Enterprise financing risk prediction method and system based on knowledge graph

Abstract

The invention provides an enterprise financing risk prediction method and system based on a knowledge graph, and relates to the technical field of financial risk assessment, and the method comprises the steps: carrying out the continuous risk monitoring of a target enterprise according to a preset index set in a preset monitoring window, and generating K risk index feature value sets; performing risk fluctuation identification, and generating K risk coefficients and K target risk index feature values; performing derivative expansion on the K risk indexes according to the K risk coefficients to generate K derivative risk index sets; acquiring derivative index feature values, and generating K derivative risk index feature value sets; constructing a target risk prediction knowledge graph; and calling M sample risk prediction knowledge maps in the knowledge map database, carrying out risk identification, and generating a target risk result of the target enterprise. The technical problem that the risk cannot be dynamically predicted by integrating various external factors in the prior art is solved, and the accuracy and efficiency of risk prediction are improved.

Description

Technical Field

[0001] The present application relates to the technical field of financial risk assessment, and specifically to a method and system for predicting corporate financing risks based on a knowledge graph. Background Art

[0002] Prediction of corporate financing risk is an important topic in the financial field. Corporate financing activities involve the borrowing and repayment of funds, as well as related interest payments. When making financing decisions, companies need to evaluate their own repayment ability, while financial institutions need to evaluate the company's credit status and debt repayment ability to determine whether to approve the loan and the loan amount and interest rate. Therefore, accurate prediction of corporate financing risks is crucial for both companies and financial institutions.

[0003] Traditional methods for predicting corporate financing risks mainly rely on financial ratio analysis, credit scoring models, etc. These methods can help companies or financial institutions assess risks to a certain extent, but they have certain limitations. First, these methods can usually only be analyzed based on the company's historical financial data, and cannot fully consider external factors such as the company's operating environment, industry conditions, and market changes. Secondly, these methods are inefficient when processing large amounts of data and have difficulty capturing the complex relationships between data. The risk prediction model they use is often a static framework that is difficult to adapt to the dynamic changes in the company's risk status. Summary of the invention

[0004] This application provides a method and system for predicting enterprise financing risks based on knowledge graphs, which solves the technical problems that the existing technology is unable to comprehensively consider the impact of multiple external factors on financing risks, and is difficult to update data in real time and dynamically predict risks, resulting in low accuracy and reliability of enterprise financing risk prediction results, and achieves the technical effect of improving the accuracy and efficiency of risk prediction.

[0005] In view of the above problems, on the one hand, the present application provides a method for predicting corporate financing risks based on a knowledge graph, the method comprising: within a preset monitoring window, continuously monitoring the risks of a target enterprise according to a preset indicator set, generating K risk indicator characteristic value sets, wherein the preset indicator set includes K risk indicators; performing risk fluctuation identification, generating K risk coefficients and K target risk indicator characteristic values, wherein the K risk coefficients are coefficients obtained by performing fluctuation identification on the K risk indicator characteristic value sets; deriving and expanding the K risk indicators according to the sizes of the K risk coefficients, generating K derived risk indicator sets; performing derived indicator characteristic value collection, generating K derived risk indicator characteristic value sets, wherein the The K derived risk indicator characteristic value sets are characteristic value sets obtained after data collection for the target enterprise according to the K derived risk indicator sets within the preset monitoring window; constructing a target risk prediction knowledge graph for the target enterprise based on the K risk indicators, the K derived risk indicator sets, the K target risk indicator characteristic values ​​and the K derived risk indicator characteristic value sets; calling M sample risk prediction knowledge graphs in the knowledge graph database, using the M sample risk prediction knowledge graphs to perform entity alignment on the target risk prediction knowledge graphs, performing risk identification based on the alignment results, and generating a target risk result for the target enterprise, wherein the M sample risk prediction knowledge graphs have M risk result identifiers.

[0006] On the other hand, the present application provides an enterprise financing risk prediction system based on a knowledge graph, and the system includes: a risk monitoring module: the risk monitoring module is used to continuously monitor the risks of the target enterprise according to a preset indicator set within a preset monitoring window, and generate K risk indicator characteristic value sets, wherein the preset indicator set includes K risk indicators; a fluctuation identification module: the fluctuation identification module is used to perform risk fluctuation identification, generate K risk coefficients and K target risk indicator characteristic values, wherein the K risk coefficients are coefficients obtained by performing fluctuation identification on the K risk indicator characteristic value sets; a derivative expansion module: the derivative expansion module is used to derive and expand the K risk indicators according to the sizes of the K risk coefficients, and generate K derived risk indicator sets; a characteristic value acquisition module: the characteristic value acquisition module is used to perform derivative indicator characteristic value acquisition, generate K derivative risk indicators, and generate K derivative risk indicators. Indicator characteristic value set, wherein the K derivative risk indicator characteristic value sets are characteristic value sets obtained after data collection for the target enterprise according to the K derivative risk indicator sets within the preset monitoring window; knowledge graph construction module: the knowledge graph construction module is used to construct a target risk prediction knowledge graph for the target enterprise based on the K risk indicators, the K derivative risk indicator sets, the K target risk indicator characteristic values ​​and the K derivative risk indicator characteristic value sets; risk identification module: the risk identification module is used to call M sample risk prediction knowledge graphs in the knowledge graph database, use the M sample risk prediction knowledge graphs to perform entity alignment on the target risk prediction knowledge graphs, perform risk identification based on the alignment results, and generate the target risk results for the target enterprise, wherein the M sample risk prediction knowledge graphs have M risk result identifiers.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: By continuously monitoring the risks of the target enterprise according to the preset indicator set within the preset monitoring window, K risk indicator characteristic value sets are generated, risk fluctuation identification is performed, and K risk coefficients and K target risk indicator characteristic values ​​are generated; the K risk indicators are derived and expanded according to the sizes of the K risk coefficients to generate K derived risk indicator sets; derivative indicator characteristic value collection is performed to generate K derived risk indicator characteristic value sets, and real-time data of various external factors affecting the financing risks of enterprises are collected, and based on these real-time data, a target risk prediction knowledge graph of the target enterprise is constructed; M sample risk prediction knowledge graphs in the knowledge graph database are called, and the target risk prediction knowledge graph is entity aligned using the M sample risk prediction knowledge graphs, and risk identification is performed according to the alignment results to generate the target risk results of the target enterprise, and a large amount of data is converted into an easy-to-understand graphical representation, so as to better reveal the correlation and regularity between the data.

