Policy information management system

Abstract

The invention relates to the technical field of information management, and provides a policy information management system, which comprises a decision support module used for performing multi-scene simulation deduction based on historical policy implementation data, historical adjustment strategies and current deduction conditions in a pre-stored policy library, and generating a plurality of new policies according to deduction results; the policy issuing and pushing module is used for performing new policy matching on the generated multiple indexes and user tags and pushing matched new policies to the corresponding users; the implementation process monitoring module is used for carrying out real-time monitoring and display; and the policy evaluation module is used for evaluating the implementation condition of the new policy based on the called external related data and a preset evaluation index, generating a new adjustment policy according to an evaluation result and feeding back the new adjustment policy to the decision support module. Through the synergistic effect of each module and the policy evaluation module, the full-process closed-loop unified intelligent management of policies from seeking, issuing, implementation to evaluation is realized, and the policy information management efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of information management, and in particular to a policy information management system. Background Art

[0002] As governments at all levels demand greater efficiency and transparency in policy management, the digitalization and intelligentization of full-process policy management have become crucial for enhancing government governance capabilities. Complete and efficient full-process policy management encompasses not only policy formulation, release, application, review, implementation, and evaluation, but also requires unified information collection, real-time monitoring, and scientific evaluation to ensure fair, impartial, and efficient policy implementation and facilitate the successful achievement of policy objectives.

[0003] However, policy information management in related technologies, such as information collation and analysis, generally relies on manual operations, resulting in fragmented distribution of policy information files, irregular archiving and management, and difficulties in querying, tracing, and analyzing them. This manual information management method relies too much on the subjectivity of operators when performing information collation and analysis, which is prone to errors and inefficient. Furthermore, new policies are generally released in a comprehensive manner before implementation. If users lack attention to relevant information, this can lead to problems in the implementation of the new policy, as well as lags in process monitoring during the implementation of the new policy, which in turn affects the efficiency and effectiveness of the new policy implementation. Summary of the Invention

[0004] The technical problem to be solved by the present invention is how to improve the efficiency of policy information management.

[0005] To solve the above problems, the present invention provides a policy information management system, comprising: A decision support module is configured to perform multi-scenario simulations based on historical policy implementation data, historical adjustment strategies, and current simulation conditions stored in a pre-stored policy library, and generate multiple new policies based on the simulation results, wherein the historical adjustment strategies are derived from the historical policy implementation data, and the current simulation conditions include user information for the current period; a policy publishing and pushing module, configured to store the plurality of new policies in the policy library and generate a plurality of indexes corresponding to each of the new policies, match the new policies based on the similarity between the indexes and user tags, and push the matching new policies to corresponding users, wherein the user tags are determined based on the user's behavioral data, attribute characteristics, and / or custom tags input by the user; An implementation process monitoring module is used to monitor and display the stage process of the user executing the new policy in real time, and to archive the generated real-time policy implementation data to the policy library; The policy evaluation module is used to evaluate the implementation of the new policy based on the retrieved external relevant data, preset evaluation indicators and real-time policy implementation data, generate a new adjustment strategy according to the evaluation results, and feed the new adjustment strategy back to the decision support module.

[0006] Optionally, the multi-scenario simulation based on historical policy implementation data, historical adjustment strategies and current simulation conditions in a pre-stored policy library includes: Based on a natural language processing model, label extraction is performed on the historical policy implementation data to obtain multiple historical labels; Using a Cartesian product combination algorithm, combining multiple historical tags according to the types of the historical tags to generate multiple pre-deduction policies; Using a pre-built execution deduction model, based on the current deduction conditions, and in accordance with the plurality of pre-deduction policies, single policy deduction or combined policy deduction is performed to obtain a plurality of deduction results; Evaluate the multiple deduction results, and obtain the pre-deduction policy with the best evaluation result to obtain the new policy.

[0007] Optionally, evaluating the plurality of deduction results and obtaining the pre-deduction policy with the best evaluation result to obtain the new policy includes: Classify and parameterize the deduction results according to preset division rules; The pre-deduction policy corresponding to the deduction result of the parameterized representation that meets the preset combination standard is retrieved, and a pre-trained integration model is used to rationally combine multiple pre-deduction policies to obtain the new policy, wherein the integration model is obtained by training a pre-constructed sorting model using the historical implementation data.

[0008] Optionally, matching the new policy based on the similarity between the index and the user tag, and pushing the matched new policy to the corresponding user, includes: Acquire the user whose similarity between the user tag and the index in the new policy is a preset maximum value, and push the corresponding new policy to the user; Alternatively, a weight is assigned to each of the indexes, a similarity score is performed on each of the user tags of the user according to the index, a final score of the user corresponding to the new policy is obtained according to the weight corresponding to each index and the similarity score, and the corresponding new policy is pushed to the user whose final score is greater than a preset value.

[0009] Optionally, the policy publishing and pushing module is further used to: The policy clause adjustment content related to the new policy and / or new policies of the same type as the new policy and / or policy declaration time limit information will be pushed to the corresponding users through the reserved contact channels, where the reserved contact channels include SMS, email, and in-site messages.

[0010] Optionally, the evaluating the implementation of the new policy based on the retrieved external relevant data, preset evaluation indicators and real-time policy implementation data includes: Based on the historical policy implementation data, the historical policy, the new policy and the real-time policy implementation data, a time series prediction algorithm and a comparative analysis model are used to analyze the implementation of the new policy to obtain influencing factors; Based on the implementation objectives of the new policy, weighting the evaluation indicators; Obtaining a real-time score based on the external relevant data, the real-time policy implementation data, and the weight corresponding to each evaluation indicator; Obtaining a standard score based on the deduction results of the new policy and the weight corresponding to each evaluation indicator; Optionally, obtaining a real-time score based on the external relevant data, the real-time policy implementation data, and a weight corresponding to each evaluation indicator includes: Using a Min-Max normalization algorithm, the real-time score is obtained based on the expected target data of the new policy, the external relevant data, and the real-time policy implementation data; The standard score is obtained based on the deduction results of the new policy and the weight corresponding to each evaluation indicator, including: The standard score is obtained according to the deduction result of the new policy using the Min-Max normalization algorithm.

