Enterprise relation graph construction method and device

Abstract

The invention is suitable for the technical field of data processing, and provides an enterprise relationship atlas construction method and device.The method comprises the steps that basic information of multiple enterprises and relationship information between the enterprises are obtained from an enterprise relationship database, the basic information is used for representing individual information of the enterprises, and the relationship information between the enterprises is used for representing the individual information of the enterprises; the inter-enterprise relationship information is used for representing an association relationship between the two enterprises; determining an aggregation weight value of each enterprise pair according to the relationship information between the enterprises; wherein the two enterprises with the incidence relation form an enterprise pair; according to the basic information of the enterprises, the relationship information between the enterprises and the aggregation weight value of each enterprise pair, generating an enterprise relationship graph; wherein the nodes in the enterprise relation graph are used for representing the enterprises, and the connecting edges in the enterprise relation graph are used for representing the incidence relation of the enterprise pairs. According to the method, omnibearing mapping of the enterprise relationship is realized, so that the constructed enterprise relationship graph is more accurate and comprehensive.

Description

Technical Field

[0001] This application belongs to the field of data processing technology, and in particular relates to a method and apparatus for constructing an enterprise relationship graph. Background Technology

[0002] With the rapid development of big data technology, enterprise relationship graphs, as a tool for intuitively displaying complex relationships between enterprises, are playing an increasingly important role in fields such as industry analysis, risk control, and investment attraction. By constructing a network structure with enterprises as nodes and inter-enterprise relationships as edges, enterprise relationship graphs can clearly display various relationships between enterprises, providing data support for enterprise decision-making.

[0003] Currently, the construction of enterprise relationship graphs is mainly based on basic enterprise information and inter-enterprise relationship information. However, existing query platforms primarily provide explicit strong relationships such as business registration, capital, supply chain, and intellectual property, making it difficult to cover weaker relationships in more dimensions such as talent, space, industry, and field. These weaker relationships are also important in inter-enterprise cooperation and competition, but due to the difficulty in data acquisition and the challenge in quantifying relationship strength, existing technologies cannot effectively incorporate them into enterprise relationship graphs. This results in inaccurate and error-prone enterprise relationship graphs, which can easily lead to failed investment promotion recommendations or recommendations that do not meet actual business needs, thus missing potential investment leads. Summary of the Invention

[0004] This application provides a method and apparatus for constructing an enterprise relationship graph, which can solve the technical problem that in the existing enterprise investment promotion process, the publicly available data of enterprises often only covers explicit relationships such as business registration, investment and financing, and contracts, while a large number of implicit relationships such as industry, talent, intellectual property, and space are not quantified, resulting in inaccurate enterprise relationship graphs with large errors.

[0005] In a first aspect, embodiments of this application provide a method for constructing an enterprise relationship graph, the method comprising: The basic information and inter-enterprise relationship information of multiple enterprises are obtained from an enterprise relationship database. The basic information is used to represent the individual information of the enterprise, and the inter-enterprise relationship information is used to represent the association between two enterprises. Based on the inter-enterprise relationship information, the aggregation weight value of each enterprise pair is determined; wherein, two enterprises with an association relationship constitute one enterprise pair; Based on the basic information of the enterprises, the relationship information between the enterprises, and the aggregation weight value of each enterprise pair, an enterprise relationship graph is generated; wherein, the nodes in the enterprise relationship graph are used to represent the enterprises, and the edges in the enterprise relationship graph are used to represent the association relationship between the enterprise pairs.

[0006] In one possible implementation of the first aspect, the inter-enterprise relationship information includes multiple major categories of relationships, and each major category of relationship includes at least one minor category of relationship; The step of determining the aggregation weight value for each enterprise pair based on the inter-enterprise relationship information includes: Based on the inter-enterprise relationship information, determine the weight values ​​of multiple sub-category relationships for each enterprise on multiple sub-category relationships; Based on the type of the major category relationship and the weight values ​​of the minor category relationships in each major category relationship, the weight values ​​of the major category relationships for each enterprise are calculated. Based on the types of multiple major category relationships and the weight values ​​of multiple major category relationships, the multiple major category relationships of each enterprise pair are aggregated across categories to obtain the aggregation weight value of each enterprise pair.

[0007] In one possible implementation of the first aspect, determining the weight values ​​of multiple sub-category relationships for each enterprise on multiple sub-category relationships based on the inter-enterprise relationship information includes: Based on the inter-enterprise relationship information, determine the feature value of each enterprise pair in each sub-category relationship; From the feature values ​​of all the enterprise pairs, determine the maximum feature value in each of the subcategories; The subcategory relation weight value of each enterprise pair in each subcategory relation is determined based on the ratio of the feature value of each enterprise pair to the maximum feature value in each subcategory relation.

[0008] In one possible implementation of the first aspect, the inter-enterprise relationship information includes first relationship information and second relationship information. The first relationship information is used to characterize the legal or economic relationship that has occurred between the two enterprises and can be directly verified by the original voucher. The second relationship information is used to characterize the relationship between the two enterprises that has no direct legal or economic relationship and shows statistically significant similarity on a preset feature dimension. The step of calculating the weight values ​​of multiple major category relationships for each enterprise in multiple major category relationships based on the type of the major category relationship and the weight values ​​of multiple minor category relationships in each major category relationship includes: If the type of the major category relationship is the first relationship information, then the weighted weight value corresponding to the multiple minor category relationships in the major category relationship is determined, and the weight value of each enterprise pair in the multiple major category relationships is calculated based on the weighted weight value and the multiple minor category relationship weight values. If the type of the major category relationship is the second relationship information, then the dynamic compensation weight value corresponding to the multiple minor category relationships in the major category relationship is determined, and the weight value of each enterprise on the multiple major category relationships is calculated based on the dynamic compensation weight value and the multiple minor category relationship weight values.

[0009] In one possible implementation of the first aspect, when the type of the major category relationship is the second relationship information, determining the weighted weight value corresponding to the multiple minor category relationships in the major category relationship as a dynamic compensation weight value includes: When the type of the major category relationship is the second relationship information, the information entropy of each minor category relationship is calculated based on the probability distribution value of each minor category relationship in all enterprise pairs and the preset information entropy calculation formula; According to the preset conversion formula, the information entropy of each subcategory relationship is converted to obtain the dynamic compensation weight value.

[0010] In one possible implementation of the first aspect, the step of cross-category aggregation of the multiple major category relationships of each enterprise pair based on the types of the multiple major category relationships and the weight values ​​of the multiple major category relationships to obtain the aggregation weight value of each enterprise pair includes: The weight values ​​of multiple major categories of relationships for each enterprise pair are normalized to obtain normalized weight values. Based on the type of each major category relationship, determine the fusion importance value corresponding to each major category relationship; Based on the normalized weight value and the fusion importance value corresponding to each major category relationship, cross-category aggregation is performed on multiple major category relationships of each enterprise pair to calculate the aggregation weight value of each enterprise pair.

