LLM-based cross-border supply chain optimization method and device, electronic equipment and medium

Abstract

The invention relates to the technical field of cross-border supply chain optimization based on LLM, and discloses an LLM-based cross-border supply chain optimization method and device, electronic equipment and a medium, the method comprises the steps of performing hierarchical analysis on related legal texts, accurately extracting legal rule features, mapping the legal rule features into a preset mathematical space, forming mathematical constraints, and establishing a mathematical model; and automatically identifying and optimizing the initial suppliers which do not conform to the mathematical constraints, and optimizing the supply chain. The method has the advantages that legal risks are reduced, transparency and flexibility of the supply chain are improved, cost is saved, efficiency is improved, legal requirements can be automatically applied to supply chain management, and efficiency and accuracy of compliance auditing are greatly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of LLM-based cross-border supply chain optimization, and in particular to an LLM-based cross-border supply chain optimization method, device, electronic equipment and medium. BACKGROUND

[0002] In the context of globalization trade, supply chain management faces increasingly complex legal and compliance requirements. With countries increasingly strict regulations on product origin, trade regulations and other issues, businesses need to respond to challenges from laws and regulations in real time, which not only affects operational efficiency, but also may lead to legal disputes and economic losses. Traditional supply chain management methods have been unable to adapt to the complexity of modern trade.

[0003] Currently, effective legal text analysis technology is lacking, often relying on manual review, which is inefficient and prone to error. In addition, existing supply chain management systems often fail to synchronize with supply chain levels when dealing with legal compliance issues, and are unable to effectively assess the compliance and risk of suppliers. SUMMARY

[0004] Therefore, it is necessary to propose an LLM-based cross-border supply chain optimization method, device, electronic equipment and medium for the existing LLM-based cross-border supply chain optimization problem.

[0005] An LLM-based cross-border supply chain optimization method, the method comprising: obtaining a hierarchical supply chain of a specified product and related legal texts; performing hierarchical analysis on the legal texts through an LLM model to obtain legal rule features; mapping the legal rule features to a preset mathematical space to generate mathematical constraints of the preset mathematical space; obtaining initial suppliers of each level in the hierarchical supply chain, and obtaining corresponding supplier data according to the level of each initial supplier; mapping the supplier data into the preset mathematical space and verifying that the supplier data meets the requirements of the mathematical constraints; optimizing the initial suppliers whose verification results do not meet the requirements of the mathematical constraints, so that the optimized hierarchical supply chain meets the requirements of the mathematical constraints.

[0006] Further, the step of obtaining a hierarchical supply chain of a specified product comprises: performing structural decomposition on the specified product to obtain a BOM tree structure of the specified product; wherein the BOM tree structure comprises a plurality of nodes, each node corresponding to a product component of the specified product; obtaining initial suppliers of each node in the BOM tree structure and importance of each node; generating a hierarchical supply chain of the specified product according to the initial suppliers of each node and the corresponding importance.

[0007] Further, the step of structurally decomposing the specified product to obtain the BOM tree structure of the specified product further comprises: obtaining product components corresponding to each node; obtaining a plurality of alternative suppliers corresponding to the node according to the product components; calculating a similarity score of each of the initial suppliers of the node and each of the alternative suppliers; sorting each of the alternative suppliers according to the size of each of the similarity scores.

[0008] Further, the step of obtaining a plurality of alternative suppliers corresponding to the node according to the product components further comprises: receiving a target function and data information of each of the alternative suppliers; inputting each of the data information and the target function into a preset mathematical space to optimize and solve the target function to obtain a target supply chain; comparing the target supply chain with the hierarchical supply chain to obtain comparison information; sending the comparison information to a specified terminal.

[0009] Further, the step of hierarchically analyzing the legal text by the LLM model to obtain the legal rule features comprises: adopting a preset semantic segmentation technology to hierarchically analyze the legal text to cut the legal text into a plurality of paragraph modules; obtaining category attributes of each of the paragraph modules and extracting the paragraph modules as clause paragraph modules; constructing a three-dimensional rule feature matrix based on the clause paragraph modules to serve as the legal rule features; wherein the three-dimensional rule feature matrix comprises a validity dimension, a time effectiveness dimension, and a regional dimension.

[0010] Further, the step of obtaining initial suppliers of each level in the hierarchical supply chain and obtaining corresponding supplier data according to the level of each of the initial suppliers further comprises: obtaining regulation information of a production location of the specified product and a location of the initial supplier; mapping the regulation information to the preset mathematical space to generate a regulation constraint of the preset mathematical space; mapping the supplier data into the preset mathematical space, and verifying whether the supplier data meets the requirements of the prescribed constraints; optimizing the initial supplier whose verification result does not meet the requirements of the prescribed constraints, so that the optimized initial supplier meets the requirements of the prescribed constraints.

[0011] Further, the step of optimizing the initial supplier whose verification result does not meet the requirements of the mathematical constraints, so that the optimized hierarchical supply chain meets the requirements of the mathematical constraints, further comprises: obtaining the locations of each target supplier of the optimized hierarchical supply chain and the production location of the specified product; obtaining corresponding intelligent compliance certification templates according to the production location and each of the locations, respectively; obtaining compliance data corresponding to each of the intelligent compliance certification templates and filling the intelligent compliance certification templates to obtain intelligent compliance certifications corresponding to each target supplier, respectively.

[0012] An LLM-based cross-border supply chain optimization device, the device comprising: a legal text acquisition module configured to acquire a hierarchical supply chain of a specified product and related legal texts; a legal rule feature acquisition module configured to hierarchically analyze the legal texts by an LLM model to obtain legal rule features; a mathematical constraint generation module configured to map the legal rule features to a preset mathematical space to generate mathematical constraints of the preset mathematical space; a supplier data acquisition module configured to acquire initial suppliers of each level of the hierarchical supply chain and obtain corresponding supplier data according to the level of each initial supplier; a supplier data mapping module configured to map the supplier data into the preset mathematical space and verify whether the supplier data meets the requirements of the mathematical constraints; a first optimization module configured to optimize the initial supplier whose verification result does not meet the requirements of the mathematical constraints, so that the optimized hierarchical supply chain meets the requirements of the mathematical constraints.