[0011] Optionally, generating a new adjustment strategy according to the evaluation result includes: Based on the influencing factors, the real-time scores, the standard scores and the real-time policy implementation data, a pre-trained strategy generation model is used to obtain the new adjustment strategy, wherein the strategy generation model is obtained by training a pre-constructed generation model using historical real-time scores, historical influencing factors, historical standard scores and the historical adjustment strategies.

[0012] Optionally, the weighting of the evaluation indicators based on the execution objectives of the new policy includes: According to the execution cycle of the new policy, the weight division of the evaluation indicators is dynamically adjusted based on the real-time policy implementation data and the execution goals in the current cycle using the trained reinforcement learning model.

[0013] Optionally, the real-time monitoring and display of the stage process of the user executing the new policy includes: Detailed description of online application materials, collection of said online application materials, total amount monitoring based on said real-time policy implementation data, progress tracking, fund disbursement monitoring, application time warning, application data warning and fund disbursement warning, and visual display.

[0014] The beneficial effects of the policy information management system of the present invention are: A decision-making support module uses a policy database compiled from historical policy implementation data to simulate multiple scenarios based on current conditions. New policies are derived based on these simulation results, eliminating biases in policymaking caused by subjective judgment. This improves the scientific and rational nature of new policymaking, avoids subjective blindness, and enhances the intelligence of policymaking. Fully electronic management enhances the efficiency and transparency of policy design. A policy release and push module establishes a unified policy release platform and structured management, enabling standardized collection and centralized management of policy documents and ensuring the authority and consistency of policy content. An index of new policies improves the efficiency of policy information retrieval. Intelligent policy matching and automatic push based on the similarity between user tags and indexes ensures precise reach of policy information, reducing the time and effort required for manual screening and the lack of user attention caused by comprehensive releases in related technologies. This improves the efficiency of policy information dissemination and service accuracy, ensuring effective coverage. An implementation process monitoring module digitizes and unifies the policy implementation process through online and unified management, eliminating offline application steps and the circulation of paper materials, reducing manual intervention and the risk of error, and improving policy implementation efficiency. A policy evaluation module is set up, and a gradient boosting tree model is used to collect multi-dimensional data based on external relevant data. The evaluation indicators and real-time policy implementation data are combined to realize a scientific and intelligent evaluation of the implementation of new policies, replacing the traditional subjective evaluation, and providing a scientific basis for subsequent policy adjustments. At the same time, the strategy is adjusted according to the evaluation results and fed back to the decision support module to form a closed loop of policy management, which facilitates the continuous optimization and improvement of policies, improves the scientificity and rationality of policy formulation, and promotes the smooth realization of policy goals. The policy information management system of this embodiment, through the synergy of the decision support module, the policy release and push module, the implementation process monitoring module and the policy evaluation module, realizes the closed-loop unified intelligent management of the entire process of policy from planning, promulgation, implementation to evaluation, improves the scientificity and rationality of policy information management, and improves the efficiency of policy implementation and policy information management. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 Schematic diagram of the structure of the policy information management system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0016] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings. Although certain embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as being limited to the embodiments described herein. Instead, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0017] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0018] The term "including" and its variations used in this document are open inclusions, that is, "including but not limited to"; the term "based on" means "based at least in part on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one other embodiment"; the term "some embodiments" means "at least some embodiments"; the term "optionally" means "optional embodiments". The relevant definitions of other terms will be given in the following description. It should be noted that the concepts of "first", "second", etc. mentioned in the present invention are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0019] It should be noted that the modifications of "one" and "multiple" mentioned in the present invention are illustrative rather than restrictive. Those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0020] The names of the messages or information exchanged between multiple devices in the embodiments of the present invention are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0021] It is understood that any part of this application related to data acquisition or collection has been authorized by the user.

[0022] like Figure 1 As shown, an embodiment of the present invention provides a policy information management system, including: The decision support module is used to perform multi-scenario simulation and deduction based on the historical policy implementation data, historical adjustment strategies and current deduction conditions in the pre-stored policy library, and generate multiple new policies according to the deduction results, wherein the historical adjustment strategies are obtained according to the historical policy implementation data, and the current deduction conditions include user information of the current time period.

[0023] Specifically, in the early policy planning stage, a decision-making support module is used to establish a policy clause formulation process. First, the historical policy implementation data of previous years are sorted and classified to form a classification storage with dimensions including implementation objects (individuals / enterprises), implementation standards, implementation effects, coverage, and fund usage to form a structured database. For example, the policy subsidy data of previous years are integrated, including personal policies: focusing on conditions such as education, age, and social security payments; corporate policies: covering industry types, revenue scale, honorary qualifications and other information. Relying on the directory management function of the policy knowledge base, a directory system is built according to the "superior policy-municipal policy-district policy" level to support real-time updating, adjustment and deletion of invalid policies, and upload through the document management function. The data is structured by ensuring data integrity and including attachments to historical policy implementation data, including original policy texts, subsidy disbursement records, and effectiveness evaluation reports. This data can be presented in a structured database. Trends and distribution patterns in historical policy implementation data can also be visualized through various chart formats, such as line and bar graphs. For example, a bar graph can display the number of times frequently used historical policies are used, a pie chart can show the proportion of each type of policy, and a line graph or geographic information map can be used to display the number and coverage of policies by department, along with their issuance and impact areas. This facilitates clear understanding of complex data and allows for quick identification of key issues and trends. Multiple scenarios are then simulated for different policy options, predicting their coverage, funding requirements, and potential effects to ensure comprehensiveness and feasibility. Historical adjustment strategies are retrieved based on historical policy implementation data. For example, if historical data indicates a low number of applications for graduate student subsidies in 2020, the historical adjustment strategy for 2021 is to increase graduate student subsidies to 3,000 yuan per month. The current simulation is conditioned on user information for the current period, incorporating the latest social realities into policy formulation to ensure that the resulting new policies are more aligned with current realities. For example, simulation and deduction are conducted for user screening for subsidy distribution. Users may include individual users and corporate users. The current deduction conditions for individual users (i.e., personal user information) may include personal basic information, social security, education and other valid conditions information for the current period, thereby screening beneficiaries who meet specific policy conditions, and efficiently screening out the number of beneficiaries who meet individual policy conditions; the current deduction conditions for corporate users (i.e., corporate user information) may include valid corporate data such as corporate basic information, honorary qualifications, financial report revenue, and the industry to which the company belongs, thereby accurately screening out companies that meet policy conditions, and thus screening out the number of beneficiaries who meet the policy that benefits enterprises.Based on historical policy texts and historical adjustment strategies from historical implementation data, random or intentional matching is performed against current simulation conditions. Pre-built and trained simulation models, for example, are used to conduct multi-scenario simulations. After each simulation, relevant simulation data is recorded. Simulation records are generated for system results obtained after operation screening, and key data from different simulation results is stored to ensure traceability of all simulation processes. Visualization tools (such as bar charts and pie charts) can also be used to display simulation results in real time. The simulation model can be trained using historical policy implementation data, such as an estimate model or LSTM model. The final screening result is the best. For example, the policy combination corresponding to the highest proportion of users who are high-tech enterprises (the relevant data of users who are high-tech enterprises can be directly retrieved according to the associated platform) generates a new policy. Among them, the new policy is not a combination of the entire historical policy, but a combination of partial policy information in multiple different historical policies. For example, historical policy A "2020 first-level enterprise subsidies of 100,000 yuan", historical policy B "2021 second-level enterprise subsidies of 200,000 yuan", combine part of the policy information "first-level enterprises" in historical policy A and part of the policy information "subsidy of 200,000 yuan" in historical policy B with the current year, and generate the policy "2025 first-level enterprise subsidies of 200,000 yuan". After simulation and deduction, the output deduction result "Subsidy for enterprise R&D investment to increase by 25%" is the best, and then the new policy "2025 first-level enterprise subsidies of 200,000 yuan" is generated. It should be noted that after the new policy is generated, the standardized policy clause text and funding estimate report, etc., will be electronically signed to ensure the openness and transparency of the approval process, realizing full electronic management from policy design to financial approval.

