Enterprise associated conduction risk monitoring method and device

Abstract

The invention relates to the field of financial risk monitoring, and particularly provides an enterprise association conduction risk monitoring method and device, and the method comprises the following steps: S1, building an enterprise association relationship graph; s2, constructing an enterprise associated risk system; s3, constructing an enterprise association index model; s4, constructing an enterprise association conduction risk evaluation model; and S5, establishing an enterprise associated conduction risk monitoring system. Compared with the prior art, the method has the advantages that real-time monitoring of risks can be realized by introducing a big data real-time framework such as SparkStreaming, and the flexibility and timeliness of the system are improved.

Description

Technical Field

[0001] The present invention relates to the field of financial risk monitoring, and specifically provides a method and device for monitoring the risk of enterprise associated conduction. Background Art

[0002] With the acceleration of the processes of globalization and informatization, enterprises are facing risk challenges from various aspects such as the market and operation. Enterprise risk monitoring is particularly important in the current complex and changeable economic environment, and the demand for enterprise risk monitoring in the financial credit field is even more urgent. Currently, financial technology technologies such as machine learning and graph databases have been widely applied in the field of financial risk monitoring. Machine learning algorithms can extract key features from a large amount of financial data, build accurate risk assessment models, and achieve risk early warning and control. The graph database, with its unique data structure and efficient query capabilities, plays an important role in processing complex relational data and realizing real-time risk monitoring and early warning. The analytic hierarchy process decomposes the elements related to decision-making into multiple levels such as goals, criteria, and solutions, and conducts qualitative and quantitative analysis on this basis. It is a systematic, simple, flexible, and effective decision-making method. This algorithm is a multi-index comprehensive evaluation algorithm, which is simple and practical and is widely used in the construction of models in the financial industry for index weight determination, quantitative solution selection, etc.

[0003] Enterprise post-loan risk monitoring is an important part of the entire process of enterprise financial credit. By establishing a perfect risk monitoring system, potential risks of enterprises can be discovered in a timely manner, the impact of risks can be evaluated, and corresponding prevention and control measures can be taken to ensure the smooth progress of business activities.

[0004] Currently, the monitoring of enterprise risks often only focuses on real-time monitoring of the risks that have already occurred in the enterprise itself. When risks occur in related enterprises such as the shareholders, investment institutions, and upstream and downstream suppliers of the enterprise, the risks will be transmitted to the target enterprise due to the associated relationship.

[0005] How to apply technologies such as machine learning, graph databases, and big data to the monitoring of enterprise associated conduction risks and expand the scope of enterprise risk monitoring is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0006] The present invention aims at the deficiencies of the above-mentioned existing technologies and provides a method for monitoring the risk of enterprise associated conduction with strong practicability.

[0007] A further technical task of the present invention is to provide a device for monitoring the risk of enterprise associated conduction with reasonable design, safety and applicability.

[0008] The technical solution adopted by the present invention to solve its technical problems is as follows:

[0009] A method for monitoring the risk of enterprise associated conduction has the following steps:

[0010] S1. Construction of enterprise association relationship graph;

[0011] S2. Construction of enterprise association risk system;

[0012] S3. Construction of enterprise association index model;

[0013] S4. Construction of enterprise association conduction risk evaluation model;

[0014] S5. Construction of enterprise association conduction risk monitoring system.

[0015] Furthermore, in step S1, it includes:

[0016] S1-1. Types of enterprise association relationships;

[0017] It includes direct enterprise relationships and suspected relationships. Direct enterprise relationships include branches, external investments, enterprise shareholders, natural person shareholders, directors, supervisors, senior managers, and historical directors, supervisors, senior managers; enterprise suspected relationships include the same actual controller relationship, the same telephone number, the same email, the same communication address, the same registered address, the same domain name information, the same judgment documents, the same enterprise name, the same patent information, and the same software copyright.

[0018] S1-2. Extraction of enterprise association relationships;

[0019] Using the multi-source data of its own enterprises as the extraction source, determining the extraction rules for each relationship type. The enterprise multi-source data is stored in the distributed relational database TiDB. The relationship type extraction rules include the extraction source table, relationship extraction logic, and extraction fields. Each type of relationship extraction forms a <enterprise, relationship, enterprise> ternary data structure, and each enterprise is determined by the enterprise unique identifier, and the relationship is determined by the relationship name.

[0020] S1-3. Construction of enterprise association relationship graph;

[0021] Using the graph database technology NebulaGraph to construct the enterprise association relationship graph. The association relationship graph consists of two types: nodes and edges. Nodes represent each enterprise, and edges represent the relationships between two enterprise nodes. The arrow of the edge indicates the direction of the relationship. Among them, direct relationships are all one-way relationships, and suspected relationships are all two-way relationships.

[0022] Furthermore, in step S2, it includes enterprise risk categories and enterprise risk events under each category;

[0023] The enterprise risk categories include severe warnings, warnings, and attention. There are a total of 23 risk events for severe warnings, 16 risk events for warnings, and 28 risk events for attention;

[0024] Enterprise risk events under various categories consist of risk objects and risk judgment rules. The risk object is the name of the risk event, and the risk judgment rules include the indicators for risk extraction.

[0025] Furthermore, when constructing the associated risk system, based on the multi-source data of the enterprise itself, a three-layer enterprise associated risk system is established. The first layer is the risk category of the enterprise, the second layer is the source dimension of each risk event of the enterprise, and the third layer is the specific risk event.