[0013] An electronic device comprising a memory and a processor, the memory storing a computer program, the computer program being executed by the processor to cause the processor to perform the following steps: acquiring a hierarchical supply chain of a specified product and related legal texts; hierarchically analyzing the legal texts by an LLM model to obtain legal rule features; mapping the legal rule features to a preset mathematical space to generate mathematical constraints of the preset mathematical space; obtaining initial suppliers of each level in the hierarchical supply chain, and obtaining corresponding supplier data according to the level of each initial supplier; mapping the supplier data into the preset mathematical space, and verifying that the supplier data meets the requirements of the mathematical constraints; optimizing the initial suppliers whose verification results do not meet the requirements of the mathematical constraints, so that the optimized hierarchical supply chain meets the requirements of the mathematical constraints.

[0014] A computer-readable storage medium stores a computer program, which, when executed by a processor, causes the processor to perform the following steps: obtaining a hierarchical supply chain of a specified product and related legal texts; performing hierarchical analysis on the legal texts by an LLM model to obtain legal rule features; mapping the legal rule features to a preset mathematical space to generate mathematical constraints of the preset mathematical space; obtaining initial suppliers of each level in the hierarchical supply chain, and obtaining corresponding supplier data according to the level of each initial supplier; mapping the supplier data into the preset mathematical space, and verifying that the supplier data meets the requirements of the mathematical constraints; optimizing the initial suppliers whose verification results do not meet the requirements of the mathematical constraints, so that the optimized hierarchical supply chain meets the requirements of the mathematical constraints.

[0015] The beneficial effects of the present application are: by performing hierarchical analysis on related legal texts, accurately extracting legal rule features, and then mapping them into a preset mathematical space to form mathematical constraints, automatically identifying and optimizing initial suppliers that do not meet the mathematical constraints, ensuring the compliance and stability of the entire supply chain. Thus, legal risks are reduced, the transparency and flexibility of the supply chain are improved, cost savings and efficiency improvements are achieved, legal requirements can be automatically applied to supply chain management, and the efficiency and accuracy of compliance audits are greatly improved. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0017] wherein: Figure 1 An application environment diagram of the LLM-based cross-border supply chain optimization method in an embodiment; Figure 2 A flowchart of the LLM-based cross-border supply chain optimization method in an embodiment; Figure 3 A structural block diagram of the LLM-based cross-border supply chain optimization device in an embodiment; Figure 4 A structural block diagram of an electronic device in an embodiment. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0019] Figure 1 An application environment diagram of the LLM-based cross-border supply chain optimization method in an embodiment. Referring to Figure 1 , the LLM-based cross-border supply chain optimization method is applied to an LLM-based cross-border supply chain optimization system. The LLM-based cross-border supply chain optimization system includes a terminal 110 and a server 120. The terminal 110 and the server 120 are connected through a network, and the terminal 110 can be a desktop terminal or a mobile terminal, and the mobile terminal can be at least one of a mobile phone, a tablet computer, a notebook computer, etc. The server 120 can be implemented by an independent server or a server cluster composed of multiple servers. The terminal 110 is used to obtain a hierarchical supply chain of a specified product, and the server 120 is used to optimize the hierarchical supply chain.

[0020] As shown in Figure 2 , in an embodiment, an LLM-based cross-border supply chain optimization method is provided. The method can be applied to a terminal or a server, and the present embodiment is exemplified by application to a terminal. The LLM-based cross-border supply chain optimization method specifically includes the following steps: S1: obtaining a hierarchical supply chain of a specified product and related legal texts; S2: hierarchically analyzing the legal texts through an LLM model to obtain legal rule features; S3: mapping the legal rule features to a preset mathematical space to generate mathematical constraints of the preset mathematical space; S4: obtaining initial suppliers of each level in the hierarchical supply chain, and obtaining corresponding supplier data according to the level of each initial supplier; S5: Map the supplier data into the pre-set mathematical space and verify that the supplier data meets the requirements of the mathematical constraints; S6: Optimize the initial suppliers that do not meet the requirements of the mathematical constraints, so that the optimized tiered supply chain meets the requirements of the mathematical constraints.

[0021] This LLM-based cross-border supply chain optimization method can be widely applied to manufacturing and distribution industries in cross-border export. In manufacturing, there are three major pain points in origin verification: supply chain penetration difficulty: multi-level suppliers (such as secondary chip suppliers of A region mobile phone assembly plants) material traceability difficulty; dynamic adaptation of rules: it is difficult to respond to version updates of "cumulative rules" in agreements in real time; complex standard judgment: when meeting the dual standards of cost classification change and regional value component, there is a lack of intelligent decision support, and enterprises face complex supply chain management and legal compliance risks. This method helps enterprises develop supply chain strategies that meet compliance requirements by hierarchical analysis of legal texts, extraction of legal rule features, and mapping to mathematical space. For example, for a specific product, enterprises can identify and optimize initial suppliers that do not meet legal constraints, thereby reducing compliance risks and improving supply chain flexibility and reliability. In the distribution industry, the optimized supply chain can ensure timely delivery of goods while complying with industry regulations, improving customer satisfaction.

[0022] As described in step S1 above, the tiered supply chain for a specified product and related legal texts are obtained. The product to be optimized is clearly defined, and the tiered supply chain information related to the product is extracted by connecting with the enterprise's internal ERP (Enterprise Resource Planning) system or other databases. The tiered supply chain usually includes all relevant nodes from raw material suppliers to final product assembly plants, i.e. Tier1 (direct suppliers), Tier2, Tier3, and more downstream suppliers. This structured information acquisition is a basic element of supply chain management, which helps to identify the function and responsibility of each node in the tiered supply chain. In addition, the system will collect all legal texts related to the product, including trade agreements, regional laws and regulations, industry standards and international treaties. By collecting these legal texts, the system lays a data foundation for subsequent analysis and application.

[0023] As described in step S2 above, the legal text is hierarchically parsed by an LLM model to obtain legal rule features. The obtained legal text is hierarchically parsed using natural language processing (NLP) techniques. Specifically, methods such as semantic segmentation and dependency syntax analysis are used to gradually split the legal text into modules such as clause sections, annotation sections, and example sections. Through this hierarchical parsing, the system can extract key information in the legal clauses, such as mandatory clauses, reference clauses, and their accompanying exceptions. For example, the system can identify core constraints related to country of origin certification and regional value content. After the legal text is parsed, the system will generate a document containing all legal rule features, which will provide practical basis for subsequent mapping to mathematical space and further optimization of decision-making. Specifically, a large language model (LLM) can be used to hierarchically parse legal text, effectively extracting and understanding the hidden legal rule features in the text. Legal texts are often complex in structure, specialized in terminology, and highly abstract, making traditional manual parsing methods inefficient and prone to errors. LLMs, due to their powerful natural language processing capabilities, can fully utilize contextual information and rich knowledge bases to conduct in-depth analysis of legal texts. The process of hierarchical parsing usually includes sentence segmentation, segmenting, and semantic understanding of legal texts. In this process, LLMs identify key information in the text, such as legal clauses, applicable scope, obligations, and rights. This information can be further summarized into a hierarchical structure, such as legal rules, sub-rules, and their applicable conditions, which helps to clearly show the relationship between legal rules. By obtaining legal rule features, LLMs can support more advanced legal research and compliance analysis. For example, when developing compliance policies and supply chain strategies, enterprises can rely on the extracted information to ensure compliance with relevant laws and regulations, reducing legal risks. In addition, the extracted hierarchical features can also provide important basis for subsequent legal document generation, judgment analysis, and related decision-making, promoting the intelligent and automated process of legal services. Therefore, the hierarchical parsing of LLM not only improves the processing efficiency of legal texts, but also provides strong support for intelligent decision-making in legal practice.