[0024] It should be noted that due to the different purposes of implementing new policies, the types of current implementation conditions will also change accordingly. For example, in this embodiment, the implementation is based on the subsidy issuance policy. Therefore, relevant economic data or relevant condition data of users will be retrieved, such as GDP, honorary qualifications, financial report revenue, industry ranking, education level, whether they are high-tech enterprises, etc., to screen users who meet the policy support requirements. Correspondingly, the types of current implementation conditions can be adaptively adjusted according to the purpose of implementing new policies. It should be noted that in order to cooperate with the implementation of new policies, the decision support module cooperates with relevant platforms to obtain the corresponding current implementation conditions.

[0025] A policy publishing and push module is used to store multiple new policies in the policy library and generate multiple indexes corresponding to each new policy, match the new policies based on the similarity between the index and the user tag, and push the matched new policies to the corresponding users, wherein the user tag is determined based on the user's behavioral data, attribute characteristics and / or custom tags entered by the user.

[0026] Specifically, in the mid-term policy release stage, the policy release and push module improves the efficiency of dissemination of new policy information and the accuracy of services by building a unified policy release platform and an efficient matching mechanism. First, the text of the new policy is structured and managed, and key elements such as policy content (policy name, issuing agency, release date, validity period), applicable objects (enterprises / individuals), application conditions (such as revenue scale, educational requirements), subsidy standards (amount / ratio), handling procedures (application channels, list of materials), labeling elements (associated with "financial support", "talent incentives", "scientific and technological innovation support" and other system labels and custom labels (such as "small and micro enterprises", "high-tech enterprises"), etc. are extracted using natural language processing models, and indexes are established according to unified rules to facilitate subsequent retrieval and matching, so as to achieve standardized collection and dynamic index management of policy documents. Among them, index construction can also adopt a multi-level labeling system and semantic analysis technology to ensure the accurate extraction and association of policy elements. On this basis, the system matches users with new policies based on user tags, such as calculating the similarity between the new policy and the user tag to determine the degree of match between the user and the new policy, and then displays relevant new policies according to priority. When a user enters the keyword "R&D subsidy", it quickly obtains and locates all new policies and / or historical policies with a similarity to the "science and technology innovation" and "subsidy" labels that is above the preset value, helping users quickly locate applicable projects. At the same time, it also provides an automatic push function, which uses existing user tags, such as user behavior data in historical policy implementation data (the behavior data of corporate users is historical policy application records, fund redemption status, and policy consultation frequency; the behavior data of individual users is policy query history and consultation records) and attribute characteristics (the attribute characteristics of corporate users are enterprise scale, number of employees, revenue, industry type, registered address, high-tech enterprise certification, number of patents, and honorary qualifications; the attribute characteristics of individual users are education level, age, social security payment location, employment status, professional qualification certificates, and awards) to match with new policies, and actively push the new policies with the highest matching degree to users, so as to achieve accurate delivery of policy information. The policy release and push module supports users to actively query and also ensures effective coverage of policy information.

[0027] The implementation process monitoring module is used to monitor and display the stage process of the user executing the new policy in real time, and archive the generated real-time policy implementation data to the policy library.

[0028] Specifically, during the mid-term policy implementation phase, the implementation process monitoring module digitizes the policy implementation process, improving efficiency and oversight capabilities. It also fully digitizes processes such as application notification issuance, document submission, review processing, and fund disbursement, enabling a "one-stop" approach to the entire process. Application material acquisition is processed online, allowing users to submit relevant information through an online platform. Each stage of the new policy implementation process is monitored in real time. For example, application deadlines are monitored, and if a user fails to submit their documents by the set deadline, a deadline reminder is sent. Detailed information about each process is also displayed, such as a visual monitoring interface that dynamically displays information such as the total number of applications, review progress, pending tasks, and fund disbursement status. Policy implementation data, such as the application, approval, and disbursement records of each subsidy, is fully archived and stored in the policy database, ensuring traceability and verifiable results. It is important to note that the implementation process monitoring module can also connect with external systems such as finance and banks to enable automated fund disbursement and receipt feedback, improving disbursement efficiency while ensuring fund security and transparency.

[0029] The policy evaluation module is used to evaluate the implementation of the new policy based on the retrieved external relevant data, preset evaluation indicators and real-time policy implementation data, generate a new adjustment strategy according to the evaluation results, and feed the new adjustment strategy back to the decision support module.