[0026] Furthermore, in step S3, the AHP (Analytic Hierarchy Process) is used to establish an enterprise association index model, which includes two parts: using the AHP method to determine the weights of enterprise association relationships and calculating the enterprise association index based on the enterprise association relationship map.

[0027] (1) Determine the weights of enterprise association relationships;

[0028] The AHP is used to determine the weights of enterprise association relationships. The steps include constructing the hierarchical structure of the AHP, constructing the judgment matrices of each layer, single sorting and consistency test of each matrix layer, total sorting and consistency test of the layer, and calculating the weights of each indicator.

[0029] According to the importance degree of enterprise association relationships, a two-layer hierarchical structure model is established with the direct relationships and suspected relationships of the enterprise as the sources. Judgment matrices are constructed respectively for the relationship types in the first layer and the relationships in the second layer under each relationship type. Since the first layer contains 3 types of relationships, a 3*3 first-layer judgment matrix is constructed. Since the second-layer suspected relationships contain 9 types of relationships, a 9*9 second-layer judgment matrix is constructed. During the construction of the judgment matrix, pairwise comparisons are made between the relationship indicators, and different values between 1-9 are assigned according to the importance degree of the two indicators.

[0030] For each of the constructed first-layer and second-layer judgment matrices, a consistency test is carried out to find out whether there are logical biases in the assigned values among the indicators. After multiple rounds of adjustment, the judgment matrices that pass the consistency test are finally obtained. During the consistency test of the judgment matrix, the arithmetic mean method is used to calculate the normalized weights, and the weights of each relationship in the first layer and the relationships under each category in the second layer are obtained respectively.

[0031] The weights of each finally obtained relationship indicator are equal to the secondary weights of each type of relationship multiplied by the primary weights of the first-layer relationships to which they belong, and the sum of the indicator weights of all relationship types is 1.

[0032] (2) Calculation of enterprise association index: Calculate the association index between each enterprise node with the enterprise association relationship map as the carrier. The shareholding penetration relationship between enterprises is a one-way relationship. If there is an external investment relationship between enterprise A and enterprise B, that is, enterprise A is a shareholder of enterprise B, then there is an edge pointing from the enterprise A node to the enterprise B node in the relationship map, and the weight of the edge is the weight of the external investment relationship WeightTZ(A, B). There is an edge pointing from the enterprise B node to the enterprise C node, and the weight of the edge is the weight of the external investment relationship WeightTZ(B, C);

[0033] The relationship of having the same actual controller and the suspected relationship between enterprises are both two-way relationships. If enterprise A and enterprise B have the same actual controller relationship, then there is a two-way relationship between enterprise A and enterprise B. There is a two-way pointing edge between the enterprise A node and the enterprise B node in the relationship map, and the weight of the edge is the weight of the same actual controller relationship WeightSK(A, B). If enterprise B and enterprise C have the same telephone number relationship, then there is a two-way relationship between enterprise B and enterprise C. There is a two-way pointing edge between the enterprise B node and the enterprise C node in the relationship map, and the weight of the edge is the weight of the same telephone number relationship WeightTel(B, C);

[0034] The association index is calculated by weighted summation among various relationships at the same level. For the association index between different levels, first calculate the association index between two enterprises at the upper level, and then multiply it by the result of the weighted summation of all relationships at the current level to obtain the association index between two enterprises across levels.

[0035] The calculation of the association index between enterprise A and enterprise B is as follows:

[0036] RelationZS(A,B) = WeightTZ(A, B) + WeightSK(A, B)

[0037] The calculation of the association index between enterprise A and enterprise C is as follows:

[0038] RelationZS(A,C) = RelationZS(A,B) * (WeightTZ(B, C) +

[0039] WeightTel(B,C))

[0040] = (WeightTZ(A, B) + WeightSK(A, B)) * (WeightTZ(B, C) + WeightTel(B, C)).

[0041] Furthermore, in step S4, an enterprise association conduction risk evaluation model is established using the AHP method, which includes two parts: determining the enterprise association risk weights using the AHP (Analytic Hierarchy Process) and calculating the enterprise association conduction risk with the relationship map as the carrier;

[0042] (1) Determine the enterprise association risk weight: The AHP (Analytic Hierarchy Process) is used to determine the enterprise association risk weight. The steps include constructing the hierarchical structure of the AHP, constructing the judgment matrix for each level, single sorting and consistency test for each matrix level, total sorting and consistency test for the hierarchy, and calculating the weight of each index;

[0043] Enterprise risks include severe warning, warning, and attention. The first level is the risk type of the enterprise, and the second level is the specific risk events under each type of risk. Judgment matrices are constructed respectively for the risk types in the first layer and the risk events in the second layer under each risk type. During the construction of the judgment matrix, pairwise comparisons are made between risk indicators, and different values between 1 - 9 are assigned according to the importance of the two indicators;

[0044] For each of the constructed judgment matrices in the first and second layers, a consistency test is carried out to find out whether there are logical biases in the assigned values between indicators. After multiple rounds of adjustment, the judgment matrices that pass the consistency test are finally obtained; During the consistency test of the judgment matrix, the arithmetic average method is used to calculate the normalized weight, and the weights of each risk type in the first layer and the relationships between risks under each category in the second layer are obtained respectively;

