AI-based enterprise benefit policy intelligent matching and declaration auxiliary system

Abstract

The invention discloses an AI-based enterprise benefit policy intelligent matching and declaration auxiliary system, and relates to the technical field of policy declaration auxiliary systems. The data collection module collects policy and enterprise data in a multi-modal fusion manner, and dynamically updates according to activeness; the preprocessing module performs cleaning through a high-performance cluster, analyzes a policy through sentiment analysis and performs AES encryption; the feature extraction module uses GAN enhancement and cognitive calculation mining to update a knowledge graph; the matching calculation module calculates the matching degree in combination with a complex network, and the declaration auxiliary module demonstrates the process through VR and AR and answers questions through intelligent customer service. The risk assessment module assesses the risk by using a block chain and the like; the user interaction module supports a brain-computer interface and the like and adaptively adjusts the interface; and the data updating and feedback module records feedback by means of the block chain and optimizes a prediction mechanism. The system is advantaged in that multi-channel accurate acquisition is facilitated, enterprise matching policy bonus is facilitated, a declaration link is simplified, the cost is reduced, data authenticity is guaranteed, cooperation is promoted, continuous optimization is carried out according to feedback, and enterprise benefit policy landing is promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of policy declaration assistance systems, and particularly to an intelligent matching and declaration assistance system for enterprise-benefiting policies based on AI. Background Art

[0002] Under the current economic development pattern, in order to promote the development of enterprises and stimulate market vitality, numerous enterprise-benefiting policies have been introduced. These policies cover multiple fields such as tax incentives, financial subsidies, and project support, aiming to help enterprises reduce costs, enhance innovation capabilities, and market competitiveness. However, enterprises face many challenges in obtaining and utilizing these policies.

[0003] On the one hand, the release channels of enterprise-benefiting policies are scattered, including official websites of departments at all levels, government affairs APPs, offline notifications, etc.

[0004] On the other hand, the situations of enterprises vary. Enterprises of different scales, industries, and development stages have different policy requirements. Therefore, it is urgent to develop a system that can intelligently match enterprise-benefiting policies and provide declaration assistance. Summary of the Invention

[0005] The intelligent matching and declaration assistance system for enterprise-benefiting policies based on AI proposed by the present invention is to solve the problems mentioned in the above prior art.

[0006] To achieve the above object, the present invention adopts the following technical solutions: An intelligent matching and declaration assistance system for enterprise-benefiting policies based on AI, comprising:

[0007] Data acquisition module: Adopting a multi-modal fusion data acquisition method, obtaining enterprise-benefiting policy data from government official websites and policy release platforms through web crawlers, and using natural language processing NLP technology to parse information; collecting basic enterprise information through interface docking, and at the same time collecting voice information of enterprise leaders in meetings and interviews through speech recognition technology and converting it into text to extract data; obtaining geographical environment and infrastructure spatial data of the enterprise's location using satellite remote sensing and geographic information system GIS; introducing an enterprise activity factor β to calculate the data update cycle T, and the formula is β is calculated based on the recent business transaction frequency and employee recruitment activities of the enterprise;

[0008] Data preprocessing module: Using a computing cluster to assist in data cleaning and processing of policy and enterprise data; for policy texts, adopting a preprocessing method based on situation analysis to identify positive, negative, and neutral expressions in the policies, and using encryption algorithms for data encryption;

[0009] Feature extraction module: Construct a feature enhancement model based on the Generative Adversarial Network (GAN) to generate virtual features to enrich the feature spaces of policies and enterprises; Combine cognitive computing technology to simulate the human brain's ability to abstract and associate information, and mine the associations between policy and enterprise features. The knowledge graph construction introduces a dynamic update mechanism to automatically update the relationships and attributes in the graph according to real-time data and events;

[0010] Matching calculation module: In addition to the cosine similarity algorithm, introduce a matching algorithm based on network theory. Regard policies and enterprises as nodes in a network, and calculate the matching degree Sim network (A, B) through the connection strength C(A, B) and path length L(A, B) between nodes. The formula is Consider the time series characteristics of policies, and use time series analysis methods to predict the change in the matching degree of policies in the future, providing forward-looking matching suggestions for enterprises; The final matching degree formula is optimized to Sim total = w1×Sim′(A, B) + w2×Sim network (A, B) + w3×Sim time (A, B), where Sim time (A, B) is the matching degree considering time series, and w1, w2, and w3 are weight coefficients optimized through genetic algorithms.