[0011] In one possible implementation of the first aspect, generating an enterprise relationship graph based on the basic information of the enterprises, the inter-enterprise relationship information, and the aggregation weight value of each enterprise pair includes: Each of the aforementioned enterprises is designated as a node in the enterprise relationship graph; wherein the information of the node is the basic information of the enterprise. The subcategories of relationships in the relationship information of each enterprise pair are used as subcategories of edges in the enterprise relationship graph, and the subcategories of relationship weights on each subcategories of relationships for each enterprise pair are used as subcategories of edge weights for the subcategories of edges in each node pair; wherein each node pair corresponds one-to-one with each enterprise pair. The major category relationship in the relationship information of each enterprise pair is used as the major category edge of the node pair in the enterprise relationship graph, and the major category relationship weight value of each enterprise pair on each major category relationship is used as the major category edge weight value of the major category edge in the node pair. The aggregated relationship obtained by aggregating multiple major category relationships of each enterprise pair is used as the fusion edge of the node pair in the enterprise relationship graph, and the aggregated weight value of each enterprise pair is used as the fusion edge weight value of the fusion edge in the node pair. Based on each node and the sub-category edge weight value of each sub-category edge, the major category edge weight value of each major category edge, and the fusion edge weight value of each fusion edge, an initial enterprise relationship graph is generated. The initial enterprise relationship graph is pruned to obtain the enterprise relationship graph.

[0012] In one possible implementation of the first aspect, the pruning of the initial enterprise relationship graph to obtain the enterprise relationship graph includes: By using a low-weight filtering method, the fusion edges whose weight values ​​are sorted according to a preset ratio are removed from all the fusion edges to obtain the filtered map. By using a redundancy filtering method, all the major category edges in each node pair in the filtered graph are pruned to obtain the enterprise relationship graph. In one possible implementation of the first aspect, after generating the enterprise relationship graph based on the basic information of the enterprises, the inter-enterprise relationship information, and the aggregate weight value of each enterprise pair, the method includes: Based on the weight values ​​of the multiple fused edges in the enterprise relationship graph, the shortest path between the enterprise and any other enterprise is determined by the shortest path algorithm.

[0013] Secondly, embodiments of this application provide an apparatus for constructing an enterprise relationship graph, comprising: The acquisition module retrieves basic information and inter-enterprise relationship information of multiple enterprises from an enterprise relationship database. The basic information is used to represent the individual information of the enterprises, and the inter-enterprise relationship information is used to represent the association relationship between two enterprises. The calculation module is used to determine the aggregation weight value of each enterprise pair based on the inter-enterprise relationship information; wherein two enterprises with an association relationship constitute one enterprise pair; The generation module is used to generate an enterprise relationship graph based on the basic information of the enterprises, the relationship information between the enterprises, and the aggregation weight value of each enterprise pair; wherein, the nodes in the enterprise relationship graph are used to represent the enterprises, and the edges in the enterprise relationship graph are used to represent the association relationship between the enterprise pairs.

[0014] Thirdly, embodiments of this application provide a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the enterprise relationship graph construction method described in any of the above claims.

[0015] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the enterprise relationship graph construction method described in any of the preceding claims.

[0016] Fifthly, embodiments of this application provide a computer program product that, when run on a terminal device, causes the terminal device to execute the enterprise relationship graph construction method described in any one of the first aspects above.

[0017] The beneficial effects of the embodiments in this application compared with the prior art are: This application provides a method for constructing an enterprise relationship graph, comprising: first, obtaining basic information and inter-enterprise relationship information of multiple enterprises from an enterprise relationship database, wherein the basic information is used to represent the individual information of an enterprise, and the inter-enterprise relationship information is used to represent the association between two enterprises; then, determining the aggregation weight value of each enterprise pair based on the inter-enterprise relationship information; wherein two enterprises with an association relationship constitute an enterprise pair; finally, generating an enterprise relationship graph based on the basic information of the enterprises, the inter-enterprise relationship information, and the aggregation weight value of each enterprise pair; wherein nodes in the enterprise relationship graph are used to represent the enterprises, and edges in the enterprise relationship graph are used to represent the association between enterprise pairs. This method generates an enterprise relationship graph based on the basic information of the enterprises, the inter-enterprise relationship information, and the aggregation weight value of each enterprise pair, achieving a comprehensive mapping of enterprise relationships, making the constructed enterprise relationship graph more accurate and comprehensive. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating a method for constructing an enterprise relationship graph according to an embodiment of this application; Figure 2 This is a schematic diagram of a method for recommending the shortest path provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an apparatus for constructing an enterprise relationship graph according to an embodiment of this application; Figure 4 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application. Detailed Implementation

[0020] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0021] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0022] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0023] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0024] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0025] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0026] With the rapid development of big data technology, enterprise relationship graphs, as a tool for intuitively displaying complex relationships between enterprises, are playing an increasingly important role in fields such as industry analysis, risk control, and investment attraction. By constructing a network structure with enterprises as nodes and inter-enterprise relationships as edges, enterprise relationship graphs can clearly display various relationships between enterprises, providing data support for enterprise decision-making.

[0027] Against the backdrop of deep integration of industrial digitalization, "precise investment attraction" has become the most pressing common need across various regions. However, existing query platforms mainly provide four types of explicit strong relationships: business registration, capital, supply chain, and intellectual property. They struggle to cover weaker relationships across multiple dimensions, such as talent, space, industry, and field. These weaker relationships are equally important in inter-enterprise cooperation and competition, but due to the difficulty in data acquisition and the challenge in quantifying relationship strength, existing technologies cannot effectively incorporate them into enterprise relationship graphs. This results in inaccurate and error-prone enterprise relationship graphs, making targeted investment attraction impossible. Furthermore, traditional graph algorithms fail to recommend the shortest path when strong relationship data is sparse, missing potential investment leads.

[0028] This application addresses the aforementioned issues by proposing a method for constructing a multidimensional relationship graph that integrates the weights of strong and weak relationships, thereby building a more accurate and comprehensive enterprise relationship graph. On one hand, it can unify 21-dimensional relationships across 9 major categories, including business, capital, supply chain, intellectual property, talent, space, industry, field, and industrial chain, into the graph model, enabling precise investment attraction targeting nodes within the industrial chain. On the other hand, in scenarios where strong relationships are lacking, it utilizes the dynamic compensation weights based on the information entropy differences of weak relationships, significantly improving the recall and interpretability of path recommendations, thus effectively supporting precise investment attraction applications with a "small entry point, large scenario" approach.

[0029] Please see Figure 1 , Figure 1 This is a flowchart illustrating a method for constructing an enterprise relationship graph according to an embodiment of this application. The method includes: S11. Obtain basic information and inter-enterprise relationship information of multiple enterprises from the enterprise relationship database. The basic information is used to represent the individual information of an enterprise, and the inter-enterprise relationship information is used to represent the relationship between two enterprises.

[0030] S12. Determine the aggregation weight value of each enterprise pair based on the inter-enterprise relationship information; where two enterprises with a relationship constitute an enterprise pair.

[0031] S13. Generate an enterprise relationship graph based on the basic information of enterprises, the relationship information between enterprises, and the aggregation weight value of each enterprise pair; wherein, the nodes in the enterprise relationship graph are used to represent enterprises, and the edges in the enterprise relationship graph are used to represent the association relationship between enterprise pairs.