[0024] As described in step S3, the legal rule features are mapped to a pre-set mathematical space to generate mathematical constraints of the pre-set mathematical space. By mapping the extracted legal rule features to the pre-set mathematical space, effective mathematical constraints are generated, defining different dimensional quantification standards. The system can describe various constraint conditions in legal clauses. For example, regional value components, product localization rates, and other indicators can be expressed through mathematical formulas, forming a multi-dimensional mathematical constraint model. This model not only provides compliance check standards for enterprises, but also enables rapid adjustment and optimization when laws and regulations are updated in real time, ensuring that the supply chain always meets current regulatory requirements. In addition, this mapping process can use more complex mathematical tools such as tensor analysis to effectively handle and optimize multiple constraint conditions, providing more accurate compliance assessments for enterprises. The pre-set mathematical space is a multi-dimensional space based on vector or tensor structure, used to represent the quantification constraints of legal rules.

[0025] As described in step S4, the initial suppliers of each level in the hierarchical supply chain are obtained, and the corresponding supplier data is obtained according to the level of each initial supplier. It is necessary to delve into the hierarchical supply chain and identify and classify each supplier. According to the obtained hierarchical supply chain information, the database related to each supplier is accessed in turn, and its basic information and related data are extracted, including the geographical location of the supplier, the main supply product, the price, the delivery cycle, the compliance record, etc. By layering the data of the initial suppliers, the system can conduct more detailed evaluations on each node. For example, the violation of Tier1 suppliers may directly affect the compliance of finished products, so a deeper compliance audit will be conducted. The information of Tier2 and Tier3 suppliers is also important, and their compliance status and relative cost-effectiveness will affect the optimization effect of the hierarchical supply chain.

[0026] As described in step S5 above, the supplier data is mapped into the pre-set mathematical space and verified to meet the mathematical constraints. The resulting supplier data is mapped into the pre-set mathematical space and compared with the corresponding mathematical constraints to accurately assess the compliance of each supplier. The specific operation includes substituting all collected supplier data into the established mathematical model to calculate the compliance of the supplier in terms of legal compliance, such as calculating the regional value component and compliance effectiveness of the supplier. In this process, if it is found that the data of a certain supplier does not meet the pre-set mathematical constraints, the system will automatically mark it as an unqualified supplier and generate a detailed compliance report. This function not only quickly identifies potential risk suppliers, but also provides accurate data for subsequent optimization, while ensuring the transparency of the tiered supply chain, allowing enterprises to monitor compliance at each stage in real time and adjust business strategies to minimize legal risks.

[0027] As described in step S6 above, the initial suppliers that do not meet the mathematical constraints are optimized to make the optimized tiered supply chain meet the requirements of the mathematical constraints. The initial suppliers that do not meet the mathematical constraints are optimized to ensure that the final tiered supply chain meets all legal compliance requirements. The optimization process can take various forms, such as replacing non-compliant suppliers with new compliant supply chain members, or guiding and supporting existing suppliers to improve their compliance. In addition, the audit of these non-compliant suppliers can also be increased, feedback can be collected, and targeted improvement measures can be implemented. After optimization, the system will re-evaluate the entire tiered supply chain structure to ensure that all nodes participating in the tiered supply chain meet the latest mathematical constraint conditions. This optimization step not only helps to reduce legal risks, but also improves the competitiveness of enterprises by reducing procurement costs and improving product quality. At the same time, the optimization results and their impact will be incorporated into the system's feedback mechanism to continuously improve and optimize the supply chain management process in the future, forming an efficient and compliant supply chain ecosystem.

[0028] In one embodiment, the step S1 of obtaining the tiered supply chain of the specified product comprises: S101: structurally decomposing the specified product to obtain a BOM tree structure of the specified product; wherein the BOM tree structure comprises a plurality of nodes, each node corresponding to a product component of the specified product; S102: obtaining the initial supplier of each node in the BOM tree structure and the importance of each node; S103: generating the tiered supply chain of the specified product according to the initial supplier of each node and the corresponding importance.

[0029] As described in step S101, the specified product is structurally decomposed to obtain the BOM tree structure of the specified product. The BOM tree structure is a graphical representation of the components of a product and their hierarchical relationships, usually displayed in a tree form, with each node representing a component or part of the product, for example, a BOM tree of a smartphone may contain display screen, processor, battery, shell and other levels, in this process, the system will extract all the components of the product by integrating with the enterprise's ERP system or other related database, and identify the physical and logical relationship of each component. The generation of this tree structure helps to clearly show the hierarchical relationship of the product, and clearly shows the position and function of each component in the overall product. At the same time, the construction of BOM tree also provides basic data for subsequent hierarchical supply chain analysis, which can make the system easier to track each component, and then evaluate the influence of each node in the hierarchical supply chain on the quality and compliance of the final product.

[0030] As described in step S102, the initial supplier of each node in the BOM tree structure and the importance of each node are obtained. The initial supplier refers to the enterprise responsible for providing a specific component or component, for example, the display screen of a smartphone may be provided by a liquid crystal display manufacturer, which is the initial supplier of the display screen node. The importance of the node is evaluated based on multiple factors, such as the impact of the component or component on the performance, cost, compliance, etc. of the final product. The general evaluation principle is that the importance of key components is relatively high, therefore, the suppliers of core components such as processors and batteries will be marked as high priority nodes; while packaging materials or other auxiliary components have less impact on the product, and can be considered to be less important. Through this importance evaluation method, the system can finally determine which nodes occupy a key position in the hierarchical supply chain relationship, and better support subsequent hierarchical supply chain optimization measures.