[0045] The weight of each finally obtained risk indicator is equal to the secondary weight of each type of risk multiplied by the primary weight of the risk type in the first layer to which it belongs, and the sum of the indicator weights of all risk types is 1;

[0046] (2) Calculate the enterprise association conduction risk: Calculate the association conduction risk of each enterprise node with the relationship graph as the carrier. The NebulaGraph graph database is used to calculate the association conduction risk of each enterprise node in the relationship graph. The risks of each enterprise node in the relationship graph include two parts: its own risk and conduction risk. The enterprise's own risk is calculated by weighted summation, which is obtained by multiplying the value of each risk by the weight of each risk and then adding them together; When calculating the association conduction risk of the current enterprise node, this node is the main target enterprise, and the enterprises with an association relationship with this enterprise are the object enterprises of this target enterprise. The own risk of each associated object enterprise is calculated by multiplying the value of each type of risk by the weight of each type of risk. The conduction risk of each associated object enterprise relative to the main enterprise is calculated by multiplying the association index between each object enterprise and the main enterprise by the own risk of each object enterprise. Node A of enterprise is the main enterprise, and this enterprise has three types of risks, Risk1, Risk2, Risk3. The weight values of each type of risk have been obtained by the AHP method as Weight1, Weight2, Weight3. The calculation of the own risk SelfRiskA of node A of enterprise is as follows:

[0047] SelfRiskA = Risk1 * Weight1 + Risk2 * Weight2 + Risk3 * Weight3;

[0048] Enterprise A is associated with Enterprise B, Enterprise C, and Enterprise D in the relationship graph respectively. The self-risks of each object enterprise are SelfRiskB, SelfRiskC, and SelfRiskD respectively. The association indices between Enterprise A and Enterprise B, Enterprise C, and Enterprise D are RelationZS(A, B), RelationZS(A, C), and RelationZS(A, D) respectively. Then the conduction risk obtained by Enterprise A from other associated object enterprises is calculated as follows:

[0049] BroadcastRiskA = SelfRiskB * RelationZS(A, B) + SelfRiskC * RelationZS(A, C) + SelfRiskD * RelationZS(A, D)

[0050] The associated conduction risk RealationBroadRiskA of Enterprise A node is calculated as follows:

[0051] RealationBroadRiskA = SelfRiskA + BroadcastRiskA.

[0052] Furthermore, in step S5, it includes three modules: enterprise associated conduction risk calculation, risk monitoring data transmission, and risk display. A real-time enterprise risk monitoring system is built using the big data real-time framework, and the two real-time frameworks of kafka and Spark are used to realize the real-time transmission and real-time analysis of the enterprise associated conduction risk in the monitoring system;

[0053] (1) Enterprise associated conduction risk calculation: The enterprise association graph database is used as the source for extracting enterprise association relationships and the calculation carrier for enterprise association indices and associated conduction risks. Combining graph algorithms, calculate the self-risks and conduction risks of each enterprise node in the association relationship graph; The enterprise association data in the enterprise association graph database will be updated regularly;

[0054] (2) Risk monitoring data transmission: Monitor the enterprise risk situation in real time through monitoring rules. The monitoring rules include enterprise monitoring indicators, indicator thresholds, and indicator logics. The enterprise monitoring indicators monitor two parts: the self-risk of the enterprise and the conduction risk of the enterprise. Set the monitoring thresholds and indicator calculation logics for different monitoring indicators respectively. When the enterprise monitoring indicator value exceeds the indicator threshold, a risk warning will be sent to the risk display platform;

[0055] (3) Risk display platform: The risk display platform provides the setting of enterprises to be monitored and monitoring rules. For the enterprises added to the watch list, the associated transmission risks of this batch of enterprises will be calculated through regular batch processing. The regular batch processing is implemented through the Spark big data framework. When the monitoring index value of an enterprise exceeds the set threshold, the risk prompt will be transmitted to the risk display platform in real time. The real-time transmission of data is achieved through real-time frameworks such as Spark Streaming and Kafka. The risk display platform displays in real time the monitoring index values of enterprises, the enterprise association relationships, and the risk situations of enterprises and their associated enterprises.

[0056] An enterprise associated transmission risk monitoring device, comprising: at least one memory and at least one processor;

[0057] The at least one memory is used to store machine-readable programs;

[0058] The at least one processor is used to call the machine-readable program to execute an enterprise associated transmission risk monitoring method.

[0059] Compared with the prior art, an enterprise associated transmission risk monitoring method and device of the present invention have the following outstanding beneficial effects:

[0060] 1. Compared with the traditional method that only monitors the risks of enterprises themselves, this application quantifies enterprise risks into two parts: the risks of the enterprises themselves and the transmission risks of other associated enterprises relative to the enterprise. While monitoring the risks of the enterprises themselves, it monitors the risks of associated enterprises related to the enterprise and the transmission risks relative to the target enterprise, and can timely discover various risks in the upstream link of risk spread, greatly expanding the breadth and depth of enterprise risk monitoring, and expanding the scope and comprehensiveness of enterprise risk monitoring.

[0061] 2. Compared with the traditional enterprise association relationship analysis method that does not rely on graph databases, the calculation of the association index between enterprise nodes is realized based on graph database technology. Applying graph database technology greatly improves the timeliness of enterprise association relationship calculation, realizes the specific quantification of enterprise association relationships in the graph, and further calculates the enterprise associated transmission risks based on the enterprise association index, expanding the application scenarios of graph database technology and AHP method in the field of financial credit risk control, and deepening the specific application of fintech in the financial field.

