the core entry for data collection, can ensure the integrity and timeliness of data.

[0023] In step S2, the particle swarm search algorithm is used to perform specific matching strategy identification processing according to the declaring entity attribute data and the policy clause text data to generate target matching strategy data; including: The particle swarm search algorithm includes: Parameter setting: Set the number of particles, the number of iterations, the initial inertia weight , the acceleration constant and ; Particle swarm initialization: Each particle represents a complete set of target matching strategy data. Set the feasible region interval of the target matching strategy data. The initial position of the particle is randomly generated in the feasible region, and the initial velocity is 0 or a small range of random values. The individual optimal is the initial position of the particle, and the global optimal is the position of the particle with the highest fitness; Construct a policy constraint condition keyword set based on the declaring entity attribute data. The policy constraint condition keywords include keywords such as blacklisted enterprises, dishonest executors, parties to major tax violation cases, environmental protection violation records, work safety accident records, and major legal lawsuits; then use the BERT embedding model to calculate the semantic similarity between the declaring entity attribute data and the policy constraint condition keywords in the policy constraint condition keyword set to generate an adaptability score; Construct a policy classification label data set; the policy classification label data includes scientific and technological innovation, energy conservation and emission reduction, rural revitalization, employment and entrepreneurship, and financial services; match the attribute characteristics according to the applicant entity attribute data and the policy classification label data set, use the graph neural network algorithm to model the multi-dimensional characteristics of the applicant entity attribute data, and combine the topological structure of the policy classification label data to obtain the matching probability between the applicant entity attribute data and the policy classification label data; Fitness function and iteration function design: Define the fitness function, particle velocity iteration function, and position iteration function by combining the fitness score and the matching probability. If the fitness of the current particle position is better than the historical value, then update , if the current particle position is better than all other particle positions, then update ; The fitness function and iteration function include: Fitness function calculation formula: ; Particle velocity iteration function calculation formula: ; Particle position iteration function calculation formula: ; Among them, and are weight coefficients, and = 1, is the target matching strategy, is the fitness score. If the semantic similarity between the applicant entity attribute data and the policy constraint condition keywords exceeds the preset threshold, then is 0. If it does not exceed the preset threshold, then is 1, is the matching probability between the applicant entity attribute data and the policy classification label data output by the graph neural network, and are random numbers between, increasing the search randomness, and are respectively the th particle's velocity at the th and th iterations, and are respectively the th particle's position at the th and th iterations; Termination condition definition: Reach the maximum number of iterations or the global optimum is less than the threshold for N consecutive times; Output the target matching strategy data; The particle swarm optimization algorithm is divided into three parts: the global exploration stage, the local optimization stage, and the convergence stage; In the global exploration stage, the inertia weight is initialized to 0.9, and the acceleration constants and are both set to 2 to quickly cover the search space and avoid falling into local optima; in the local optimization stage, the inertia weight is reduced to 0.4, and the acceleration constant is set to 2.5, is set to 1.5 to focus on high-potential regions and improve the accuracy of the target matching strategy data; in the convergence stage, the inertia weight is fixed at 0.3, allowing only minor perturbations. If the fitness does not improve for 10 consecutive iterations, the search is terminated early. This stage is used to stabilize the optimal target matching strategy data and prevent oscillations.

[0024] Step S3 performs a matching result feedback operation based on the applicant entity attribute data, the policy clause text data, and the target matching strategy data; it includes: The applicant entity attribute data, the policy clause text data, and the target matching strategy data are transmitted to the terminal device through the cloud computing network and displayed and output in an orderly layered manner from left to right using a visual interface; the applicant entity attribute data is located on the left side of the interface, the policy clause text data is in the middle, and the target matching strategy data is on the right side; in addition, matching strategy adjustment selection button modules are set at the left and right ends of the bottom of the visual interface, and the matching strategy adjustment selection button modules include a matching strategy needs adjustment selection button module and a matching strategy does not need adjustment selection button module; users can intuitively view the matching results through this interface and decide whether to adjust the matching strategy according to their own needs.