[0008] As mentioned above, this application is based on the knowledge graph, collects a large amount of real-time data to conduct a comprehensive analysis of various external factors, captures the enterprise risk status more comprehensively and dynamically, adapts to changes in external risks, improves the accuracy and reliability of enterprise financing risk forecasts, and helps financial institutions and enterprises to more effectively identify and manage financing risks, thereby optimizing financing decisions.

[0009] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A schematic diagram of the process of predicting the enterprise financing risk based on the knowledge graph provided in the embodiment of the present application; Figure 2 A schematic diagram of a process for generating K risk coefficients and K target risk indicator characteristic values ​​in a method for predicting corporate financing risks based on a knowledge graph provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of the enterprise financing risk prediction system based on the knowledge graph provided in an embodiment of the present application.

[0011] Explanation of the accompanying drawings: risk monitoring module 10, fluctuation identification module 20, derivative expansion module 30, feature value acquisition module 40, knowledge graph construction module 50, risk identification module 60. DETAILED DESCRIPTION

[0012] The embodiment of the present application provides a method and system for predicting corporate financing risks based on a knowledge graph, continuously monitors the risks of target enterprises according to a preset indicator set, collects multiple risk indicator data, constructs a target risk prediction knowledge graph for the target enterprise, and calls multiple sample risk prediction knowledge graphs in the knowledge graph database for risk identification, thereby generating target risk results for the target enterprise. This achieves dynamic prediction of financing risks by comprehensively considering multiple external factors, thereby improving the accuracy and efficiency of risk prediction.

[0013] Embodiment 1, as Figure 1 As shown, the embodiment of the present application provides a method for predicting corporate financing risks based on a knowledge graph. It should be noted in advance that in the embodiment of the present application, K is an integer greater than 1. The method includes: S1: within a preset monitoring window, continuously monitor the risks of the target enterprise according to a preset indicator set, and generate K risk indicator feature value sets, wherein the preset indicator set includes K risk indicators.

[0014] Specifically, the preset monitoring window is a time frame for continuous monitoring of enterprise risks. The preset monitoring window can be any period of time, such as a month, a quarter or a year, and is pre-customized according to the actual monitoring needs and frequency. The target enterprise refers to the monitored enterprise entity, which may be a company in a specific industry, or an enterprise that needs to assess credit risk for loan, investment or other financial decisions. The preset indicator set is a set of K risk indicators, which is used to assess the financial and operating conditions of the enterprise, including revenue, profit margin, debt repayment ability, debt-to-asset ratio, enterprise scale, structural design, number of benchmark enterprises, etc. The risk indicator characteristic value set is a data set obtained after real-time monitoring and quantification of each risk indicator within the preset monitoring window, and each risk indicator will have one or more corresponding characteristic values. By continuously monitoring the risks of the target enterprise according to the K risk indicators in the preset indicator set, the corresponding K risk indicator characteristic value sets can be obtained. For example, if the monitoring window is a quarter, the risk indicator characteristic value of revenue is the total revenue of the enterprise in this quarter. By generating K risk indicator characteristic value sets, real-time risk data of the target enterprise can be collected to establish a data foundation for subsequent risk analysis and prediction.

[0015] S2: Execute risk fluctuation identification to generate K risk coefficients and K target risk indicator characteristic values, wherein the K risk coefficients are coefficients obtained by performing fluctuation identification on the K risk indicator characteristic value sets.

[0016] Specifically, risk fluctuation identification refers to identifying the changes in risk indicators over time, that is, risk fluctuations, by analyzing the set of risk indicator characteristic values. The risk coefficient is a coefficient generated to quantify risk fluctuations when performing risk fluctuation identification. It is a mathematical expression of the changes in the characteristic values ​​of risk indicators. It can be a standard deviation, coefficient of variation, or other statistics used to indicate the magnitude of the risk. The target risk indicator characteristic value refers to the specific value obtained after risk fluctuation identification for each risk indicator. It reflects the average or specific status of the target enterprise on this indicator during the monitoring period.