[0062] 3. Users can log in to the platform to automatically configure the enterprises to be monitored and the rules for enterprise risk monitoring, enabling users participating in the formulation of enterprise risk monitoring rules. This can better ensure that the risk monitoring rule set meets the requirements of users for measuring enterprise risks, improve the accuracy of risk monitoring rules, and enhance the effectiveness of enterprise risk monitoring.

[0063] 4. This application applies the big data real-time framework to the construction process of the enterprise risk monitoring system, realizing the real-time transmission and automatic parsing of enterprise risk data and enterprise associated data, greatly improving the flexibility of the enterprise risk monitoring method and the timeliness of enterprise risk monitoring.

[0064] In addition, the enterprise associated conduction risk monitoring system proposed in this application can also be flexibly expanded. By introducing big data real-time frameworks such as SparkStreaming, real-time risk monitoring can be achieved, improving the flexibility and timeliness of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0066] Attached Figure 1 is a schematic flowchart of an enterprise associated conduction risk monitoring method;

[0067] Attached Figure 2 is a schematic diagram of enterprise associated index calculation in an enterprise associated conduction risk monitoring method. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0068] In order to enable those skilled in the art to better understand the solution of the present invention, the following will further elaborate on the present invention in combination with specific embodiments. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0069] The following gives a best embodiment:

[0070] As Figure 1 shown, the enterprise associated conduction risk monitoring method in this embodiment has the following steps:

[0071] S1. Construction of the enterprise associated relationship graph;

[0072] Including:

[0073] S1-1. Types of corporate affiliation: mainly include direct corporate relationships and suspected relationships. Direct corporate relationships include branches, overseas investments, corporate shareholders, natural person shareholders, directors, senior managers and supervisors, and historical directors, senior managers and supervisors. Suspected corporate relationships include relationships with the same actual controller, same telephone number, same email address, same mailing address, same registered address, same domain name information, same judicial documents, same corporate name, same patent information, and same software copyright.

[0074] S1-2. Enterprise association relationship extraction: Taking the multi-source data of the company as the extraction source, determine the extraction rules of each relationship type. The multi-source data of the company is mainly stored in the distributed relational database TiDB. The relationship type extraction rules mainly include the extraction source table, relationship extraction logic, and extraction fields. The extraction of various relationships forms a <company, relationship, enterprise> ternary data structure. Each company is identified by the company's unique identifier, and the relationship is identified by the relationship name;

[0075] S1-3. Construction of enterprise association relationship map: The graph database technology NebulaGraph is used to build the enterprise association relationship map. The association relationship map is mainly composed of two types: nodes and edges. Nodes represent each enterprise, and edges represent the relationship between two enterprise nodes. The arrows on the edges indicate the direction of the relationship. Direct relationships such as branches, foreign investments, and corporate shareholders are all unidirectional relationships, and suspected relationships such as the same phone number and the same email address are all bidirectional relationships.

[0076] S2. Construction of enterprise-related risk system;

[0077] Including enterprise risk categories and enterprise risk events under each category:

[0078] (1) Enterprise risks include three types: severe warning, warning, and attention. Severe warning includes case filing information (plaintiff / appellant / applicant), case filing information (defendant / respondent / executor), etc., totaling 23 risk events; warning includes environmental protection penalties, tax arrears information, liquidation information, etc., totaling 16 risk events; attention mainly includes court announcements, equity freezes, administrative penalties, etc., totaling 28 risk events;

[0079] (2) Each risk event is mainly composed of a risk object and a risk determination rule. The risk object is mainly the name of the risk event, such as the case filing information plaintiff / appellant / applicant. The risk determination rule includes risk extraction indicators, such as the case identity of the enterprise, risk determination logic, such as the case identity of the enterprise is the plaintiff, appellant or applicant, and risk determination threshold, such as once a case is filed with the case identity of the enterprise as the plaintiff, appellant or applicant, it is determined that this type of risk has occurred;

[0080] (3) Construction of associated risk system: Based on the multi-source data of the enterprise itself, including different types such as industry and commerce, justice, and operation, a three-layer enterprise associated risk system is established. The first layer is the risk category of the enterprise, the second layer is the source dimension of each risk event of the enterprise, such as judgment documents, and the third layer is the specific risk event, including information such as the name of the risk event, the extracted indicators of the risk event, the indicator values, and the specific time when the risk event occurred.

[0081] S3. Construction of enterprise association index model;

[0082] The AHP (Analytic Hierarchy Process) method is used to establish an enterprise association index model, which includes two parts: determining the weights of enterprise association relationships using the AHP method and calculating the enterprise association index based on the enterprise association relationship graph.