[0031] As described in step S103 above, based on the importance information of the initial suppliers and nodes in the BOM tree structure obtained in the previous steps, a complete hierarchical supply chain is generated, which involves integrating all initial suppliers and their relationships with various nodes of the product, and forming a hierarchical supply chain model. In the hierarchical supply chain, Tier 1 suppliers are those who directly provide components or assemblies to the product manufacturer, while Tier 2, Tier 3, etc. suppliers provide secondary components or raw materials. The generated hierarchical supply chain not only shows the position of different suppliers in the hierarchical supply chain, but also clearly shows their functional and responsibility relationship with the final product. In this way, the system can clearly express the supply relationship and dependency relationship between each level, and provide the system with how to evaluate and adjust the target of each link in the hierarchical supply chain. Finally, when the overall hierarchical supply chain is formed, the enterprise can adjust according to the actual situation to ensure the scientificity of supplier selection and management, effectively reduce the delivery risk, compliance cost, and improve the overall production efficiency and market competitiveness.

[0032] In one embodiment, after the step S101 of structurally decomposing the designated product to obtain the BOM tree structure of the designated product, the method further comprises: S1021: obtaining the product component corresponding to each node; S1022: obtaining a plurality of alternative suppliers corresponding to the node according to the product component; S1023: calculating the similarity score of each of the initial suppliers of the node and each of the alternative suppliers; S1024: sorting each of the alternative suppliers according to the size of each of the similarity scores.

[0033] As described in step S1021 above, the product components corresponding to each node are obtained. The system extracts the product components corresponding to each node from the previously generated BOM tree structure, regardless of whether it is an electronic product, a mechanical device, or other types of products, each component represents the specific material or component required, and the information of these product components is very important because it is directly related to the performance, reliability and compliance of the product. By analyzing the BOM tree, the system can obtain the specific materials required by each node and combine them with the information of the initial suppliers to further evaluate the effectiveness of the hierarchical supply chain. For example, the mainboard assembly of a mobile phone may contain CPU, memory, wireless module and other product components. This process usually involves the integration of multiple data sources, including ERP (Enterprise Resource Planning) systems, product design documents and supplier files, etc. After extracting the product components, the system can well understand which components are key components and which are secondary components, which not only lays the foundation for subsequent alternative supplier acquisition and similarity evaluation, but also ensures the scientificity and effectiveness of the entire hierarchical supply chain in the material selection and supplier management process.

[0034] As described in step S1022 above, according to the product components of each node extracted, a plurality of alternative suppliers are found and obtained, the alternative suppliers refer to suppliers who can provide similar or similar product components, have comparable technical and supply capabilities with the initial suppliers, and the introduction of alternative suppliers is mainly to enhance the flexibility and resilience of the hierarchical supply chain, and reduce the risk caused by a single supplier. Through the query of various information sources such as industry databases, market research, historical procurement data, etc., the system can effectively identify a plurality of alternative suppliers that match each product component. In most cases, this process also needs to consider factors such as the geographical location, quality certification, delivery capacity and price of the alternative suppliers to ensure that the selected alternative suppliers can meet the actual needs of production. Through such an alternative supplier acquisition mechanism, enterprises can quickly adjust when encountering supplier interruptions, product quality problems or price fluctuations and other unexpected situations, reducing the impact on overall production and compliance. Alternative suppliers can be obtained by "querying a supplier database or an external market data platform".

[0035] As described in step S1023, the similarity score of each node's initial supplier and each alternative supplier is calculated. The system needs to calculate the similarity between the initial supplier of each node and its corresponding alternative supplier, which is a key step in quantitatively evaluating the ability and adaptability of the supplier, aiming to provide data support for subsequent ranking and decision-making. The calculation of the similarity score involves multiple dimensional factors, including the historical performance of the supplier, product quality, delivery speed, price competitiveness, etc. For example, a weighted scoring model can be used to quantitatively evaluate these factors and obtain a comprehensive similarity score. Specifically, the system will first set the weight of each evaluation index, and then score each supplier in detail to obtain the similarity score calculation formula.

[0036] As described in step S1024, the alternative suppliers are ranked according to the calculated similarity scores. The purpose of ranking is to determine which alternative suppliers have the most potential to become the initial supplier, thereby providing support for decision-making. The ranking results will make it easier for managers to identify the most suitable alternative suppliers and make quick adjustments according to actual conditions. The specific implementation usually includes inputting all alternative suppliers and their similarity scores into a ranking algorithm, which can be a simple quicksort or mergesort, etc. The system will first check the similarity scores of all suppliers, and then output the ranked supplier list in descending order. In the final ranking, the supplier with the highest score will be given priority as a candidate or alternative for the initial supplier. This process not only enhances the flexibility of the hierarchical supply chain, but also ensures that it can respond to various changes in the hierarchical supply chain in actual operation, such as certain legal compliance risks, unstable market dynamics, or sudden supply disruptions, etc. This enables the enterprise to maintain competitiveness in a highly competitive market environment.

[0037] In one embodiment, after the step S1022 of obtaining a plurality of alternative suppliers corresponding to the node according to the product components, the method further comprises: S10231: receiving a target function and data information of each alternative supplier; S10232: inputting each data information and the target function into a predetermined mathematical space to optimize and solve the target function, and obtaining a target supply chain; S10233: comparing the target supply chain with the hierarchical supply chain to obtain comparison information; S10234: sending the comparison information to a designated terminal.

[0038] As described in step S10231 above, the objective function and data information of each alternative supplier are received. The objective function is a mathematical expression used to measure system performance during optimization, often involving indicators such as cost, delivery time, compliance, or risk level of the target supply chain. For example, a company may want to minimize total procurement costs while maximizing product quality stability. To effectively optimize, the system needs to clearly define the objective function and ensure it meets business needs. The received alternative supplier data information usually includes relevant characteristics of each alternative supplier, such as delivery time, price, historical performance, compliance, and credibility. These data will be used to analyze the impact of different alternative suppliers on the objective function, providing a basis for subsequent optimization. In most cases, this data information is stored in a structured form in a database, so it can be quickly retrieved and processed directly from the database. By integrating the objective function and alternative supplier data, the system lays a solid foundation for further optimization steps, ensuring that the best supply chain solution can be found. The objective function is a target function set by relevant personnel, such as a total cost minimization function.