[0017] The K risk indicator characteristic value sets generated in step S1 are subjected to risk fluctuation identification to obtain K risk coefficients and K target risk indicator characteristic values, where each risk indicator corresponds to a risk coefficient and a target risk indicator characteristic value. By performing risk fluctuation identification, the dynamic changes of risks can be grasped, and the risk coefficients and target risk indicator characteristic values ​​can be obtained for subsequent financing risk assessment and prediction.

[0018] S3: deriving and expanding the K risk indicators according to the sizes of the K risk coefficients to generate a set of K derived risk indicators.

[0019] Specifically, the original risk indicators are derived and expanded based on the size of the K risk coefficients obtained in step S2. For larger risk coefficients, more derivative indicators need to be expanded to assist risk prediction, and K derived risk indicator sets are obtained. The derived risk indicator set is a new risk indicator set generated by deriving and expanding the original risk indicators. These derived indicators may include the rate of change of the original indicator, a combination with other indicators, or a new indicator obtained after time series analysis of the original indicator. Each risk indicator corresponds to a derived risk indicator set, and each derived risk indicator contains one or more derived indicators. For example, if the revenue of an enterprise fluctuates greatly and has a high risk coefficient, new indicators are derived from the risk indicator of revenue, such as revenue growth trend, market share changes, or comparison with the overall performance of the industry, so as to more comprehensively assess the revenue risk of the enterprise.

[0020] By deriving and expanding K risk indicators, analyzing the risk status of the enterprise from multiple perspectives and generating a more comprehensive set of derivative risk indicators, we can analyze the factors behind the risks more deeply and consider the relationship between risk indicators, which will help to more accurately predict the financing risks of the enterprise and improve the accuracy and reliability of risk prediction results.

[0021] S4: Execute derivative indicator feature value collection to generate K derivative risk indicator feature value sets, wherein the K derivative risk indicator feature value sets are feature value sets obtained after data collection on the target enterprise according to the K derivative risk indicator sets within the preset monitoring window.

[0022] Specifically, the collection of characteristic values ​​of derivative risk indicators refers to the process of collecting data on the set of derivative risk indicators generated by the target enterprise in step S3 within the preset monitoring window. According to the definition of the derivative risk indicators, relevant data are collected, which may include financial statements, market research reports, macroeconomic data, etc. The collected data are processed and analyzed to calculate the characteristic value of each derivative risk indicator. For example, it is assumed that the target enterprise is an e-commerce platform, and new risk indicators, namely customer churn rate and market expansion rate, have been derived based on the size of the risk coefficient. Within the preset monitoring window, data related to these derivative risk indicators are collected. For example, customer transaction data on the platform is collected to calculate the customer churn rate, and market research data is collected to evaluate the market expansion rate. By processing these data, the characteristic value of each derivative risk indicator is calculated. For example, the customer churn rate is 5% and the market expansion rate is 20% per year.

[0023] By collecting the characteristic values ​​of derivative indicators for the target enterprises, we obtain a set of K characteristic values ​​of derivative risk indicators. Each characteristic value corresponds to a derivative risk indicator, which can provide more fine-grained and comprehensive data support for financing risk prediction, and help to more accurately identify and predict the financing risks that enterprises may face.

[0024] S5: Based on the K risk indicators, the K derived risk indicator sets, the K target risk indicator characteristic values ​​and the K derived risk indicator characteristic value sets, construct a target risk prediction knowledge graph for the target enterprise.

[0025] Specifically, the K risk indicators, K derived risk indicator sets, K target risk indicator feature values, and K derived risk indicator feature value sets in the above steps S1 to S5 are integrated into a unified framework, the risk indicators and their feature values ​​are taken as entities, and the relationships between entities are identified, such as causal relationships, correlation relationships, and impact levels. Using a graph database or graph processing tool, such as Neo4j, ArangoDB, StellarDB, Gephi, KAPO, etc., the entities and their relationships are constructed into a knowledge graph, in which nodes represent risk indicators or their feature values, and edges represent the relationships between them, so as to obtain a target risk prediction knowledge graph for the target enterprise.

[0026] By integrating K risk indicators, K derivative risk indicator sets, K target risk indicator characteristic values ​​and K derivative risk indicator characteristic values ​​into the knowledge graph, we can clearly grasp the relationship between different risk indicators and understand the mutual influence between different risk indicators, so as to more accurately predict the financing risk of the target enterprise.

[0027] S6: Call M sample risk prediction knowledge graphs in the knowledge graph database, use the M sample risk prediction knowledge graphs to perform entity alignment on the target risk prediction knowledge graph, perform risk identification based on the alignment results, and generate the target risk results of the target enterprise, wherein the M sample risk prediction knowledge graphs have M risk result identifiers.

[0028] Specifically, the knowledge graph database is a database that stores multiple risk prediction knowledge graphs, each of which may correspond to historical risk data and prediction models of different enterprises or industries. The sample risk prediction knowledge graph refers to the risk prediction knowledge graph that has been constructed. They are used as reference samples for comparison and analysis with the risk prediction knowledge graph of the target enterprise. Entity alignment is the process of matching the risk prediction knowledge graph of the target enterprise with the corresponding entity in the sample knowledge graph. The risk result identifier is the risk result identified in the sample risk prediction knowledge graph, which is used to indicate the risk level of the corresponding entity or situation. It can be a risk level such as high risk, medium risk, low risk, etc., or a risk score, risk label, risk loss amount, etc. M sample risk prediction knowledge graphs similar to or related to the target enterprise are selected from the knowledge graph database, and each sample risk prediction knowledge graph has 1 risk result identifier. Using predefined rules or algorithms, the risk prediction knowledge graph of the target enterprise is entity aligned with the selected sample knowledge graph, including matching the attributes, structure and relationships of nodes. According to the results of entity alignment, the similarities and differences between the target enterprise and the sample are analyzed to identify the risks that the target enterprise may face. The alignment results and risk indicators are comprehensively analyzed to generate risk results for the target enterprise, such as risk assessment reports or risk levels.