[0083] (1) Determining the weights of enterprise association relationships: This patent uses the AHP method to determine the weights of enterprise association relationships. The main steps include constructing the hierarchical structure of the AHP method, constructing the judgment matrices of each layer, single-layer sorting and consistency test of each matrix layer, total-layer sorting and consistency test of the layer, and calculating the weights of each indicator. According to the importance degree of enterprise association relationships, a two-layer hierarchical structure model is established with the direct relationships and suspected relationships of the enterprise as the sources. Judgment matrices are constructed respectively for the relationship types in the first layer and the relationships in the second layer under each relationship type. If the first layer contains 3 types of relationships, a 3*3 judgment matrix for the first layer is constructed. If the suspected relationship in the second layer contains 9 types of relationships, a 9*9 judgment matrix for the second layer is constructed. During the construction of the judgment matrix, pairwise comparisons are made between the relationship indicators to judge, and different values between 1-9 are assigned according to the importance degree of the two indicators; for each constructed judgment matrix of the first layer and the second layer, a consistency test is carried out to find out whether there are logical biases in the assigned values between the indicators. After multiple rounds of adjustment, the judgment matrices that pass the consistency test are finally obtained; in the process of consistency test of the judgment matrix in this patent, the arithmetic average method is used to calculate the normalized weights, and the weights of each relationship in the first layer and the relationships under each category in the second layer are obtained respectively; the finally obtained weights of each relationship indicator are equal to the secondary weights of each type of relationship multiplied by the primary weights of the first-layer relationships to which they belong, and the sum of the indicator weights of all relationship types is 1;

[0084] (2) Calculation of enterprise association index: The association index between each enterprise node is calculated based on the enterprise association relationship graph. The shareholding penetration relationship between enterprises is a one-way relationship. If there is an external investment relationship between enterprise A and enterprise B, that is, enterprise A is a shareholder of enterprise B, then there is an edge pointing from the node of enterprise A to the node of enterprise B in the relationship graph, and the weight of the edge is the weight of the external investment relationship WeightTZ(A, B). The node of enterprise B has an edge pointing to the node of enterprise C, and the weight of the edge is the weight of the external investment relationship WeightTZ(B, C). The relationships of having the same actual controller and suspected relationship between enterprises are both two-way relationships. If enterprise A and enterprise B have the same actual controller, then there is a two-way relationship between enterprise A and enterprise B. There is a two-way pointing edge between the node of enterprise A and the node of enterprise B in the relationship graph, and the weight of the edge is the weight of the relationship of having the same actual controller WeightSK(A, B). If enterprise B and enterprise C have the same telephone number relationship, then there is a two-way relationship between enterprise B and enterprise C. There is a two-way pointing edge between the node of enterprise B and the node of enterprise C in the relationship graph, and the weight of the edge is the weight of the same telephone number relationship WeightTel(B, C). The association index is calculated by weighted summation among various relationships at the same level. For cross-level relationships between two enterprises, the association index between the two enterprises at the upper level is first calculated and then multiplied by the result of the weighted summation of all relationships at the current level to obtain the association index between the two cross-level enterprises.

[0085] As Figure 2 shown, the calculation of the association index between enterprise A and enterprise B is as follows:

[0086] RelationZS(A,B) = WeightTZ(A, B) + WeightSK(A, B)

[0087] The calculation of the association index between enterprise A and enterprise C is as follows:

[0088] RelationZS(A,C) = RelationZS(A,B) * (WeightTZ(B, C) + WeightTel(B,C))

[0089] = (WeightTZ(A, B) + WeightSK(A, B)) * (WeightTZ(B, C) + WeightTel(B,C)).

[0090] S4. Construction of enterprise association conduction risk evaluation model;

[0091] The AHP method is used to establish an enterprise association conduction risk evaluation model, which mainly includes two parts: determining the enterprise association risk weight by using the AHP analytic hierarchy process and calculating the enterprise association conduction risk based on the relationship graph.

[0092] (1) Determine the enterprise association risk weight: This patent uses the AHP (Analytic Hierarchy Process) to determine the enterprise association risk weight. The main steps include constructing the hierarchical structure of the AHP, constructing the judgment matrices at each level, single ranking and consistency test of each matrix level, overall ranking and consistency test of the hierarchy, and calculating the weights of each index. Enterprise risks mainly include severe warning, warning, and attention. The first level is the risk type of the enterprise, and the second level is the specific risk events under each type of risk. Judgment matrices are constructed respectively for the risk types at the first level and the risk events at the second level under each risk type. During the construction of the judgment matrix, pairwise comparison and judgment are made between risk indicators, and different values between 1 and 9 are assigned according to the importance of the two indicators; for each of the constructed judgment matrices at the first and second levels, a consistency test is carried out to find out whether there are logical biases in the assigned values between the indicators. After multiple rounds of adjustment, the judgment matrices that pass the consistency test are finally obtained; in the process of the consistency test of the judgment matrix in this patent, the arithmetic average method is used to calculate the normalized weights, and the weights of each risk type at the first level and the relationships of each risk under each category at the second level are obtained respectively; the weights of each final risk index obtained are equal to the secondary weights of each type of risk multiplied by the primary weights of the risk types at the first level to which they belong, and the sum of the index weights of all risk types is 1;

[0093] (2) Calculate the enterprise association conduction risk: Calculate the association conduction risk of each enterprise node with the relationship graph as the carrier. This patent uses the NebulaGraph database to calculate the association conduction risk of each enterprise node in the relationship graph. The risks of each enterprise node in the relationship graph mainly include its own risk and conduction risk. The enterprise's own risk is calculated by weighted summation, which is obtained by multiplying the value of each risk by the weight of each risk and then adding them up; when calculating the association conduction risk of the current enterprise node, this node is the main target enterprise, and each enterprise with an association relationship with this enterprise is the object enterprise of this target enterprise. The own risk of each associated object enterprise is calculated by multiplying the value of each type of risk by the weight of each type of risk. The conduction risk of each associated object enterprise relative to the main enterprise is calculated by multiplying the association index between each object enterprise and the main enterprise by the own risk of each object enterprise. Node A of enterprise is the main enterprise, and this enterprise has three types of risks, Risk1, Risk2, and Risk3. The weight values of each type of risk have been obtained by the AHP method as Weight1, Weight2, and Weight3. The calculation of the own risk SelfRiskA of node A of enterprise is as follows:

[0094] SelfRiskA = Risk1 * Weight1 + Risk2 * Weight2 + Risk3 * Weight3

[0095] Enterprise A is associated with Enterprise B, Enterprise C, and Enterprise D in the relationship graph respectively. The self-risks of each target enterprise are SelfRiskB, SelfRiskC, and SelfRiskD respectively. The association indices between Enterprise A and Enterprise B, Enterprise C, and Enterprise D are RelationZS(A,B), RelationZS(A,C), and RelationZS(A,D) respectively. Then the conduction risk of Enterprise A obtained from other associated target enterprises is calculated as follows:

[0096] BroadcastRiskA = SelfRiskB * RelationZS(A,B) + SelfRiskC * RelationZS(A,C) + SelfRiskD * RelationZS(A,D)

[0097] The association conduction risk RealationBroadRiskA of Enterprise A node is calculated as follows:

[0098] RealationBroadRiskA = SelfRiskA + BroadcastRiskA.

[0099] S5. Construction of enterprise association conduction risk monitoring system;

[0100] It includes three modules: enterprise association conduction risk calculation, risk monitoring data transmission, and risk display. A real-time enterprise risk monitoring system is constructed using a big data real-time framework, and the real-time transmission and real-time analysis of enterprise association conduction risk in the monitoring system are realized using two real-time frameworks, kafka and Spark.

[0101] (1) Enterprise association conduction risk calculation: Using the enterprise association graph database as the source for extracting enterprise association relationships and the calculation carrier for enterprise association indices and association conduction risks, combined with graph algorithms to calculate the self-risks and conduction risks of each enterprise node in the association relationship graph; the enterprise association data in the enterprise association graph database will be updated regularly;

[0102] (2) Risk monitoring data transmission: Real-time monitoring of enterprise risk situations through monitoring rules. The monitoring rules include enterprise monitoring indicators, indicator thresholds, and indicator logics. The enterprise monitoring indicators mainly monitor two parts: the self-risk of the enterprise and the conduction risk of the enterprise. Different monitoring thresholds and indicator calculation logics are set respectively. When the value of the enterprise monitoring indicator exceeds the indicator threshold, a risk warning will be sent to the risk display platform;

[0103] (3) Risk Display Platform: The risk display platform can provide the setting of enterprises to be monitored and monitoring rules. For the enterprises added to the monitoring list, the associated transmission risks of this batch of enterprises will be calculated through regular batch processing. In this patent, the regular batch processing is implemented through the Spark big data framework. When the enterprise monitoring index value exceeds the set threshold, the risk prompt will be transmitted to the risk display platform in real time. In this patent, the real-time transmission of data is achieved through real-time frameworks such as Spark Streaming and Kafka. The risk display platform displays the enterprise monitoring index value, enterprise association relationship, and the risk situation of the enterprise and its associated enterprises in real time.

[0104] Based on the above method, an enterprise associated transmission risk monitoring device in this embodiment includes: at least one memory and at least one processor;

[0105] The at least one memory is used to store machine-readable programs;

[0106] The at least one processor is used to call the machine-readable program to execute an enterprise associated transmission risk monitoring method.

[0107] The above specific implementation manners are only specific cases of the present invention. The patent protection scope of the present invention includes but is not limited to the above specific implementation manners. Any technical solution that conforms to the above specific implementation manners recorded in the present invention and any appropriate changes or substitutions made by those of ordinary skill in the art shall fall within the patent protection scope of the present invention.

[0108] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring the risk of enterprise association conduction, characterized in that, The steps are as follows: S1. Construction of enterprise association relationship graph; S2. Construction of enterprise association risk system; S3. Construction of enterprise association index model; S4. Construction of enterprise association conduction risk evaluation model; S5. Construction of enterprise association conduction risk monitoring system.

2. The enterprise associated conduction risk monitoring method according to claim 1, characterized in that In step S1, it includes: S1-1. Types of enterprise association relationships; It includes direct enterprise relationships and suspected relationships. Direct enterprise relationships include branches, foreign investments, enterprise shareholders, natural person shareholders, directors, supervisors, senior managers, and legal representatives, historical directors, supervisors, senior managers, and legal representatives. Enterprise suspected relationships include the same actual controller relationship, the same phone number, the same email, the same communication address, the same registered address, the same domain name information, the same judgment documents, the same enterprise name, the same patent information, and the same software copyright; S1-2. Extraction of enterprise association relationships; Using the multi-source data of its own enterprises as the extraction source, determining the extraction rules for each relationship type. The multi-source data of enterprises is stored in the distributed relational database TiDB. The relationship type extraction rules include the extraction source table, relationship extraction logic, and extraction fields. Each type of relationship extraction forms a <enterprise, relationship, enterprise> ternary data structure, and each enterprise is determined by the enterprise unique identifier, and the relationship is determined by the relationship name; S1-3. Construction of enterprise association relationship graph; Using the graph database technology NebulaGraph to construct the enterprise association relationship graph. The association relationship graph consists of two types: nodes and edges. Nodes represent each enterprise, and edges represent the relationships between two enterprise nodes. The arrow of the edge indicates the direction of the relationship. Among them, direct relationships are all one-way relationships, and suspected relationships are all two-way relationships.

