AI-based attribute feature matrix industrial policy intelligent processing method
By constructing an AI-based attribute feature matrix, the entire process from data collection to application implementation is automated, solving the problems of lag and precise matching in industrial policy processing and improving the efficiency and accuracy of policy processing.
Patent Information
- Application Number
- CN202511547877.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-28
- Publication Date
- 2026-02-10
AI Technical Summary
Existing technologies suffer from problems such as delayed policy implementation, missing features, non-standard expressions, and inability to achieve accurate multi-dimensional matching and dynamic evaluation when dealing with massive and continuously updated industrial policies. They cannot meet the needs for efficient policy implementation, accurate matching, and dynamic optimization.
By adopting an AI-based attribute feature matrix approach, through data collection and preprocessing, constructing an industrial policy attribute feature matrix, and training and optimizing the AI model, the entire process from policy data acquisition to application implementation is automated, including data cleaning, standardization processing, feature extraction, precise matching of enterprises and policies, and policy effect evaluation.
It has achieved high efficiency and accuracy in policy processing, broken down data silos and reliance on manual intervention, ensured real-time response and dynamic updates of policy information, and provided high-quality policy processing results.
Smart Images
Figure CN121504356A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of artificial intelligence technology and industrial policy analysis, in particular to an attribute feature matrix industrial policy intelligent processing method based on AI. BACKGROUND
[0002] With the increasing support of the state and local governments for industrial development, the industrial policy documents covering tax incentives, financial subsidies and other fields have shown explosive growth. However, the current processing mainly relies on manual processing, which has significant shortcomings. Not only does it need to read and extract key information from each document, but it is difficult to achieve real-time processing in the face of a large amount of and continuously updated policies, resulting in a lag in policy implementation. In addition, due to the lack of a unified attribute feature system, the extraction relies on personal experience, which is prone to feature omission and non-standard expression, and cannot form a standardized data set, which is not conducive to subsequent analysis. At the same time, the matching of enterprise needs and policies is extensive, without a multi-dimensional and accurate mechanism, which is prone to "two-way unsmoothness" and waste of resources, and it is difficult to dynamically evaluate policy effectiveness and mine correlation rules, which cannot provide intelligent support for decision-making.
[0003] Although some existing technologies attempt to introduce intelligent algorithms to process industrial policies, they mainly focus on single links such as text classification, without building a "data collection-feature modeling-intelligent processing-application feedback" whole process system, and without combining attribute feature matrix to achieve multi-dimensional and high-precision processing, which cannot meet the actual needs of efficient implementation, accurate matching and dynamic optimization of policies. Therefore, developing an attribute feature matrix industrial policy intelligent processing method based on AI has become the key to solving the current problems. SUMMARY
[0004] The purpose of the present application is to solve the problems in the prior art and to provide an attribute feature matrix industrial policy intelligent processing method based on AI.
[0005] In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions:
[0006] The attribute feature matrix industrial policy intelligent processing method based on AI, characterized in that S1. Industrial policy data collection and preprocessing, collecting multi-source industrial policy data, cleaning and standardizing the policy data to obtain a structured policy data set;
[0007] S2. Constructing an attribute feature matrix of industrial policy, determining the attribute dimensions of the policy based on the policy business logic, extracting the feature values of each policy in each attribute dimension in the structured policy data set, and constructing a standardized attribute feature matrix of industrial policy;
[0008] S3. AI model training and optimization, taking the attribute feature matrix of industrial policy as the core input, combining historical policy application data to construct a training sample set, training and optimizing the AI processing model;
[0009] S4. Industrial policy intelligent processing, through the AI processing model completed by training, intelligent analysis of newly acquired industrial policy, enterprise-policy precise matching, policy effect evaluation and policy pushing, output policy processing result, through the integration of multi-link intelligent processing, the limitations of "data island" and "manual dependence" in traditional industrial policy processing are broken, and the whole process automation from policy data acquisition to application landing is realized. Among them, the formation of structured policy data set is the basis for subsequent feature extraction and model training, which can provide high-quality input data for AI model, support the accuracy and efficiency of policy processing, and is suitable for government policy management departments at all levels, industrial park service platforms and enterprise policy service systems and other scenarios.