[0030] Specifically, in the later policy performance evaluation phase, multi-dimensional data collection and analysis are used to scientifically assess the effectiveness of new policy implementation and generate optimization recommendations. Policy implementation data from the new policy implementation process, such as various business data, including basic information such as subsidy recipients, amounts, and timeframes, is compiled into a list of policy subsidy recipients. This data is then analyzed using a gradient boosting tree model, combining relevant historical policy implementation data with current external data such as business operating data, social security contributions, and GDP, as well as real-time policy implementation data generated by the new policy. For example, the model evaluates the factors influencing real-time policy implementation data to form a comprehensive policy performance evaluation. For example, the 2024 "R&D subsidy" funding will drive a 15% increase in R&D investment by target users compared to 2023. Furthermore, based on the evaluation results, new policy adjustment strategies are generated, such as increasing the subsidy amount for small and medium-sized enterprises (SMEs) to increase their application rate. These policy adjustment strategies are then fed back to the decision support module as a key reference for the next round of new policy formulation. The evaluation results can be compiled into a structured report, including basic information such as a policy overview (goals, implementation period, funding scale), data collection scope, and evaluation methodology; evaluation information such as the completion of core indicators, trend analysis, and problem identification (e.g., "the application rate for SMEs is only 30%, lower than the 50% for large enterprises"); new adjustment strategies such as policy adjustment suggestions (e.g., "expanding the scope of relaxed application conditions for SMEs") and optimization directions for the next phase. This provides data support for policy adjustments.

[0031] In this embodiment, a decision support module is provided. Based on a policy library obtained by sorting and classifying historical policy implementation data from previous years, a multi-scenario simulation is performed based on current deduction conditions. New policies are derived based on the deduction results, avoiding bias in policy formulation caused by human subjective judgment, improving the scientificity and rationality of new policy formulation, avoiding subjective blindness, and enhancing the intelligence of policy formulation. Furthermore, the full electronic management process improves the efficiency and transparency of policy design. A policy release and push module is provided to build a unified policy release platform and structured management, achieving standardized collection and centralized management of policy documents, ensuring the authority and consistency of policy content; extracting indexes for new policies improves the efficiency of policy information retrieval; intelligent policy matching and automatic push based on the similarity between user tags and indexes achieves accurate access to policy information, reducing the time and effort consumed by users in manual screening and the lack of user attention caused by comprehensive release in related technologies, improving the efficiency of policy information dissemination and service accuracy, and ensuring effective coverage of policy information. An implementation process monitoring module is provided to digitize and unify the management of the policy implementation process online, reducing offline application links and the circulation of paper materials, reducing manual intervention and the risk of error, and improving policy implementation efficiency. A policy evaluation module is set up, and a gradient boosting tree model is used to collect multi-dimensional data based on external relevant data. The evaluation indicators and real-time policy implementation data are combined to realize a scientific and intelligent evaluation of the implementation of new policies, replacing the traditional subjective evaluation, and providing a scientific basis for subsequent policy adjustments. At the same time, the strategy is adjusted according to the evaluation results and fed back to the decision support module to form a closed loop of policy management, which facilitates the continuous optimization and improvement of policies, improves the scientificity and rationality of policy formulation, and promotes the smooth realization of policy goals. The policy information management system of this embodiment, through the synergy of the decision support module, the policy release and push module, the implementation process monitoring module and the policy evaluation module, realizes the closed-loop unified intelligent management of the entire process of policy from planning, promulgation, implementation to evaluation, improves the scientificity and rationality of policy information management, and improves the efficiency of policy implementation and policy information management.

[0032] Optionally, the multi-scenario simulation is performed based on historical policy implementation data, historical adjustment strategies, and current simulation conditions in a pre-stored policy library, and multiple new policies are generated according to the simulation results, including: Based on a natural language processing model, labels are extracted from the historical policy implementation data to obtain multiple historical labels.

[0033] Specifically, a natural language processing model is trained using a subset of annotated policies to obtain a trained natural language processing model for policy label extraction. The trained natural language processing model is then used to label historical policy implementation data. This involves breaking down policy elements and extracting them into corresponding historical labels, facilitating the subsequent reconstruction and combination of these labels. It should be noted that historical policy implementation data can be textual data, such as historical policy texts, or digital data representing policy implementation results, such as the number of 200 historical policy applications. However, the labeling process is performed on textual data. Historical labels can be system labels, which include specific conditions (e.g., "high-tech enterprise" and "Master's degree or above"), which are pre-defined in areas such as "financial support," "talent incentives," and "scientific and technological innovation support." Furthermore, historical labels can include custom labels, which are added based on business department needs (e.g., "affected enterprises" and "employees in remote areas"). Logical rules (e.g., "enterprise revenue decreased by ≥20% year-on-year and the number of employees was ≤50") can be configured using the conditional labeling function.

[0034] A Cartesian product combination algorithm is used to combine multiple historical tags according to the types of the historical tags to generate multiple pre-deduction policies.

[0035] Specifically, for different policy directions (such as subsidy amount adjustments and coverage expansion), a Cartesian product combination algorithm is used to combine historical labels to generate new pre-simulation policies for policy simulations targeting different objectives. When using the Cartesian product combination algorithm to combine labels, the type of historical labels should be considered. For example, if each pre-simulation policy contains three types of labels—attributes, solutions, and specific values—then the label combination should ensure that each pre-simulation policy contains only one historical label of each type. Furthermore, before conducting a policy simulation, simulation conditions must be set, including condition annotation, condition comparison, and condition management. Condition annotation can be done using a label management tool. By specifying new label names and corresponding logic, the requirements are communicated to relevant staff, who then configure the labels. After the simulation is complete, users can select different simulation results and visually compare the differences between the filter conditions and the resulting data under different results, enabling comparison of simulation conditions. Deduction condition management includes setting, adding, updating and deleting the conditions required for the deduction through the operation page, including but not limited to the management of non-custom conditions such as education, family, age, etc. According to relevant standards and policy simulation results, it supports the simulation calculation of policy redemption funds, and records and saves the simulation results to form a query list.