[0032] It should be noted that in this embodiment, the executing entity can be a terminal device such as a server, and there is no specific limitation thereto. The enterprise relationship graph constructed in this embodiment can be applied to multiple application scenarios such as enterprise investment promotion, industry analysis, and risk prevention and control.

[0033] In step S11, the enterprise relationship database is a database that stores basic enterprise information and inter-enterprise relationship information. Basic enterprise information represents the individual characteristics of an enterprise, including but not limited to enterprise name, unified social credit code, legal representative, registered capital, and industry classification. This basic information uniquely identifies the enterprise and can be displayed as a node attribute in the enterprise relationship graph, helping to distinguish different enterprises. Inter-enterprise relationship information represents the connections between each enterprise and other enterprises, including but not limited to key personnel relationships, financial relationships, supply chain relationships, intellectual property relationships, senior management relationships, spatial relationships, industry relationships, domain relationships, and industrial chain relationships. This inter-enterprise relationship information reflects the strength and direction of the connections between enterprises and can be used to define the type and weight of edges in the enterprise relationship graph.

[0034] To balance "directly quantifiable legal and economic connections" with "potential similarities obtained through data mining," this application employs an "explicit-implicit" dual-dimensional determination method, uniformly classifying inter-enterprise relationships into first relationships (i.e., strong relationships) and second relationships (i.e., weak relationships). In other words, inter-enterprise relationship information can include both first-relationship and second-relationship information. First-relationship information characterizes existing legal or economic connections between two enterprises that can be directly verified by original documents. This information can include multiple broad categories such as key personnel relationships, financial relationships, supply chain relationships, and intellectual property relationships. Second-relationship information characterizes relationships between two enterprises that lack direct legal or economic connections but exhibit statistically significant similarity across preset feature dimensions. This information can include multiple broad categories such as executive relationships, spatial relationships, industry relationships, domain relationships, and supply chain relationships. In other words, first relationships are direct relationships between enterprises, while second relationships are indirect relationships with a certain degree of similarity across a specific dimension.

[0035] In one possible implementation, before retrieving basic information about multiple enterprises and inter-enterprise relationships from an enterprise relational database, the enterprise relational database needs to be constructed. The construction of the enterprise relational database includes: The original information data of each enterprise is cleaned and the coordinates are standardized to obtain the basic information of each enterprise.

[0036] Based on the basic information of each enterprise, determine the primary or secondary relationship information between any two enterprises.

[0037] A database of enterprise relationships is constructed based on the basic information of each enterprise and the primary or secondary relationship information between any two enterprises.

[0038] Raw data refers to the unprocessed initial data of a company. This raw data may include information from various sources such as business registration, investment and financing, patents, software copyrights, bidding announcements, product information on the company's official website, industry classification, and supply chain grading. This data may originate from channels such as business registration systems, news reports, company websites, and third-party data platforms. The data formats are diverse and may contain issues such as duplication, missing data, or errors. Therefore, it is necessary to perform data cleaning and coordinate unification processing on this raw data.

[0039] Data cleaning involves performing operations such as error correction, deduplication, filling in missing values, and standardizing the format of raw data to improve data quality. Coordinate standardization maps enterprise-related data to a unified standard or dimension, ensuring data comparability and consistency. Typically, data cleaning and coordinate standardization are performed on the raw data using the enterprise's "Unified Social Credit Code" as the primary key and the "Standardized Enterprise Name" as the secondary key to obtain the basic information identifying the enterprise. It should be understood that by performing data cleaning and coordinate standardization on the raw data and supplementing missing data, it is possible to ensure that the enterprise relationship graph covers the key characteristics and relationships of the enterprise and reduce the impact of noisy data on the accuracy of the enterprise relationship graph.

[0040] Then, based on the basic information of the identified companies, the primary or secondary relationship information between the two companies can be extracted. For example, extracting the legal person information of the companies from the basic information can determine the key personnel relationship between the two companies; extracting the bidding relationship and specific amounts between the companies from the basic information can determine the supply chain relationship between the two companies, etc.

[0041] After obtaining the basic information of each enterprise and the first or second relationship information between each pair of enterprises, this information is aggregated and stored to build an enterprise relationship database that stores information about multiple enterprises.

[0042] It should be understood that by constructing an enterprise relationship database, scattered raw enterprise data can be transformed into structured data, providing a high-quality data foundation for complex enterprise relationship analysis and reliable input for subsequent weight calculation and generation of enterprise relationship graphs.

[0043] In step S12, two related companies form a company pair. The aggregation weight value quantifies the degree of association between the two related companies.

[0044] Inter-enterprise relationship information includes multiple major categories of relationships, each of which contains at least one sub-category of relationships. Table 1 shows the classification of inter-enterprise relationship information. In Table 1, inter-enterprise relationship information can include primary relationship information and secondary relationship information. Primary relationship information can include multiple primary major categories of relationships, namely key personnel relationships, financial relationships, supply chain relationships, and intellectual property relationships. Each primary major category can further include at least one primary sub-category of relationships; for example, the financial relationship type can include three primary sub-categories: investment relationships, branch office relationships, and subsidiary relationships. Secondary relationship information can include multiple secondary major categories of relationships, namely executive relationships, spatial relationships, industry relationships, sectoral relationships, and supply chain relationships. Each secondary major category of relationships can further include at least one secondary sub-category of relationships; for example, the executive relationship type can include three secondary sub-categories: alumni relationships, hometown relationships, and colleague relationships.

[0045] Table 1. Classification of Inter-firm Relationship Information

[0046] In one possible implementation, the aggregation weight value for each pair of firms is determined based on inter-firm relationship information, including: Based on the inter-enterprise relationship information, determine the weight values ​​of each enterprise for multiple sub-category relationships across multiple sub-category relationships.

[0047] Based on the type of major category relationship and the weight values ​​of multiple minor category relationships within each major category relationship, the weight values ​​of multiple major category relationships for each enterprise are calculated.

[0048] Based on the types and weight values ​​of multiple major category relationships, the multiple major category relationships of each enterprise pair are aggregated across categories to obtain the aggregation weight value of each enterprise pair.

[0049] Specifically, since the inter-enterprise relationship information of each enterprise pair can contain multiple major categories of relationships, and each major category of relationship may also contain multiple minor categories of relationships, it is necessary to calculate the minor category relationship weight value corresponding to each minor category relationship and the major category relationship weight value corresponding to each major category relationship when determining the aggregation weight value of each enterprise pair.

[0050] In one possible implementation, based on inter-firm relationship information, the weight values ​​of each firm for multiple sub-categories of relationships across multiple major categories are determined, including: Based on the inter-firm relationship information, determine the feature values ​​of each firm pair in the subcategories of relationships.

[0051] From the eigenvalues ​​of all firm pairs, determine the largest eigenvalue in each subcategory relationship.

[0052] The subcategory relation weight value of each enterprise pair is determined based on the ratio of the eigenvalue to the maximum eigenvalue of each enterprise pair in each subcategory relation.