[0010] As a further scheme of the present application, in step S1, "multi-source industrial policy data" includes policy files published on government official websites, historical policies stored in policy databases, and policy implementation data fed back by government service platforms. The preprocessing includes removing redundant data, correcting data format, and unifying term expression. Regular expressions and semantic similarity algorithms are used to realize data standardization.
[0011] As a further scheme of the present application, in step S2, "policy attribute dimensions" include policy basic attributes, policy applicable attributes, policy content attributes, and policy correlation attributes. The feature values are obtained through NLP semantic extraction algorithm combined with artificial rule checking. The attribute feature matrix takes policies as rows, attribute dimensions as columns, and the cells store the feature values of the corresponding policies in the attribute dimensions. NLP semantic extraction can use spaCy toolkit for entity recognition and relationship extraction. Artificial rule checking is performed by policy domain experts to sample and review key feature values (such as support amount threshold) to ensure the accuracy of the feature matrix.
[0012] As a further scheme of the present application, in step S2, attribute feature weight calculation is also included. The weights of each attribute dimension are determined by using the analytic hierarchy process combined with the entropy weight method, and are integrated into the attribute feature matrix to form an industrial policy attribute feature matrix with weights.
[0013] As a further scheme of the present application, in step S3, "AI processing model" includes policy analysis sub-model, matching sub-model, and evaluation sub-model. The policy analysis sub-model uses a CNN-LSTM hybrid neural network to extract the attribute features of new policies and update them to the attribute feature matrix. The matching sub-model uses a collaborative filtering algorithm based on cosine similarity to realize two-way precise matching of enterprise attribute feature matrix and industrial policy attribute feature matrix. The evaluation sub-model uses a gradient boosting tree algorithm to evaluate the implementation effect of the policy based on policy implementation data and industrial economic data.
[0014] As a further scheme of the present application, the enterprise attribute feature matrix comprises enterprise basic attributes, operation attributes, and demand attributes, which correspond to the dimensions of the industrial policy attribute feature matrix.
[0015] As a further scheme of the present application, the policy intelligent processing in step S4 further comprises policy dynamic updating, when a policy revision, invalidation, or new policy is detected, the industrial policy attribute feature matrix is automatically updated, and the AI model processing is retriggered, to ensure the timeliness of the policy processing result. The policy dynamic updating is realized by deploying a network crawler to monitor the update information of the official website and the policy publishing platform of the government in real time, combining a text comparison algorithm to identify the policy revision content, setting an automatic marking mechanism for invalid policies, and automatically extracting features of the new policy through an analysis submodel and writing the features into the matrix. After the update, the model only needs to perform incremental training on the changed part, to reduce the consumption of computing resources, ensure the completion of result update within 1-2 hours, and avoid the acquisition of outdated policy information by enterprises or the evaluation based on the old policy by the government.
[0016] As a further scheme of the present application, the policy processing result in step S4 comprises a policy analysis report, an enterprise-policy matching list, and a policy effect evaluation report, and is pushed to the government affair platform and the enterprise service platform through an API interface.
[0017] The present application has the following beneficial effects:
[0018] 1. Compared with the existing mode of relying on manual extraction of policy information from each article and being unable to respond to massive policy updates in real time, the present application efficiently forms a structured data set through S1 multi-source data collection (covering the official website of the government, historical databases, and feedback of government affairs execution) and standardized processing of regular expressions and semantic similarity algorithms, and realizes the full-process automation of “data collection-feature modeling-intelligent analysis-matching evaluation-result pushing” by combining S2-S4 feature matrix construction, AI model training, and intelligent processing, to greatly improve the policy processing efficiency and avoid the problem of policy lag.