[0036] Utilizing the pre-built execution deduction model, based on the current deduction conditions, single policy deduction or combined policy deduction is performed according to the plurality of pre-deduction policies to obtain a plurality of deduction results.

[0037] Specifically, the execution prediction model can be trained, for example, using historical policy implementation data, based on current prediction conditions. Single-policy prediction involves inputting specific conditions (e.g., "10% subsidy for R&D investment for small and micro enterprises") into the execution prediction model to obtain information such as the number of benefited enterprises, the amount of funding required, and the expected growth rate of R&D investment. Combined policy prediction involves inputting multiple policy scenarios (e.g., "talent subsidies plus tax exemptions") into the execution prediction model to identify the affected enterprise groups and obtain data such as the industry distribution of covered enterprises, total funding needs, and employment impact. This data can then be used to generate a "Policy Combination Effect Forecast Report."

[0038] Evaluate the multiple deduction results, and obtain the pre-deduction policy with the best evaluation result to obtain the new policy.

[0039] Specifically, after obtaining multiple simulation results, a visual comparison of the simulation results of different scenarios is performed to evaluate and analyze the results. Visualization software, such as tables and charts, can be used to display the simulation results, including core indicators such as the number of people covered, capital costs, and input-output. Line charts can also be used to compare the enterprise impact rates under different subsidy policies (e.g., the difference in application volume for a 10% vs. 15% subsidy). The pre-simulated policy corresponding to the optimal simulation result (e.g., a line chart confirming that pre-simulated policy A has the highest enterprise impact rate) is then identified as the new policy. Alternatively, individual parameters can be optimized based on the visualization results to improve the overall implementation of the new policy.

[0040] Optionally, evaluating the plurality of deduction results and obtaining the pre-deduction policy with the best evaluation result to obtain the new policy includes: Classify and parameterize the deduction results according to preset division rules; The pre-deduction policy corresponding to the deduction result of the parameterized representation that meets the preset combination standard is retrieved, and a pre-trained integration model is used to rationally combine multiple pre-deduction policies to obtain the new policy, wherein the integration model is obtained by training a pre-constructed sorting model using the historical implementation data.

[0041] Specifically, first obtain the division rules, for example, divide the target users of subsidy distribution into high benefit, medium benefit and low benefit according to their turnover, and make a unified parameterized representation of the deduction results based on the division rules. For example, when the turnover of the target users of subsidy distribution reaches 10 million yuan, the deduction results are parameterized as "high benefit". Automatic sorting can be performed based on data analysis tools to recommend parameter combinations of "high coverage, low cost and high benefit". In order to avoid the irrationality of the new policy generated by simply combining the pre-deduction policies corresponding to each parameter combination, the corresponding pre-deduction policies are input into the pre-trained integration model for combination processing to obtain the new policy, so as to improve the scientificity and rationality of the new policy generation. Among them, the pre-built sorting model is trained by the historical implementation data of the integration model. The sorting model can be a Transformer model, a recurrent neural network, etc.

[0042] In addition, after obtaining the deduction results, the policy clause text can also be generated based on the optimal deduction results obtained from the evaluation, including policy goals (such as "driving a 20% increase in R&D investment of high-tech enterprises in 2025"); application conditions (based on the conversion of labeled conditions into natural language, such as "must meet the relevant requirements of the "High-tech Enterprise Certification and Management Measures"); subsidy standards (such as "subsidies will be given at 12% of actual R&D expenses, with an annual upper limit of 500,000 yuan per enterprise"); implementation process (explanation of the entire process of application-review-publicity-redemption); expected distribution of redemption funds in each district and county (with a geographic information map); quarterly capital expenditure plan; risk reserve provision ratio (reserved at 10% of the total budget), etc.

[0043] Optionally, matching the new policy based on the similarity between the index and the user tag, and pushing the matched new policy to the corresponding user, includes: Acquire the user whose user tag satisfies all the indexes in the new policy, and push the corresponding new policy to the user; Alternatively, a weight is assigned to each of the indexes, a similarity score is performed on each of the user tags of the user according to the index, a final score of the user corresponding to the new policy is obtained according to the weight corresponding to each index and the similarity score, and the corresponding new policy is pushed to the user whose final score is greater than a preset value.

[0044] Specifically, the matching of users and new policies may include hard condition matching or soft condition matching. Hard condition matching means that the applicant user must meet all the application conditions of the new policy before the new policy can be searched or the new policy can be pushed to the applicant user, that is, the similarity between the user tag and the index in the new policy is 1. Soft condition matching means that a comprehensive scoring and rating can be performed based on the applicant user's conditions and the relevant information of the new policy. When its level reaches the preset application level, the new policy can be searched or pushed to the applicant user. It can be understood that even if some of the applicant user's information does not meet the content of the new policy, after the applicant user is scored, it is considered that the applicant user belongs to the key assistance target of the new policy, and the new policy will be pushed to the applicant user. When performing hard condition matching, if the similarity between the user's user tag and the index in the new policy is 1, it is considered that the user meets the hard conditions, and the corresponding new policy will be pushed to the user when the user searches for his own user tag on the relevant platform. When performing soft condition matching, we first assign weights to each index in the new policy, that is, clarify the importance of each detailed rule in the new policy, and then perform similarity scoring on the index and user tag, that is, calculate the similarity between the index and the user tag, and then integrate the corresponding similarity score with the weight of the corresponding index to obtain the final score of the new policy for the user. When the user searches for his own user tags on the relevant platform or actively pushes to the user a user with a score greater than the preset value, such as a high-tech enterprise (user tags are "science and technology innovation" and "high-tech enterprise"), priority will be given to displaying or pushing new policies with a score higher than 0.8 points, such as "additional deduction for R&D expenses" and "subsidy for high-tech enterprise recognition" to the user; talents with a master's degree (user tags are "talent" and "master") will be given priority to push new policies with a score higher than 0.8 points, such as "talent apartment application" and "education improvement subsidy".

[0045] Alternatively, after obtaining the final scores of users corresponding to the new policy, the final scores are sorted in descending order to obtain a recommendation list, so as to display the priority-sorted policy recommendation list to the user.

