Enterprise risk calculation method and system based on industry chain knowledge graph

Abstract

The application provides an enterprise risk calculation method and system based on an industry chain knowledge graph, relates to the technical field of knowledge graph application, and comprises the following steps: in the industry chain knowledge graph, taking a risk source wave as a starting enterprise, taking an enterprise to be calculated for an event total risk value as a terminal enterprise, and connecting edges between enterprises as a type of edge; traversing a type of edge with the same relationship type in the knowledge graph, and enumerating a risk transmission path between the starting enterprise and the terminal enterprise; obtaining an initial risk intensity according to a risk source type; obtaining a conduction weight scalar, a comprehensive attenuation value and a correlation closeness based on node attributes and edge attributes of enterprises on the risk transmission path; calculating a path transmission risk intensity of each risk transmission path based on the conduction weight scalar, the comprehensive attenuation value, the initial risk intensity and the correlation closeness; and superimposing probabilities of multiple risk transmission paths to obtain an event total risk value of the terminal enterprise, so that the event total risk value is close to a real state.

Description

Technical Field

[0001] This invention relates to the field of knowledge graph application technology, and in particular to a method and system for enterprise risk estimation based on industry chain knowledge graph. Background Technology

[0002] With the deepening of globalization and informatization, business collaboration and data interaction among enterprises are becoming increasingly frequent, and the industrial chain structure exhibits highly dynamic and strongly coupled characteristics. The collaborative relationships between enterprises in the industrial chain are a key foundation for product production and value realization. However, each link in the industrial chain generally faces various internal and external risks, including but not limited to supplier delivery delays, raw material price fluctuations, sharp drops in product demand, overdue customer payments, and production equipment failures. These risks not only affect the operational stability of individual enterprises but may also be transmitted and amplified along the industrial chain, ultimately leading to partial or even systemic industrial chain disruptions. Therefore, industrial chain knowledge graphs have emerged as an effective tool for structurally representing various entities and their relationships within the industrial chain, supporting the analysis and assessment of the propagation of risk source impacts among enterprises.

[0003] Currently, the mainstream method for calculating and assessing risk impact intensity based on industry chain knowledge graphs is as follows: First, risk propagation paths are generated by traversing fixed node connections in the knowledge graph according to predefined rules; then, the risk value is quantified using several simple indicators (such as node attribute values, path hop count, etc.). However, the above technical solutions have the following significant shortcomings in practical applications. For example, traditional risk quantification methods generally only focus on the node's own attributes (such as enterprise size, credit rating) or path topology characteristics (such as propagation hop count), lacking comprehensive modeling of the impact of multiple factors. Specifically, existing methods do not consider dynamic related factors such as time decay effects, redundancy between paths, availability of alternative resources, and business dependence intensity, leading to a significant deviation between the assessed risk impact intensity and the actual impact range. Summary of the Invention

[0004] This invention provides a method and system for enterprise risk estimation based on industry chain knowledge graphs, in order to solve the technical problem that the assessment results of risk impact intensity have significant deviations from the actual impact range due to the simplicity of the indicators.

[0005] This invention provides a method for enterprise risk estimation based on a supply chain knowledge graph, comprising: Construct entities for the industry chain knowledge graph and establish relationships between entities. Extract knowledge from various defined data sources to form the industry chain knowledge graph. Starting with the enterprises affected by the risk source, and ending with the enterprise whose total risk value of the event to be measured, the edges connecting enterprises are considered as a type of edge. Traversing along the edges of the same type in the industrial chain knowledge graph, the risk propagation path between the starting enterprise and the ending enterprise is enumerated. The relationship type of the edge is one of the following: competitive relationship, cooperative relationship, supply relationship, and innovation relationship. Based on the stated risk source type, the initial risk intensity is obtained; Based on the node attributes and edge attributes of enterprises on the risk propagation path, the transmission weight scalar, the comprehensive attenuation value and the correlation tightness are obtained. The transmission weight scalar is used to quantify the impact of risk transmission corresponding to a certain type of edge relationship. Based on the transmission weight scalar, the comprehensive attenuation value, the initial risk intensity, and the correlation tightness, the path transmission risk intensity of each risk propagation path is calculated; The probabilities of multiple risk propagation paths are summed to obtain the total risk value of the event for the endpoint enterprise.

[0006] In one embodiment of the present invention, the entity includes a type of node and a type of node, wherein the type of node is an enterprise, and the type of node includes at least one of products, technology or talent. The edges formed between the first type of nodes are called first-type edges, and the edges formed between the first type of nodes or the second type of nodes are called second-type edges. When a new event is imported into the industry chain knowledge graph, the corresponding first-type edges are updated according to the second-type edges.

[0007] In one embodiment of the present invention, after constructing entities of a supply chain knowledge graph and establishing relationships between entities, extracting knowledge from various set data sources to form a supply chain knowledge graph, traversing along edges of the same relationship type in the knowledge graph, and enumerating risk propagation paths that satisfy both the starting and ending points of enterprises, the process includes: Capture various real-time information related to entities; Extract the publisher information from the real-time information; Based on the publisher information, the information introduction degree is obtained; The information introduction degree threshold and the preset introduction degree are compared. If the information introduction degree is greater than the preset introduction degree threshold, the real-time information is imported into the industry chain knowledge graph and the industry chain knowledge graph is updated. If the information introduction degree is less than or equal to the preset introduction degree threshold, the real-time information is discarded.

[0008] In one embodiment of the present invention, after constructing entities of a supply chain knowledge graph and establishing relationships between entities, extracting knowledge from various set data sources to form a supply chain knowledge graph, and before enumerating the risk propagation path between the starting enterprise and the ending enterprise, starting from the enterprises affected by the risk source, traversing along the edges of similar relationships in the knowledge graph, the process includes: The real-time information captured is processed based on the aforementioned industry chain knowledge graph, and the relationship between the relevant entities and events is marked. The real-time information includes at least enterprise operation change information, talent change information, patent technology change information, and upstream and downstream supply and demand change information. Based on the annotation results, four types of data are distinguished: newly added entities, newly added associated edges, changes in the attributes of existing entities, and invalid existing relationships. The annotated entity data and relationship data are then classified and stored in the database to complete the incremental update and existing correction of the industry chain knowledge graph.

[0009] In one embodiment of the present invention, the probabilities of multiple risk propagation paths are superimposed to obtain the total event risk value of the enterprise, including: Calculate the betweenness centrality and degree centrality of the starting enterprise on the topology of the industry chain knowledge graph. Based on the betweenness centrality and degree centrality, obtain the importance correction parameter to correct the total event risk value of the enterprise.

[0010] In one embodiment of the present invention, based on the node attributes and edge attributes of enterprises along the risk propagation path, a transmission weight scalar, a comprehensive attenuation value, and a correlation tightness are obtained, including: The transmission weight scalar is obtained based on the relationship type of the edge attributes in the risk transmission path and the mapping relationship between the relationship type and the preset transmission weight scalar.

[0011] In one embodiment of the present invention, based on the node attributes and edge attributes of enterprises along the risk propagation path, a transmission weight scalar, a comprehensive attenuation value, and a correlation tightness are obtained, including: Based on the difference between the generation timestamp of the relation edges in the aforementioned industry chain knowledge graph and the current time, the time-decrease factor is obtained; Based on the relationship type of the edge attributes in the risk transmission path, the relationship type risk transmission coefficient is obtained; Obtain the geographical coordinates of the starting enterprise and the ending enterprise, calculate the straight-line distance between the starting enterprise and the ending enterprise based on the geographical coordinates, count the number of hops in the link network between the starting enterprise and the ending enterprise in the industry chain knowledge graph topology, and obtain a dual distance attenuation factor based on the number of hops and the straight-line distance. Based on the time-decrease factor, the relationship type risk transmission coefficient, and the dual distance decrease factor, a comprehensive decrease value is obtained.