[0011] Furthermore, it also includes:

[0012] Declaration assistance module: Use virtual reality and augmented reality technologies to provide enterprises with a demonstration of the declaration process; Enterprises experience the declaration process through equipment, and the equipment provides real-time guidance and prompts during actual operations; Introduce a robot customer service that realizes autonomous learning and evolution based on reinforcement learning algorithms;

[0013] Risk assessment module: Adopt a method combined with federated learning for risk assessment; The data of different enterprises are used for model training locally. Introduce chaos theory to analyze the uncertain risks in the policy declaration process, and construct a chaos model mapping model x n+1 = μx n (1 - x n ) to predict the evolution of risks, where μ is the control parameter, x n is the current risk state, and x n+1 is the risk state at the next moment;

[0014] User interaction module: Develop an interaction method based on the principle of brain-computer interface. Enterprise users can perform system operations and information queries through brain wave signals. The interface design adopts an adaptive learning algorithm to automatically adjust the interface layout and function display according to the user's usage habits;

[0015] Data Update and Feedback Module: Utilize the immutable characteristic to record the feedback information of the enterprise, introduce a predictive feedback mechanism, and based on historical feedback data and system operation conditions, predict the problems encountered by the enterprise through a time series prediction model and provide corresponding solutions.

[0016] Policy Insight Module: Use a method combining deep reinforcement learning and meta-learning to analyze and predict policies, use semantic web technology to perform semantic annotation and reasoning on policies, and mine the logic and impacts behind policies; adopt the topic model Latent Dirichlet Allocation to extract topics from policy texts, with the formula where w i is a word, z i is a topic, β is a hyperparameter, D is the number of documents, N d is the number of words in document d, V is the size of the vocabulary, and predict policy priorities and trend changes through topic analysis.

[0017] Furthermore, it also includes:

[0018] Ecological Collaboration Module: Build an enterprise-benefiting policy ecosystem involving government departments, enterprises, financial institutions, and scientific research institutions and universities; achieve data sharing and collaborative cooperation among all parties through contracts; adopt a game theory model extension model, set the cooperation payoff matrix R and the betrayal payoff matrix S, and calculate the payoffs of all parties under different strategies where p ij is the probability that the other party chooses cooperation, n is the number of participants, and optimize the cooperation strategies of all parties in the ecosystem.

[0019] Furthermore, when the data collection module collects enterprise data, it uses drone inspection technology to obtain images and video data of the enterprise's factory building appearance and equipment operation status, enrich the visualization data dimension of the enterprise, and for the obtained image and video data, use image recognition algorithms to perform object detection and analysis, and extract information such as the number of enterprise equipment, equipment models, and factory building construction conditions.

[0020] Furthermore, when the feature extraction module constructs a knowledge graph, it ensures the authenticity of the knowledge graph data; uses the graph convolutional neural network GCN to perform feature learning on the knowledge graph, with the formula where H (l) is the node feature matrix of the l-th layer, is the adjacency matrix with self-connections added, is 's degree matrix, W (l) is the weight matrix of the l-th layer, σ is the activation function, and mine the relationships between nodes through graph convolution operations.

[0021] Further, when calculating the matching degree, the matching calculation module introduces the social network analysis method, considering the status and influence of the enterprise in the industry social network and the dissemination and diffusion effect of the policy in the policy network; by analyzing the social relationships and partner information of the enterprise, it calculates the centrality indexes of the enterprise such as degree centrality where n is the total number of network nodes, V is the node set, and [u~v] indicates whether nodes u and v are connected, and betweenness centrality where σ st is the number of shortest paths from node s to t, and σ st (v) is the number of shortest paths from s to t passing through node v. The matching degree is adjusted according to the centrality index to make the matching result conform to the actual social resources and development potential of the enterprise.