[0029] By using M sample risk prediction knowledge graphs to identify the target risk prediction knowledge graph and generate the target risk results of the target enterprise, we can better understand the financing risk status of the target enterprise and make corresponding decisions.

[0030] Furthermore, if Figure 2 As shown, step S2 of the embodiment of the present application also includes: S21: Calculate the mean of the K risk indicator characteristic value sets respectively to generate K risk indicator characteristic value means.

[0031] S22: Construct K first risk intervals, wherein the K first risk intervals are K intervals obtained by searching the K risk indicator characteristic value sets with the mean of the K risk indicator characteristic values ​​as the interval center and a preset identification step as the interval radius, and the K first risk intervals have K first clustering factors.

[0032] S23: constructing K first left risk intervals and K first right risk intervals, wherein the K first left risk intervals have K first left clustering factors, and the K first right risk intervals have K first right clustering factors.

[0033] S24: Determine K fluctuation identification directions based on the K first left-side clustering factors and the K second right-side clustering factors, perform risk fluctuation identification of K risk indicator characteristic value sets based on the K fluctuation identification directions, and generate K risk coefficients and K target risk indicator characteristic values.

[0034] Preferably, step S24 of the embodiment of the present application further includes: S241: The risk intervals corresponding to the K fluctuation identification directions are taken as K stage risk intervals.

[0035] S242: Determine whether the K stage clustering factors of the K stage risk intervals are greater than or equal to the K first clustering factors. If so, perform risk fluctuation identification of K risk indicator characteristic value sets according to the K stage risk intervals and the K fluctuation identification directions until the preset identification constraints are met, and use the K stage risk intervals corresponding to the maximum values ​​of the clustering factors in the identification process as K target risk intervals.

[0036] S243: Calculate the mean of each of the multiple risk indicator characteristic values ​​in the K target risk intervals to generate K target risk indicator characteristic values.

[0037] Preferably, step S24 of the embodiment of the present application further includes: S244: performing variance calculations on the multiple risk indicator characteristic values ​​in the K target risk intervals respectively, and generating the K risk coefficients according to the calculation results.

[0038] Specifically, the mean of all the eigenvalues ​​in each set of K risk indicator eigenvalue sets generated in step S1 is calculated to obtain the mean of K risk indicator eigenvalues. A recognition step is preset, and this recognition step refers to the eigenvalue difference between the centers of two adjacent intervals. The first risk interval is an interval retrieved from the corresponding risk indicator eigenvalue set with the mean of each risk indicator eigenvalue as the interval center and the preset recognition step as the interval radius. This interval is used to represent the range of variation of the risk indicator eigenvalue. The first clustering factor refers to the ratio of the number of eigenvalues ​​in each first risk interval to the total number of eigenvalues ​​in the corresponding risk indicator eigenvalue, reflecting the density of eigenvalue clustering in the first risk interval. With the mean of the K risk indicator eigenvalues ​​calculated above as the interval center and the preset recognition step as the interval radius, a search is performed in the K risk indicator eigenvalue sets to construct K first risk intervals. These K first risk intervals correspond to K first clustering factors.

[0039] On the basis of the K first risk intervals, a certain range is expanded to the left and right respectively to obtain K first left risk intervals and K first right risk intervals, which are used to identify the changing trend of the characteristic value of the risk indicator. The K first left risk intervals correspond to K first left aggregation factors, and the K first right risk intervals correspond to K first right aggregation factors. The calculation method of the first left aggregation factor and the first right aggregation factor is similar to the first aggregation factor. The first left aggregation factor reflects the degree of aggregation of the characteristic value on the left side of the mean, while the second right aggregation factor reflects the degree of aggregation of the characteristic value on the right side of the mean.

[0040] Compare the first left-side clustering factor and the first right-side clustering factor corresponding to each first risk interval, and take the direction corresponding to the risk interval with a larger clustering factor as the fluctuation identification direction to obtain K fluctuation identification directions. According to the determined fluctuation identification direction, the K risk intervals corresponding to the K fluctuation identification directions are taken as K stage risk intervals. These intervals represent the direction and range of the fluctuation of the characteristic value of the risk indicator, and are used to analyze the evolution process of the risk in more detail. The K stage clustering factors are the clustering factors corresponding to the stage risk interval, which are used to measure the degree of aggregation of the characteristic value of the risk indicator in the interval.

[0041] The K stage clustering factors of the K stage risk intervals are compared with the K first clustering factors respectively to determine whether the K stage clustering factors of the K stage risk intervals are greater than or equal to the K first clustering factors. If so, risk fluctuation identification of the K risk indicator characteristic value sets is performed according to the K stage risk intervals and the K fluctuation identification directions until the preset identification constraints are met, and the K stage risk intervals corresponding to the maximum values ​​of the clustering factors in the identification process are taken as K target risk intervals. The K target risk intervals represent the most significant parts of the risk fluctuations. The preset identification constraints refer to a series of conditions or standards set in advance in order to ensure the accuracy and effectiveness of the identification results during the risk fluctuation identification process. It can be that the preset number of identifications is reached, or the difference in the clustering factors of two adjacent identifications is greater than the preset difference, which proves that the boundary of the interval with a relatively dense distribution has been reached at this time.