[0046] Additionally, matching results can be displayed in tables or charts. For example, a matching radar chart can show the degree of fit between user tags and policy tags (such as company size, industry, and qualification matching scores). A condition list comparison can list all policy requirements, marking those that the user has met (green) and those that have not (red), and provide correction suggestions (such as "It is recommended to apply for high-tech enterprise certification to meet qualification requirements"). Application guidance can also be provided, such as the automatically generated "Policy Application Guide," which includes key information such as the list of materials, application entry point, and deadline. This improves the convenience of user policy applications.

[0047] Optionally, the policy publishing and pushing module is further used to: The policy clause adjustment content related to the new policy and / or new policies of the same type as the new policy and / or policy declaration time limit information will be pushed to the corresponding users through the reserved contact channels, where the reserved contact channels include SMS and email.

[0048] Specifically, in subsequent policy updates, if the policy clauses related to the new policy and / or new policies of the same type as the new policy (such as the newly added "Cultural Industry Support Policy", which pushes notifications to companies registered in the Cultural Industry Park), are also pushed to users who have previously applied for the policy, so that users can understand the policy changes and achieve comprehensive policy coverage. Alternatively, after the user applies for a new policy, the relevant policy application time limit information will be pushed to the user's bound terminal so that the user can understand the progress of the policy application. Users can reserve contact information when searching for policies or applying for policies, which can be a contact number, email address, etc. When pushing subsequent policies, the reserved contact channels such as SMS, email or phone can be used for push.

[0049] Optionally, the evaluating the implementation of the new policy based on the retrieved external relevant data, preset evaluation indicators and real-time policy implementation data includes: Based on the historical policy implementation data, the historical policy, the new policy and the real-time policy implementation data, a time series prediction algorithm and a comparative analysis model are used to analyze the implementation of the new policy to obtain influencing factors.

[0050] Specifically, long short-term memory networks (LSTMs) are used to process time series data, capturing the lag and cyclical characteristics of policy implementation data (e.g., "peak effect of enterprise revenue growth three months after subsidy issuance" and "seasonal fluctuations in annual policy effectiveness"). Decision tree models, such as the XGBoost model, are used to rank the importance of features in historical and real-time policy implementation data, identifying key factors influencing policy effectiveness (e.g., "enterprise size is more sensitive to R&D subsidies than industry type"). K-means clustering algorithms are used to group historical and similar real-time policy implementation data, locating differences (e.g., "the enterprise benefit rate from talent subsidies in 2024 increased by 12% compared to 2022, but regional coverage balance decreased by 5%"). Attribution analysis is conducted based on the resulting data, using the propensity score matching (PSM) algorithm to separate the impact of policy factors from the external environment (e.g., market fluctuations), quantifying the actual contribution of the policy (e.g., "after excluding industry growth factors, the policy's net pull on enterprise R&D investment was 18%)." Ultimately, the factors influencing policy implementation data are identified.

[0051] Based on the implementation objectives of the new policy, the evaluation indicators are weighted.

[0052] A real-time score is obtained according to the external relevant data, the real-time policy implementation data and the weight corresponding to each evaluation indicator.

[0053] Specifically, a real-time score of the current implementation of the new policy is obtained by scoring based on the external relevant data obtained (such as funding coverage rate, enterprise benefit rate, regional balance), real-time policy implementation data and corresponding weights.

[0054] A standard score is obtained based on the deduction results of the new policy and the weight corresponding to each evaluation indicator. The standard score of the new policy under ideal conditions during the deduction simulation is obtained, wherein the deduction results include corresponding external related data and policy implementation data.

[0055] A new adjustment strategy is obtained based on the real-time score, the standard score and the influencing factors.

[0056] Specifically, the standard score represents the policy implementation effect of the new policy under ideal conditions, the real-time score represents the policy implementation effect of the new policy under actual conditions, and the influencing factors represent the influencing factors that cause the difference between the real-time effect and the expected effect under ideal conditions when the policy was implemented in previous years. Therefore, based on the standard score, the influencing factors and the real-time score are combined to obtain a new adjustment strategy. After adjusting the newly added side according to the new adjustment strategy, the policy implementation effect under actual conditions obtained in the next time period will be closer to the policy implementation effect under ideal conditions, thereby realizing real-time adjustment of the new policy execution effect. For example, a pre-trained policy generation model can be used to obtain a new adjustment strategy, wherein the policy generation model uses historical real-time scores, historical influencing factors, historical standard scores, and historical adjustment strategies to train a pre-constructed generation model.

[0057] The real-time scoring is obtained based on the external relevant data, the real-time policy implementation data, and the weight corresponding to each evaluation indicator, including: Using a Min-Max normalization algorithm, the real-time score is obtained based on the expected target data of the new policy, the external relevant data, and the real-time policy implementation data; The standard score is obtained based on the deduction results of the new policy and the weight corresponding to each evaluation indicator, including: The standard score is obtained according to the deduction result of the new policy using the Min-Max normalization algorithm.

[0058] Specifically, the expected target data is the target that the new policy hopes to achieve, for example, the new policy's expected declaration rate is 70%. To ensure the uniformity of the scores, the Min-Max normalization algorithm is used to calculate the scores.

[0059] Optionally, generating a new adjustment strategy according to the evaluation result includes: Based on the influencing factors, the real-time scores, the standard scores and the real-time policy implementation data, a pre-trained strategy generation model is used to obtain the new adjustment strategy, wherein the strategy generation model is obtained by training a pre-constructed generation model using historical real-time scores, historical influencing factors, historical standard scores and the historical adjustment strategies.

[0060] Specifically, the generative model can be constructed using models such as convolutional neural networks, recurrent neural networks, etc. In addition, expert experience can be used to analyze influencing factors to obtain new adjustment strategies.

[0061] Optionally, policy effectiveness analysis and evaluation may also include: The corresponding external related data is retrieved according to the evaluation index.

[0062] Specifically, evaluation indicators represent attributes related to the expected effects of new policy implementation. These include mandatory indicators such as "subsidy fund utilization rate" and "number of benefited enterprises," as well as customized indicators tailored to policy characteristics, such as "growth rate of R&D investment by high-tech enterprises" and "talent retention rate." Evaluation indicators can include a name (such as "talent retention rate") and a calculation basis (number of talents employed for at least one year after receiving the subsidy / total number of subsidized employees × 100%). Based on these evaluation indicators, external data can be accessed through various systems. For example, business-related data can be obtained through market supervision users, including revenue, tax payments, and changes in employee numbers. Social security data can be synchronized from human resources and social security management users, including individual employment status and social security contribution bases. Economic data can be accessed through statistics users, including macroeconomic data such as regional GDP, employment rate, and industrial value added. Public opinion data can be obtained by crawling policy feedback and evaluations from channels such as government new media and enterprise survey questionnaires.