[0022] Further, when providing the application guide, the application assistance module uses holographic projection technology to display the application process and material requirements for the enterprise, develops an application assistant, and has the function of automatically correcting and optimizing the application materials based on natural language processing and machine learning technologies; through learning historical application data and policy requirements, it constructs a language model BERT to review and modify the materials submitted by the enterprise.

[0023] Further, when assessing risks, the risk assessment module combines machine learning integration algorithms to improve the analysis and prediction capabilities of risk factors. At the same time, it introduces a contract mechanism to automatically warn of and handle risks.

[0024] Further, the user interaction module supports gesture recognition and eye movement tracking technologies, providing users with diverse interaction methods; it uses a convolutional neural network CNN to recognize gesture images, and by analyzing eye movement data, it judges the user's focus and operation intention, and adjusts the service strategy and interface style according to the information.

[0025] Further, it also includes:

[0026] Security audit module: By recording all operations and data of the system, it realizes the traceability and non-tampering of system operations; by establishing an operation behavior model, it monitors the real-time data of system operations.

[0027] Compared with the existing technologies, the beneficial effects of the present invention are:

[0028] For enterprises, the system widely and accurately collects policy and enterprise own information through multi-modal data acquisition, solving the problem of information asymmetry. Using advanced AI technologies, it can quickly and accurately perform intelligent matching between enterprises and applicable enterprise-benefiting policies, greatly improving the efficiency and accuracy of policy acquisition, and enabling enterprises not to miss the policy dividends suitable for their own development.

[0029] In the declaration process, the system uses VR and AR technologies to provide an immersive demonstration of the declaration process, combines intelligent robot customer service to answer questions in real time, and can also automatically review and optimize the declaration materials, helping enterprises easily cope with the cumbersome declaration process, reducing the probability of declaration errors and failures, and saving a large amount of time and labor costs.

[0030] From the perspective of policy implementation, the system promotes the effective implementation of policies beneficial to enterprises. By using blockchain technology to record the policy matching and declaration processes, it ensures the authenticity and traceability of data, and improves the transparency and fairness of policy implementation. The ecological collaboration module connects multiple parties such as the government, enterprises, financial institutions, and scientific research institutions and universities, promotes the collaborative cooperation of all parties, optimizes resource allocation, and enables policies to better play the role of promoting enterprise development and industrial upgrading. At the same time, the self-learning and optimization functions of the system can continuously improve according to enterprise feedback and policy changes, continuously enhance service quality, and provide long-term and stable support for enterprise development and policy implementation. Brief Description of the Drawings

[0031] Figure 1 It is a schematic block diagram of the intelligent policy matching and declaration assistance system for enterprises beneficial to enterprises based on AI proposed by the present invention;

[0032] Figure 2 It is a schematic block diagram for comparing the policy matching accuracy rates of the intelligent policy matching and declaration assistance system for enterprises beneficial to enterprises based on AI proposed by the present invention;

[0033] Figure 3 It is a schematic block diagram for comparing the declaration success rates of the intelligent policy matching and declaration assistance system for enterprises beneficial to enterprises based on AI proposed by the present invention. Detailed Embodiment

[0034] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0035] In the description of the present invention, it should be understood that the orientation or positional relationships indicated by the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", etc. are based on the orientation or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation to the present invention.

[0036] In addition, the terms "first" and "second" are only used for descriptive purposes and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of the said features. In the description of the present invention, "a plurality of" means two or more, unless otherwise specifically defined. In addition, the terms "mounted", "connected" and "coupled" should be construed broadly. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium, and it may be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. The present invention will be further described in detail below with reference to the accompanying drawings.

[0037] Refer to Figures 1 - 3 : A specific implementation of an AI-based intelligent matching and application assistance system for enterprise-benefiting policies.

[0038] System Overview

[0039] The AI-based intelligent matching and application assistance system for enterprise-benefiting policies is a comprehensive solution designed to solve the problems faced by enterprises in the process of obtaining, matching, and applying for enterprise-benefiting policies. The system mainly consists of a data collection module, a data preprocessing module, a feature extraction module, a matching calculation module, an application assistance module, a risk assessment module, a user interaction module, and a data update and feedback module. The specific implementation processes of each module will be elaborated in detail below.