[0042] For example, there is a target enterprise with three risk indicators (K=3), namely A, B, and C, and the identification step is set to 1.

[0043] Risk indicator characteristic value set: The eigenvalue set of risk indicator A: {1,2,2,3,3,3,4,4,5} The eigenvalue set of risk indicator B: {2,2,3,3,3,3,4,5,5} The eigenvalue set of risk indicator C: {1,1,2,2,3,3,3,4,4,5} Compute the mean of each set: The mean value of risk indicator A:

[0044] The mean value of risk indicator B is:

[0045] The mean value of risk indicator C:

[0046] With each mean as the interval center and the identification step length as the radius, the first risk interval is constructed: The interval of risk indicator A: [2,4] The interval of risk indicator B: [2.33, 4.33] The interval of risk index C: [1.8,3.8] Calculate the first clustering factor: The first clustering factor of risk indicator A: There are 7 eigenvalues ​​in the interval [2,4], the total number of eigenvalues ​​is 9, and the clustering factor is 0.78; The first clustering factor of risk indicator B: there are 6 eigenvalues ​​in the interval [2.33, 4.33], the total number of eigenvalues ​​is 9, and the clustering factor is 0.67; The first clustering factor of risk indicator C: There are 8 eigenvalues ​​in the interval [1.8,3.8], the total number of eigenvalues ​​is 10, and the clustering factor is 0.8.

[0047] Construct the first left risk interval and the first right risk interval and calculate the clustering factor Risk Indicator A: First left risk interval: [1,3], first left clustering factor 0.67 First right risk interval: [3,5], first right clustering factor 0.67 Risk Indicator B: First left risk interval: [1.33, 3.33], first left clustering factor 0.89 First right risk interval: [3.33,5.33], first right clustering factor 0.33 Risk Indicator C: First left risk interval: [0.8, 2.8], first left aggregation factor 0.6 First right risk interval: [2.8,4.8], first right aggregation factor 0.7 Compare the first left clustering factor and the first right clustering factor to determine the fluctuation direction: Risk indicator A: The left side aggregation factor = the right side aggregation factor, so the fluctuation direction can be selected arbitrarily. Here, the right side is selected, and the stage risk range is [3,5] Risk indicator B: The left side aggregation factor > the right side aggregation factor, so the fluctuation direction is on the left side, and the stage risk range is [1.33, 3.33] Risk indicator C: The right side aggregation factor is greater than the left side aggregation factor, so the fluctuation direction is to the right, and the stage risk range is [2.8,4.8].

[0048] Compare the stage aggregation factor and the first aggregation factor: Risk indicator A: The stage aggregation factor 0.67 < the first aggregation factor 0.78, the condition is not met, and the target risk interval is not updated.

[0049] Risk indicator B: stage aggregation factor 0.89>first aggregation factor 0.67, updated target risk interval is [1.33,3.33].

[0050] Risk indicator C: The stage aggregation factor 0.7 < the first aggregation factor 0.8, the condition is not met, and the target risk interval is not updated.

[0051] For risk indicator B, according to its fluctuation identification direction, expand one step to the left to obtain a new interval [0.33, 2.33]. The corresponding interval aggregation factor is calculated to be 0.22. Compare the new aggregation factor with the stage aggregation factor. If 0.22 < 0.89, the target risk interval will not be updated.

[0052] Final target risk range: Risk indicator A: Target risk interval is [2,4] Risk indicator B: Target risk interval is [1.33, 3.33] Risk indicator C: Target risk interval is [1.8,3.8] The mean of the characteristic values ​​of multiple risk indicators in the K target risk intervals is calculated respectively, and the mean is used as the K target risk indicator characteristic values. The variance of the characteristic values ​​of multiple risk indicators in the K target risk intervals is calculated respectively, and the calculation results are used as K risk coefficients.

[0053] Through the above process, the risk indicator characteristic value set is deeply mined to find the target risk range that best represents the risk fluctuation direction and degree, thereby determining the risk coefficient and target risk indicator characteristic value, providing valuable reference data values ​​for the subsequent financing risk forecast of the target enterprise, and ensuring that the forecast results are accurate and effective.

[0054] Optionally, step S1 in the embodiment of the present application further includes: S11: Construct a first subset of preset indicators, wherein the first subset of preset indicators includes revenue, profit margin, debt-paying ability, and debt-to-asset ratio.

[0055] S12: Construct a second subset of preset indicators, wherein the second subset of preset indicators includes enterprise scale, structural design, and the number of benchmarked enterprises.

[0056] S13: Construct the preset indicator set according to the first preset indicator subset and the second preset indicator subset.

[0057] Specifically, the first preset indicator subset is a set of key financial indicators used to assess the financial health of an enterprise. These indicators are usually related to the profitability, debt-paying ability and financial stability of the enterprise. For example, revenue, profit margin, debt-paying ability and debt-to-asset ratio. The second preset indicator subset is a set of non-financial indicators used to assess other key factors of an enterprise, such as its size, organizational structure and industry position. These indicators can help to gain a more comprehensive understanding of the operation of an enterprise. For example, enterprise size, structural design and the number of benchmarked enterprises. The first preset indicator subset and the second preset indicator subset are combined to form a preset indicator set, which is used to comprehensively assess the risk status of an enterprise.