[0053] When calculating the weight value of a subcategory relationship, a percentage-based method can be used. Regardless of whether the subcategory relationship belongs to the first or second category of relationship information, the enterprise assigns weights to each subcategory relationship. Minor class relation weight values The ratio of the eigenvalue of the enterprise pair to the largest eigenvalue among all corresponding eigenvalues ​​of all enterprise pairs can be used for calculation.

[0054] Specifically, firstly, based on the inter-firm relationship information, the feature value corresponding to each firm pair is determined for each sub-category relationship. Then, for each sub-category relationship, the feature value with the largest value is selected from all the feature values ​​of firm pairs; this is the maximum feature value. Finally, the ratio of the feature value of each firm pair to its corresponding maximum feature value for each sub-category relationship is used as the sub-category relationship weight value for each firm pair in each sub-category relationship.

[0055] For example, subcategories If the relationship is based on the same legal entity, then the weight of the sub-category relationship corresponding to the current enterprise pair's same legal entity relationship is the ratio obtained by dividing the number of the same legal entity in the current enterprise pair by the maximum number of the same legal entity among all enterprise pairs. For example, if the number of the same legal entity in the current enterprise pair is 2, and the maximum number of the same legal entity among all enterprise pairs is 5, then the weight of this sub-category relationship is 0.4. For example, sub-category relationship... If the relationship is an investment relationship, then the weight of the sub-relationship corresponding to the current investment relationship is the ratio obtained by dividing the total investment amount of the current investment relationship by the largest total investment amount among all investment relationships. For example, if the total investment amount of the current investment relationship is 2 million and the largest total investment amount among all investment relationships is 10 million, then the weight of the sub-relationship is 0.2.

[0056] In one possible implementation, based on the type of major category relationship and the weight values ​​of multiple minor category relationships within each major category relationship, the weight values ​​of multiple major category relationships for each enterprise on multiple major category relationships are calculated, including: If the type of major category relationship is first relationship information, then the weighted weight values ​​corresponding to multiple minor category relationships in the major category relationship are determined, and the weight values ​​of multiple major category relationships for each enterprise are calculated based on the weighted weight values ​​and the weight values ​​of multiple minor category relationships.

[0057] When the type of major category relationship is second relationship information, the dynamic compensation weight value corresponding to multiple minor category relationships in the major category relationship is determined, and the weight value of each enterprise for multiple major category relationships in multiple major category relationships is calculated based on the dynamic compensation weight value and the weight values ​​of multiple minor category relationships.

[0058] Following the example above, after obtaining the subcategory relation weight value for each enterprise pair in each subcategory relation, we continue to calculate the major relation weight value for each enterprise pair in each major relation.

[0059] Since inter-enterprise relationship information includes primary relationship information and secondary relationship information, primary relationship information represents an existing legal or economic connection between two enterprises that can be directly verified by original documents, while secondary relationship information represents a connection between two enterprises that has no direct legal or economic connection but exhibits statistically significant similarity on preset feature dimensions. Therefore, the calculation methods for the weight values ​​of different types of relationship information differ. When calculating the weight value of a major category of relationship, it is necessary to calculate it according to the type of relationship information to which the major category of relationship belongs.

[0060] Specifically, when the major category relationship is of the first-level relationship type, since the first-level relationship information (i.e., strong relationship) directly contains both the relationship type and strength, the weight value corresponding to the first-level relationship information can be directly reflected by the relative strength of the actual relationship between the enterprises. This is equivalent to the weighted weight values ​​corresponding to multiple minor category relationships. In this embodiment, the weighted weight value corresponding to the minor category relationship can be... Relationships within the same broad category Relationships between the various subcategories Minor class relation weight value Perform weighted calculations to obtain the weight values ​​of the major categories of relations under the first relation information. ,Right now: ; in, This represents the weight value of the major category relation when j is the first relation information. This represents the weight value of the minor class relation. j represents the minor category relationship, and j represents the major category relationship. These are the weighted weight values.

[0061] When the major category relationship is of the second-order relation information type, a dynamic compensation weight value is used to reflect the rarity compensation of the second-order relation information. Perform a weighted summation to obtain the major category relation weights under the second relation information. ,Right now: ; in, This represents the weight value of the major relation when j is the second relation information. This represents the weight value of the minor class relation. j represents the minor category relationship, and j represents the major category relationship. This is for dynamically compensated weight values.

[0062] In one possible implementation, when the type of the major category relationship is second-level relationship information, the dynamic compensation weight values ​​corresponding to multiple minor category relationships within the major category relationship are determined, including: When the type of major category relationship is second-level relationship information, the information entropy of each minor category relationship is calculated based on the probability distribution value of each minor category relationship in all enterprise pairs and the preset information entropy calculation formula.

[0063] Based on the preset conversion formula, the information entropy of each subcategory relationship is converted to obtain a dynamic compensation weight value.

[0064] It should be noted that when the type of the major category relationship is second-level relationship information, the dynamic compensation weight value of each minor category relationship can be calculated based on the information entropy of each minor category relationship, specifically as follows: For each subcategory of relation in the second relation information Collect the minor class relation weights of all its enterprise pairs. Given a set, calculate the probability distribution of each unique value in that set. , probability distribution value Substituting into the preset information entropy calculation formula, the relationship of each subcategory is calculated. Information entropy .

[0065] The preset formula for calculating information entropy is: ; Then, according to the preset conversion formula, the relationships of each subcategory are... Information entropy The transformation yields the dynamic compensation weight value. In order to achieve rare features (i.e., low...) High compensation target, dynamic compensation weight value Need and information entropy Negative correlation.

[0066] The preset conversion formula is: ; ; ; in, For all subcategories of relationships The maximum value in the information entropy; Relationships for each subcategory The percentage of information entropy.

[0067] In one possible implementation, based on the types and weight values ​​of multiple major category relationships, the multiple major category relationships of each enterprise pair are aggregated across categories to obtain the aggregation weight value of each enterprise pair, including: The weight values ​​of multiple major relationship categories for each enterprise are normalized to obtain normalized weight values.

[0068] Based on the type of each major category relationship, determine the fusion importance value corresponding to each major category relationship.

[0069] Based on the normalized weight value and fusion importance value corresponding to each major category relationship, cross-category aggregation is performed on multiple major category relationships of each enterprise pair to calculate the aggregation weight value of each enterprise pair.

[0070] Following the example above, after obtaining the weight values ​​of multiple major category relationships for each enterprise pair, the aggregation weight value for each enterprise pair can be calculated, aggregating different major category relationships between the same enterprise pair into a single aggregation relationship. Specifically: First, considering that the weight values ​​of different major relationship categories have different meanings and units, relationships of different major categories cannot be directly included in the fusion calculation under different dimensions. Therefore, normalization is required to obtain weight values ​​under a unified dimension. This involves normalizing the weight values ​​of the major relationship categories corresponding to multiple major relationships for each enterprise pair. The weight values ​​of the major categories of relations , Perform normalization processing (such as min-max normalization) separately to obtain the final normalized weight values. , .

[0071] Then, based on the type of each major category relationship, the fusion importance value corresponding to each major category relationship is determined. Specifically, when a major category relationship belongs to the first category of relationship information, the fusion importance value corresponding to that major category relationship is... When the major category relationship belongs to the second category of relationship information, the fusion importance value corresponding to that major category relationship is... In this embodiment, it can be set to... =2, =1.