[0039] As described in step S10232 above, the received objective function and alternative supplier data information are input into a pre-set mathematical space for optimization solution. This process usually involves using mathematical modeling techniques such as linear programming, integer programming, or other suitable algorithms to achieve optimization of cost, efficiency, or other performance indicators. In this way, the system can generate an ideal target supply chain, i.e. a supply chain structure with optimal performance under specific constraints. In specific operations, the system first converts the objective function into a standard mathematical form and inputs the alternative supplier data information, which may involve supplier delivery capacity, price, compliance record, etc. Then, the system will perform calculations through mathematical algorithms to find the best combination of the target supply chain based on all business conditions and legal compliance requirements. The result of the optimization solution will be a specific supply chain configuration scheme, including the selected alternative suppliers and related ordering strategies.

[0040] As described in step S10233 above, the generated target supply chain is compared with the original hierarchical supply chain to obtain comparison information. Relevant parameters of the target supply chain and the original hierarchical supply chain are extracted, including cost, delivery time, compliance, supplier risk, and other indicators, and are compared one by one. Through comparison, the system can identify the advantages and disadvantages of the target supply chain and understand to what extent the optimization scheme can improve the efficiency and compliance of the overall supply chain. For example, if the total cost of the target supply chain is significantly lower than that of the hierarchical supply chain, or the delivery speed is significantly improved, the system will record these comparison results. In some cases, if the compliance of the target supply chain decreases, the system will also record this negative information. The comparison results will form a detailed comparison report and be provided to decision makers for reference.

[0041] As described in step S10234 above, the obtained comparison information is sent to a designated terminal for viewing and decision making by relevant personnel. The terminal can be a computer, a mobile device, or a terminal interface in an integrated information management system of the management layer. In this way, key decision makers can quickly obtain evaluation results and recommendations about the target supply chain and make more accurate strategic judgments. The comparison information sent usually includes specific comparison data and charts of the target supply chain and the hierarchical supply chain in terms of cost, efficiency, compliance, and other aspects. These intuitive information displays help decision makers quickly understand the pros and cons of different schemes and choose the best scheme to execute.

[0042] In one embodiment, the step S2 of hierarchically analyzing the legal text by the LLM model to obtain legal rule features comprises: S201: hierarchically analyzing the legal text by using a preset semantic segmentation technology to cut the legal text into multiple paragraph modules; S202: obtaining the category attributes of each paragraph module and extracting the paragraph module as a clause paragraph module; S203: constructing a three-dimensional rule feature matrix based on the clause paragraph module as the legal rule feature; wherein the three-dimensional rule feature matrix comprises a validity dimension, a time effectiveness dimension, and a regional dimension.

[0043] As described in step S201 above, the legal text is hierarchically parsed using a pre-set semantic segmentation technique. This process is to effectively cut the legal text into multiple manageable and analyzable paragraph modules. Legal texts are often relatively complex and contain multiple levels of information. Direct analysis of the entire text may lead to low efficiency and misunderstanding. Therefore, the semantic segmentation technique can be used to divide the legal document into paragraphs with independent meaning, such as clauses, definitions, exceptions, and annotations, according to the context and semantics. This technique is usually based on natural language processing (NLP) algorithms, which can identify the logical relationship between sentences and extract important information. In this way, different legal provisions and conditions in the text are effectively extracted to form multiple independent and clearly structured paragraph modules. This not only helps subsequent rule extraction and analysis, but also facilitates the citation and search of specific legal provisions in subsequent work. For example, the provisions of the law on subsidies may be a paragraph module, and the context information is complete. This improves the efficiency of legal text processing and reduces errors that may occur during manual processing, ensuring that the final legal rule features are accurate and reliable.

[0044] As described in step S202 above, the class attribute of each paragraph module is obtained, and the clause paragraph module is further extracted. By analyzing the content and semantics of each paragraph, they are classified into specific legal categories, such as defining clauses, mandatory clauses, reference clauses, and exception clauses. Obtaining the class attribute involves content analysis of each paragraph module, usually using text classification methods, such as machine learning-based classifiers or simple classification methods based on keywords and context rules. Determining whether a paragraph belongs to a specific type of legal clause may involve comparison with previous cases, semantic analysis, and reasoning of context relationships. At the same time, extracting clause paragraph modules from paragraphs will help the system understand the specific meaning and application scenarios of legal rules. Through such classification, the system not only improves the ability to analyze legal texts, but also provides strong data support for optimizing supply chain compliance.

[0045] As described in step S203 above, the extracted clause paragraph module is used to construct a three-dimensional rule feature matrix, which will serve as a specific representation of the acquired legal rule features, aiming to effectively organize and quantify complex legal clause information. The three-dimensional rule feature matrix generally includes dimensions such as validity dimension, time dimension, and regional dimension. Each dimension is a classification and quantification of legal clause attributes, reflecting the characteristics and application of legal rules in different aspects. Validity dimension: reflects the mandatory or reference nature of the clause, such as some clauses in regulations are mandatory clauses that must be followed, while others are recommended clauses, and some clauses may have exceptions. Time dimension: shows the effective time and applicable time of different clauses, such as some legal clauses are only valid in a specific year, or there are transitional period provisions before the effective date of a new revised clause. Regional dimension: represents the geographical scope of the application of legal clauses, which may involve the unique requirements of different countries, regions or economic zones, for example, a clause is only valid in a specific country or region. The validity dimension represents the strength and degree of constraint of the rule, and its value and coding are as follows: Value: Mandatory (Mandatory): This rule must be followed and cannot be violated. Recommended (Recommended): This rule is recommended and whether to follow it can be decided according to specific circumstances. Exceptional (Exceptional): This rule applies to specific situations or individuals and is not generally applicable. Coding: Mandatory = 1; Recommended = 2; Exceptional = 3. The time dimension represents the time limit of the rule's application, and its value and coding are as follows: Value: Perpetual (Perpetual): The rule is valid for the entire validity period. Temporary (Temporary): The rule is valid within a specific time range, usually limited to start and end dates. Expired (Expired): The rule has expired and is no longer applicable. Coding: Perpetual = 1; Temporary = 2; Expired = 3. The regional dimension represents the geographical scope of the rule's application, and its value and coding are as follows: Value: National (National): The rule applies to all regions worldwide. Regional (Regional): The rule applies to a specific region or province. Local (Local): The rule only applies to a specific city or organization. Coding: Global = 1; Regional = 2; Local = 3.

[0046] In this way, the three-dimensional rule feature matrix constructed by the system not only organizes the legal clauses in a data structure, but also provides basic data for subsequent automated compliance checking and decision making. Using this matrix, the system can parallel process information in different dimensions, perform intelligent analysis and reasoning, and achieve efficient application of legal rules in the hierarchical supply chain optimization process.