[0058] Through the above steps, combining financial and non-financial indicators, we can obtain a more comprehensive set of preset indicators, which will provide a more comprehensive understanding of the overall situation of the enterprise, cover as many indicators as possible that affect the financing risks of the enterprise, and facilitate the collection of comprehensive and complete data based on these risk indicators.

[0059] Furthermore, step S6 of the embodiment of the present application also includes: S61: Constructing preset alignment rules.

[0060] S62: Perform entity alignment on the M sample risk prediction knowledge graphs and the target risk prediction knowledge graphs according to preset alignment rules to obtain M sample alignment entity statistics.

[0061] S63: Perform risk prediction on the target enterprise based on the M sample aligned entity statistics and the M risk result identifiers to generate a target risk result for the target enterprise.

[0062] Preferably, step S63 of the embodiment of the present application further includes: S631: The ratios of the M sample alignment entity statistics to the total number of entities in the target risk prediction knowledge graph are respectively used as M first prediction weights.

[0063] S632: Perform weighted calculation on the M risk result identifiers using the M first prediction weights to generate the target risk result.

[0064] Specifically, the preset alignment rules refer to a set of rules or standards used to match and associate target and sample enterprise entities during the entity alignment process. These rules may include similarity calculation methods, entity matching strategies, etc. to ensure the accuracy and consistency of the alignment. The sample alignment entity statistics are the number of target entities that meet the preset alignment rules after matching and associating each sample entity with the target entity during the entity alignment process. Applying the preset alignment rules, the entities in the target risk prediction knowledge graph are matched with the entities in each sample knowledge graph, and the nodes that represent the same concept or entity in different graphs are identified to obtain M sample alignment entity statistics.

[0065] Calculate the ratio of the statistics of the M sample aligned entities to the total number of entities in the target risk prediction knowledge graph to obtain M first prediction weights, which reflect the relative importance or influence of each sample graph in predicting the risk of the target enterprise. Use the M first prediction weights to perform weighted calculation on the M risk result identifiers to generate the risk prediction result of the target enterprise.

[0066] Through the above steps, the risk prediction information of multiple sample enterprises is used to predict the risk status of the target enterprise. Different weights are set according to the degree of similarity, and the prediction results of multiple samples are integrated to make full use of historical sample data, improve the accuracy and reliability of risk prediction, and obtain more accurate target risk results.

[0067] In summary, the enterprise financing risk prediction method based on knowledge graph provided in the embodiment of the present application has the following technical effects: By continuously monitoring the risks of the target enterprise according to the preset indicator set within the preset monitoring window, generating K risk indicator characteristic value sets, performing risk fluctuation identification, and generating K risk coefficients and K target risk indicator characteristic values ​​by constructing risk intervals and performing iterative optimization; deriving and expanding the K risk indicators according to the size of the K risk coefficients to generate K derivative risk indicator sets; performing derivative indicator characteristic value collection to generate K derivative risk indicator characteristic value sets, collecting real-time data of various external factors affecting the financing risk of the enterprise, and constructing the target risk prediction knowledge graph of the target enterprise based on these real-time data; revealing the relationship between different risk indicators and understanding the mutual influence between different risk indicators. Calling M sample risk prediction knowledge graphs in the knowledge graph database, using the M sample risk prediction knowledge graphs to perform entity alignment on the target risk prediction knowledge graph, setting different weights for each sample risk prediction knowledge graph according to the entity alignment results, and generating the target risk results of the target enterprise through weighted calculation.

[0068] As mentioned above, this application collects a large amount of real-time data to conduct a comprehensive analysis of various external factors, comprehensively considers the impact of various factors on corporate financing risks, draws a target risk prediction knowledge graph for target enterprises, and deeply mines historical data, calls multiple samples in the knowledge graph database for risk identification and prediction, ensures the reliability of risk prediction results, eliminates the influence of accidental factors, and captures corporate risk status more comprehensively and dynamically, adapts to changes in external risks, improves the accuracy and reliability of corporate financing risk prediction, and helps financial institutions and enterprises to more effectively identify and manage financing risks, thereby optimizing financing decisions.

[0069] Embodiment 2, as Figure 3 As shown, the embodiment of the present application provides a corporate financing risk prediction system based on a knowledge graph, and the system includes: Risk monitoring module 10: The risk monitoring module 10 is used to continuously monitor the risks of the target enterprise according to a preset indicator set within a preset monitoring window, and generate K risk indicator feature value sets, wherein the preset indicator set includes K risk indicators; Fluctuation identification module 20: The fluctuation identification module 20 is used to perform risk fluctuation identification, generate K risk coefficients and K target risk indicator characteristic values, wherein the K risk coefficients are coefficients obtained by performing fluctuation identification on the K risk indicator characteristic value sets; Derivative expansion module 30: The derivative expansion module 30 is used to derive and expand the K risk indicators according to the magnitudes of the K risk coefficients to generate a set of K derived risk indicators; Characteristic value collection module 40: The characteristic value collection module 40 is used to perform derivative indicator characteristic value collection and generate K derivative risk indicator characteristic value sets, wherein the K derivative risk indicator characteristic value sets are characteristic value sets obtained after data collection of the target enterprise according to the K derivative risk indicator sets within the preset monitoring window; Knowledge graph construction module 50: The knowledge graph construction module 50 is used to construct a target risk prediction knowledge graph of the target enterprise based on the K risk indicators, the K derived risk indicator sets, the K target risk indicator feature values ​​and the K derived risk indicator feature value sets; Risk identification module 60: The risk identification module 60 is used to call M sample risk prediction knowledge graphs in the knowledge graph database, use the M sample risk prediction knowledge graphs to perform entity alignment on the target risk prediction knowledge graph, perform risk identification based on the alignment results, and generate the target risk results of the target enterprise, wherein the M sample risk prediction knowledge graphs have M risk result identifiers.