[0019] 2. In view of the defects of the prior art, such as lack of a unified feature system and reliance on personal experience to easily cause feature omission and non-standardized expression, S2 of the present application determines four attribute dimensions of “basic+applicable+content+association” based on the policy business logic, acquires feature values through NLP semantic extraction combined with artificial rule checking, more innovatively adopts the analytic hierarchy process and entropy weight method to calculate attribute weights, forms an industrial policy attribute feature matrix with weights, eliminates feature expression differences, takes into account both subjective business logic and objective data rules, provides a high-precision feature basis for subsequent AI processing, and solves the problem of missing standardized data set. BRIEF DESCRIPTION OF DRAWINGS
[0020] Fig. 1This is a schematic diagram of the overall technical process structure of the AI-based attribute feature matrix intelligent processing method for industrial policy proposed in this invention;
[0021] Fig. 2 This is a detailed flowchart illustrating the core module structure of the AI-based intelligent processing method for industrial policy based on attribute feature matrices proposed in this invention.
[0022] Fig. 3 This is a schematic diagram of the AI model training and application process structure of the AI-based attribute feature matrix intelligent processing method for industrial policy proposed in this invention. Detailed Implementation
[0023] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0024] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.
[0025] Reference Figs. 1-3 This paper presents an AI-based intelligent processing method for industrial policies using attribute feature matrices. The method comprises two steps: S1. Industrial policy data collection and preprocessing: collecting industrial policy data from multiple sources, cleaning and standardizing the data to obtain a structured policy dataset. This systematically integrates data collection and preprocessing, breaking down the barriers of scattered and disorganized data in traditional industrial policy processing. This provides a high-quality data foundation for subsequent intelligent analysis based on attribute feature matrices, enabling AI models to more accurately capture core policy information. S2. Constructing an industrial policy attribute feature matrix: determining policy attribute dimensions based on policy business logic, extracting feature values of each policy in the structured policy dataset under each attribute dimension, and constructing a standardized industrial policy attribute feature matrix.
[0026] S3. AI Model Training and Optimization: Using the industrial policy attribute feature matrix as the core input, and combining historical policy application data to construct a training sample set, the AI processing model is trained and optimized.
[0027] S4. Intelligent processing of industrial policies: Through a trained AI processing model, newly acquired industrial policies are intelligently analyzed, enterprises are accurately matched with policies, policy effects are evaluated, and policies are pushed out, outputting policy processing results.
[0028] In this embodiment, the "multi-source industrial policy data" in step S1 includes policy documents published on government websites, historical policies stored in policy databases, and policy implementation data fed back by government service platforms. Preprocessing includes removing redundant data, correcting data formats, and standardizing terminology. Regular expressions and semantic similarity algorithms are used to achieve data standardization. The fusion of multi-source data ensures the comprehensiveness of policy information—government website data guarantees authority, historical policy data provides a basis for longitudinal comparison, and implementation feedback data reflects the effectiveness of policy implementation. The combination of regular expressions and semantic similarity algorithms not only efficiently solves the problem of inconsistent data formats, but also eliminates ambiguity caused by differences in terminology through semantic calibration, laying a rigorous foundation for data structuring.
[0029] In this embodiment, the "policy attribute dimension" in step S2 includes basic policy attributes, policy applicability attributes, policy content attributes, and policy association attributes. The feature values are obtained through NLP semantic extraction algorithms combined with manual rule verification. The attribute feature matrix is arranged with policies as rows and attribute dimensions as columns. Each cell stores the feature value of the corresponding policy under that attribute dimension. The division of policy attribute dimensions covers all elements of the policy from basic information to actual application—basic attributes clarify the policy's "identity information," applicability attributes define the policy's coverage, content attributes reveal the policy's core measures, and association attributes reflect the policy's collaborative relationship. The combination of NLP algorithms and manual verification ensures the efficiency of feature value extraction and reduces semantic understanding bias through manual rule correction, enabling the attribute feature matrix to accurately depict the essence of the policy.