[0072] Finally, based on the normalized weight value and fusion importance value corresponding to each major category relationship, cross-category aggregation is performed on the multiple major category relationships of each enterprise pair to calculate the aggregation weight value of each enterprise pair, i.e.: ; in, Indicates the company's The aggregate weight value, This indicates the major category relationship belonging to the first relation information. The corresponding fusion importance value, This indicates a major category of relation belonging to the second relation information. The corresponding fusion importance value, This indicates the major category relationship belonging to the first relation information. The corresponding normalized weight values, This indicates a major category of relation belonging to the second relation information. The corresponding normalized weight values.

[0073] In step S13, the enterprise relationship graph is a data model that uses a graph structure (nodes + edges) to represent enterprises and the relationships between them. Nodes represent enterprises, edges represent the relationship types between two enterprises, and edge weights reflect the strength of each relationship type between enterprises. The enterprise relationship graph can intuitively display the ecological network of various enterprises and can support various application scenarios such as risk propagation analysis, enterprise investment attraction, and supply chain optimization.

[0074] In one possible implementation, an enterprise relationship graph is generated based on the basic information of the enterprises, the relationship information between enterprises, and the aggregation weight value of each enterprise pair, including: Each enterprise is treated as a node in an enterprise relationship graph; the information of each node is the basic information of the enterprise.

[0075] The subcategories of relationships in the relationship information of each enterprise pair are used as subcategories of edges in the enterprise relationship graph. The weight values ​​of subcategories of relationships in each enterprise pair on each subcategorie are used as the weight values ​​of subcategories of edges in each node pair. Each node pair corresponds one-to-one with each enterprise pair.

[0076] The major category relationships in the relationship information of each enterprise pair are used as the major category edges of the node pairs in the enterprise relationship graph, and the major category relationship weight value of each enterprise pair on each major category relationship is used as the major category edge weight value of the node pairs.

[0077] The aggregated relationship obtained by aggregating multiple major category relationships of each enterprise pair is used as the fusion edge of the node pair in the enterprise relationship graph, and the aggregated weight value of each enterprise pair is used as the fusion edge weight value of the fusion edge in the node pair.

[0078] An initial enterprise relationship graph is generated based on each node and the weight values ​​of the minor category edges connected to each node, the major category edges connected to each major category, and the fusion edge weight values ​​of each fusion edge.

[0079] The initial enterprise relationship graph is pruned to obtain the new enterprise relationship graph.

[0080] In a business relationship graph, nodes represent each business, and each node stores basic information about the business. Different businesses can be intuitively identified and distinguished through these nodes. An edge is a line connecting two nodes in the graph, indicating a relationship between the two businesses. The edge weight is a value labeled on the edge, used to quantify the strength of the relationship between the businesses represented by the edge.

[0081] In constructing a business relationship graph, edges can include sub-category edges, major category edges, and fusion edges. Sub-category edges represent sub-category relationships within the relationship information of each business pair; major category edges represent major category relationships within the relationship information of each business pair; and fusion edges represent aggregated relationships obtained by aggregating multiple major category relationships for each business pair. In the final business relationship graph, there is a fusion edge between any two nodes. Correspondingly, edge weight values ​​can include sub-category edge weight values, major category edge weight values, and fusion edge weight values. Sub-category edge weight values ​​represent the weight of sub-category relationships for each business pair within each sub-category relationship; major category edge weight values ​​represent the weight of major category relationships for each business pair within each major category relationship; and fusion edge weight values ​​represent the aggregated weight obtained by aggregating multiple major category relationships for each business pair. The integrated edge weight value can be understood as a comprehensive weight value obtained by considering the weight values ​​of all major categories of edges connected to each node in the enterprise relationship graph, and calculating it using a certain method (such as adding the weight values ​​of all connected major categories of edges, taking the average, or weighted summation). The integrated edge weight value can reflect the degree of association between two enterprises in the graph. For example, if the weight value of the investment relationship between enterprise A and enterprise B is 0.6, and the weight value of the spatial relationship is 0.3, then in the enterprise relationship graph, the weight value of the investment relationship edge between node A (representing enterprise A) and node B (representing enterprise B) is set to 0.6, and the weight value of the spatial relationship edge is set to 0.3.

[0082] Specifically, each enterprise is treated as a node, and the basic information of each enterprise is used as the information of the corresponding node. Each enterprise pair is treated as a node pair.

[0083] Then, the minor category relationships in the enterprise pair relationship information are used as minor category edges for each node pair, and the minor category relationship weight value of each enterprise pair on each minor category relationship is used as the minor category edge weight value of each node pair's minor category edge. This yields a minor category edge enterprise relationship network graph. Correspondingly, the major category relationships in the enterprise pair relationship information are used as major category edges for each node pair, and the major category relationship weight value of each enterprise pair on each major category relationship is used as the major category edge weight value of each node pair's major category edge. This yields a major category edge enterprise relationship network graph. Finally, the aggregation relationships obtained by aggregating multiple major category relationships of an enterprise pair are used as fusion edges for each node pair, and the aggregation weight value of each enterprise pair on the aggregation relationship is used as the fusion edge weight value of each node pair's fusion edge. This constructs an initial enterprise relationship network graph, i.e., an initial enterprise relationship graph.

[0084] The initial enterprise relationship graph contains raw information about the relationships between various enterprises. However, it may contain some redundant or less important information. Therefore, the initial enterprise relationship graph needs to be pruned to remove unimportant or redundant edges, resulting in the final enterprise relationship graph. This final graph more concisely and accurately reflects the key and core relationships between enterprises.

[0085] It is understood that the enterprise relationship graph construction method provided in this application includes: first, obtaining basic information and inter-enterprise relationship information of multiple enterprises from an enterprise relationship database, wherein the basic information is used to represent the individual information of an enterprise, and the inter-enterprise relationship information is used to represent the association relationship between two enterprises; then, determining the aggregation weight value of each enterprise pair based on the inter-enterprise relationship information; wherein two enterprises with an association relationship constitute an enterprise pair; finally, generating an enterprise relationship graph based on the basic information of the enterprises, the inter-enterprise relationship information, and the aggregation weight value of each enterprise pair; wherein the nodes in the enterprise relationship graph are used to represent the enterprises, and the edges in the enterprise relationship graph are used to represent the association relationship between enterprise pairs. This method generates an enterprise relationship graph based on the basic information of the enterprises, the inter-enterprise relationship information, and the aggregation weight value of each enterprise pair, achieving a comprehensive mapping of enterprise relationships, making the constructed enterprise relationship graph more accurate and comprehensive.

[0086] In one possible implementation, the initial enterprise relationship graph is pruned to obtain the enterprise relationship graph, including: By using a low-weight filtering method, fusion edges whose weight values ​​are sorted according to a preset ratio are removed from all fusion edges, resulting in a filtered graph.

[0087] By using a redundancy filtering method, repeated edge clipping is performed on all major category connections in each node pair of the filtered graph to obtain the enterprise relationship graph.