[0012] In one embodiment of the present invention, based on the node attributes and edge attributes of enterprises along the risk propagation path, a transmission weight scalar, a comprehensive attenuation value, and a correlation tightness are obtained, including: When the relationship type of one edge in the risk propagation path is an innovation relationship, the number of shared patents between the starting enterprise and the ending enterprise and the shortest patent citation path length from the starting enterprise to the ending enterprise are extracted from the industrial chain knowledge graph. Based on the number of shared patents and the patent citation path length, the correlation tightness is obtained. The statistics of the shortest patent citation path length are obtained by relying on the citation association data of the intermediate enterprises and their patents along the risk propagation path. When the relationship type of one edge in the risk propagation path is a competitive relationship, the overlapping sub-product categories between the starting enterprise and the ending enterprise are extracted from the industry chain knowledge graph, and the correlation tightness is obtained based on the operating indicators of the sub-product categories; wherein, the operating indicators are any one of operating scale, revenue, production capacity, output or shipment amount; When the relationship type of one edge on the risk propagation path is a supply relationship, the adjacent upstream and downstream cooperative enterprises on the risk propagation path are located, where the supplier is the upstream enterprise and the purchaser is the downstream enterprise; the target product purchased by the downstream enterprise from the corresponding upstream is extracted from the industry chain knowledge graph, and based on the annual purchase amount of the target product, the total annual purchase amount of the target product by the downstream enterprise for the target product, and the total number of upstream enterprises that can supply the target product to the downstream enterprise; based on the annual purchase amount, the total purchase amount, and the total number of upstream enterprises, the supply chain coupling tightness coefficient between two adjacent enterprises on the risk propagation path is obtained, and based on the supply chain coupling tightness coefficient, the association tightness is obtained; When the relationship type of one type of edge in the risk propagation path is a cooperative relationship, the quantitative value of strong contractual binding cooperation between the starting enterprise and the ending enterprise and the quantitative value of business cooperation in industrialization projects are extracted from the risk propagation path in the industrial chain knowledge graph. Based on the quantitative value of strong contractual binding cooperation and the quantitative value of business cooperation in industrialization projects, the correlation tightness is calculated.

[0013] In one embodiment of the present invention, after superimposing the probabilities of multiple risk propagation paths to obtain the total event risk value of the endpoint enterprise, the method includes: Calculate the total risk value of the events for the relevant enterprises in the aforementioned industry chain knowledge graph; Based on the total risk value of the event and the preset risk threshold, high-risk enterprises in the industry chain knowledge graph are screened and identified, and the screened high-risk enterprises are visualized.

[0014] This invention also provides a business risk estimation system based on a supply chain knowledge graph, comprising: The graph construction module is used to construct entities in the industry chain knowledge graph and establish relationships between entities. It extracts knowledge from various set data sources to form the industry chain knowledge graph. Path selection module: It is used to start from the enterprises affected by the risk source and end from the enterprises whose total risk value of the event to be measured. Enterprises are connected by a type of edge. It traverses along the same type of edge in the knowledge graph and enumerates the risk propagation paths that satisfy both the starting point and the ending point as enterprises. The relationship type of the edge is one of the following: competitive relationship, cooperative relationship, supply relationship and innovation relationship. Analysis and Calculation Module: Based on the risk source type, the analysis and calculation module is used to obtain the initial risk intensity; based on the node attributes and edge attributes of enterprises on the risk propagation path, the analysis and calculation module is used to obtain the transmission weight scalar, the comprehensive attenuation value, and the correlation tightness, whereby the transmission weight scalar is used to quantify the impact of risk transmission corresponding to a certain type of edge relationship; based on the transmission weight scalar, the comprehensive attenuation value, the initial risk intensity, and the correlation tightness, the analysis and calculation module is used to calculate the path transmission risk intensity of each of the risk propagation paths; by superimposing the probabilities of multiple risk propagation paths, the analysis and calculation module is used to obtain the total event risk value of the endpoint enterprise.

[0015] The beneficial effects of this invention are as follows: The enterprise risk estimation method and system based on the industrial chain knowledge graph proposed in this invention calculates the path transmission risk intensity of each risk propagation path by introducing initial risk intensity, transmission weight scalar, comprehensive attenuation value and correlation tightness. By introducing multiple parameters, the calculation of the path transmission risk intensity can take into account multiple influencing factors, making the value of the path transmission risk intensity closer to the actual state, and obtaining a more accurate total event risk value for the terminal enterprise. This improves the accuracy of the enterprise's calculation of the risk impact intensity of different risk sources, so as to predict enterprise risks in advance and intervene in a timely manner. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0017] In the attached diagram: Figure 1 A flowchart illustrating a method for estimating enterprise risk based on a supply chain knowledge graph, provided in an embodiment of the present invention; Figure 2 A flowchart illustrating another enterprise risk estimation method based on supply chain knowledge graph provided in an embodiment of the present invention; Figure 3 A flowchart illustrating another enterprise risk estimation method based on supply chain knowledge graph provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of an enterprise risk estimation system based on a supply chain knowledge graph, provided in an embodiment of the present invention. Detailed Implementation

[0018] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0019] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. The drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the shape, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0020] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0021] Please see Figure 1 An embodiment of the present invention provides a method for enterprise risk estimation based on a supply chain knowledge graph, comprising: Step S110: Construct the entities of the industry chain knowledge graph and establish the relationships between entities. Extract knowledge from various set data sources to form the industry chain knowledge graph.

[0022] In this application, entities include type I nodes and type II nodes. Type I nodes are enterprises, and type II nodes include at least one of products, technologies, or talents. The edges formed between type I nodes are type I edges, and the edges formed between type I nodes or type II nodes are type II edges. When a new event is imported into the industry chain knowledge graph, the corresponding type I edges are updated according to the type II edges.

[0023] Each of the four entities has its own corresponding attribute diagram. The following are examples of the attribute diagrams for the four entities: Enterprises: They belong to a type of entity, with the Unified Social Credit Code (unique identifier of the enterprise) as the primary key. The attributes of an enterprise entity include its unique identifier, enterprise name, region, industry, position in the industrial chain, revenue scale, listing status, core products, production capacity data, compliance status, etc. Products: These belong to the second category of entities. The primary key is product model + product category. The product entity is the core carrier of supply and demand in the industrial chain. The attributes of the product entity include product category, product name, product model, industrial chain link, technical parameters, manufacturer, application scenario, market share, etc. Technology: It belongs to the second type of entity, with the patent application number as the primary key. Technology entities are the core elements of innovation in the industrial chain. The attributes of technology entities include patent / technology name, patent application number, IPC classification number, technical field, legal status, R&D entity, citation relationship, application / authorization time, etc. Talent: It belongs to the second type of entity, with the ID card number as the primary key. The talent entity is the core subject of innovation in the industrial chain. The attributes of the talent entity include the person's name, affiliated enterprise / institution, research field, patent achievements, industry qualifications, core positions, etc.

[0024] This application uses a first-type edge to carry structured relationships between enterprises that can be used for risk transmission, and a second-type edge to carry the original underlying relationships between enterprises and heterogeneous entities such as products, patents, and talents. Based on the original data of the second-type edge, a first-type edge is derived to generate relationships between various types of enterprises. The specific definitions of the four types of relationships are as follows: Supply Relationship: This is one of the core relationship types of Category I edges, used to characterize the structured supply and demand relationships between upstream and downstream enterprises in the industrial chain. The underlying structure relies on the original Category II edge relationships between enterprises and products, and between enterprises and raw materials. These Category II edges accumulate raw data such as supply targets, transaction scenarios, and upstream and downstream matching relationships. Based on this, Category I supply edges between enterprises are derived. Category I edge attributes include structured fields such as supplier, purchaser, transaction amount, cooperation period, and supply share. Data is mainly extracted from bidding announcements, enterprise annual reports, and supply and demand matching data, accurately representing the strength and characteristics of supply and demand binding between upstream and downstream enterprises in the industrial chain.

[0025] Cooperative Relationships: This is one type of edge relationship used to characterize the structured connections between enterprises in joint operations, joint R&D, and strategic cooperation. The underlying structure relies on the original second-order edge connections between enterprises and projects, and between enterprises and R&D achievements. These second-order edges retain underlying information such as cooperative projects, R&D content, and cooperation scenarios. First-order edges for inter-enterprise cooperation are derived based on the second-order edge topology. These first-order edges include structured fields such as the cooperating parties, cooperation type, cooperation content, cooperation time, and cooperation field. Data is primarily extracted from joint patents, strategic cooperation announcements, and joint bidding information to quantify the closeness of commercial and R&D cooperation between enterprises.

[0026] Competitive Relationship: This is one type of edge relationship, used to depict the structured horizontal competitive relationships between companies in the same industry chain or product track. The underlying structure relies on the original second-type edge relationships between companies and products, and between companies and industry tracks. These second-type edges record basic information such as company product layout, industry classification, and market layout. First-type competitive edges between companies are derived by matching the homogeneous topological features of these second-type edges. The attributes of these first-type edges include structured fields such as the competing parties, their respective tracks, competing products, and market share, used to characterize the substitution relationship and market competition intensity between companies.

[0027] Innovation Relationships: This is one type of edge relationship used to characterize the structured connections between enterprises in the industrial chain regarding technological research and development and talent innovation. The underlying topology is based on multiple innovation dimensions of type II edges, including the R&D ownership relationship between enterprises and patents, the technology citation relationship between patents, and the employment and R&D relationship between enterprises and R&D personnel. Type II edges accumulate original innovation data such as R&D fields, technological achievements, patent citation paths, and personnel employment information. Based on the innovation element paths and the connectivity of type II edges, type I innovation edges between enterprises are derived. Type I edges are used to characterize the strength of connections in joint R&D, technological cooperation, and implicit technology transfer between enterprises. The data is mainly extracted from patent databases and academic achievement data.