[0047] In one embodiment, after the step S4 of obtaining the initial suppliers of each level in the hierarchical supply chain and obtaining the corresponding supplier data according to the level of each initial supplier, the method further comprises: S501: Obtain the specified information of the production location of the specified product and the location of the initial supplier; S502: Map the specified information to the preset mathematical space to generate the specified constraints of the preset mathematical space; S503: Map the supplier data to the preset mathematical space and verify whether the supplier data meets the specified constraints; S504: Optimize the initial supplier whose verification result does not meet the specified constraints so that the optimized initial supplier meets the specified constraints.

[0048] As described in step S501, the specified information of the production location of the specified product and the location of each initial supplier is collected and integrated. This information is crucial for understanding the compliance risks, trade barriers and legal environment of different geographical regions. The rule factors directly affect the stability and reliability of the hierarchical supply chain. The system needs to evaluate which regions' risks may affect the delivery and compliance of the product. In actual operation, the system can obtain this information through multiple channels, including rule information analysis databases, relevant reports published by specified agencies, industry analysis data and market research of professional consulting agencies, etc. The system collects information such as whether there is a compliance risk in the country where the initial supplier is located. These information helps to form a comprehensive risk assessment of each supplier, providing support for subsequent optimization decisions, and ensuring that the risks affecting major supply decisions are fully considered in the process of optimizing the hierarchical supply chain, thereby improving the safety and compliance of the overall hierarchical supply chain.

[0049] As described in step S502, the collected regulatory information is mapped into the pre-set mathematical space, and then the regulatory constraints are generated. This process involves quantifying and standardizing the regulatory information risk, so that it can be effectively managed and analyzed in the established mathematical framework. Generally, the mapping of regulatory information creates a set of variables, which can include geopolitical risk scores, trade rule compliance, diplomatic stability scores, etc. To achieve this mapping, data modeling methods such as multiple linear regression analysis or support vector machines (SVM) can be used to convert the regulatory information risk into a form suitable for processing in the mathematical space. Specifically, a regional risk scoring model can be used to score the regulatory information, which is then converted into a form in the mathematical space. For example, a risk scoring system can be established, which uses the weighted combination of various information to form a comprehensive score, and these scores correspond to specific constraint conditions in the mathematical space. In this way, the system can determine which regions where the initial suppliers are located have a higher compliance risk in the current context, and integrate this information as regulatory constraints into the subsequent hierarchical supply chain management and optimization strategies. This provides stronger data orientation for hierarchical supply chain decision-making, ensuring that relevant policy and risk factors are considered when selecting partners, thereby protecting the interests of enterprises in the international trade environment.

[0050] As described in step S503, the supplier data is mapped into the pre-set mathematical space, and the supplier data is verified to meet the requirements of the regulatory constraints. The collected supplier data is mapped into the pre-set mathematical space to verify whether it meets the previously set regulatory constraints. Specifically, the system first extracts relevant supplier data, such as the geographical location of the supplier, trade history, stability of past cooperation, etc. Then, the system uses the established regulatory constraints to verify these data in real time. This process may include quantitative analysis and logical judgment to determine whether the regulatory information risk of the region where the supplier is located will affect its delivery capability. For example, if a supplier is located in a high-risk region, the system will mark it based on this and warn the user that the supplier has a high geopolitical risk. Through this verification, the system not only effectively filters compliant suppliers, but also makes timely adjustments to ensure the overall health of the hierarchical supply chain in terms of legal and policy aspects.

[0051] As described above in step S504, the initial suppliers whose verification results show that they do not meet the prescribed constraints are optimized, which includes evaluating the characteristics of the unqualified suppliers and determining specific strategies that need to be replaced, modified or improved to ensure that a final compliance level supply chain that meets the prescribed constraints is formed. The specific method of optimization can include various strategies, such as finding new alternative suppliers, modifying supplier management processes, or communicating with existing suppliers to improve their compliance. At the same time, the system can use historical data analysis techniques to evaluate the background, stability and past delivery capabilities of potential alternative suppliers, so as to select the most suitable alternative. In addition, by establishing closer cooperation with suppliers who meet the regulatory requirements, enterprises can not only reduce risks, but also improve the flexibility and resilience of the tiered supply chain.

[0052] In one embodiment, the step S6 of optimizing the initial suppliers whose verification results do not meet the requirements of the mathematical constraints so that the optimized tiered supply chain meets the requirements of the mathematical constraints further comprises: S701: obtaining the locations of each target supplier of the optimized tiered supply chain and the production location of the specified product; S702: obtaining the corresponding intelligent compliance certification templates according to the production location and each of the locations; S703: obtaining the compliance data corresponding to each of the intelligent compliance certification templates and filling it into the intelligent compliance certification templates to obtain the intelligent compliance certification corresponding to each target supplier respectively.

[0053] As described above in step S701, the locations of each target supplier of the optimized tiered supply chain and the production location of the specified product are obtained. The geographic positions of all target suppliers in the optimized tiered supply chain are obtained, and the production location of the specified product is determined. Obtaining the geographic information of the target suppliers involves extracting relevant data from the enterprise's ERP system or supply chain management system, which usually includes the detailed address, country, region, etc. of the target suppliers. On the other hand, the information of the production location also needs to be extracted from the corresponding product data. For example, regulations, trade barriers, etc. in different countries or regions may affect the requirements of manufacturers in terms of compliance. By organizing and listing these geographic information, the system can provide necessary background support for the generation of compliance certification templates in the following step, laying a foundation for ensuring the compliance of each target supplier.

[0054] As described in step S702 above, according to the previously obtained production location and target supplier location, the corresponding intelligent compliance certification template is obtained respectively. Obtaining the compliance certification template usually needs to be combined with legal databases, industry standards and designated agency departments, etc. The system can quickly identify the most suitable compliance certification template based on the information of the production location and the supplier location. For example, if a product is produced in region A and the target supplier is located in region B, the system can refer to the compliance certification template under the free trade agreement between region A and region B. These templates usually contain standard legal texts and structures to effectively present the required compliance information.

[0055] As described in step S703 above, the compliance data corresponding to each intelligent compliance certification template is obtained, and these data are filled into the obtained template to generate the intelligent compliance certification of each target supplier. Specifically, the system first collects applicable compliance data from company databases, financial records, audit reports, etc. This may include product of origin certification, material compliance, production standards, supply chain transparency, etc. These data will be filled into the corresponding compliance certification template by the system to generate a formatted proof file. When processing data, the system needs to ensure the accuracy and consistency of the filled information to avoid legal and compliance risks due to data errors. Through the interface or crawler technology, the relevant data is automatically extracted from the compliance database and filled into the template.