[0088] To enable rapid recommendations using the enterprise relationship graph, the initial graph needs to be pruned. This involves removing redundant nodes or edges to create a lighter network while maintaining its original performance and architecture. In this embodiment, low-weight filtering and redundancy filtering methods can be used to prune the initial enterprise relationship graph. The low-weight filtering method sorts the weight values ​​of all merged edges in a preset order and then removes merged edges whose weight values ​​fall within a preset proportion. The redundancy filtering method addresses redundant edges in the filtered graph by pruning duplicate edges of all major categories in each node pair, retaining only the edge with the highest weight value among multiple duplicate edges.

[0089] Specifically, firstly, the weight values ​​of all merged edges in the initial relation graph (i.e., The edges are sorted, either from lowest to highest or highest to lowest. In this embodiment, all fused edge weights are sorted in descending order. Then, the number of edges to be removed is determined according to a preset ratio, and these detached fused edges are removed from the initial relationship graph to obtain the filtered graph. The preset ratio is a pre-defined percentage; in this embodiment, the preset ratio can be 20%. That is, if there are 100 fused edges in the initial relationship graph and the preset ratio is 20%, then the fused edge weights of these 100 fused edges are sorted from highest to lowest, and the last 20 fused edges are removed to obtain the filtered graph.

[0090] It should be understood that by removing low-weighted edges and eliminating loose edges that have little impact on the enterprise relationship network and have weak relationships, the graph can focus more on the important relationships between enterprises, reduce unnecessary interference information, and improve the efficiency and accuracy of subsequent analysis.

[0091] Furthermore, based on the filtered graph, all major category edges in each node pair are traversed. If the number of major category edges in the current node pair is greater than 1, the major category edge with the largest weight value is retained, and the others are deleted. If the weight values ​​of these major category edges are all equal, one of them is randomly retained.

[0092] In this way, after pruning the initial enterprise relationship graph using low-weight screening and redundancy screening methods, the final enterprise relationship graph can be obtained. In this graph, there is only one connecting edge between each pair of nodes, i.e., the merged connecting edge.

[0093] It should be understood that by eliminating free edges and redundant information in the graph, the relationships between enterprises are presented in a simpler and clearer way, avoiding increased graph complexity due to repeated edges. At the same time, it can more accurately reflect the true relationship structure between enterprises, facilitating more in-depth enterprise relationship analysis.

[0094] In one possible implementation, after generating an enterprise relationship graph based on the basic information of the enterprises, inter-enterprise relationship information, and the aggregate weight value of each enterprise pair, the method includes: Based on the weight values ​​of multiple fused edges in the enterprise relationship graph, the shortest path between an enterprise and any other enterprise is determined using the shortest path algorithm.

[0095] After generating the enterprise relationship graph, since the generated enterprise relationship graph is a weighted graph, the shortest path algorithm can be used to find the shortest path from a node to any other node in the graph (i.e. the shortest path between a company and any other company). By fusing the first relationship information and the second relationship information (i.e. strong and weak relationships), users can track and understand all possible connection paths between various companies, thereby revealing how companies interact with each other.

[0096] The shortest path algorithm is used to find the shortest path between two nodes in a graph structure. In a graph, there may be multiple paths from one node to another. The shortest path algorithm can find the path with the fewest edges or the smallest sum of weights. The process of determining the shortest path using the shortest path algorithm can be as follows: Take one node (i.e., one enterprise) in the enterprise relationship graph as the starting point and another node (i.e., another enterprise) as the ending point. The merged edge between the node pairs is the path, and the weight value of the merged edge is the path length. Starting from the starting point, explore the nodes connected to it step by step, and calculate the path length to each node based on the merged edge weight value. Continuously update the shortest path information to each node until the shortest path from the starting point to the ending point is found.

[0097] In this embodiment, by using the shortest path algorithm based on the weight value of each fusion edge in the enterprise relationship graph, the most direct and closely related path between an enterprise and any other enterprise can be determined. In practical applications, the shortest path can represent the most convenient cooperation channel, the lowest-cost supply chain path, or the most promising business expansion direction, which is of great significance for an enterprise's strategic planning, resource allocation, and risk management.

[0098] like Figure 2 As shown, Figure 2 This is a schematic diagram illustrating a method for recommending shortest paths according to an embodiment of this application. Figure 2First, based on eight categories of raw data—including business registration, investment and financing, patents, software copyrights, bidding announcements, product information from company websites, industry classifications, and supply chain grading—the database uses the "Unified Social Credit Code" as the primary key and the "Standardized Company Name" as the secondary key. Data cleaning, deduplication, missing value completion, and coordinate unification are performed to form a corporate information database. Then, to balance "directly quantifiable legal / economic connections" with "potential similarities obtained through data mining," an explicit-implicit dual-dimensional judgment method is adopted to uniformly classify inter-company relationships into first-level relationship information (strong relationships) and second-level relationship information (weak relationships). The weight value of each type of relationship is calculated separately, and finally, the weights are fused to obtain a corporate relationship graph. Low-weight pruning and redundancy pruning are then performed on the corporate relationship graph to recommend the shortest path from one company to any other company.

[0099] It should be understood that determining the shortest path between a firm and any other firm can help analyze the most direct and effective ways in which firms are connected.

[0100] It is understood that the enterprise relationship graph construction method provided in this application embodiment has the following advantages compared with existing enterprise relationship graph construction methods: (1) Dimensional expansion. For the first time, the 21-dimensional relationships of 9 major categories, namely "industry, capital, supply, talent, intellectual property, space, industry, field, and industrial chain", are unified into a single graph model, which improves the relationship coverage by more than 5 times compared with the existing 3-4 dimensional graph model platform.

[0101] (2) Sparsity enhancement. In scenarios where strong relationships are missing, Shannon entropy is used to dynamically compensate for weak relationship weights, making the enterprise relationship graph more accurate and comprehensive, thereby significantly improving the path recall rate and reducing the loss of investment attraction caused by data blind spots.

[0102] (3) Targeted investment attraction along the industrial chain. By integrating hierarchical industrial chain tag data, it is possible to “locate enterprises by link and recommend paths by link”, and directly output a targeted investment attraction chain of “venture capital - alumni fund - target enterprise”.

[0103] (4) Explainability. Each recommended path can also be accompanied by three-dimensional explanatory indicators such as "strong / weak ratio, weight source, and industry chain coverage", which makes it easier for investment promotion personnel to quickly assess credibility and commercial value.

[0104] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0105] A method for constructing an enterprise relationship graph corresponding to the above embodiment, Figure 3The diagram shows a structural schematic of an apparatus for constructing an enterprise relationship graph according to an embodiment of this application. For ease of explanation, only the parts related to the embodiment of this application are shown.

[0106] Reference Figure 3 The enterprise relationship graph construction apparatus 2 of this embodiment includes: The acquisition module 21 is used to acquire basic information and inter-enterprise relationship information of multiple enterprises from the enterprise relational database. The basic information is used to represent the individual information of an enterprise, and the inter-enterprise relationship information is used to represent the relationship between two enterprises.

[0107] The calculation module 22 is used to determine the aggregation weight value of each enterprise pair based on the inter-enterprise relationship information; wherein two enterprises with an association relationship constitute an enterprise pair.