[0028] In an exemplary embodiment, after constructing entities of the industry chain knowledge graph and establishing relationships between entities, extracting knowledge from various set data sources to form the industry chain knowledge graph, traversing along a type of edge with the same relationship type in the knowledge graph, and enumerating risk propagation paths that satisfy both the starting point and the ending point as enterprises, the process includes steps S210, S220, S230, and S240.

[0029] Step S210: Capture various types of real-time information related to entities.

[0030] Step S220: Extract publisher information from real-time information.

[0031] Step S230: Obtain the information attraction degree based on the publisher information; Step S240: Compare the information introduction degree with the preset introduction degree threshold. If the information introduction degree is greater than the preset introduction degree threshold, import the real-time information into the industry chain knowledge graph and update the industry chain knowledge graph. If the information introduction degree is less than or equal to the preset introduction degree threshold, discard the real-time information.

[0032] By extracting the publisher information of real-time information to calculate the information introduction degree, and combining it with threshold filtering to achieve optimal data inclusion in the map, on the one hand, invalid and redundant information is filtered out, reducing the computing power and storage losses caused by frequent map updates; on the other hand, the value of information is quantified based on the publisher dimension to ensure the effectiveness of the data included in the map, avoiding interference from inferior data while iterating the map in real time, stabilizing the quality of map data, and providing reliable data support for subsequent dynamic correction of first-class edges and accurate measurement of industrial chain risks.

[0033] In one exemplary embodiment, different preset information attraction degrees are predefined for different publisher information, and the information attraction degree is obtained based on the mapping relationship between publisher information and preset information attraction degrees. The scope of real-time information capture includes: Government regulatory data includes market supervision and business registration data, tax data, industrial policy documents, import and export trade data, industry qualification and licensing data, and enterprise administrative penalty information, providing authoritative basic attribute and compliance status information for entities in the industrial chain; Enterprise operating data includes annual reports of listed companies, public financial reports of small and medium-sized enterprises, and information disclosed on corporate websites, extracting operating information such as enterprise revenue structure, upstream and downstream cooperation, core products, and production capacity. Technological innovation data includes the State Intellectual Property Office's patent database and academic paper database, extracting technological innovation information such as patent applicants, technology classifications, legal status, citation relationships, and R&D personnel. Supply chain transaction data includes government procurement announcements, public resource trading center bidding announcements, and supply and demand matching data of the industrial chain, extracting supply chain related information such as purchasers, suppliers, transaction targets, transaction amounts, and cooperation periods; The public opinion and emergency data includes mainstream financial news, industry media reports, and official emergency announcements, extracting risk-related information such as abnormal business operations, industry emergencies, policy changes, and market fluctuations.

[0034] In this application, the information incorporation degree of government information > the information incorporation degree of enterprise public information > the information incorporation degree of public opinion data.

[0035] In an exemplary embodiment, after constructing entities of the industry chain knowledge graph and establishing relationships between entities, extracting knowledge from various set data sources to form the industry chain knowledge graph, traversing along the edges of similar relationships in the knowledge graph, and enumerating risk propagation paths that satisfy both the starting point and the ending point as enterprises, the process includes steps S310 and S320.

[0036] Step S310: Process the captured real-time information based on the industry chain knowledge graph, and label the relationship between relevant entities and events. The real-time information includes at least enterprise operation change information, talent change information, patent technology change information, and upstream and downstream supply and demand change information.

[0037] Based on the standardized and organized data completed through preprocessing and entity extraction, and combined with the entity architecture and association system of the industrial chain knowledge graph, we can accurately identify various entities and dynamic events covered in real-time information, and label the association logic between entities and events, and between entities, so as to realize the transformation of disordered information text into structured entity association data. This is conducive to capturing various triggering events of industrial chain risk transmission more completely.

[0038] Step S320: Based on the annotation results, distinguish between four data types: newly added entities, newly added associated edges, changes in the attributes of existing entities, and invalid existing relationships. Classify and store the annotated entity data and relationship data in the database to complete the incremental update and existing correction of the industry chain knowledge graph.

[0039] For the annotated structured and related data, a refined graph update scenario is implemented. New entities correspond to newly emerging industry entities such as enterprises, technologies, talents, and products within the industrial chain; new related edges correspond to newly formed supply and demand, cooperation, R&D, and competition relationships between existing entities; changes in the attributes of existing entities correspond to dynamic changes in their inherent attributes such as operating status, technical parameters, employment information, and business scale; and the expiration of existing relationships correspond to the termination, expiration, or dissolution of existing cooperation, R&D, and supply and demand relationships. By classifying and storing these four types of data, the graph is simultaneously expanded incrementally and existing data is corrected, ensuring that the graph's topology, entity attributes, relationships, and the actual dynamics of the industrial chain are synchronized in real time.

[0040] In the industry chain knowledge graph of this application, enterprises are set as Class I nodes, and products, technologies, and talents are set as Class II nodes. The connection edges formed between enterprises are defined as Class I edges, and the connection edges formed between Class I nodes and Class II nodes, and between Class II nodes themselves, are defined as Class II edges. Meta-paths corresponding to product supply, product competition, investment and financing, talent employment, and patent application are pre-configured. Based on the connectivity topology of Class II edges, Class I edges labeled with corresponding relationship types are generated according to the matching rules of each meta-path. When a new industry event is added to the industry chain knowledge graph, only the local Class II edges affected by the event are updated, and the corresponding Class I edges are dynamically and synchronously corrected according to the updated Class II edges. When performing industry chain risk transmission calculation, an operation mask is configured based on the relationship type labeled by the Class I edges to block Class I edges that are not the target relationship, as well as all Class II nodes and edges. Only the filtered and allowed Class I edges are used to carry out risk transmission calculation, without the need to split and construct entity subgraphs.

[0041] The aforementioned industry chain knowledge graph achieves hierarchical decoupling, optimizing data iteration efficiency. This application divides the industry chain knowledge graph into a two-layer architecture of type I nodes / edges and type II nodes / edges. Type I edges between various enterprises are derived from the underlying type II edges combined with preset meta-paths. At least part of the original industry basic data is stored in the type II edges. When new industry events occur, such as new product launches, equity changes, patent licensing, or talent mobility, only a small number of disturbed type II edges need to be locally updated, and the corresponding type I edges can be adaptively corrected in conjunction with these updates. There is no need to reconstruct the entire graph, which significantly reduces the computational overhead and data maintenance costs of incremental graph updates and improves the real-time update capability of industry chain data.

[0042] Meanwhile, by abandoning the physical storage of multiple isomorphic graphs, storage resources are saved. Compared to the traditional approach of pre-constructing multiple independent isomorphic graphs based on different relationships, this application maintains only one complete heterogeneous base graph of the industrial chain. It distinguishes various relationships such as supply relationships, competitive relationships, innovation relationships, and cooperative relationships by relying on the relationship type label of a type of edge, eliminating the redundant storage overhead of multiple subgraphs and simplifying the graph storage architecture and database management logic. Furthermore, by relying on mask logic to isolate the calculation range, it achieves independent risk calculation across dimensions. The risk transmission calculation stage adopts a calculation mask mechanism, eliminating the need for physical splitting, generating, and storing temporary subgraphs. By masking irrelevant type I edges, all type II nodes and edges, and using type I edges of a single relationship type as needed to carry out risk calculation, it effectively isolates the cross-interference problem of risks from different dimensions such as supply chain, industry competition, equity investment and financing, and technology research and development, ensuring independent calculation of various risk transmission paths and improving the accuracy of risk quantification results for different scenarios in the industrial chain.

[0043] The industry chain knowledge graph in this application also enables the complete retention of underlying raw data, facilitating the fine-grained quantification of edge weights. The original associations related to products, technologies, and talents are uniformly stored in the second-class edges. The second-class edges do not participate in risk transmission but only serve as the data foundation for calculating the coefficients of the first-class edges. When calculating the association strength and risk transmission coefficients of various first-class edges, the full range of detailed features carried by the second-class edges, such as detailed product information, patent data, talent positions, and technology references, can be retrieved at any time. Based on the original fine-grained data, the edge weights can be finely assigned, improving the scientificity and credibility of the risk transmission coefficient quantification.

[0044] For example, if a type II node is talent A, and talent A works for company A, then company A and talent A are connected by a type II edge to reflect the relationship between talent A and company A. When real-time information discloses that talent A has left company A and started working for company B, the type II edge between talent A and company A is broken, and a new type II edge is created between talent A and company B.

[0045] For example, by capturing real-time information, a new type II node, patent A, is added to the industry chain knowledge graph. Patent A is a shared patent, and the joint applicants are company A and company B. Then, there is a type II edge connecting company A and patent A, and there is a type II edge connecting company B and patent A. Through the above two type II edges, there is a type I edge connecting company A and company B. The edge attribute of this type I edge is innovation relationship.