[0056] Referring to Figure 3 The application also provides a cross-border supply chain optimization device based on LLM, which comprises: A legal text acquisition module 902 is configured to acquire the hierarchical supply chain of a specified product and related legal texts; A legal rule feature acquisition module 904 is configured to hierarchically analyze the legal texts by an LLM model to acquire legal rule features; A mathematical constraint generation module 906 is configured to map the legal rule features to a preset mathematical space to generate mathematical constraints of the preset mathematical space; A supplier data acquisition module 908 is configured to acquire initial suppliers of each level in the hierarchical supply chain and acquire corresponding supplier data according to the level of each initial supplier; A supplier data mapping module 910 is configured to map the supplier data into the preset mathematical space and verify whether the supplier data meets the requirements of the mathematical constraints; A first optimization module 912 is configured to optimize the initial suppliers whose verification results do not meet the requirements of the mathematical constraints, so that the optimized hierarchical supply chain meets the requirements of the mathematical constraints.

[0057] In an embodiment, the legal text acquisition module 902 comprises: a structure decomposition submodule configured to perform structure decomposition on the specified product to acquire a BOM tree structure of the specified product, wherein the BOM tree structure comprises a plurality of nodes, and each node corresponds to a product component of the specified product; an initial supplier acquisition submodule configured to acquire initial suppliers of each node in the BOM tree structure and importance of each node; a hierarchical supply chain submodule configured to generate a hierarchical supply chain of the specified product according to the initial suppliers of each node and the corresponding importance.

[0058] In an embodiment, the legal text acquisition module 902 further comprises: a product component acquisition submodule configured to acquire the product component corresponding to each node; a substitute supplier acquisition submodule configured to acquire a plurality of substitute suppliers corresponding to each node according to the product component; a similarity score calculation submodule configured to calculate a similarity score between the initial supplier of each node and each substitute supplier; a sorting submodule configured to sort each substitute supplier according to the size of each similarity score.

[0059] In an embodiment, the legal text acquisition module 902 further comprises: a data information receiving submodule configured to receive a target function and data information of each substitute supplier; a target supply chain acquisition submodule configured to input each data information and the target function into a preset mathematical space to optimize and solve the target function to obtain a target supply chain; a comparison information acquisition submodule configured to compare the target supply chain with the hierarchical supply chain to obtain comparison information; a comparison information sending submodule configured to send the comparison information to a specified terminal.

[0060] In an embodiment, the legal rule feature acquisition module 904 comprises: a legal text cutting submodule configured to use a preset semantic segmentation technique to hierarchically analyze the legal text to cut the legal text into a plurality of paragraph modules; a category attribute acquisition submodule configured to acquire category attributes of each paragraph module and extract the paragraph module as a clause paragraph module; The three-dimensional rule feature matrix construction submodule is configured to construct a three-dimensional rule feature matrix based on the clause paragraph module, so as to serve as the legal rule feature; and the three-dimensional rule feature matrix comprises a validity dimension, a time limit dimension, and a region dimension.

[0061] In one embodiment, the LLM-based cross-border supply chain optimization device further comprises: The regulation information acquisition module is configured to acquire regulation information of the production location of the specified product and the location of the initial supplier. The regulation information mapping module is configured to map the regulation information to the preset mathematical space, so as to generate a regulation constraint of the preset mathematical space. The verification module is configured to map the supplier data to the preset mathematical space, and verify whether the supplier data meets the regulation constraint. The second optimization module is configured to optimize the initial supplier that does not meet the regulation constraint, so that the optimized initial supplier meets the regulation constraint.

[0062] In one embodiment, the LLM-based cross-border supply chain optimization device further comprises: The production location acquisition module is configured to acquire the location of each target supplier of the optimized hierarchical supply chain and the production location of the specified product. The intelligent compliance certification template acquisition module is configured to acquire a corresponding intelligent compliance certification template according to the production location and each location. The compliance data acquisition module is configured to acquire compliance data corresponding to each intelligent compliance certification template, and fill the intelligent compliance certification template, so as to obtain an intelligent compliance certification corresponding to each target supplier.

[0063] Figure 4 An internal structure diagram of an electronic device in one embodiment is shown. The electronic device can be a terminal or a server, and can be a computer device. As shown in the figure, Figure 4 The electronic device includes a processor, a memory, and a network interface connected by a system bus. The memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the electronic device stores an operating system, and can also store a computer program. When the computer program is executed by the processor, the processor can implement the LLM-based cross-border supply chain optimization method. The internal memory can also store a computer program. When the computer program is executed by the processor, the processor can execute the LLM-based cross-border supply chain optimization method. Those skilled in the art can understand, Figure 4The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the electronic device to which the scheme of the present application is applied. The specific electronic device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0064] In one embodiment, an electronic device is provided, comprising a memory and a processor, the memory storing a computer program, the computer program being executed by the processor to cause the processor to perform the following steps: obtain a hierarchical supply chain of a specified product and related legal texts; perform hierarchical analysis on the legal texts by an LLM model to obtain legal rule features; map the legal rule features to a preset mathematical space to generate mathematical constraints of the preset mathematical space; obtain initial suppliers of each level in the hierarchical supply chain, and obtain corresponding supplier data according to the level of each initial supplier; map the supplier data into the preset mathematical space, and verify that the supplier data meets the requirements of the mathematical constraints; optimize the initial suppliers whose verification results do not meet the requirements of the mathematical constraints, so that the optimized hierarchical supply chain meets the requirements of the mathematical constraints.

[0065] By performing hierarchical analysis on related legal texts, accurately extracting legal rule features, and then mapping them into a preset mathematical space to form mathematical constraints, automatically identifying and optimizing initial suppliers that do not meet the mathematical constraints, the compliance and stability of the entire supply chain are ensured. Thus, legal risks are reduced, the transparency and flexibility of the supply chain are improved, cost savings and efficiency improvements are achieved, legal requirements can be automatically applied to supply chain management, and the efficiency and accuracy of compliance audits are greatly improved.

[0066] In one embodiment, a computer-readable storage medium is provided, storing a computer program, the computer program being executed by a processor to cause the processor to perform the following steps: obtain a hierarchical supply chain of a specified product and related legal texts; perform hierarchical analysis on the legal texts by an LLM model to obtain legal rule features; map the legal rule features to a preset mathematical space to generate mathematical constraints of the preset mathematical space; obtain initial suppliers of each level in the hierarchical supply chain, and obtain corresponding supplier data according to the level of each initial supplier; mapping the supplier data into the preset mathematical space, and verifying that the supplier data meets the requirements of the mathematical constraints; optimizing an initial supplier that does not meet the requirements of the mathematical constraints, so that the optimized hierarchical supply chain meets the requirements of the mathematical constraints.