[0070] Furthermore, the fluctuation identification module 20 in the embodiment of the present application is also used to perform the following steps: Mean calculation unit: used to perform mean calculation on the K risk indicator characteristic value sets respectively to generate the mean values ​​of the K risk indicator characteristic values.

[0071] The first risk interval construction unit is used to construct K first risk intervals, wherein the K first risk intervals are K intervals obtained by searching the K risk indicator characteristic value sets with the mean of the K risk indicator characteristic values ​​as the interval center and a preset identification step as the interval radius, and the K first risk intervals have K first clustering factors.

[0072] The second risk interval construction unit constructs K first left risk intervals and K first right risk intervals, wherein the K first left risk intervals have K first left clustering factors, and the K first right risk intervals have K first right clustering factors.

[0073] Risk fluctuation identification unit: determines K fluctuation identification directions based on the K first left-side aggregation factors and the K second right-side aggregation factors, performs risk fluctuation identification of K risk indicator characteristic value sets based on the K fluctuation identification directions, and generates K risk coefficients and K target risk indicator characteristic values.

[0074] Furthermore, the risk fluctuation identification unit of the embodiment of the present application further includes: A stage risk interval generating unit is used to use the risk intervals corresponding to the K fluctuation identification directions as K stage risk intervals.

[0075] Target risk interval generation unit: used to determine whether the K stage clustering factors of the K stage risk intervals are greater than or equal to the K first clustering factors. If so, risk fluctuation identification of K risk indicator characteristic value sets is performed according to the K stage risk intervals and the K fluctuation identification directions until the preset identification constraints are met, and the K stage risk intervals corresponding to the maximum values ​​of the clustering factors in the identification process are used as K target risk intervals.

[0076] The target risk indicator characteristic value generating unit is configured to respectively calculate the mean of a plurality of risk indicator characteristic values ​​in the K target risk intervals to generate K target risk indicator characteristic values.

[0077] Furthermore, the risk fluctuation identification unit of the embodiment of the present application further includes: The risk coefficient generating unit is used to perform variance calculation on a plurality of risk indicator characteristic values ​​in the K target risk intervals respectively, and generate the K risk coefficients according to the calculation results.

[0078] Furthermore, the risk monitoring module 10 in the embodiment of the present application further includes: The first preset indicator unit is used to construct a first preset indicator subset, wherein the first preset indicator subset includes revenue, profit margin, debt repayment ability, and debt-to-asset ratio.

[0079] The second preset indicator unit is used to construct a second preset indicator subset, wherein the second preset indicator subset includes enterprise scale, structural design, and the number of benchmarked enterprises.

[0080] Preset indicator set unit: used to construct the preset indicator set according to the first preset indicator subset and the second preset indicator subset.

[0081] Furthermore, the risk identification module 60 in the embodiment of the present application also includes: Rule building unit: used to build preset alignment rules.

[0082] Entity alignment unit: used to perform entity alignment on the M sample risk prediction knowledge graphs and the target risk prediction knowledge graphs according to preset alignment rules to obtain M sample alignment entity statistics.

[0083] Prediction unit: used for performing risk prediction on the target enterprise based on the M sample aligned entity statistics and the M risk result identifiers, and generating a target risk result of the target enterprise.

[0084] Furthermore, the prediction unit in the embodiment of the present application also includes: A weight generating unit is used to respectively use the ratios of the M sample alignment entity statistics to the total number of entities in the target risk prediction knowledge graph as M first prediction weights.

[0085] A weighted calculation unit is used to perform weighted calculation on the M risk result identifiers using the M first prediction weights to generate the target risk result.

[0086] Through the above detailed description of the enterprise financing risk prediction method based on knowledge graph in this specification, those skilled in the art can clearly understand the enterprise financing risk prediction system based on knowledge graph in this embodiment. For the system disclosed in Example 2, since it corresponds to the method disclosed in Example 1, it has corresponding functional modules and beneficial effects. For relevant matters, please refer to the description of the method part.