[0025] Step S4 collects the user feedback opinion data and performs a matching strategy adjustment judgment process to generate strategy adjustment judgment data; it includes: Online collects the user's feedback opinions on the matching results through the matching strategy adjustment selection button module of the visual interface; when the user clicks the matching strategy needs adjustment selection button module, the generated user feedback opinion data is a need to adjust instruction; when the user clicks the matching strategy does not need adjustment selection button module, the generated user feedback opinion data is a does not need adjustment instruction; If the user feedback data indicates a need to adjust the instruction, the output policy adjustment judgment data is "needs adjustment"; at the same time, the user can re-enter online, through the natural language processing interface, the matching policy adjustment text information for the target matching policy data, and generate the optimized matching policy data; the natural language processing interface supports natural language processing technology such as GPT-4 or PaLM, and can accurately understand the user's adjustment intention and generate the corresponding optimized matching policy data; if the user feedback data indicates no need to adjust the instruction, the output policy adjustment judgment data is "does not need adjustment", and this policy matching task is directly ended at this time.

[0026] Embodiment 2 Please refer to Figure 3 , this embodiment provides a policy resource intelligent matching system that executes the above-mentioned policy resource intelligent matching method. The system includes a declaration subject attribute collection module, a matching policy generation module, a matching result feedback module, and a policy optimization module; The declaration subject attribute collection module is used to input enterprise attribute information and policy clause text information online to generate declaration subject attribute data; the declaration subject attribute collection module includes a declaration subject information entry unit, a policy clause information entry unit, a policy constraint condition storage unit, and a policy suitability judgment unit; The matching policy generation module is used to perform specific matching policy identification processing based on the declaration subject attribute data and the policy clause text data, and generate target matching policy data through data identification; The matching result feedback module executes the matching result based on the declaration subject attribute data, the policy clause text data, and the target matching policy data, and hierarchically feedbacks and displays it through a visual interface; the matching result feedback module includes a matching result feedback unit, a user feedback opinion collection unit, a policy adjustment judgment unit, and a matching policy optimization unit; The policy optimization module is used to perform a policy optimization task on the target matching policy data and generate optimized matching policy data when the user feedback indicates a need for adjustment.

[0027] Embodiment 3 Please refer to Figure 1-2 , in order to better enable relevant personnel in the technical field to fully understand and implement the present invention, the following further supplements and explains the specific implementation principle of the present invention in combination with a specific application scenario: During the deployment of the policy resource intelligent matching system in a certain province, the government service platform serves as the core data collection entry for the declaration entity and policy clause data. First, a high-tech enterprise inputs its attribute information online through the platform, including enterprise type as "high-tech enterprise", industry field as "information technology", registered location as "City A, Province X", employee scale as "500 people", R & D investment ratio as "8%", etc. At the same time, the enterprise also uploads relevant policy clause text information, such as a R & D expense subsidy policy for high-tech enterprises. These data are stored separately through the declaration entity information input unit and the policy clause information input unit of the government service platform and transmitted to the subsequent processing module. Subsequently, in the matching strategy identification stage, the policy adaptability judgment unit constructs a set of policy constraint condition keyword sets, including keywords such as "blacklisted enterprises", "dishonest executors", and "parties involved in major tax violation cases". Using the BERT embedding model, semantic similarity calculations are performed one by one between the declaration entity attribute data and the keywords in the policy constraint condition keyword sets to generate adaptability scores. The policy classification label storage unit pre-stores policy classification label data such as those for scientific and technological innovation, energy conservation and emission reduction, and rural revitalization. The policy category analysis unit, by invoking the graph neural network algorithm module, conducts multi-dimensional feature modeling and topological structure matching between the declaration entity attribute data and the label data in the policy classification label data set to obtain the matching probability. Then, based on the adaptability scores and matching probabilities, the fitness function, particle velocity iteration function, and position iteration function of the particle swarm search algorithm are defined, and the target matching strategy data is generated. In the global exploration stage, the particle swarm search algorithm performs global rapid coverage search to avoid falling into local optima. In the local optimization stage, it then focuses on high-potential areas to improve the search accuracy of the target matching strategy data. In the convergence stage, it is used to stabilize the optimal target matching strategy data to prevent oscillations. Thus, the globally optimal target matching strategy data is accurately generated. In the visual hierarchical feedback stage, the matching result feedback unit transmits the declaration entity attribute data, policy clause text data, and target matching strategy data to the terminal device through the cloud computing network and displays them in a visual interface. For example, the enterprise's attribute information is shown on the left side of the interface, the "scientific and technological innovation" policy category it matches is shown in the middle, and the specific operation rules of the R & D expense subsidy policy are shown on the right side. Users can decide whether to adjust the matching strategy through the matching strategy adjustment selection button module at the bottom of the interface. This hierarchical display method is intuitive and clear, facilitating users to quickly understand the matching results.