[0046] Step S130: Obtain the initial risk intensity based on the risk source type.

[0047] The initial risk intensity is determined based on the type of risk source and can be obtained through a pre-set mapping table. The initial risk intensity is set as a quantitative value in the range of 0-1 according to the risk event level. Step S140: Based on the node attributes and edge attributes of enterprises on the risk propagation path, obtain the transmission weight scalar, the comprehensive attenuation value and the correlation tightness. The transmission weight scalar is used to quantify the impact of risk transmission corresponding to a certain type of edge relationship.

[0048] In an exemplary embodiment, based on the node attributes and edge attributes of enterprises on the risk propagation path, the transmission weight scalar, the comprehensive attenuation value, and the correlation tightness are obtained, including step S410.

[0049] Step S410: Based on the relationship type of the edge attributes in the risk transmission path and the mapping relationship between the relationship type and the preset transmission weight scalar, obtain the transmission weight scalar. .

[0050] Transmission weight scalar This represents the transmission weight scalar corresponding to the relationship type (supply / cooperation / competition / innovation) of the path k from starting company i to ending company j. The transmission weight scalar corresponding to the relationship type (supply / cooperation / competition / innovation) can be preset.

[0051] Because supply chain disruptions directly lead to production halts, the risk transmission effect is the strongest and most direct. For example, when the relationship type of k is a supply relationship... .

[0052] The disruption of joint research and development may lead to project delays, with a moderate impact. For example, when the relationship type of k is a collaboration... .

[0053] Because technological dependence creates barriers, although the transmission speed may not be as fast as that of the supply chain, its impact is profound and difficult to replace. For example, when k belongs to an innovation relationship... 7.

[0054] Because the transmission of risk from competitors can be hedging (a production halt by one party may benefit the other), or it can be indirectly transmitted through changes in market share. For example, when k belongs to a competitive relationship... 3.

[0055] In an exemplary embodiment, based on the node attributes and edge attributes of enterprises on the risk propagation path, the transmission weight scalar, the comprehensive attenuation value and the correlation tightness are obtained, including steps S510, S520, S530 and S540.

[0056] Step S510: Based on the difference between the generation timestamp of the relation edge in the industry chain knowledge graph and the current time, obtain the time decay factor.

[0057] Step S520: Based on the relationship type of the edge attributes in the risk transmission path, obtain the relationship type risk transmission coefficient.

[0058] Step S530: Obtain the geographical coordinates of the starting enterprise and the ending enterprise, calculate the straight-line distance between the starting enterprise and the ending enterprise based on the geographical coordinates; count the number of hops in the link network topology of the starting enterprise and the ending enterprise in the industry chain knowledge graph; obtain the dual distance attenuation factor based on the number of hops and the straight-line distance.

[0059] Step S540: Based on the time-related attenuation factor, the relationship type risk transmission coefficient, and the dual distance attenuation factor, obtain the comprehensive attenuation value.

[0060] In an exemplary embodiment, the overall attenuation value The calculation method is as follows: or

[0061] in, The time-decrease factor corresponding to path k is calculated based on the difference between the generation timestamp of the relation edge in the industry chain knowledge graph and the current time. The calculation formula is: , This is the duration of the relationship (i.e., the difference between the generation timestamp of the relationship edge in the industry chain knowledge graph and the current time). Industry experience coefficient; This represents the risk transmission coefficient for the relationship type corresponding to path k. Different links in the industrial chain have varying abilities to block risk transmission (such as the barrier effect of core technologies). This coefficient is jointly determined by the "enterprise qualification attributes" and "relationship type" in the dynamic graph. Because the four types of relationships have completely different characteristics in terms of risk transmission and blocking: Innovation relationships (technological barriers): These relationships have strong exclusivity and barrier effects, and risks are not easily replaced or bypassed. Therefore, the risk transmission coefficient of this type of relationship is relatively low (high transmission resistance, and risks are not easily penetrated). Supply relationships: Risk can be mitigated by finding alternative suppliers; the risk transmission coefficient for this relationship type is moderate. Competitive Relationship: When the starting enterprise suffers setbacks, declining production capacity, and insufficient supply, downstream customers will turn to competing end-point enterprises. The end-point enterprises will benefit from the revenue and have their risks offset. There is no risk transmission in the same direction in the competitive chain. Instead, there is risk diversion and reverse benefit effect. Therefore, the risk transmission coefficient of the competitive relationship side is relatively low. Cooperative Relationship: When the starting enterprise experiences risk events such as business setbacks, project stagnation, or abnormal performance, the terminal enterprise bound to it faces problems such as interruption of supporting services, project delays, and revenue losses. The risks are transmitted in the same direction along the cooperative link, without the effect of order transfer or risk hedging. Therefore, the risk transmission coefficient corresponding to the cooperative relationship edge is too high.

[0062] For example, each relationship type pre-determines a basic risk propagation weight. For example, the basic risk transmission weight for cooperative relationships is 0.9, for innovative relationships it is 0.3, for supply relationships it is 0.6, and for competitive relationships it is 0.3. The higher the value, the weaker the barrier capability and the easier it is for risks to be transmitted.

[0063]

[0064] in, The risk transmission coefficient for the relation type corresponding to path k; Basic risk resistance weight; Starting point: Enterprise qualification correction factor; For the endpoint enterprise j qualification correction factor; By extracting the type labels of a type of edge and the attribute labels of nodes in the knowledge graph of the industrial chain, matching the corresponding correction parameters by looking up the table, and performing product or nonlinear fusion calculation, dynamic quantification based on business reality can be achieved.

[0065] The attribute tags of the aforementioned enterprise nodes include "multiple supplier qualifications" and "ISO supply chain certification," indicating that the enterprises possess strong risk buffering and substitution capabilities, which will enhance the risk resistance coefficient of this path (i.e., (Growing larger, blocking risks).

[0066] The dual distance decay factor corresponding to path k is a nonlinear function that combines "graph structure distance decay" and "geospatial distance decay".

[0067] The attenuation of supply chain risk transmission depends not only on network hop count but also on the influence of physical space (longer physical distances often mean higher logistics costs, longer response times, and greater difficulty in substitution). This is a dual-distance attenuation factor. The calculation logic is as follows:

[0068] in, Distance decay in graph structures is based on the number of hops in the graph network. calculate; The attenuation coefficient of the graph structure is an empirical value. This represents the number of hops along the path.

[0069]

[0070] in, The geospatial distance attenuation is calculated based on the physical straight-line distance d between the starting enterprise i and the ending enterprise j. The spatial attenuation coefficient is an empirical value. The physical straight-line distance between the registered address / office address of the starting company i and the ending company j.

[0071]

[0072] in, It is a dual distance attenuation factor; For graph structure distance attenuation; Attenuation due to geographic distance; is the spatial attenuation coefficient, which is a preset value; This is the geographical attenuation coefficient, which is a preset value. This represents the number of hops in the path. The physical straight-line distance between the registered address / office address of the starting company i and the ending company j.

[0073] Dual Distance Attenuation Factor The introduction of this allows even two companies that are directly connected in the graph structure (hop count) to... But if the physical distance spans half a country ( (Although the risk transmission is extremely high), the actual risk transmission attenuation will still be significant, which can improve the authenticity of risk transmission in the industrial chain.

[0074] The path hop count n mentioned above refers to the number of edges of one type traversed along the path from starting company i to ending company j on path k. For example, if starting company i → intermediate company l → ending company j, then the path hop count n is 2; if starting company i → ending company j, then the path hop count n is 1.

[0075] In an exemplary embodiment, based on the node attributes and edge attributes of enterprises on the risk propagation path, the transmission weight scalar, the comprehensive attenuation value, and the correlation tightness are obtained, including step S610.

[0076] Step S610: When the relationship type of one edge in the risk propagation path is an innovation relationship, extract the number of shared patents between the starting enterprise and the ending enterprise and the shortest patent citation path length from the starting enterprise to the ending enterprise from the industry chain knowledge graph. Based on the number of shared patents and the patent citation path length, obtain the correlation tightness. The statistics of the shortest patent citation path length are obtained by relying on the citation correlation data of the intermediate enterprises and their patents along the risk propagation path. When the relationship type of one edge in the risk propagation path is a competitive relationship, the overlapping sub-product categories between the starting enterprise and the ending enterprise are extracted from the industry chain knowledge graph, and the correlation tightness is obtained based on the operating indicators of the sub-product categories; among them, the operating indicators can be any one of operating scale, revenue, capacity, output or shipment amount. When the relationship type of an edge on the risk propagation path is a supply relationship, the upstream and downstream cooperative enterprises adjacent to each other on the risk propagation path are located, where the supplier is the upstream enterprise and the purchaser is the downstream enterprise. The target product purchased by the downstream enterprise from the corresponding upstream is extracted from the industry chain knowledge graph, and based on the annual purchase amount of the target product, the total annual purchase amount of the target product by the downstream enterprise for the target product, and the total number of upstream enterprises that can supply the target product to the downstream, the supply chain coupling tightness coefficient between two adjacent enterprises on the risk propagation path is obtained based on the annual purchase amount, the total purchase amount, and the total number of upstream enterprises, and the association tightness is obtained based on the supply chain coupling tightness coefficient. When the relationship type of one edge in the risk propagation path is a cooperative relationship, the quantitative values ​​of strong contractual binding cooperation between the starting enterprise and the ending enterprise and the quantitative values ​​of business cooperation in industrialization projects are extracted from the risk propagation path of the industrial chain knowledge graph. Based on the quantitative values ​​of strong contractual binding cooperation and business cooperation in industrialization projects, the correlation tightness is calculated.