[0040] Specific Implementations of Each Module of the System

[0041] Data Collection Module: This module adopts a multi-modal fusion data collection method. Using web crawler programs, it regularly crawls enterprise-benefiting policy data from channels such as government official websites and policy release platforms, including information such as policy names, issuing departments, applicable industries, policy terms, support methods, and application conditions. At the same time, by docking with the enterprise's financial system, human resources system, etc. through interfaces, basic information of the enterprise is obtained, such as enterprise name, unified social credit code, registered address, industry affiliation, business scope, enterprise scale, tax payment situation, number of employees, etc. To enrich the enterprise data dimension, speech recognition technology is also used to collect the speech information of enterprise principals in meetings and interviews, and convert it into text to extract key data. Satellite remote sensing and geographic information system (GIS) are used to obtain spatial data such as the geographical environment and infrastructure of the enterprise's location. The calculation of the data update period T introduces an enterprise activity factor β, and the formula is where, T maxis the maximum allowed update period, α is the importance coefficient (set according to the importance of policies or enterprise information), f is the historical update frequency of the data source, and β is calculated based on the recent business transaction frequency, employee recruitment, and other activities of the enterprise. For example, for an enterprise with frequent recent business and large-scale recruitment activities, the value of \(\beta\) is relatively high, and the data update period will be correspondingly shortened to ensure more timely data updates for active enterprises.

[0042] Data preprocessing module: The collected data will first be cleaned on a high-performance computing cluster. Duplicate, incorrect, and incomplete data will be removed. For policy texts, natural language processing (NLP) techniques such as lexical analysis and syntactic analysis will be used for text standardization, unifying term expressions, and clarifying ambiguous expressions in policy clauses. At the same time, a preprocessing method based on sentiment analysis will be adopted to identify positive, negative, and neutral expressions in policies to help enterprises better understand policy tendencies. For enterprise data, format unification and data type conversion will be carried out. For example, the enterprise scale will be converted from a text description to a specific numerical range. A data filling algorithm will be used to handle missing values. For key missing information in policy data, if it cannot be supplemented, it will be marked as pending verification; for enterprise data, it will be reasonably filled according to similar data of other enterprises in the same industry. In terms of data encryption, an advanced AES (Advanced Encryption Standard) algorithm will be used to encrypt and store sensitive data to ensure the security of data transmission and storage.

[0043] Feature extraction module: Build a feature enhancement model based on the generative adversarial network (GAN). The generator attempts to generate additional virtual features to enrich the feature spaces of policies and enterprises, and the discriminator judges the authenticity of the features. Through the adversarial training of the two, the diversity and quality of the features are improved. Combining cognitive computing technology, it simulates the human brain's ability to abstract and associate information. Through in-depth analysis of policy and enterprise data, deep-seated and potential correlation relationships between features are mined. For example, the correlation relationships between a certain policy and multiple aspects such as tax incentives and technological innovation of enterprises in a specific industry are discovered. The construction of the knowledge graph introduces a dynamic update mechanism. The relationships and attributes in the graph are automatically updated according to real-time data and events to ensure the timeliness and accuracy of the knowledge graph. At the same time, a graph convolutional neural network (GCN) is used for feature learning of the knowledge graph, and the formula is where H (l) is the node feature matrix of the l-th layer, is the adjacency matrix with self-connections added, is 's degree matrix, W (l) is the weight matrix of the l-th layer, σ is the activation function, and the hidden relationships between nodes are mined through graph convolutional operations to improve the accuracy of feature relationship mining.

[0044] Matching calculation module: This module uses a variety of algorithms to calculate the matching degree between policies and enterprises. First, the cosine similarity algorithm is used to calculate the basic matching degree. The formula is A and B represent the characteristic vectors of policy and enterprise respectively, and n is the dimension of the characteristic vector. In order to more accurately reflect the matching degree between enterprise and policy, the weight adjustment coefficient w is introduced i , the adjusted matching formula is Weight w i According to the actual situation of policies and enterprises, it is obtained through training of machine learning algorithms. In addition to the cosine similarity algorithm, a matching algorithm based on complex network theory is also introduced. Policies and enterprises are regarded as nodes in a complex network, and the matching degree Sim is calculated by the connection strength C(A, B) and path length L(A, B) between nodes. network (A, B), the formula is At the same time, considering the time series characteristics of the policy, the time series analysis method is used to predict the change in the matching degree of the policy in the future, and to provide forward-looking matching suggestions for enterprises. The final matching formula is optimized to Sim total =w1×Sim′(A,B)+w2×Sim network (A, B)+w3×Sim time (A, B), where Sim time (A, B) is the matching degree of the time series considered, and w1, w2, and w3 are weight coefficients obtained by optimizing the genetic algorithm.