[0028] In the stage of policy adjustment and optimization, the user feedback collection unit collects user feedback online through the matching policy adjustment selection button module of the visualization interface; for example, if the user believes that some policy terms in the matching result do not meet the actual needs, the user can click the "Matching policy needs to be adjusted" button to generate user feedback data as an adjustment instruction; if the user is satisfied with the matching result, the user clicks the "Matching policy does not need to be adjusted" button to generate user feedback data as a non-adjustment instruction; the policy adjustment judgment unit generates policy adjustment judgment data based on the user feedback data. If it is a non-adjustment, the current matching task is directly ended. When the policy adjustment judgment data indicates a need for adjustment, the user inputs online again the adjustment text information for the target matching policy data through the natural language processing interface; for example, the user may propose "adding an additional subsidy policy for enterprises with a R & D investment ratio higher than 10%". The natural language processing interface uses technologies such as GPT-4 or PaLM to accurately understand the user's adjustment intention and generate optimized matching policy data; the policy optimization unit receives the optimized matching policy data and updates it to the system for subsequent matching tasks.

[0029] Through the implementation of the above specific application scenarios, the present invention effectively solves the contradictions between policy dynamics and system staticity, semantic complexity and simplicity of human-computer interaction, and the accuracy of resource supply and service efficiency, and significantly improves the scientificity and practicality of intelligent policy resource matching.

[0030] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present application shall be included within the protection scope of the present application.

Claims

1. An intelligent matching method for policy resources, characterized in that: It includes the following steps: S1. Data collection: Collect the declared entity attribute data and policy clause text data; S2. Matching strategy identification: Use the particle swarm search algorithm to perform specific matching strategy identification processing according to the fitness score and matching probability, and generate the target matching strategy data; Use the BERT embedding model to calculate the semantic similarity between the declared entity attribute data and the policy constraint condition keywords to generate the fitness score; Use the graph neural network algorithm to match the declared entity attribute data and the policy clause text data to obtain the matching probability; S3. Visual hierarchical feedback: Execute the matching result based on the declared entity attribute data, policy clause text data and target matching strategy data, and display it in a hierarchical manner through the visual interface; S4. Strategy adjustment and optimization: Collect the user feedback, and generate the strategy adjustment judgment data according to the matching strategy adjustment instruction. When the user feedback is that no adjustment is required, directly end the current policy matching task; When the user feedback is that adjustment is required, perform the strategy optimization task on the target matching strategy data, and generate the optimized matching strategy data.

2. The intelligent matching method of policy resources according to claim 1, wherein: The data collection includes the steps: Input the enterprise attribute information online through the government service platform and generate the declared entity attribute data; The declared entity attribute data includes enterprise type, industry field, registered place, registered capital, employee scale, qualification certificate, R & D investment ratio and credit rating; Input the policy clause text information online through the government service platform and generate the policy clause text data; The policy clause text information includes policy name, applicable scope and support direction.

3. The intelligent matching method for policy resources according to claim 1, wherein: The generation of the fitness score includes the steps: Construct a set of policy constraint condition keywords based on the declared entity attribute data; The policy constraint condition keywords include blacklisted enterprises, dishonest executors, parties to major tax violation cases, environmental protection violation records, work safety accident records and major legal lawsuits; Use the BERT embedding model to calculate the semantic similarity between the declared entity attribute data and the policy constraint condition keywords to generate the fitness score.

4. The intelligent matching method for policy resources according to claim 1, wherein: The use of the graph neural network algorithm to match the declared entity attribute data and the policy clause text data includes the steps: Construct a set of policy classification label data; The policy classification label data includes scientific and technological innovation, energy conservation and emission reduction, rural revitalization, employment and entrepreneurship, and financial services; Perform attribute feature matching according to the declared entity attribute data and the set of policy classification label data, use the graph neural network algorithm to model the multi-dimensional features of the declared entity attribute data, and combine the policy clause text data to obtain the matching probability between the declared entity attribute data and the policy classification label data.