[0077] In one exemplary embodiment, The calculation formula is as follows:

[0078] in, For the degree of correlation; , , and These are the adjustment coefficients, and when the relationship type of one type of edge attribute of path k is a supply relationship, , , and All are 0; when the relation type of one type of edge attribute of path k is an innovative relation, , , and All are 0; when the relation type of one type of edge attribute of path k is a competitive relationship. , , and All are 0; when the relation type of one type of edge attribute of path k is a cooperative relationship. , , and All are 0; among them, .

[0079] The supply chain coupling tightness coefficient represents the greater the supply chain coupling tightness coefficient, the stronger the dependence of the end enterprise j on the upstream starting enterprise i, the fewer alternative suppliers, and the higher the transmission efficiency of upstream risks to downstream through the supply chain. The inter-firm technology dependency coupling coefficient is used to characterize the degree to which the starting firm i transmits risk to the ending firm j through joint R&D and patent citation. The industry market overlap competition coefficient represents the degree of risk transmission between the starting enterprise i and the ending enterprise j due to the industry market overlap. The inter-firm cooperation coefficient represents the degree of closeness formed by cooperation between the starting firm i and the ending firm j.

[0080] In one exemplary embodiment, The calculation method is as follows:

[0081] in, The product category procurement dependence coefficient represents the degree of financial dependence of downstream end-point enterprise j on starting-point enterprise i (upstream supplier) for procurement within the same product category. The higher the starting point, the better for the enterprise. Supply disruptions affect end-point companies The greater the impact on cash flow.

[0082] The total annual purchase amount of the target product from the starting point enterprise i by the endpoint enterprise j; The total annual purchase amount of the target product for end-point company j;

[0083] The coefficient representing the difficulty of supply substitution is used to quantify the ease or difficulty of replacing upstream sources of goods for the corresponding product category of downstream end enterprise j. The value ranges from 0 to 1. The closer the value is to 1, the fewer the alternative suppliers, the higher the difficulty of replacing the source of goods after the supply is cut off, and the greater the intensity of supply chain risk transmission. The coefficient is used for smoothing to avoid the divisor being zero in the above formula; The upstream supplier indices for the same product category refer to the total number of upstream supplier companies in the industry chain knowledge graph that can provide the end-point company j with products of the same major product category; the upstream supplier indices for the same product category are used to calculate the substitution difficulty coefficient. The basic topology parameters.

[0084] when (Exclusive Supply): A value close to 1 indicates complete irreplaceability and the highest level of risk.

[0085] when (Multiple sources of supply): The decrease as the number of suppliers increases indicates that substitution is easy and the risk tightness is reduced.

[0086] The final supply tightness is obtained by weighting and integrating the procurement dependence ratio coefficient and the supply substitution difficulty coefficient of the above categories. The integration process typically employs product or weighted summation to reflect the "weakest link effect." When either indicator is low, it lowers the overall supply chain tightness and inhibits the transmission of risks through the supply chain; only when both indicators are simultaneously in a high range will high supply chain dependence and high-intensity transmission risks occur.

[0087] In an exemplary embodiment, the formula for combining the category procurement dependence ratio coefficient and the supply substitution difficulty characterization coefficient is as follows:

[0088] in, This represents the supply chain coupling tightness coefficient. This represents the product category procurement dependence ratio. The coefficient representing the difficulty of supply substitution; As a base fixed coefficient, Its function is to make replacement difficult =0 (unlimited supply, completely replaceable). The lower bound of basic supply dependence will still be maintained to ensure =0 It won't go directly to zero; it represents the basic supply chain risk inherent in the procurement ratio. To replace the difficulty scaling factor, right Perform amplitude scaling.

[0089] and The value can be a fixed value, or it can be adjusted according to the characteristics of the industry.

[0090] In another exemplary embodiment, the formula for combining the category procurement dependence ratio coefficient and the supply substitution difficulty characterization coefficient is as follows:

[0091] For example, path k only has a starting company i (a cathode material supplier) and an ending company j (a power battery company), with no other intermediate companies. The power battery company (ending company j) needs to purchase 1 billion yuan worth of cathode materials annually, of which 750 million yuan needs to be purchased from the cathode material supplier (starting company i). According to the supply chain knowledge graph, starting company i is the only cooperative supplier of cathode materials for ending company j. Therefore, the in-degree of upstream suppliers of the same category of cathode materials for ending company j in the supply chain knowledge graph is... ,at this time Approaching 1.0.

[0092] (The product category procurement dependence ratio is 75%, which is considered high dependence) (Extremely high density, extremely high risk) For example, a certain power battery company (the end company) The company needs to purchase 1 billion yuan worth of cathode materials annually, of which the cost will be paid to cathode material suppliers (starting point companies). The company purchased 750 million yuan worth of cathode materials. According to the supply chain knowledge graph, the endpoint company J has three other alternative suppliers for cathode materials. Therefore, the entry degree of upstream suppliers of the same category of cathode materials for endpoint company J in the supply chain knowledge graph is... (Including the starting company i).

[0093] (The product category procurement dependence ratio is 75%, which is considered high dependence)

[0094] (The supply chain coupling tightness coefficient has decreased significantly, which means that the supply chain coupling between the starting enterprise i and the ending enterprise j is relatively low. In the cathode material supply link, the starting enterprise i is highly substitutable.) The above example only considers the case where the destination company j and the origin company i are directly connected. If the destination company j and the origin company i are connected through an intermediate company q, then the supply chain coupling tightness coefficient between the origin company i and the intermediate company q needs to be calculated. Then calculate the supply chain coupling tightness coefficient between intermediate enterprise q and terminal enterprise j. ,Right now In this system, the product supplied by starting firm i to intermediate firm q is product A, and the product supplied by intermediate firm q to ending firm j is product B. Product A must be an input factor of product B. If the route between starting firm i and ending firm j involves one intermediate firm q and one intermediate firm l, then the supply chain coupling tightness coefficient between starting firm i and intermediate firm q needs to be calculated. Then calculate the supply chain coupling tightness coefficient between intermediate enterprise q and intermediate enterprise l. It is also necessary to calculate the supply chain coupling tightness coefficient between intermediate enterprise l and terminal enterprise j. ,Right now The calculation method for intermediate companies along the route between the starting company i and the ending company j is as described above, and is performed sequentially based on the number of nodes of the intermediate companies.

[0095] In the industry chain knowledge graph, secondary nodes include patents. Patents are connected to their corresponding applicants (i.e., enterprise nodes) via second-class edges, and patents are also connected to cited patents via second-class edges, representing the citation relationships between patent nodes, thus forming a "patent citation network." Since some patents are jointly developed by multiple enterprises, these jointly developed patents are connected to their respective enterprises via second-class edges, forming a "joint R&D network" in the industry chain knowledge graph. In the aforementioned "patent citation network" and "joint R&D network," two core indicators—the number of shared patents and the length of the core patent citation path—are extracted for quantification. The core logic is that the number of shared patents formed by joint enterprise R&D measures the breadth and depth of technology binding, while the length of the core patent citation path measures the difficulty of circumventing the technology. Specifically, the inter-enterprise technology dependency coupling coefficient... The calculation process is as follows:

[0096] in, This refers to the percentage of shared patents. The number of patents held by the starting company i and the ending company j as joint applicants; The number of patents held by starting company i; The number of patents held by the endpoint company j.

[0097] Shared patent percentage The higher the value, the deeper the technological integration between the starting enterprise i and the ending enterprise j, and the easier it is for risks to be transmitted through the type of edge in the relationship between the enterprises, which is an innovation relationship.

[0098] Starting point enterprise i uses patents for type II edges and type II nodes The connection indicates that the starting enterprise i holds patent x, and the ending enterprise j is connected to the second-class node patent v through a second-class edge, indicating that the ending enterprise j and the starting enterprise i jointly hold patent v.

[0099] density coefficient The calculation method is as follows:

[0100] in, The core technology affinity coefficient between starting company i and ending company j; This is the path attenuation coefficient, which is a fixed value. This represents the number of hops in the patent citation path.