[0063] Based on the implementation objectives of the new policy, the evaluation indicators are weighted.

[0064] Specifically, since new policies can correspond to multiple evaluation indicators, the importance of corresponding evaluation indicators will vary based on the different implementation objectives of the new policies. Therefore, weighting of multiple evaluation indicators can be performed based on historical experience or using neural network models (such as a Transformer model trained with relevant data) to improve evaluation accuracy. For example, if the implementation objectives include expanding R&D investment, improving innovation output, and ensuring financial benefits, after weighting, the weight of expanding R&D investment accounts for 40%, of which the corresponding evaluation indicators R&D investment growth rate accounts for 30% and R&D subsidy application rate accounts for 10%, the weight of improving innovation output accounts for 35%, the corresponding evaluation indicators patent output efficiency accounts for 25%, high-tech product revenue accounts for 10%, and the weight of ensuring financial benefits accounts for 25%. The corresponding evaluation indicators unit subsidy-driven output value growth accounts for 15% and funding coverage rate accounts for 10%.

[0065] A comprehensive score is obtained based on the external relevant data, the real-time policy implementation data and the weight corresponding to each of the evaluation indicators.

[0066] The implementation of the new policy is graded according to the preset grade and the comprehensive score, and the implementation of the new policy below the preset average grade is analyzed to obtain influencing factors.

[0067] Specifically, scoring can be performed based on external relevant data, and a trained scoring model can be used for scoring, and then a comprehensive score for the implementation of the new policy can be obtained based on the weights of the corresponding evaluation indicators. Among them, the scoring model can be constructed by models such as convolutional neural networks and recurrent neural networks, and trained using historical policy implementation data. After obtaining the comprehensive score, the implementation of the new policy is graded according to the preset grade division and the comprehensive score. It can be divided into excellent, indicating that the policy effect is significant and good, indicating that the policy effect has reached the expected goal and is qualified, indicating that some goals need to be improved and are unqualified, indicating that the policy implementation effect is poor. After obtaining the grade division, the gradient boosting tree model can be used to analyze and evaluate the implementation of new policies that are below the preset average grade, and obtain the influencing factors that affect the policy implementation effect, so as to facilitate subsequent policy adjustments. It should be noted that the gradient boosting tree model is trained using historical policy implementation data and the corresponding historical influencing factors that are marked.

[0068] Optionally, obtaining a comprehensive score based on the external relevant data and the weight corresponding to each evaluation indicator includes: The Min-Max normalization algorithm is used to obtain a single indicator score for each evaluation indicator according to the expected target data of the new policy, the external related data and the real-time policy implementation data.

[0069] Specifically, the expected target data is the goal that the new policy expects to achieve. For example, the expected declaration rate of the new policy is 70%. External relevant data is used to calculate the evaluation indicators corresponding to the expected target data, and then the single indicator score is obtained based on the ratio of the evaluation indicators and the expected target data. In order to ensure the uniformity of the score, the Min-Max normalization algorithm is used to calculate the score.

[0070] The comprehensive score is obtained according to the weight corresponding to each evaluation indicator and the score of the single indicator.

[0071] Specifically, the individual indicator scores are weighted according to the corresponding weights of the aforementioned evaluation indicators to obtain a final comprehensive score for the new policy implementation. Additionally, the individual indicator scores and the comprehensive score can be visualized to more intuitively demonstrate the new policy implementation and compare it with historical policy implementation, thereby facilitating the subsequent generation of new adjustment strategies.

[0072] Optionally, generating a new adjustment strategy according to the evaluation result includes: Based on the influencing factors and the real-time policy implementation data, a pre-trained policy generation model is used to obtain the new adjustment strategy, wherein the policy generation model is obtained by training a pre-constructed generation model using historical policy implementation data, historical influencing factors and the historical adjustment strategy.

[0073] Optionally, the weighting of the evaluation indicators based on the execution objectives of the new policy includes: According to the execution cycle of the new policy, the weight division of the evaluation indicators is dynamically adjusted based on the real-time policy implementation data and the execution goals in the current cycle using the trained reinforcement learning model.

[0074] Specifically, since new policies have a long implementation cycle, to ensure real-time monitoring of their implementation, a trained reinforcement learning model is used to dynamically adjust and allocate the weights of evaluation indicators based on real-time performance. For example, if a new policy has an 18-month implementation cycle, with an initial implementation period of 0-6 months, a mid-implementation period of 6-12 months, and a final implementation period of 12-18 months, evaluation indicator weights are allocated for each period. The reinforcement learning model is trained using pre-annotated weights and historical policy implementation data.

[0075] Optionally, the real-time monitoring and display of the stage process of the user executing the new policy includes: Detailed description of online application materials, collection of said online application materials, total amount monitoring based on said real-time policy implementation data, progress tracking, fund disbursement monitoring, application time warning, application data warning and fund disbursement warning, and visual display.

[0076] Specifically, total amount monitoring includes the current declared amount, reviewed amount, and amount to be reviewed, etc., which can all be updated and displayed using visualizations such as bar charts. Progress tracking includes the time taken for each link, and fund disbursement progress monitoring includes the real-time disbursement of funds. Declaration time warning includes the countdown designed for the review link. For example, the initial review period is 3 working days, and a reminder is sent to the reviewer when there is 1 day left. Declaration data warning includes triggering an alarm when abnormal data occurs. For example, if the same user submits 10 declaration documents in a single day, it can be marked as a "high-frequency declaration" and paid special attention to. Fund disbursement warning includes warning of fund disbursement status. For example, when the policy fund redemption exceeds 80% of the budget, a warning will be issued to prompt adjustment of the budget or suspension of acceptance.