3. The enterprise-related conduction risk monitoring method according to claim 2, wherein In step S2, it includes enterprise risk categories and enterprise risk events under each category; The enterprise risk categories include serious warnings, warnings, and attention. There are a total of 23 risk events for serious warnings, 16 risk events for warnings, and 28 risk events for attention; Enterprise risk events under each category consist of a risk object and a risk judgment rule. The risk object is the name of the risk event, and the risk judgment rule includes the indicators for risk extraction.

4. The enterprise association conduction risk monitoring method according to claim 3, characterized in that, When constructing the association risk system, based on the multi-source data of its own enterprises, a three-layer enterprise association risk system is established. The first layer is the risk category of the enterprise, the second layer is the source dimension of each enterprise risk event, and the third layer is the specific risk event.

5. The enterprise-related conduction risk monitoring method according to claim 4, wherein In step S3, the AHP (Analytic Hierarchy Process) method is used to establish an enterprise association index model, which includes two parts: using the AHP method to determine the weights of enterprise association relationships and calculating the enterprise association index with the enterprise association relationship graph as the carrier; (1) Determine the weights of enterprise association relationships; Using the AHP method to determine the weights of enterprise association relationships. The steps include constructing the hierarchical structure of the AHP method, constructing the judgment matrix of each layer, single sorting and consistency test of each matrix layer, total sorting and consistency test of the layer, and calculating the weights of each index; According to the importance of enterprise association relationships, a two-layer hierarchical structure model is established with the direct relationships and suspected relationships of enterprises as the sources. Judgment matrices are constructed respectively for the relationship types in the first layer and the relationships in the second layer under each relationship type. Since the first layer contains 3 types of relationships, a 3*3 judgment matrix for the first layer is constructed. Since the suspected relationships in the second layer contain 9 types of relationships, a 9*9 judgment matrix for the second layer is constructed. During the construction of the judgment matrix, pairwise comparisons are made between the relationship indicators, and different values between 1 and 9 are assigned according to the importance of the two indicators; For each of the constructed judgment matrices in the first and second layers, a consistency test is carried out to find out whether there are logical biases in the assigned values among the indicators. After multiple rounds of adjustment, the judgment matrices that pass the consistency test are finally obtained; During the consistency test of the judgment matrix, the arithmetic average method is used to calculate the normalized weights, and the weights of each relationship in the first layer and the relationships under each category in the second layer are obtained respectively; The weights of the final obtained relationship indicators are equal to the secondary weights of each type of relationship multiplied by the primary weights of the first-layer relationships to which they belong, and the sum of the indicator weights of all relationship types is 1; (2) Calculation of enterprise association index: Using the enterprise association relationship graph as a carrier, calculate the association index between each enterprise node. The shareholding penetration relationship between enterprises is a one-way relationship. If there is an external investment relationship between enterprise A and enterprise B, that is, enterprise A is a shareholder of enterprise B, then there is an edge pointing from the enterprise A node to the enterprise B node in the relationship graph, and the weight of the edge is the weight of the external investment relationship WeightTZ(A, B). There is an edge pointing from the enterprise B node to the enterprise C node, and the weight of the edge is the weight of the external investment relationship WeightTZ(B, C); The relationship of having the same actual controller and suspected relationships between enterprises are both two-way relationships. If enterprise A and enterprise B have the same actual controller relationship, then there is a two-way relationship between enterprise A and enterprise B. There is a two-way pointing edge between the enterprise A node and the enterprise B node in the relationship graph, and the weight of the edge is the weight of the same actual controller relationship WeightSK(A, B). If enterprise B and enterprise C have the same telephone number relationship, then there is a two-way relationship between enterprise B and enterprise C. There is a two-way pointing edge between the enterprise B node and the enterprise C node in the relationship graph, and the weight of the edge is the weight of the same telephone number relationship WeightTel(B, C); The association index is calculated by weighted summation among various relationships at the same level. For the association index between different levels, first calculate the association index between two enterprises at the upper level, and then multiply it by the result of the weighted summation of all relationships at the current level to obtain the association index between two enterprises across levels. The calculation of the association index between enterprise A and enterprise B is as follows: RelationZS(A,B) = WeightTZ(A, B) + WeightSK(A, B) The calculation of the association index between enterprise A and enterprise C is as follows: RelationZS(A,C) = RelationZS(A,B) * (WeightTZ(B, C) + WeightTel(B, C)) =(WeightTZ(A, B) + WeightSK(A, B)) * (WeightTZ(B, C) + WeightTel(B, C)).