[0101] In the "Patent Citation Network", patent nodes are among the second type of nodes. The core technology was patented. The more technical references a patent makes, and the shorter the reference path (even a direct reference), the better. Patents The more directly a company relies on core technologies, i.e., patents... The corresponding companies have patents The greater a company's reliance on core technologies, the more patents it will have. Corresponding companies can substitute for or circumvent patents. The technologies mentioned above are extremely difficult for the corresponding companies.

[0102] For example, starting company i holds patent x, and ending company j holds patent v. In a "patent citation network", the patents... →Patent →Patent In this context, the type II edge between two adjacent patent nodes is a referencing edge, which has a direction, with the referrer pointing to the referenced node. Then, the patent node... There are two referencing edges between the patent node v and the patent node v. The path hop count is... The value is 2.

[0103] In an exemplary embodiment, the inter-firm technology dependency coupling coefficient The calculation formula is as follows:

[0104] in, and The weighting coefficient is a constant. The coupling coefficient representing the technological interdependence between enterprises; This refers to the percentage of shared patents. The coefficient representing the closeness of core technologies between starting company i and ending company j.

[0105] In another exemplary embodiment, the inter-firm technology dependency coupling coefficient The calculation formula is as follows:

[0106] in, The coupling coefficient representing the technological interdependence between enterprises; This refers to the percentage of shared patents. The coefficient representing the closeness of core technologies between starting company i and ending company j.

[0107] For example, a power battery company in the city (endpoint company j) and a core motor control technology provider (starting company j) The transmission of technological risks occurs, and calculations A knowledge graph analysis of the industry chain reveals that a power battery company and a core motor control technology provider jointly applied for 5 patents related to "high-efficiency thermal management system." The power battery company holds a total of 50 patents, and the core motor control technology provider holds a total of 100 patents. If the power battery company owns patent x, and the core motor control technology provider's patent v directly references patent x, then the shortest reference path is... .

[0108]

[0109] Assuming ,but

[0110] Assuming a simple average weight is used

[0111]

[0112] For example, if patent v owned by a core motor control technology provider does not directly reference patent x, but patent v references patent y from another company, and patent y references patent v, then the shortest reference path is... .

[0113]

[0114] Assuming a simple average weight is used

[0115]

[0116] Through the above calculations, when a core motor control technology provider's patent v directly references patent x, although the shared patent percentage between the core motor control technology provider and the power battery company is not high, the direct reference relationship between the core patents (reference path length is 1) makes technological substitution extremely difficult. This calculation method highlights the "technological bottleneck" risk hidden in the "patent citation network".

[0117] In an exemplary embodiment, the industry market overlap competition coefficient The calculation method is as follows:

[0118] in, The coefficient representing the overlap and competition in the same industry market; To further subdivide product categories (for example, representing subdivided product categories such as ternary cathode materials and lithium iron phosphate materials); The operating scale of enterprise i under the subdivided product category s can be quantified by revenue, production capacity, output or shipment amount. The operating scale of the endpoint enterprise j in the subdivided product category s can be quantified by revenue, capacity, output or shipment amount. This represents the total number of all product categories involved by both the starting company i and the ending company j.

[0119] Industry overlap competition coefficient In the calculation formula, The minimum operating scale of each of the sub-categories is taken and summed to represent the overlapping competitive business volume of the two companies. Since both companies target the same downstream customers for the same product, the total market demand is limited. They compete for the same batch of customer orders, and the upper limit of their overlapping competitive business is constrained by the company with the smaller production capacity. That is, the entire 600 million unit capacity of the endpoint company j must compete with the starting company i for the market. The starting company i only directly competes with the endpoint company j for the market corresponding to its 600 million unit capacity; the orders corresponding to its additional 400 million unit capacity do not directly compete with the endpoint company j. Therefore, the overlapping competitive volume of a single product category is taken as the minimum of the two.

[0120] Overlapping competition coefficient in the same industry market In the calculation formula Take the maximum of the total operating amount of the two companies across all categories; the coefficient is in the range of [0,1]. The higher the value, the higher the overlap of product categories between the starting company i and the ending company j, and the stronger the risk transmission coupling brought about by competition in the same industry.

[0121] The calculation method is as follows:

[0122] in, This represents the supply chain coupling tightness coefficient. For weighted score conversion coefficients, The conversion factor is the base score. and The unit for all values ​​is 1 / ten thousand yuan (AHP / entropy weight method assignment). ); To quantify the strong contractual binding cooperation, the statistical content of the strong contractual binding cooperation quantification value includes the paid-in capital of equity participation by both starting enterprise i and ending enterprise j under the strategic cooperation agreement, the paid-in capital of joint venture factory construction, and the subject matter of the joint venture contract; the value of ordinary strategic cooperation framework agreements is converted according to the agreement. The unit is RMB 10,000. When there is no equity or large contractual amount... =0; This is a quantified value for business cooperation in industrialization projects. The statistical content includes the total amount of projects jointly won by starting company i and ending company j, the total amount of projects jointly constructed for industrialization, and the total amount of contracts for commissioned development projects. The unit is ten thousand yuan. (Note: The last sentence appears to be incomplete and possibly refers to a separate, unrelated statement.) =0.

[0123] Piecewise function for equity addition:

[0124] Where S represents the shareholding ratio of the endpoint company j in the starting company i.

[0125] In an exemplary embodiment, the probabilities of multiple risk propagation paths are superimposed to obtain the total event risk value of the enterprise, including step S710.

[0126] Step S710: Calculate the betweenness centrality and degree centrality of the starting enterprise on the topology of the industry chain knowledge graph. Based on the betweenness centrality and degree centrality, obtain the importance correction parameter to correct the total event risk value of the enterprise.

[0127] In one exemplary embodiment, the formula for calculating the importance correction parameter is as follows:

[0128] or,

[0129] in: The value of the importance correction parameter for the starting enterprise i in the industry chain knowledge graph is larger, indicating that the starting enterprise i is more important in the industry chain knowledge graph and that its operating status has a greater impact on the overall operation of the industry chain. Let represent the degree centrality of the starting enterprise i. Degree centrality is a numerical indicator used to describe the ending enterprise j, reflecting the number of upstream and downstream enterprises (or the weight of transaction relationships and cooperation relationships) directly connected to the starting enterprise i in the industry chain knowledge graph. The higher the value, the more direct trading partners the company has, indicating that it is a local "trading hub". Let denot be the betweenness centrality of the starting firm i. Betweenness centrality is a numerical value that reflects the ability of the starting firm i as a "bridge" node to control the flow of information / materials. The higher the value, the more likely the company plays a crucial role as a "gateway" between different industrial clusters or supply chain sub-networks, controlling key channels for the flow of information or materials. This is the main attribute coefficient of the industrial chain, if the endpoint enterprise For enterprises marked as "chain leader enterprises" in the graph, the coefficient is significantly greater than 1, indicating that the impact on them will have a magnified effect on the entire industry chain. In this application, a chain leader enterprise refers to a leading enterprise that occupies a core and dominant position in an industry chain, and can lead and drive the coordinated development of the entire industry chain through its own strong comprehensive strength, and has a decisive impact on the industrial ecosystem. , and The weighting coefficients are fixed values. , and The size can be adjusted according to specific needs.

[0130] In a supply chain knowledge graph, supply and innovation relationships between enterprises are represented by directed edges, while cooperation and competition relationships are considered as undirected edges. That is, each cooperation and competition relationship has a degree centrality. When the endpoint enterprise j is considered, it is regarded as an equivalent directed edge of two first-class edges. The calculation method is as follows:

[0131] in, The total degree of starting enterprise i; The number of incoming edges is a type of entry edge for the starting enterprise i. The number of incoming edges includes the number of incoming edges from supply relationships and innovation relationships (upstream suppliers) as well as the equivalent incoming edges from cooperative relationships and competitive relationships (since cooperative relationships and competitive relationships are undirected edges, there is an undirected edge between the starting enterprise i and its adjacent enterprise nodes under the above relationships, which can be regarded as the above undirected edge containing an equivalent incoming edge and an equivalent outgoing edge). That is, an undirected edge provides one number of incoming edges. Let i be the starting enterprise. The number of outgoing edges is calculated in the same way as the number of incoming edges.

[0132]

[0133] in, Normalized degree centrality of enterprise i as the starting point; The total degree of starting enterprise i; Let i be the number of edges of the type that can theoretically connect the starting enterprise i to the other enterprises. In some cases, there may be multiple edges of different relationship types between two enterprises. In the above formula, it is assumed that there is only one relationship type between the two enterprises.