[0045] Application assistance module: Use virtual reality (VR) and augmented reality (AR) technologies to provide enterprises with an immersive application process demonstration. Enterprises can experience the application process in person through VR devices and understand the specific operations and requirements of each link. AR devices can provide real-time guidance and prompts in actual operations. For example, when filling out the application form, the AR device can identify the form content and provide relevant filling specifications and examples. Introduce intelligent robot customer service, and realize autonomous learning and evolution based on reinforcement learning algorithms. Intelligent robot customer service can continuously improve service quality through interaction with enterprises and answer questions encountered by enterprises in the application process. For some complex problems, the system can also automatically transfer to manual customer service for processing. Develop an intelligent application assistant based on natural language processing and machine learning technology, with the function of automatic error correction and optimization of application materials. By learning from historical application data and policy requirements, build language models such as the Transformer-based BERT model, and intelligently review and modify the materials submitted by enterprises to improve the quality of application materials.

[0046] Risk Assessment Module: Risk assessment is carried out by combining blockchain and federated learning. The data of different enterprises are used for model training locally, and the model parameters are securely shared through blockchain technology to avoid data leakage. After the data of each enterprise is encrypted locally, machine learning algorithms such as random forest algorithm are used for model training, and then the trained model parameters are uploaded to the blockchain for aggregation. The chaos theory is introduced to analyze the uncertainty risks in the policy declaration process, and a chaos model such as the Logistic mapping model x n+1 = μx n (1 - x n )(where μ is the control parameter, x n is the current risk state, and x n+1 is the risk state at the next moment) is used to predict the mutation and evolution of risks. Through the quantitative assessment of risks, the risk levels (such as low risk, medium risk, high risk) are output. For different risk levels, corresponding risk response suggestions are provided for enterprises, such as supplementing materials, adjusting the declaration strategy, etc. At the same time, the intelligent contract mechanism of blockchain is introduced, and when the risk reaches a certain threshold, the corresponding response measures are automatically triggered.

[0047] User Interaction Module: An interaction method based on the principle of brain-computer interface is developed, and enterprise users can perform system operations and information queries through specific brain wave signals. For example, users can select different function menus through different brain wave states such as concentrating and relaxing, improving the convenience and efficiency of interaction. The interface design adopts an adaptive learning algorithm, which automatically adjusts the interface layout and function display according to the user's usage habits and preferences. At the same time, gesture recognition and eye movement tracking technologies are supported to provide users with more diverse interaction methods. A convolutional neural network (CNN) is used to recognize gesture images, and by analyzing eye movement data such as fixation point trajectories and fixation times, the user's attention focus and operation intention are judged, and the service strategy and interface style are adjusted according to this information.

[0048] The Data Update and Feedback Module uses the immutable feature of blockchain to record the feedback information of enterprises, ensuring the authenticity and traceability of feedback. After an enterprise completes the policy declaration, it can feedback information such as problems encountered during the declaration process and the declaration results through the system. These feedback information will be encrypted and stored on the blockchain to form a complete feedback record. A predictive feedback mechanism is introduced. According to historical feedback data and system operation conditions, through a time series prediction model such as the ARIMA model, the problems that enterprises may encounter are predicted and corresponding solutions are provided. For example, if it is found that a certain type of enterprise often fails due to material problems when declaring a certain policy, the system can provide targeted material preparation suggestions for this type of enterprise in advance.