[0108] The generation module 23 is used to generate an enterprise relationship graph based on the basic information of enterprises, the relationship information between enterprises, and the aggregation weight value of each enterprise pair; wherein, the nodes in the enterprise relationship graph are used to represent enterprises, and the edges in the enterprise relationship graph are used to represent the association relationship between enterprise pairs.

[0109] Furthermore, the inter-enterprise relationship information includes multiple major categories of relationships, and each major category of relationship includes at least one minor category of relationship; Calculation module 22 includes: The sub-module for calculating minor categories is used to determine the weight values ​​of each enterprise for multiple minor category relationships based on the relationship information between enterprises.

[0110] The category calculation submodule is used to calculate the weight values ​​of multiple major category relationships for each enterprise based on the type of major category relationship and the weight values ​​of multiple minor category relationships in each major category relationship.

[0111] The aggregation calculation submodule is used to aggregate the multiple major category relationships of each enterprise pair across categories based on the type of multiple major category relationships and the weight values ​​of multiple major category relationships, so as to obtain the aggregation weight value of each enterprise pair.

[0112] Furthermore, the sub-module for calculating minor categories includes: The feature value determination unit is used to determine the feature value of each enterprise pair in each subcategory relationship based on the inter-enterprise relationship information.

[0113] The maximum value determination unit is used to determine the maximum eigenvalue in each subcategory relationship from the eigenvalues ​​of all enterprise pairs.

[0114] The sub-category weight value determination unit is used to determine the sub-category relationship weight value of each enterprise pair in each sub-category relationship based on the ratio of the feature value to the maximum feature value of each enterprise pair in each sub-category relationship.

[0115] Furthermore, the inter-enterprise relationship information includes first relationship information and second relationship information. The first relationship information is used to characterize the legal or economic relationship that has occurred between two enterprises and can be directly verified by the original vouchers. The second relationship information is used to characterize the relationship between two enterprises that has no direct legal or economic relationship and shows statistically significant similarity on the preset feature dimensions.

[0116] The major category calculation submodule includes: The first major calculation unit is used to determine the weighted weight values ​​corresponding to multiple minor category relationships in the major category relationship when the type of major category relationship is the first relationship information. Based on the weighted weight values ​​and the weight values ​​of multiple minor category relationships, it calculates the weight values ​​of multiple major category relationships for each enterprise in multiple major category relationships.

[0117] The second major calculation unit is used to determine the dynamic compensation weight values ​​corresponding to multiple minor category relationships in the major category relationship when the type of major category relationship is second relationship information. Based on the dynamic compensation weight values ​​and the weight values ​​of multiple minor category relationships, the unit calculates the weight values ​​of multiple major category relationships for each enterprise in multiple major category relationships.

[0118] Furthermore, the second major category of computing units includes: The information entropy calculation unit is used to calculate the information entropy of each sub-category relationship when the type of the major category relationship is second-category information, based on the probability distribution value of each sub-category relationship in all enterprise pairs and the preset information entropy calculation formula.

[0119] The dynamic compensation weight value calculation unit is used to convert the information entropy of each subcategory relationship into a dynamic compensation weight value according to a preset conversion formula.

[0120] Furthermore, the aggregation calculation submodule includes: The normalization processing unit is used to normalize the weight values ​​of multiple major categories of relationships for each enterprise pair to obtain normalized weight values.

[0121] The fusion importance value calculation unit is used to determine the fusion importance value corresponding to each major category relationship based on the type of each major category relationship; The aggregation weight value calculation unit is used to perform cross-category aggregation on multiple major category relationships of each enterprise pair based on the normalized weight value and fusion importance value corresponding to each major category relationship, and calculate the aggregation weight value of each enterprise pair.

[0122] Furthermore, the generation module 23 includes: The node determination unit is used to identify each enterprise as a node in the enterprise relationship graph; the node information is the basic information of the enterprise. The minor category edge determination unit is used to determine the minor category relationships in the relationship information of each enterprise pair as minor category edges in the enterprise relationship graph, and to determine the minor category relationship weight value of each enterprise pair on each minor category relationship as the minor category edge weight value of each node pair. Each node pair corresponds one-to-one with each enterprise pair.

[0123] The major category edge determination unit is used to take the major category relationship in the relationship information of each enterprise pair as the major category edge of the node pair in the enterprise relationship graph, and to take the major category relationship weight value of each enterprise pair on each major category relationship as the major category edge weight value of the major category edge in the node pair.

[0124] The fusion edge determination unit is used to aggregate the multiple major category relationships of each enterprise pair into the aggregated relationship, and use the aggregated relationship as the fusion edge of the node pair in the enterprise relationship graph. It also uses the aggregated weight value of each enterprise pair as the fusion edge weight value of the fusion edge in the node pair.

[0125] The initial graph generation unit is used to generate an initial enterprise relationship graph based on each node and the weight values ​​of the minor category edges connected to each node, the major category edges connected to each node, and the fusion edge weight values ​​of each fusion edge.

[0126] The graph optimization unit is used to prune the initial enterprise relationship graph to obtain the enterprise relationship graph.

[0127] Furthermore, the map optimization unit includes: The first filtering subunit is used to remove all fused edges whose weight values ​​are sorted according to a preset ratio using a low-weight filtering method, thus obtaining the filtered map.

[0128] The second filtering subunit is used to perform redundant edge trimming on all major category edges in each node pair of the filtered graph using a redundancy filtering method, thereby obtaining an enterprise relationship graph. Furthermore, the enterprise relationship graph construction device 2 includes: The shortest path determination module is used to determine the shortest path between an enterprise and any other enterprise based on the weight values ​​of multiple fused edges in the enterprise relationship graph and the shortest path algorithm.

[0129] It should be noted that the information interaction and execution process between the modules in the above-mentioned enterprise relationship graph construction device 2 are based on the same concept as the method embodiment of this application. For details on their specific functions and technical effects, please refer to the method embodiment section, which will not be repeated here.

[0130] This application also provides a terminal device, such as... Figure 4 As shown, Figure 4 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application. (Refer to...) Figure 4 The terminal device 3 in this embodiment includes a memory 31, a processor 32, and a computer program stored in the memory 31 and executable on the processor 32. When the processor 32 executes the computer program, it implements the steps in the embodiment of the enterprise relationship graph construction method of any of the above.

[0131] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps described in the various method embodiments above.

[0132] This application provides a computer program product that, when run on a mobile terminal, enables the mobile terminal to implement the steps described in the above-described method embodiments.

[0133] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. A computer-readable medium can include at least: any entity or device capable of carrying computer program code to a photographic device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0134] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0135] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0136] In the embodiments provided in this application, it should be understood that the disclosed apparatus / network devices and methods can be implemented in other ways. For example, the apparatus / network device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0137] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0138] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for constructing an enterprise relationship graph, characterized in that, include: The basic information and inter-enterprise relationship information of multiple enterprises are obtained from an enterprise relationship database. The basic information is used to represent the individual information of the enterprise, and the inter-enterprise relationship information is used to represent the association between two enterprises. Based on the inter-enterprise relationship information, the aggregation weight value of each enterprise pair is determined; wherein, two enterprises with an association relationship constitute one enterprise pair; Based on the basic information of the enterprises, the relationship information between the enterprises, and the aggregation weight value of each enterprise pair, an enterprise relationship graph is generated; wherein, the nodes in the enterprise relationship graph are used to represent the enterprises, and the edges in the enterprise relationship graph are used to represent the association relationship between the enterprise pairs.