[0134] Betweenness centrality of starting firm i The calculation method is as follows:

[0135] in, For the endpoint firm j, the original unnormalized betweenness number is given. s and t are any two enterprise nodes that are different from the starting enterprise i. The three entities representing the starting point company s, the ending point company t, and the companies along the way i do not overlap; This represents the number of shortest paths passing through enterprise i in a certain type of edge shortest path from starting enterprise s to ending enterprise t. In a knowledge graph of an industry chain consisting of only one type of edge, the total number of shortest paths from the starting enterprise s to the ending enterprise t, all edges of the same type. The shortest path mentioned above is the path with the fewest number of edges of a certain type among all possible paths, and the number of edges of that type on the path with the fewest edges of that type is used as the criterion for determining the shortest path. For example, path 1: s (enterprise) → a (enterprise) → t (enterprise), path 1 passes through 2 edges of the first type, and its length = 2; Path 2: s (enterprise) → b (enterprise) → c (enterprise) → t (enterprise). Path 2 passes through 3 first-class edges and has a length of 3.

[0136] In this embodiment, the main attribute coefficient of the industrial chain The quantification method includes the following steps: (1) Construct a basic evaluation index system. For example, select M quantitative indicators that can reflect the leading position of the enterprise in the supply chain. In this implementation, M=5, and the specific indicators are: annual operating revenue of the enterprise (unit: RMB 100 million); market share of the enterprise's products in its respective sub-market (unit: %); number of valid invention patents owned by the enterprise (unit: items); number of industry or national standards that the enterprise has participated in formulating (unit: items); number of upstream and downstream cooperative enterprises in the enterprise's direct supply chain (unit: enterprises). The original data of the above indicators come from publicly verifiable data sources, including but not limited to enterprise annual reports, patent database of the State Intellectual Property Office, national standard information public service platform, and statistical reports of industry associations.

[0137] (2) Indicator standardization. For the k-th evaluation indicator (k=1, 2, ..., M), the indicator values ​​of all enterprises in the industry chain knowledge graph are statistically analyzed. The min-max normalization method is used to map each indicator to the [0, 1] interval:

[0138] in, Let i be the normalized score of the starting enterprise i on the k-th evaluation indicator; and These are the minimum and maximum values ​​of all enterprises in the industry chain knowledge graph on the k-th indicator, respectively. Let be the original actual indicator value of enterprise i at the k-th evaluation indicator. For positive indicators (such as higher revenue being better), use the above formula directly; if... For inverse indicators, the reciprocal can be taken first, or... Conversion.

[0139] For example, in the industry chain knowledge graph, the maximum operating revenue is 56 billion yuan and the minimum is 0. Therefore, the normalized score of a company with operating revenue of 50 billion yuan is 500 / 560≈0.89, rounded to 0.9. Other evaluation indicators follow the same principle.

[0140] As an alternative, z-score normalization, quantile normalization, or sigmoid mapping can be used, and the normalization method can be adjusted according to actual needs.

[0141] (3) Weighted composite score. A weight vector is defined. ,satisfy The weights are pre-determined by the Delphi method or the Analytic Hierarchy Process (AHP) based on industry expert opinions and stored in the system configuration file.

[0142] The overall score of the starting company i is: Where M represents the total number of evaluation indicators involved in the scoring; The weight coefficient for the k-th evaluation indicator; Let i be the normalized score of the starting enterprise i on the kth evaluation indicator; The starting point is the comprehensive evaluation score of enterprise i; Obviously .

[0143] (4) Mapping to chain principal attribute coefficients. A piecewise linear mapping function is used:

[0144] in, The coefficient of the chain's main attribute; Threshold for determining the chain leader (default) ); The starting point is the comprehensive evaluation score of enterprise i; The chain leader's scaling factor (default) ).

[0145] In this application, the main attribute coefficient of the industrial chain is set. This coefficient highlights the impact of leading chain enterprises on the entire industry chain. While leading chain enterprises may not be prominent in degree centrality or betweenness centrality (e.g., core R&D-focused chain leaders that don't directly engage in large-scale transactions), their influence in setting technical standards, brand dominance, or resource allocation means that any disruption (such as production stoppage or supply disruption) can rapidly amplify the impact across the entire industry chain. This coefficient assigns these enterprises a strategic importance beyond mere "structural weight."

[0146] As a simplified alternative, the following can also be adopted directly:

[0147] in, The coefficient of the chain's main attribute; Preset magnification factor (usually) ); The starting point is the comprehensive evaluation score of enterprise i.

[0148] For example, a new energy vehicle battery company Its indicators and standardized forms are as follows: Operating revenue: 50 billion yuan, normalized to 0.9; Market share: 25%, normalized to 0.85; Number of invention patents: 320, normalized to 0.75; Number of participating standards: 5, normalized to 0.6; If we take weights of (0.3, 0.25, 0.2, 0.1), then .

[0149] use ,Pick ,but .

[0150] For another ordinary parts supplier, S=0.2, then P=2.8. This demonstrates that the coefficient of the leading company in the supply chain is much higher than that of the ordinary companies.

[0151] Step S150: Calculate the path transmission risk intensity of each risk propagation path based on the transmission weight scalar, the comprehensive attenuation value, the initial risk intensity, and the correlation tightness.

[0152] Step S160: The probabilities of multiple risk propagation paths are superimposed to obtain the total event risk value of the endpoint enterprise.

[0153] In an exemplary embodiment, the formula for calculating the total event risk value of an enterprise is as follows:

[0154] in, Let be the total risk value of events from starting firm i to ending firm j; The parameter is used to adjust the importance of endpoint firm j in the industry chain knowledge graph. The initial risk intensity of the risk source node is set as a quantitative value in the range of 0-1 according to the risk event level; The comprehensive attenuation value of path k is obtained by combining the time attenuation factor, the relationship type risk transmission coefficient, and the dual distance attenuation factor. The path The transmission weight scalar corresponding to the relationship type (supply / cooperation / competition / innovation); Let k be the degree of association on path k.

[0155] In one exemplary embodiment, The transmission weight scalar represents the relationship type (supply / cooperation / competition / innovation) corresponding to the path k from starting company i to ending company j.

[0156] For example, when the relationship type to which k belongs is a supply relationship Supply chain disruptions directly lead to production stoppages, exhibiting the strongest and most direct risk transmission effect. This applies when the relationship type of k is a cooperative relationship. Interruptions in joint R&D may lead to project delays, with a moderate impact. When the relationship type of k is an innovation relationship... 7. Technological dependence creates barriers to entry; although the transmission speed may not be as fast as the supply chain, its impact is profound and difficult to replace. This applies when the relationship type of k is an innovation relationship. 3. Risk transmission from competitors has a hedging effect (a production halt by one party may benefit the other party), or it may be indirectly transmitted through changes in market share.

[0157] In an exemplary embodiment, after superimposing the probabilities of multiple risk propagation paths to obtain the total event risk value of the endpoint enterprise, step S810 is included.

[0158] Step S810: Calculate the total event risk value of relevant enterprises in the industry chain knowledge graph.

[0159] In this application, the relevant companies are the first three companies in the risk propagation path to avoid excessive redundant calculations during the calculation process.

[0160] Step S820: Based on the preset risk threshold, identify high-risk enterprises in the industry chain knowledge graph and visualize the selected high-risk enterprises.

[0161] By selecting different colors to mark enterprise nodes, high-risk enterprises in the industry chain knowledge graph can be reflected, or a prompt can be given by increasing the radius of high-risk enterprise nodes.

[0162] like Figure 2 As shown, this application also provides an enterprise risk estimation system based on a supply chain knowledge graph, including: Graph Construction Module 410: Used to construct entities in the industry chain knowledge graph and establish relationships between entities, extract knowledge from various set data sources, and form the industry chain knowledge graph.

[0163] Path selection module 420: It is used to start with enterprises affected by the risk source and end with enterprises whose total risk value of the event to be measured. Enterprises are connected by a type of edge. It traverses along the same type of edge in the knowledge graph and enumerates the risk propagation paths that satisfy both the starting point and the ending point as enterprises. The relationship type of the edge is one of the following: competitive relationship, cooperative relationship, supply relationship and innovation relationship.

[0164] Analysis and Calculation Module 430: Based on the risk source type, the analysis and calculation module is used to obtain the initial risk intensity; based on the node attributes and edge attributes of enterprises on the risk propagation path, the analysis and calculation module is used to obtain the transmission weight scalar, the comprehensive attenuation value, and the correlation tightness. The transmission weight scalar is used to quantify the impact of risk transmission corresponding to a certain type of edge relationship; based on the transmission weight scalar, the comprehensive attenuation value, the initial risk intensity, and the correlation tightness, the analysis and calculation module is used to calculate the path transmission risk intensity of each risk propagation path; by superimposing the probabilities of multiple risk propagation paths, the analysis and calculation module is used to obtain the total event risk value of the endpoint enterprise.