[0049] Data Representation of Beneficial Effects

[0050] To verify the beneficial effects of this system, 100 enterprises were selected for a comparative experiment. Among them, 50 enterprises used this system for the matching and application of enterprise-benefiting policies, and the other 50 enterprises operated in the traditional way. The experimental results are shown in the following table:

[0051] Comparison items Traditional method This system Policy matching accuracy rate 60％ 90％ Declaration success rate 50％ 85％ Declaration time (average per time, unit: days) 15 5 Labor cost (average per time, unit: yuan) 2000 500

[0052] As can be seen from the above data, this system has significant advantages in terms of policy matching accuracy rate, application success rate, application time, and labor cost, etc. It can effectively improve the efficiency and quality of enterprises' access to enterprise-benefiting policies and reduce the application costs of enterprises.

[0053] The above is only the preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered within the protection scope of the present invention.

Claims

1. An AI-based intelligent matching and application assistance system for enterprise-benefiting policies, characterized in that, Including: Data collection module: Adopt a multi-modal fusion data collection method. Obtain enterprise-benefiting policy data from government official websites and policy release platforms through web crawlers, and use natural language processing (NLP) technology to parse information; collect basic enterprise information through interface docking, and at the same time collect the speech information of enterprise principals in meetings and interviews through speech recognition technology and convert it into text for data extraction; use satellite remote sensing and geographic information system (GIS) to obtain the geographical environment and infrastructure spatial data of the enterprise's location; introduce the enterprise activity factor β to calculate the data update cycle T, and the formula is β is calculated based on the recent business transaction frequency and employee recruitment activities of the enterprise; Data preprocessing module: Using a computing cluster to assist in data cleaning and processing of policy and enterprise data; for policy texts, adopting a preprocessing method based on situation analysis to identify positive, negative, and neutral expressions in policies, and using encryption algorithms for data encryption; Feature extraction module: Constructing a feature enhancement model based on the Generative Adversarial Network (GAN) to generate virtual features to enrich the feature spaces of policies and enterprises; combining cognitive computing technology to simulate the human brain's ability to abstract and associate information, mining the associations between policy and enterprise features, and introducing a dynamic update mechanism into the knowledge graph construction to automatically update the relationships and attributes in the graph according to real-time data and events; Matching calculation module: In addition to the cosine similarity algorithm, a matching algorithm based on network theory is introduced. Policies and enterprises are regarded as nodes in the network, and the matching degree Sim is calculated through the connection strength C(A, B) and path length L(A, B) between nodes. network (A, B), and the formula is Considering the time series characteristics of policies, the time series analysis method is used to predict the change of the matching degree of policies in the future time, providing forward-looking matching suggestions for enterprises; the final matching degree formula is optimized to Sim total = w1×Sim′(A, B)+w2×Sim network (A, B)+w3×Sim time (A, B), where Sim time (A, B) is the matching degree considering the time series, and w1, w2, and w3 are weight coefficients optimized by the genetic algorithm.

2. The AI-based intelligent matching and declaration assistance system for enterprise-benefiting policies according to claim 1, wherein Also including: Declaration assistance module: Using virtual reality and augmented reality technologies to provide enterprises with a demonstration of the declaration process; Enterprises experience the declaration process through the device, and the device provides real-time guidance and prompts during actual operation; introducing a robot customer service that realizes autonomous learning and evolution based on reinforcement learning algorithms; Risk assessment module: Adopting a method combined with federated learning for risk assessment; Data of different enterprises are used for model training locally. The chaos theory is introduced to analyze the uncertainty risks in the policy declaration process, and a chaos model mapping model x n+1 = μx n (1 - x n ) is used to predict the evolution of risks, where μ is the control parameter, and x n is the current risk state, and x n+1 is the risk state at the next moment; User interaction module: Developing an interaction method based on the principle of brain-computer interface, where enterprise users can perform system operations and information queries through brain wave signals, and the interface design adopts an adaptive learning algorithm to automatically adjust the interface layout and function display according to the user's usage habits; Data update and feedback module: Using the non-tamperable feature to record the feedback information of enterprises, introducing a predictive feedback mechanism, and predicting the problems encountered by enterprises and providing corresponding solutions through a time series prediction model based on historical feedback data and system operation conditions. Policy Insight Module: Using a method that combines deep reinforcement learning and meta-learning to analyze and predict policies, semantic web technology is used to semantically annotate and reason about policies, and the logic and impacts behind policies are mined; the Latent Dirichlet Allocation (LDA) topic model is used to extract topics from policy texts. The formula is where w i is a word, z i is a topic, β is a hyperparameter, D is the number of documents, N d is the number of words in document d, V is the size of the vocabulary, and policy priorities and trend changes are predicted through topic analysis.