5. The intelligent matching method for policy resources according to claim 1, wherein The particle swarm search algorithm includes: Parameter setting: Set the number of particles, the number of iterations, and initialize the inertia weight , acceleration constants and ; Particle swarm initialization: Each particle represents a complete set of target matching strategy data. Set the feasible region interval of the target matching strategy data. The initial positions of the particles are randomly generated within the feasible region, and the individual best is the initial position of the particle, and the global best is the position of the particle with the highest fitness; Fitness Function and Iterative Function Design: Define the fitness function, particle velocity iterative function, and position iterative function by combining the fitness score and the matching probability. If the fitness of the current particle position is better than the historical value, update , and if the current particle position is better than all other particle positions, update ; The fitness function and iteration function include: Fitness function calculation formula: ； Particle velocity iteration function calculation formula: ； Particle position iteration function calculation formula: ； Among them, and are weight coefficients, and = 1, is the target matching strategy, is the adaptability score. If the semantic similarity between the declared subject attribute data and the policy constraint condition keywords exceeds the preset threshold, then is . If it does not exceed the preset threshold, then is . is the matching probability between the declared subject attribute data output by the graph neural network and the policy classification label data, and are random numbers between, and are respectively the velocities of the nd particle at the th and th iterations, and are respectively the positions of the nd particle at the th and th iterations; Termination condition definition: Reach the maximum number of iterations or the global optimum is less than the threshold for N consecutive times; Output the target matching strategy data; Divide the particle swarm search algorithm into three parts: global exploration stage, local optimization stage and convergence stage; The global exploration phase is used to quickly cover the search space and avoid falling into local optima; the local optimization phase is used to focus on high-potential regions and improve the accuracy of the target matching strategy data; the convergence phase is used to stabilize the optimal target matching strategy data and prevent oscillations.

6. The intelligent matching method of policy resources according to claim 1, characterized in that: The visual hierarchical feedback includes the steps of: Transmitting the declared entity attribute data, policy clause text data, and target matching strategy data to the terminal device through the cloud computing network, and orderly and hierarchically displaying the declared entity attribute data, policy clause text data, and target matching strategy data through the visual interface; Synchronously generating a matching strategy adjustment selection button module on the visual interface; the matching strategy adjustment selection button module includes a matching strategy that needs to be adjusted selection button module and a matching strategy that does not need to be adjusted selection button module.

7. The intelligent matching method for policy resources according to claim 1, wherein: The strategy adjustment and optimization includes the steps of: Online collecting the user's feedback on the target matching strategy data through the matching strategy adjustment selection button module on the visual interface, and generating user feedback opinion data, where the user feedback opinion data includes an adjustment required instruction and an adjustment not required instruction; Performing keyword judgment processing on the user feedback opinion data for the matching strategy adjustment instruction information, and generating strategy adjustment judgment data; when the user feedback opinion data is an adjustment required instruction, it means that the user needs to adjust the matching rules and operation steps in the target matching strategy data, and the output strategy adjustment judgment data is adjustment required; when the user feedback opinion data is an adjustment not required instruction, it means that the user does not need to adjust the matching rules and operation steps in the target matching strategy data, and the output strategy adjustment judgment data is adjustment not required, and at this time, this policy matching operation is directly ended.

8. The intelligent matching method of policy resources according to claim 7, characterized in that: The strategy adjustment and optimization further includes the steps of: When the strategy adjustment judgment data is adjustment required, the user inputs the matching strategy adjustment text information for the target matching strategy data online through the natural language processing interface, and generates the optimized matching strategy data.

9. An intelligent policy resource matching system that executes the intelligent policy resource matching method according to any one of claims 1 to 8, characterized in that: The system includes a declared entity attribute collection module, a matching strategy generation module, a matching result feedback module, and a strategy optimization module. The declared entity attribute collection module is used to input enterprise attribute information and policy clause text information online, and generate declared entity attribute data. The matching strategy generation module is used to perform specific matching strategy identification processing based on the declared entity attribute data and the policy clause text data, and generate target matching strategy data through data identification. The matching result feedback module executes the matching result based on the declared entity attribute data, the policy clause text data, and the target matching strategy data, and hierarchically feeds back and displays it through the visual interface. The strategy optimization module is used to perform a strategy optimization operation on the target matching strategy data and generate the optimized matching strategy data when the user feedback is adjustment required.

10. The intelligent policy resource matching system according to claim 9, wherein: The declared entity attribute collection module includes a declared entity information input unit, a policy clause information input unit, a policy constraint condition storage unit, and a policy suitability judgment unit; the matching result feedback module includes a matching result feedback unit, a user feedback opinion collection unit, a strategy adjustment judgment unit, and a matching strategy optimization unit.

Images