[0165] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A method for enterprise risk estimation based on supply chain knowledge graphs, characterized in that, include: Construct entities for the industry chain knowledge graph and establish relationships between entities. Extract knowledge from various defined data sources to form the industry chain knowledge graph. Starting with the enterprises affected by the risk source, and ending with the enterprise whose total risk value of the event to be measured, the edges connecting enterprises are considered as a type of edge. Traversing along the edges of the same type in the industrial chain knowledge graph, the risk propagation path between the starting enterprise and the ending enterprise is enumerated. The relationship type of the edge is one of the following: competitive relationship, cooperative relationship, supply relationship, and innovation relationship. Based on the stated risk source type, the initial risk intensity is obtained; Based on the node attributes and edge attributes of enterprises on the risk propagation path, the transmission weight scalar, the comprehensive attenuation value and the correlation tightness are obtained. The transmission weight scalar is used to quantify the impact of risk transmission corresponding to a certain type of edge relationship. Based on the transmission weight scalar, the comprehensive attenuation value, the initial risk intensity, and the correlation tightness, the path transmission risk intensity of each risk propagation path is calculated; The probabilities of multiple risk propagation paths are summed to obtain the total risk value of the event for the endpoint enterprise.

2. The enterprise risk estimation method based on supply chain knowledge graph as described in claim 1, characterized in that: The entity includes a type I node and a type II node. The type I node is an enterprise, and the type II node includes at least one of products, technologies, or talents. The edges formed between the first type of nodes are called first-type edges, and the edges formed between the first type of nodes or the second type of nodes are called second-type edges. When a new event is imported into the industry chain knowledge graph, the corresponding first-type edges are updated according to the second-type edges.

3. The enterprise risk estimation method based on supply chain knowledge graph as described in claim 1, characterized in that, After constructing the entities in the industry chain knowledge graph and establishing the relationships between them, knowledge is extracted from various defined data sources to form the industry chain knowledge graph. Then, the graph is traversed along edges of the same relationship type, enumerating the risk propagation paths that satisfy both the starting and ending points of enterprises, including: Capture various real-time information related to entities; Extract the publisher information from the real-time information; Based on the publisher information, the information introduction degree is obtained; The information introduction degree is compared with the preset introduction degree threshold. If the information introduction degree is greater than the preset introduction degree threshold, the real-time information is imported into the industry chain knowledge graph and the industry chain knowledge graph is updated. If the information introduction degree is less than or equal to the preset introduction degree threshold, the real-time information is discarded.

4. The enterprise risk estimation method based on supply chain knowledge graph as described in claim 3, characterized in that, After constructing the entities of the industry chain knowledge graph and establishing the relationships between entities, and extracting knowledge from various defined data sources to form the industry chain knowledge graph, the system traverses along the edges of similar relationships in the knowledge graph, starting with the enterprises affected by the risk source, and enumerating the risk propagation path between the starting and ending enterprises. This includes: The real-time information captured is processed based on the aforementioned industry chain knowledge graph, and the relationship between the relevant entities and events is marked. The real-time information includes at least enterprise operation change information, talent change information, patent technology change information, and upstream and downstream supply and demand change information. Based on the annotation results, four types of data are distinguished: newly added entities, newly added associated edges, changes in the attributes of existing entities, and invalid existing relationships. The annotated entity data and relationship data are then classified and stored in the database to complete the incremental update and existing correction of the industry chain knowledge graph.

5. The enterprise risk estimation method based on supply chain knowledge graph as described in claim 1, characterized in that, The probabilities of multiple risk propagation paths are summed to obtain the total event risk value of the enterprise, including: Calculate the betweenness centrality and degree centrality of the starting enterprise on the topology of the industry chain knowledge graph. Based on the betweenness centrality and degree centrality, obtain the importance correction parameter to correct the total event risk value of the enterprise.

6. The enterprise risk estimation method based on supply chain knowledge graph as described in claim 1, characterized in that, Based on the node attributes and edge attributes of enterprises along the risk propagation path, the transmission weight scalar, the comprehensive attenuation value, and the correlation tightness are obtained, including: The transmission weight scalar is obtained based on the relationship type of the edge attributes in the risk transmission path and the mapping relationship between the relationship type and the preset transmission weight scalar.

7. The enterprise risk estimation method based on supply chain knowledge graph as described in claim 1, characterized in that, Based on the node attributes and edge attributes of enterprises along the risk propagation path, the transmission weight scalar, the comprehensive attenuation value, and the correlation tightness are obtained, including: Based on the difference between the generation timestamp of the relation edges in the aforementioned industry chain knowledge graph and the current time, the time-decrease factor is obtained; Based on the relationship type of the edge attributes in the risk transmission path, the relationship type risk transmission coefficient is obtained; Obtain the geographical coordinates of the starting enterprise and the ending enterprise, calculate the straight-line distance between the starting enterprise and the ending enterprise based on the geographical coordinates, count the number of hops in the link network between the starting enterprise and the ending enterprise in the industry chain knowledge graph topology, and obtain a dual distance attenuation factor based on the number of hops and the straight-line distance. Based on the time-decrease factor, the relationship type risk transmission coefficient, and the dual distance decrease factor, a comprehensive decrease value is obtained.

8. The enterprise risk estimation method based on supply chain knowledge graph as described in claim 1, characterized in that, Based on the node attributes and edge attributes of enterprises along the risk propagation path, the transmission weight scalar, the comprehensive attenuation value, and the correlation tightness are obtained, including: When the relationship type of one edge in the risk propagation path is an innovation relationship, the number of shared patents between the starting enterprise and the ending enterprise and the shortest patent citation path length from the starting enterprise to the ending enterprise are extracted from the industrial chain knowledge graph. Based on the number of shared patents and the patent citation path length, the correlation tightness is obtained. The statistics of the shortest patent citation path length are obtained by relying on the citation association data of the intermediate enterprises and their patents along the risk propagation path. When the relationship type of one edge in the risk propagation path is a competitive relationship, the overlapping sub-product categories between the starting enterprise and the ending enterprise are extracted from the industry chain knowledge graph, and the correlation tightness is obtained based on the operating indicators of the sub-product categories; wherein, the operating indicators are any one of operating scale, revenue, production capacity, output or shipment amount; When the relationship type of one edge on the risk propagation path is a supply relationship, the adjacent upstream and downstream cooperative enterprises on the risk propagation path are located, where the supplier is the upstream enterprise and the purchaser is the downstream enterprise; the target product purchased by the downstream enterprise from the corresponding upstream is extracted from the industry chain knowledge graph, and based on the annual purchase amount of the target product, the total annual purchase amount of the target product by the downstream enterprise for the target product, and the total number of upstream enterprises that can supply the target product to the downstream enterprise; based on the annual purchase amount, the total purchase amount, and the total number of upstream enterprises, the supply chain coupling tightness coefficient between two adjacent enterprises on the risk propagation path is obtained, and based on the supply chain coupling tightness coefficient, the association tightness is obtained; When the relationship type of one type of edge in the risk propagation path is a cooperative relationship, the quantitative value of strong contractual binding cooperation between the starting enterprise and the ending enterprise and the quantitative value of business cooperation in industrialization projects are extracted from the risk propagation path in the industrial chain knowledge graph. Based on the quantitative value of strong contractual binding cooperation and the quantitative value of business cooperation in industrialization projects, the correlation tightness is calculated.

9. The enterprise risk estimation method based on supply chain knowledge graph as described in claim 1, characterized in that, After summing the probabilities of multiple risk propagation paths to obtain the total event risk value for the endpoint enterprise, the following steps are taken: Calculate the total risk value of the events for the relevant enterprises in the aforementioned industry chain knowledge graph; Based on the total risk value of the event and the preset risk threshold, high-risk enterprises in the industry chain knowledge graph are screened and identified, and the screened high-risk enterprises are visualized.

10. A business risk estimation system based on a supply chain knowledge graph, characterized in that, include: The graph construction module is used to construct entities in the industry chain knowledge graph and establish relationships between entities. It extracts knowledge from various set data sources to form the industry chain knowledge graph. Path selection module: It is used to start from the enterprises affected by the risk source and end from the enterprises whose total risk value of the event to be measured. Enterprises are connected by a type of edge. It traverses along the same type of edge in the knowledge graph and enumerates the risk propagation paths that satisfy both the starting point and the ending point as enterprises. The relationship type of the edge is one of the following: competitive relationship, cooperative relationship, supply relationship and innovation relationship. Analysis and calculation module: Based on the risk source type, the analysis and calculation module is used to obtain the initial risk intensity; Based on the node attributes and edge attributes of enterprises along the risk propagation path, the analysis and calculation module is used to obtain the transmission weight scalar, the comprehensive attenuation value, and the correlation tightness. The transmission weight scalar is used to quantify the impact of risk transmission corresponding to a certain type of edge relationship. Based on the transmission weight scalar, the comprehensive attenuation value, the initial risk intensity, and the correlation tightness, the analysis and calculation module is used to calculate the path transmission risk intensity of each of the risk propagation paths. The probability of multiple risk propagation paths is superimposed, and the analysis and calculation module is used to obtain the total risk value of the event for the endpoint enterprise.

Images