3. The AI-based intelligent matching and declaration assistance system for enterprise-benefiting policies according to claim 1, wherein Also including: Ecological Collaboration Module: Build an enterprise-benefiting policy ecosystem involving government departments, enterprises, financial institutions, and scientific research institutions and universities; achieve data sharing and collaborative cooperation among all parties through contracts; use the game theory model extension model, set the cooperation payoff matrix R and the betrayal payoff matrix S, and calculate the payoffs of all parties under different strategies where p ij is the probability that the other party chooses cooperation, n is the number of participants, and optimize the cooperation strategies of all parties in the ecosystem.

4. The AI-based intelligent matching and application assistance system for enterprise-benefiting policies according to claim 1, wherein When the data collection module collects enterprise data, it uses drone patrol technology to obtain images and video data of the enterprise's factory building appearance and equipment operation status, enriching the visual data dimension of the enterprise. For the obtained image and video data, image recognition algorithms are used for target detection and analysis to extract information such as the number of enterprise equipment, equipment models, and factory building construction conditions.

5. The AI-based intelligent matching and application assistance system for enterprise-benefiting policies according to claim 1, characterized in that, When constructing the knowledge graph, the feature extraction module ensures the authenticity of the knowledge graph data; uses the graph convolutional neural network GCN to perform feature learning on the knowledge graph, and the formula is where H (l) is the node feature matrix of the l-th layer, is the adjacency matrix with self-connections added, is the degree matrix of, W (l) is the weight matrix of the l-th layer, σ is the activation function, and the relationship between nodes is mined through graph convolutional operations.

6. The AI-based intelligent matching and application assistance system for enterprise-benefiting policies according to claim 1, wherein, When the matching calculation module calculates the matching degree, it introduces social network analysis methods, considering the status and influence of the enterprise in the industry social network and the spread and diffusion effects of the policy in the policy network; By analyzing the social relationships and partner information of an enterprise, calculate the centrality indicators of the enterprise such as degree centrality where n is the total number of network nodes, V is the node set, and [u~v] indicates whether nodes u and v are connected, and betweenness centrality where σ st is the number of shortest paths from node s to t, and σ st (ν) is the number of shortest paths from s to t passing through node v. Adjust the matching degree according to the centrality indicators to make the matching result conform to the actual social resources and development potential of the enterprise.

7. The AI-based intelligent matching and declaration assistance system for enterprise-benefiting policies according to claim 2, characterized in that, When the declaration assistance module provides a declaration guide, it uses holographic projection technology to display the declaration process and material requirements for enterprises, develops a declaration assistant with the function of automatically correcting and optimizing declaration materials based on natural language processing and machine learning technologies; through learning historical declaration data and policy requirements, constructs a language model BERT to review and modify the materials submitted by enterprises.

8. The AI-based intelligent matching and declaration assistance system for enterprise-benefiting policies according to claim 2, wherein When the risk assessment module assesses risks, it combines machine learning integration algorithms to improve the analysis and prediction ability of risk factors. At the same time, it introduces a contract mechanism to automatically warn and handle risks.

9. The AI-based intelligent matching and declaration assistance system for enterprise-benefiting policies according to claim 2, wherein The user interaction module supports gesture recognition and eye movement tracking technologies to provide users with diverse interaction methods; uses a Convolutional Neural Network (CNN) to recognize gesture images, and judges the user's attention points and operation intentions through the analysis of eye movement data, and adjusts the service strategy and interface style according to the information.

10. The AI-based intelligent matching and declaration assistance system for enterprise-benefiting policies according to claim 1, characterized in that, Also including: Security Audit Module: By recording all operations and data of the system, it realizes the traceability and immutability of system operations; By establishing an operation behavior model, it conducts real-time data monitoring of system operations.

Images