2. The method for constructing an enterprise relationship graph as described in claim 1, characterized in that, The inter-enterprise relationship information includes multiple major categories of relationships, and each major category of relationship includes at least one minor category of relationship. The step of determining the aggregation weight value for each enterprise pair based on the inter-enterprise relationship information includes: Based on the inter-enterprise relationship information, determine the weight values ​​of multiple sub-category relationships for each enterprise on multiple sub-category relationships; Based on the type of the major category relationship and the weight values ​​of the minor category relationships in each major category relationship, the weight values ​​of the major category relationships for each enterprise are calculated. Based on the types of multiple major category relationships and the weight values ​​of multiple major category relationships, the multiple major category relationships of each enterprise pair are aggregated across categories to obtain the aggregation weight value of each enterprise pair.

3. The method for constructing an enterprise relationship graph as described in claim 2, characterized in that, The step of determining the weight values ​​of multiple subcategories of relationships for each enterprise based on the inter-enterprise relationship information includes: Based on the inter-enterprise relationship information, determine the feature value of each enterprise pair in each sub-category relationship; From the feature values ​​of all the enterprise pairs, determine the maximum feature value in each of the subcategories; The subcategory relation weight value of each enterprise pair in each subcategory relation is determined based on the ratio of the feature value of each enterprise pair to the maximum feature value in each subcategory relation.

4. The method for constructing an enterprise relationship graph as described in claim 3, characterized in that, The inter-enterprise relationship information includes first relationship information and second relationship information. The first relationship information is used to characterize the legal or economic relationship that has occurred between the two enterprises and can be directly verified by the original vouchers. The second relationship information is used to characterize the relationship between the two enterprises that has no direct legal or economic relationship and shows statistically significant similarity on a preset feature dimension. The step of calculating the weight values ​​of multiple major category relationships for each enterprise in multiple major category relationships based on the type of the major category relationship and the weight values ​​of multiple minor category relationships in each major category relationship includes: If the type of the major category relationship is the first relationship information, then the weighted weight value corresponding to the multiple minor category relationships in the major category relationship is determined, and the weight value of each enterprise pair in the multiple major category relationships is calculated based on the weighted weight value and the multiple minor category relationship weight values. If the type of the major category relationship is the second relationship information, then the dynamic compensation weight value corresponding to the multiple minor category relationships in the major category relationship is determined, and the weight value of each enterprise on the multiple major category relationships is calculated based on the dynamic compensation weight value and the multiple minor category relationship weight values.

5. The method for constructing an enterprise relationship graph as described in claim 4, characterized in that, When the type of the major category relationship is the second relationship information, the dynamic compensation weight values ​​corresponding to the multiple minor category relationships in the major category relationship are determined, including: When the type of the major category relationship is the second relationship information, the information entropy of each minor category relationship is calculated based on the probability distribution value of each minor category relationship in all enterprise pairs and the preset information entropy calculation formula; According to the preset conversion formula, the information entropy of each subcategory relationship is converted to obtain the dynamic compensation weight value.

6. The method for constructing an enterprise relationship graph as described in claim 5, characterized in that, The step of cross-category aggregation of the multiple major category relationships for each enterprise pair, based on the types of the multiple major category relationships and the weight values ​​of the multiple major category relationships, to obtain the aggregation weight value for each enterprise pair, includes: The weight values ​​of multiple major categories of relationships for each enterprise pair are normalized to obtain normalized weight values. Based on the type of each major category relationship, determine the fusion importance value corresponding to each major category relationship; Based on the normalized weight value and the fusion importance value corresponding to each major category relationship, cross-category aggregation is performed on multiple major category relationships of each enterprise pair to calculate the aggregation weight value of each enterprise pair.

7. The method for constructing an enterprise relationship graph as described in claim 6, characterized in that, The step of generating an enterprise relationship graph based on the basic information of the enterprises, the inter-enterprise relationship information, and the aggregation weight value of each enterprise pair includes: Each of the aforementioned enterprises is designated as a node in the enterprise relationship graph; wherein the information of the node is the basic information of the enterprise. The subcategories of relationships in the relationship information of each enterprise pair are used as subcategories of edges in the enterprise relationship graph, and the subcategories of relationship weights on each subcategories of relationships for each enterprise pair are used as subcategories of edge weights for the subcategories of edges in each node pair; wherein each node pair corresponds one-to-one with each enterprise pair. The major category relationship in the relationship information of each enterprise pair is used as the major category edge of the node pair in the enterprise relationship graph, and the major category relationship weight value of each enterprise pair on each major category relationship is used as the major category edge weight value of the major category edge in the node pair. The aggregation relationship obtained by aggregating multiple major category relationships of each enterprise pair across categories is used as the fusion connection edge of the node pair in the enterprise relationship graph, and the aggregation weight value of each enterprise pair is used as the fusion connection edge weight value of the fusion connection edge in the node pair. Based on each node and the sub-category edge weight value of each sub-category edge, the major category edge weight value of each major category edge, and the fusion edge weight value of each fusion edge, an initial enterprise relationship graph is generated. The initial enterprise relationship graph is pruned to obtain the enterprise relationship graph.

8. The method for constructing an enterprise relationship graph as described in claim 7, characterized in that, The step of pruning the initial enterprise relationship graph to obtain the enterprise relationship graph includes: By using a low-weight filtering method, the fusion edges whose weight values ​​are sorted according to a preset ratio are removed from all the fusion edges to obtain the filtered map. By using a redundancy filtering method, all the major category edges in each node pair in the filtered graph are pruned to obtain the enterprise relationship graph.

9. The method for constructing an enterprise relationship graph as described in claim 8, characterized in that, After generating the enterprise relationship graph based on the inter-enterprise relationship information and the aggregation weight value of each enterprise pair, the method includes: Based on the weight values ​​of the multiple fused edges in the enterprise relationship graph, the shortest path between the enterprise and any other enterprise is determined by the shortest path algorithm.

10. An apparatus for constructing an enterprise relationship graph, characterized in that, include: The acquisition module retrieves basic information and inter-enterprise relationship information of multiple enterprises from an enterprise relationship database. The basic information is used to represent the individual information of the enterprises, and the inter-enterprise relationship information is used to represent the association relationship between two enterprises. The calculation module is used to determine the aggregation weight value of each enterprise pair based on the inter-enterprise relationship information; wherein two enterprises with an association relationship constitute one enterprise pair; The generation module is used to generate an enterprise relationship graph based on the basic information of the enterprises, the relationship information between the enterprises, and the aggregation weight value of each enterprise pair; wherein, the nodes in the enterprise relationship graph are used to represent the enterprises, and the edges in the enterprise relationship graph are used to represent the association relationship between the enterprise pairs.

Images