[0030] In this embodiment, step S2 also includes the calculation of attribute feature weights. The weights of each attribute dimension are determined by combining the analytic hierarchy process (AHP) with the entropy weight method, and then integrated into the attribute feature matrix to form a weighted industrial policy attribute feature matrix. The AHP quantifies the subjective importance of attributes through expert experience, while the entropy weight method automatically calculates objective weights based on data distribution. The combination of the two effectively avoids the one-sidedness of a single weight method, so that the weighted feature matrix can reflect both business logic and the inherent laws of data, thereby improving the sensitivity of subsequent AI models to key policy features.
[0031] In this embodiment, the "AI processing model" in step S3 includes a policy analysis sub-model, a matching sub-model, and an evaluation sub-model. The policy analysis sub-model uses a CNN-LSTM hybrid neural network to extract the attribute features of new policies and update the attribute feature matrix. The matching sub-model uses a cosine similarity-based collaborative filtering algorithm to achieve bidirectional accurate matching by combining the enterprise attribute feature matrix and the industry policy attribute feature matrix. The evaluation sub-model uses a gradient boosting tree algorithm to evaluate the policy implementation effect based on policy implementation data and industry economic data. In the CNN-LSTM hybrid network, CNN is good at capturing local key features in policy text, while LSTM can handle the sequential semantic relationship of policy clauses. The combination of the two achieves efficient extraction of new policy features. Cosine similarity collaborative filtering accurately quantifies the matching degree by measuring the angle between the enterprise and policy feature vectors. The bidirectional matching mechanism considers both enterprise needs and policy requirements, avoiding "one-way adaptation" bias. The gradient boosting tree algorithm can effectively handle the nonlinear relationship in policy implementation data and achieve multi-dimensional quantitative evaluation of policy effects by combining industry economic data.
[0032] In this embodiment, the enterprise attribute feature matrix includes basic enterprise attributes, operational attributes, and demand attributes, which correspond to the dimensions of the industrial policy attribute feature matrix. The dimensional correspondence between enterprise attributes and policy attributes constructs a precise coordinate system for "policy-enterprise" matching—basic attributes correspond to the applicable geographical area and timeliness requirements of the policy; operational attributes correspond to the target industry and scale threshold of the policy; and demand attributes correspond to the policy support direction. This dimensional alignment ensures that the matching algorithm can directly calculate similarity through feature vectors, greatly improving matching efficiency and accuracy.
[0033] In this embodiment, the "policy intelligent processing" in step S4 also includes dynamic policy updates. When a policy revision, invalidation, or new policy is detected, the industrial policy attribute feature matrix is automatically updated, and the AI model processing is re-triggered to ensure the timeliness of policy processing results. The policy dynamic update mechanism is triggered by real-time monitoring of government release channels. It updates feature values for revised policies, marks the status of invalid policies, and extracts features for new policies. This avoids the problems of "policies still being recommended after expiration" and "lagging processing of new policies" in traditional static systems, ensuring that the output results of the AI model are always synchronized with the latest policy system and guaranteeing the timeliness of government decision-making and enterprise applications.
[0034] In this embodiment, the "policy processing results" in step S4 include a policy analysis report, an enterprise-policy matching list, and a policy effectiveness evaluation report. These are pushed to the government's e-government platform and enterprise service platform via API interfaces. The policy analysis report presents the core elements of the policy, such as applicable conditions and application procedures, in a structured form, reducing the understanding cost for enterprises. The matching list is sorted by matching degree and marked with appropriate points, such as "enterprise revenue meets policy scale requirements," making it easy for enterprises to quickly locate applicable policies. The evaluation report provides the government with a basis for policy adjustments. The API interface push enables seamless cross-platform integration. The government can dynamically grasp the policy implementation status through the e-government platform, and enterprises can obtain customized policy information in real time through the service platform, forming a closed-loop service of "policy release-analysis-matching-evaluation-optimization."
[0035] For ease of description, spatial relative terms such as "above," "on top of," "on the upper surface of," "above," etc., are used herein to describe the spatial positional relationship of a device or feature as shown in the figures to other devices or features. It should be understood that spatial relative terms are intended to encompass different orientations in use or operation beyond the orientation of the device as described in the figures. For example, if the device in the figures were inverted, a device described as "above" or "on top of" other devices or structures would subsequently be positioned as "below" or "under" other devices or structures. Thus, the exemplary term "above" can include both "above" and "below." The device may also be positioned in other different ways (rotated 90 degrees or in other orientations), and the spatial relative descriptions used herein will be interpreted accordingly.