6. The enterprise-related conduction risk monitoring method according to claim 5, wherein In step S4, the AHP method is used to establish an enterprise association conduction risk evaluation model, which includes two parts: determining the enterprise association risk weights by the AHP (Analytic Hierarchy Process) and calculating the enterprise association conduction risk with the relationship graph as the carrier; (1) Determine the enterprise association risk weights: The AHP is used to determine the enterprise association risk weights. The steps include constructing the hierarchical structure of the AHP, constructing the judgment matrices for each layer, single sorting and consistency test for each matrix layer, total sorting and consistency test for the hierarchy, and calculating the weights of each index; Enterprise risks include three categories: serious warning, warning, and attention. The first level is the risk type of the enterprise, and the second level is the specific risk events under each type of risk. Judgment matrices are constructed respectively for the risk types at the first level and the risk events at the second level under each risk type. During the construction of the judgment matrix, pairwise comparisons are made between risk indicators to judge, and different values between 1 and 9 are assigned according to the importance of the two indicators; For each of the judgment matrices constructed for the first and second layers, a consistency test is carried out to find out whether there are logical biases in the assigned values among the indicators. After multiple rounds of adjustment, the judgment matrices that pass the consistency test are finally obtained; During the consistency test of the judgment matrix, the arithmetic mean method is used to calculate the normalized weights, and the weights of each risk type at the first level and the relationships of each risk under each category at the second level are obtained respectively; The weights of each final risk indicator are equal to the secondary weights of each type of risk multiplied by the primary weights of the risk types at the first level to which they belong, and the sum of the indicator weights of all risk types is 1; (2) Calculate the enterprise association conduction risk: Calculate the association conduction risk of each enterprise node with the relationship graph as the carrier. The NebulaGraph database is used to calculate the association conduction risk of each enterprise node in the relationship graph. The risks of each enterprise node in the relationship graph include two parts: its own risk and the conduction risk. The enterprise's own risk is calculated by the weighted sum method, which is obtained by multiplying the value of each risk by the weight of each risk and then adding them up; When calculating the association conduction risk of the current enterprise node, this node is the main target enterprise, and each enterprise with an association relationship with this enterprise is the object enterprise of this target enterprise. The own risk of each associated object enterprise is calculated by multiplying the value of each type of risk by the weight of each type of risk. The conduction risk of each associated object enterprise relative to the main enterprise is calculated by multiplying the association index between each object enterprise and the main enterprise by the own risk of each object enterprise. Node A of the enterprise is the main enterprise, and this enterprise has three types of risks, Risk1, Risk2, and Risk3. The weight values of each type of risk have been obtained by the AHP method as Weight1, Weight2, and Weight3. The calculation of the own risk SelfRiskA of node A of the enterprise is as follows: SelfRiskA = Risk1 * Weight1 + Risk2 * Weight2 + Risk3 * Weight3; Enterprise A is associated with Enterprise B, Enterprise C, and Enterprise D in the relationship graph respectively. The self-risks of each object enterprise are SelfRiskB, SelfRiskC, and SelfRiskD respectively. The association indices between Enterprise A and Enterprise B, Enterprise C, and Enterprise D are RelationZS(A,B), RelationZS(A,C), and RelationZS(A,D) respectively. Then the conduction risk obtained by Enterprise A through the conduction of other associated object enterprises is calculated as follows: BroadcastRiskA = SelfRiskB * RelationZS(A,B) + SelfRiskC * RelationZS(A,C) + SelfRiskD * RelationZS(A,D) The associated conduction risk RealationBroadRiskA of the Enterprise A node is calculated as follows: RealationBroadRiskA = SelfRiskA + BroadcastRiskA.

7. The enterprise associated conduction risk monitoring method according to claim 6, wherein In step S5, it includes three modules: enterprise associated conduction risk calculation, risk monitoring data transmission, and risk display. A real-time enterprise risk monitoring system is built using a big data real-time framework. The real-time transmission and real-time analysis of the enterprise associated conduction risk in the monitoring system are realized using two real-time frameworks, kafka and Spark; (1) Enterprise associated conduction risk calculation: The enterprise association graph database is used as the source for extracting enterprise association relationships and the calculation carrier for enterprise association indices and associated conduction risks. Combining graph algorithms to calculate the self-risks and conduction risks of each enterprise node in the association relationship graph; The enterprise association data in the enterprise association graph database will be updated regularly; (2) Risk monitoring data transmission: The enterprise risk situation is monitored in real time through monitoring rules. The monitoring rules include enterprise monitoring indicators, indicator thresholds, and indicator logics. The enterprise monitoring indicators monitor two parts: the self-risk of the enterprise and the conduction risk of the enterprise. The monitoring thresholds and indicator calculation logics of different monitoring indicators are set respectively. When the value of the enterprise monitoring indicator exceeds the indicator threshold, a risk prompt will be sent to the risk display platform; (3) Risk display platform: The risk display platform provides the setting of enterprises to be concerned about and monitoring rules. For the enterprises added to the concern list, the associated conduction risks of this batch of enterprises will be calculated in the form of regular batch processing. The regular batch processing is realized through the Spark big data framework. When the value of the enterprise monitoring indicator exceeds the set threshold, the risk prompt will be transmitted to the risk display platform in real time. The real-time transmission of data is realized through real-time frameworks such as SparkStreaming and Kafka. The risk display platform displays the values of enterprise monitoring indicators, enterprise association relationships, and the risk situations of enterprises and associated enterprises in real time.

8. An enterprise-related conduction risk monitoring device, characterized in that, Including: At least one memory and at least one processor; The at least one memory is used for storing machine-readable programs; The at least one processor is used for calling the machine-readable program to execute the method according to any one of claims 1 to 7.

Images