[0087] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. The enterprise financing risk prediction method based on knowledge graph is characterized by: The method comprises: Within a preset monitoring window, continuously monitor the risks of the target enterprise according to a preset indicator set to generate K risk indicator feature value sets, wherein the preset indicator set includes K risk indicators; Perform risk fluctuation identification to generate K risk coefficients and K target risk indicator characteristic values, wherein the K risk coefficients are coefficients obtained by performing fluctuation identification on the K risk indicator characteristic value sets; Derivatively expanding the K risk indicators according to the magnitudes of the K risk coefficients to generate a set of K derived risk indicators; Execute derivative indicator feature value collection to generate K derivative risk indicator feature value sets, wherein the K derivative risk indicator feature value sets are feature value sets obtained after data collection on the target enterprise according to the K derivative risk indicator sets within the preset monitoring window; Based on the K risk indicators, the K derived risk indicator sets, the K target risk indicator characteristic values ​​and the K derived risk indicator characteristic value sets, construct a target risk prediction knowledge graph for the target enterprise; Call M sample risk prediction knowledge graphs in the knowledge graph database, use the M sample risk prediction knowledge graphs to perform entity alignment on the target risk prediction knowledge graph, perform risk identification based on the alignment results, and generate the target risk results of the target enterprise, wherein the M sample risk prediction knowledge graphs have M risk result identifiers.

2. The method according to claim 1, characterized in that Perform risk fluctuation identification to generate K risk coefficients and K target risk indicator characteristic values, the method comprising: Calculating the mean of the K risk indicator characteristic value sets respectively to generate the mean of the K risk indicator characteristic values; Constructing K first risk intervals, wherein the K first risk intervals are K intervals obtained by searching in the K risk indicator characteristic value sets with the mean of the K risk indicator characteristic values ​​as the interval center and a preset identification step length as the interval radius, and the K first risk intervals have K first clustering factors; Constructing K first left risk intervals and K first right risk intervals, wherein the K first left risk intervals have K first left clustering factors, and the K first right risk intervals have K first right clustering factors; Based on the K first left-side clustering factors and the K second right-side clustering factors, K fluctuation identification directions are determined, and based on the K fluctuation identification directions, risk fluctuation identification of K risk indicator characteristic value sets is performed to generate K risk coefficients and K target risk indicator characteristic values.

3. The method according to claim 2, characterized in that The method comprises: The risk intervals corresponding to the K fluctuation identification directions are used as K stage risk intervals; Determine whether the K stage clustering factors of the K stage risk intervals are greater than or equal to the K first clustering factors. If so, perform risk fluctuation identification of the K risk indicator characteristic value sets according to the K stage risk intervals and the K fluctuation identification directions until the preset identification constraints are met, and use the K stage risk intervals corresponding to the maximum values ​​of the clustering factors in the identification process as K target risk intervals; The mean values ​​of the multiple risk indicator characteristic values ​​in the K target risk intervals are respectively calculated to generate K target risk indicator characteristic values.

4. The method according to claim 3, characterized in that The method comprises: Variance calculations are performed on a plurality of risk indicator characteristic values ​​in the K target risk intervals respectively, and the K risk coefficients are generated according to the calculation results.

5. The method according to claim 1, characterized in that The method comprises: Constructing a first preset indicator subset, wherein the first preset indicator subset includes revenue, profit margin, debt-paying ability, and debt-to-asset ratio; Constructing a second preset indicator subset, wherein the second preset indicator subset includes enterprise scale, structural design, and the number of benchmarked enterprises; The preset indicator set is constructed according to the first preset indicator subset and the second preset indicator subset.

6. The method according to claim 1, characterized in that The method comprises: Construct preset alignment rules; Perform entity alignment on the M sample risk prediction knowledge graphs and the target risk prediction knowledge graph according to a preset alignment rule to obtain M sample alignment entity statistics; Based on the M sample aligned entity statistics and the M risk result identifiers, risk prediction is performed on the target enterprise to generate a target risk result for the target enterprise.

7. The method according to claim 6, characterized in that The method comprises: The ratios of the M sample alignment entity statistics to the total number of entities in the target risk prediction knowledge graph are respectively used as M first prediction weights; The M first prediction weights are used to perform weighted calculation on the M risk result identifiers to generate the target risk result.

8. The enterprise financing risk prediction system based on knowledge graph is characterized by: The system comprises: Risk monitoring module: The risk monitoring module is used to continuously monitor the risks of the target enterprise according to a preset indicator set within a preset monitoring window, and generate K risk indicator feature value sets, wherein the preset indicator set includes K risk indicators; Fluctuation identification module: The fluctuation identification module is used to perform risk fluctuation identification and generate K risk coefficients and K target risk indicator characteristic values, wherein the K risk coefficients are coefficients obtained by performing fluctuation identification on the K risk indicator characteristic value sets; Derivative expansion module: The derivative expansion module is used to derive and expand the K risk indicators according to the sizes of the K risk coefficients to generate a set of K derived risk indicators; Characteristic value collection module: the characteristic value collection module is used to perform derivative indicator characteristic value collection and generate K derivative risk indicator characteristic value sets, wherein the K derivative risk indicator characteristic value sets are characteristic value sets obtained after data collection of the target enterprise according to the K derivative risk indicator sets within the preset monitoring window; Knowledge graph construction module: the knowledge graph construction module is used to construct a target risk prediction knowledge graph of the target enterprise based on the K risk indicators, the K derived risk indicator sets, the K target risk indicator feature values ​​and the K derived risk indicator feature value sets; Risk identification module: The risk identification module is used to call M sample risk prediction knowledge graphs in the knowledge graph database, use the M sample risk prediction knowledge graphs to perform entity alignment on the target risk prediction knowledge graph, perform risk identification based on the alignment results, and generate the target risk results of the target enterprise, wherein the M sample risk prediction knowledge graphs have M risk result identifiers.

Images