[0067] By hierarchically analyzing relevant legal texts, accurately extracting legal rule features, and then mapping them into a preset mathematical space to form mathematical constraints, the initial suppliers that do not meet the mathematical constraints are automatically identified and optimized to ensure the compliance and stability of the entire supply chain. This reduces legal risks, improves the transparency and flexibility of the supply chain, saves costs and improves efficiency, and enables legal requirements to be automatically applied to supply chain management, greatly improving the efficiency and accuracy of compliance audits.

[0068] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by a computer program instructing related hardware. The program can be stored in a non-volatile computer readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiments. Any reference to memory, storage, database or other medium used in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0069] The technical features of the above embodiments can be combined in any way. To make the description concise, not all possible combinations of the technical features in the above embodiments are described, but as long as the combinations of the technical features do not exist, they should be considered as the scope of the present disclosure.

[0070] The above-described embodiments are merely illustrative of several embodiments of the present application, which are described in more detail and in a specific manner, but should not be construed as limiting the scope of the patent of the present application. It should be noted that, for those of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A cross-border supply chain optimization method based on LLM, characterized in that, The method includes: Obtain the hierarchical supply chain and related legal documents for the specified product; The legal text is hierarchically parsed using an LLM model to obtain legal rule features; The legal rule features are mapped to a preset mathematical space to generate mathematical constraints in the preset mathematical space; Obtain the initial suppliers at each level of the hierarchical supply chain, and obtain the corresponding supplier data according to the level of each initial supplier; The supplier data is mapped into the preset mathematical space, and the supplier data is verified to meet the requirements of the mathematical constraints. The initial suppliers whose verification results do not meet the requirements of the mathematical constraints are optimized so that the optimized hierarchical supply chain meets the requirements of the mathematical constraints.

2. The LLM-based cross-border supply chain optimization method according to claim 1, characterized in that, The step of obtaining the hierarchical supply chain for the specified product includes: The specified product is structurally decomposed to obtain the BOM tree structure of the specified product; wherein the BOM tree structure includes multiple nodes, and each node corresponds to a product component of the specified product; Obtain the initial suppliers and importance of each node in the BOM tree structure; A hierarchical supply chain for the specified product is generated based on the initial suppliers of each node and their corresponding importance.

3. The LLM-based cross-border supply chain optimization method according to claim 2, characterized in that, After the step of structurally decomposing the specified product to obtain the BOM tree structure of the specified product, the method further includes: Obtain the product components corresponding to each node; Based on the product composition, obtain multiple alternative suppliers corresponding to the node; Calculate the similarity score between the initial supplier and each of the alternative suppliers for each of the nodes; The alternative suppliers are sorted according to the magnitude of their respective similarity scores.

4. The LLM-based cross-border supply chain optimization method according to claim 3, characterized in that, After the step of obtaining multiple alternative suppliers corresponding to the node based on the product composition, the method further includes: Receive the objective function and data information from each of the alternative suppliers; The data and objective function are input into a preset mathematical space to optimize and solve the objective function, thereby obtaining the target supply chain. By comparing the target supply chain with the hierarchical supply chain, comparison information is obtained; The comparison information is sent to the designated terminal.

5. The LLM-based cross-border supply chain optimization method according to claim 1, characterized in that, The step of performing hierarchical parsing of the legal text using an LLM model to obtain legal rule features includes: The legal text is hierarchically parsed using a pre-defined semantic segmentation technique to divide it into multiple paragraph modules; Obtain the category attributes of each paragraph module and extract the paragraph modules as clause paragraph modules; A three-dimensional rule feature matrix is ​​constructed based on the aforementioned clause paragraph module to serve as the legal rule feature; wherein, the three-dimensional rule feature matrix includes a validity dimension, a timeliness dimension, and a geographical dimension.

6. The LLM-based cross-border supply chain optimization method according to claim 1, characterized in that, After the steps of obtaining the initial suppliers at each level of the hierarchical supply chain and obtaining the corresponding supplier data according to the level of each initial supplier, the method further includes: Obtain the specified information regarding the production location of the designated product and the location of the initial supplier; The specified information is mapped to the preset mathematical space to generate the specified constraints of the preset mathematical space; The supplier data is mapped into the preset mathematical space, and the supplier data is verified to meet the specified constraints. The initial suppliers whose verification results show that they do not meet the requirements of the specified constraints are optimized so that the optimized initial suppliers meet the requirements of the specified constraints.

7. The LLM-based cross-border supply chain optimization method according to claim 1, characterized in that, After the step of optimizing the initial suppliers whose verification results do not meet the requirements of the mathematical constraints, so that the optimized hierarchical supply chain meets the requirements of the mathematical constraints, the method further includes: Obtain the locations of each target supplier in the optimized hierarchical supply chain, as well as the production location of the specified product; Obtain the corresponding smart compliance certificate template based on the production location and each of the aforementioned locations; Obtain the compliance data corresponding to each of the aforementioned smart compliance certificate templates and fill it into the smart compliance certificate templates to obtain the smart compliance certificates corresponding to each target supplier.

8. A cross-border supply chain optimization device based on LLM, characterized in that, The device includes: The legal document acquisition module is used to acquire the hierarchical supply chain and related legal documents for a specified product. The legal rule feature acquisition module is used to perform hierarchical parsing of the legal text using an LLM model to obtain legal rule features; A mathematical constraint generation module is used to map the legal rule features to a preset mathematical space to generate mathematical constraints in the preset mathematical space. The supplier data acquisition module is used to acquire the initial suppliers at each level of the hierarchical supply chain, and to acquire the corresponding supplier data according to the level of each initial supplier. The supplier data mapping module is used to map supplier data into the preset mathematical space and verify that the supplier data meets the requirements of the mathematical constraints. The first optimization module is used to optimize the initial suppliers whose verification results do not meet the requirements of the mathematical constraints, so that the optimized hierarchical supply chain meets the requirements of the mathematical constraints.

9. A computer-readable storage medium, characterized in that, The system contains a computer program that, when executed by a processor, causes the processor to perform the steps of the LLM-based cross-border supply chain optimization method as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, The device includes a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the LLM-based cross-border supply chain optimization method as described in any one of claims 1 to 7.

Images