[0036] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.
[0037] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0038] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. An AI-based intelligent processing method for industrial policy based on attribute feature matrices, characterized in that: S1. Industrial policy data collection and preprocessing: Collect industrial policy data from multiple sources, clean and standardize the policy data to obtain a structured policy dataset; S2. Construct an industrial policy attribute feature matrix, determine the policy attribute dimensions based on policy business logic, extract the feature values of each policy in the structured policy dataset under each attribute dimension, and construct a standardized industrial policy attribute feature matrix; S3. AI Model Training and Optimization: Using the industrial policy attribute feature matrix as the core input, and combining it with historical policy application data to construct a training sample set, the AI processing model is trained and optimized. S4. Intelligent processing of industrial policies: Through a trained AI processing model, newly acquired industrial policies are intelligently analyzed, enterprises are accurately matched with policies, policy effects are evaluated, and policies are pushed out, outputting policy processing results.
2. The AI-based intelligent processing method for industrial policy based on attribute feature matrix according to claim 1, characterized in that, In step S1, "multi-source industrial policy data" includes policy documents published on government websites, historical policies stored in policy databases, and policy implementation data fed back by government service platforms. The preprocessing includes removing redundant data, correcting data formats, standardizing terminology, and using regular expressions and semantic similarity algorithms to achieve data standardization.
3. The AI-based intelligent processing method for industrial policy based on attribute feature matrix according to claim 1, characterized in that, In step S2, the "policy attribute dimension" includes basic policy attributes, policy applicability attributes, policy content attributes, and policy association attributes. The feature values are obtained through NLP semantic extraction algorithms combined with manual rule verification. The attribute feature matrix is arranged with policies as rows and attribute dimensions as columns, with each cell storing the feature value of the corresponding policy under that attribute dimension.
4. The AI-based intelligent processing method for industrial policy based on attribute feature matrix according to claim 1, characterized in that, Step S2 also includes the calculation of attribute feature weights. The weights of each attribute dimension are determined by combining the analytic hierarchy process with the entropy weight method and then incorporated into the attribute feature matrix to form a weighted industrial policy attribute feature matrix.
5. The AI-based intelligent processing method for industrial policy based on attribute feature matrix according to claim 1, characterized in that, In step S3, the "AI processing model" includes a policy analysis sub-model, a matching sub-model, and an evaluation sub-model. The policy analysis sub-model uses a CNN-LSTM hybrid neural network to extract the attribute features of new policies and update the attribute feature matrix. The matching sub-model uses a cosine similarity-based collaborative filtering algorithm to achieve bidirectional accurate matching by combining the enterprise attribute feature matrix and the industry policy attribute feature matrix. The evaluation sub-model uses a gradient boosting tree algorithm to evaluate the policy implementation effect based on policy implementation data and industry economic data.
6. The AI-based intelligent processing method for industrial policy based on attribute feature matrix according to claim 5, characterized in that, The enterprise attribute feature matrix includes basic enterprise attributes, operational attributes, and demand attributes, which correspond to the dimensions of the industrial policy attribute feature matrix.
7. The AI-based intelligent processing method for industrial policy based on attribute feature matrix according to claim 1, characterized in that, Step S4, "Intelligent Policy Processing," also includes dynamic policy updates. When a policy revision, invalidation, or new policy is detected, the industrial policy attribute feature matrix is automatically updated, and the AI model processing is re-triggered to ensure the timeliness of the policy processing results.
8. The AI-based intelligent processing method for industrial policy based on attribute feature matrix according to claim 1, characterized in that, Step S4, "Policy Processing Results," includes a policy analysis report, an enterprise-policy matching list, and a policy effectiveness evaluation report, which are then pushed to the government's e-government platform and enterprise service platform via API.