[0077] In addition, the full onlineization of processes such as the issuance of application notifications, submission of materials, review and processing, and fund redemption supports, for example, the dynamic combination of main and detailed forms and the configuration of field-level duplicate checking rules (based on key fields such as the unified social credit code and invoice number), enabling a "one-stop service" for the entire process, eliminating offline application links and the circulation of paper materials. All records of subsidy application, approval, and issuance are fully archived using blockchain technology to ensure traceability and verifiable results. It can also be connected to external systems through an API gateway to achieve, for example, the automatic issuance of fund disbursement instructions and a closed-loop feedback loop for receipt of funds, using encrypted transmission protocols to ensure fund security, and optimizing redemption efficiency and transparency through a whitelist mechanism and frequency control algorithm.

[0078] Although the present invention is disclosed as above, the protection scope of the present invention is not limited thereto. Those skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will fall within the protection scope of the present invention.

Claims

1. A policy information management system, characterized in that: include: A decision support module is configured to perform multi-scenario simulations based on historical policy implementation data, historical adjustment strategies, and current simulation conditions stored in a pre-stored policy library, and generate multiple new policies based on the simulation results, wherein the historical adjustment strategies are derived from the historical policy implementation data, and the current simulation conditions include user information for the current period; a policy publishing and pushing module, configured to store the plurality of new policies in the policy library and generate a plurality of indexes corresponding to each of the new policies, match the new policies based on the similarity between the indexes and user tags, and push the matching new policies to corresponding users, wherein the user tags are determined based on the user's behavioral data, attribute characteristics, and / or custom tags input by the user; An implementation process monitoring module is used to monitor and display the stage process of the user executing the new policy in real time, and to archive the generated real-time policy implementation data to the policy library; The policy evaluation module is used to evaluate the implementation of the new policy based on the retrieved external relevant data, preset evaluation indicators and real-time policy implementation data, generate a new adjustment strategy according to the evaluation results, and feed the new adjustment strategy back to the decision support module.

2. The policy information management system according to claim 1, characterized in that: The method involves performing multi-scenario simulations based on historical policy implementation data, historical adjustment strategies, and current simulation conditions in the pre-stored policy library, and generating multiple new policies based on the simulation results, including: Based on a natural language processing model, label extraction is performed on the historical policy implementation data to obtain multiple historical labels; Using a Cartesian product combination algorithm, combining multiple historical tags according to the types of the historical tags to generate multiple pre-deduction policies; Using a pre-built execution deduction model, based on the current deduction conditions, and in accordance with the plurality of pre-deduction policies, single policy deduction or combined policy deduction is performed to obtain a plurality of deduction results; Evaluate the multiple deduction results, and obtain the pre-deduction policy with the best evaluation result to obtain the new policy.

3. The policy information management system according to claim 2, characterized in that: The step of evaluating the plurality of deduction results and obtaining the pre-deduction policy with the best evaluation result to obtain the new policy includes: Classify and parameterize the deduction results according to preset division rules; The pre-deduction policy corresponding to the deduction result of the parameterized representation that meets the preset combination standard is retrieved, and a pre-trained integration model is used to rationally combine multiple pre-deduction policies to obtain the new policy, wherein the integration model is obtained by training a pre-constructed sorting model using the historical implementation data.

4. The policy information management system according to claim 1, characterized in that: The matching of the new policy based on the similarity between the index and the user tag, and pushing the matched new policy to the corresponding user, includes: Acquire the user whose similarity between the user tag and the index in the new policy is a preset maximum value, and push the corresponding new policy to the user; Alternatively, a weight is assigned to each of the indexes, a similarity score is performed on each of the user tags of the user according to the index, a final score of the user corresponding to the new policy is obtained according to the weight corresponding to each index and the similarity score, and the corresponding new policy is pushed to the user whose final score is greater than a preset value.

5. The policy information management system according to claim 1, characterized in that: The policy publishing and push module is also used to: The policy clause adjustment content related to the new policy and / or new policies of the same type as the new policy and / or policy declaration time limit information will be pushed to the corresponding users through the reserved contact channels, where the reserved contact channels include SMS, email, and in-site messages.

6. The policy information management system according to claim 1, characterized in that: The evaluation of the implementation of the new policy based on the retrieved external relevant data, preset evaluation indicators and real-time policy implementation data includes: Based on the historical policy implementation data, the historical policy, the new policy and the real-time policy implementation data, a time series prediction algorithm and a comparative analysis model are used to analyze the implementation of the new policy to obtain influencing factors; Based on the implementation objectives of the new policy, weighting the evaluation indicators; Obtaining a real-time score based on the external relevant data, the real-time policy implementation data, and the weight corresponding to each evaluation indicator; A standard score is obtained based on the deduction results of the new policy and the weight corresponding to each evaluation indicator.

7. The policy information management system according to claim 6, characterized in that: The real-time scoring is obtained based on the external relevant data, the real-time policy implementation data, and the weight corresponding to each evaluation indicator, including: Using a Min-Max normalization algorithm, the real-time score is obtained based on the expected target data of the new policy, the external relevant data, and the real-time policy implementation data; The standard score is obtained based on the deduction results of the new policy and the weight corresponding to each evaluation indicator, including: The standard score is obtained according to the deduction result of the new policy using the Min-Max normalization algorithm.

8. The policy information management system according to claim 6, characterized in that: Generating a new adjustment strategy according to the evaluation results includes: Based on the influencing factors, the real-time scores, the standard scores and the real-time policy implementation data, a pre-trained strategy generation model is used to obtain the new adjustment strategy, wherein the strategy generation model is obtained by training a pre-constructed generation model using historical real-time scores, historical influencing factors, historical standard scores and the historical adjustment strategies.

9. The policy information management system according to claim 6, characterized in that: The weighting of the evaluation indicators based on the execution objectives of the new policy includes: According to the execution cycle of the new policy, the weight division of the evaluation indicators is dynamically adjusted based on the real-time policy implementation data and the execution goals in the current cycle using the trained reinforcement learning model.

10. The policy information management system according to claim 1, characterized in that: The real-time monitoring and display of the stage process of the user implementing the new policy includes: Detailed description of online application materials, collection of said online application materials, total amount monitoring based on said real-time policy implementation data, progress tracking, fund disbursement monitoring, application time warning, application data warning and fund disbursement warning, and visual display.

Images