A risk control system and method for the financing guarantee industry

Abstract

The application discloses a risk control system and method for a financing guarantee industry, wherein an industry evaluation index system table module reduces the dimension of qualitative indexes, and a support vector machine model adopting recursive feature elimination is used to reduce the dimension of quantitative data to screen important indexes; a risk assessment model module adopts a multilayer Delphi method module to determine the weight of each index, sets an alarm signal, and sets a risk threshold of each index according to an industry standard; a risk value module is used to obtain each index value through a user input mode, and each business link risk value is calculated according to the risk assessment model module; a penalty factor module performs a credit rating prediction, and determines a credit rating binding penalty factor; and an early warning mechanism module monitors the system through an early warning task, triggers early warning monitoring according to early warning rules after a risk result is obtained, and performs early warning display. The application can effectively improve the risk control capability of a guarantee company.

Description

Technical Field

[0001] This invention relates to the field of intelligent information monitoring and big data analysis technology, and in particular to a risk control system and method for the financing guarantee industry. Background Technology

[0002] Throughout the project implementation process, the decision-makers at financing guarantee institutions often struggle to understand the information behind the risk assessment report: When was the investigation conducted? Who conducted the investigation? How was the investigation conducted? How significant is the risk? Who analyzed the results? How were the results analyzed? Why were these results obtained? The questions are numerous, and the decision-makers simply receive a report riddled with problems. Because of the lack of a scientifically sound risk quantification scheme, the project's risk warning range is extremely difficult to determine, let alone preventative measures after risks materialize. This leads to a situation where risk warnings are left unclear about "who should evaluate and how should they be controlled," making risk assessment and risk warnings stumbling blocks to the industry's development.

[0003] Therefore, it is necessary to use information technology to monitor essential business information during business processes, oversee all aspects of project investigation, and conduct risk assessments.

[0004] In conclusion, employing scientific and reasonable information technology solutions to monitor necessary and important business information throughout the entire business process, while simultaneously conducting risk assessments of each business segment, has become an indispensable part of the work for guarantee institutions to reduce risk. Therefore, combining scientific and reasonable quantitative methods for risk assessment, and establishing risk assessment models and early warning mechanisms for business segments within the financing guarantee risk control system, to improve the risk control capabilities of guarantee companies and enable timely response measures, is of paramount practical significance and necessity for the development of financing guarantee enterprises. Summary of the Invention

[0005] Based on the problems mentioned in the above technical background, the present invention provides a risk control solution and system for the financing guarantee industry to solve the above problems.

[0006] This invention provides a risk control solution for the financing guarantee industry, which includes an industry evaluation index system table module, a risk assessment model module, a risk value module, a penalty factor module, and an early warning mechanism module connected in sequence.

[0007] The industry evaluation index system module uses an improved risk matrix method module and a fuzzy comprehensive evaluation method module to reduce the dimensionality of qualitative indicators, and uses a support vector machine model with recursive feature elimination to reduce the dimensionality of quantitative data in order to screen out important indicators.

[0008] The risk assessment model module is used to determine the weight of each indicator based on the dimensionality-reduced indicator system table, using the multi-level DeFell method module, then setting alarm signals, and setting the risk threshold of each indicator according to industry standards.

[0009] The risk value module is used to obtain the values ​​of various indicators through user input and to calculate the risk value of each business process based on the risk assessment model module.

[0010] The penalty factor module is used to input indicator data into the random forest model module to predict the credit rating and determine the penalty factor bound to the credit rating.

[0011] The early warning mechanism module is used to input the result of multiplying the risk value module and the penalty factor module into the early warning mechanism, monitor the system through early warning tasks, and trigger early warning monitoring according to the early warning rules after obtaining the risk result, and display the early warning.

[0012] Furthermore, the improved risk matrix method module adds a prevention probability dimension to the classic risk matrix method, setting three levels of high, medium, and low for each dimension. The DeFilh method is used to evaluate the high, medium, and low levels one by one for each of the three dimensions. The fuzzy comprehensive evaluation method module performs further evaluation to determine the key risks.

[0013] Furthermore, the recursive feature elimination support vector machine model includes: the recursive feature elimination model using support vector machine as the training model initializes the previous index features, trains the model using SVM, deletes the features with the smallest absolute value weight, and then repeats the above process until the required number of features is reached.

[0014] Furthermore, the multi-layered Defield method module includes four layers: the first layer is to determine the weight of each indicator, i.e., the target layer; the second layer divides all indicators into qualitative or quantitative indicators; the third layer is the dimension to which each indicator belongs: operational risk, financial risk; and the fourth layer is the specific indicator item.

[0015] Furthermore, the risk assessment model module includes three main modules: risk identification, risk analysis, and risk measurement. The risk identification module is used to define evaluation indicators and set alarm signals. The risk analysis module is used to study how to quantitatively or qualitatively evaluate alarm signals and generate corresponding risk assessment results. The risk measurement module is used to study how to comprehensively evaluate the risk assessment results across various business dimensions and obtain the final evaluation results. Quantitative analysis results are directly incorporated into risk measurement. For qualitative risk assessment results, a multi-level analysis method is used to quantitatively analyze non-quantitative events and convert qualitative analysis into weights for final risk measurement.

[0016] Furthermore, the random forest model module needs to be trained before the indicator data is input. It uses the SME listing data in the Guotai An database as the dataset for training to achieve the prediction of credit rating. The credit rating is divided into 18 levels (referencing industry standards) and corresponds to different penalty factors.

[0017] Furthermore, the early warning mechanism module includes three parts: early warning task definition, early warning task monitoring, and early warning result display. The early warning mechanism monitors the monitoring task by defining the alarm signals associated with the early warning task, triggers early warning rules through data perception, pushes early warning messages to the corresponding business links, and displays the early warning information or generates relevant early warning reports.

[0018] The present invention provides a method for a risk control system for the financing guarantee industry, comprising the following steps:

[0019] Step 1: Construct an industry evaluation index system table module. Use an improved risk matrix method module and a fuzzy comprehensive evaluation method module to reduce the dimensionality of qualitative indicators, and use a support vector machine model with recursive feature elimination to reduce the dimensionality of quantitative data in order to screen out important indicators.

[0020] Step 2: Based on the reduced-dimensionality indicator system table, the weight of each indicator is determined using the multi-level DeFilh method module. Then, alarm signals are set, and the risk thresholds of each indicator are set according to industry standards to establish a risk assessment model module.

[0021] Step 3: Obtain the values ​​of each indicator by the user input method, calculate the risk value of each business link according to the risk assessment model module, and construct the risk value module;

[0022] Step 4: Input the indicator data into the random forest model module to predict the credit rating and determine the credit rating binding penalty factor module.

[0023] Step 5: The result of multiplying the risk value module and the penalty factor module is input into the early warning mechanism module. The early warning mechanism module monitors the system through early warning tasks, obtains the risk result, triggers early warning monitoring according to the early warning rules, and displays the early warning.

[0024] Furthermore, step 1 specifically includes:

[0025] The indicator system is established using a multi-dimensional mechanism. Through enterprise profiling, the basic situation of the enterprise and the business information of the enterprise in various dimensions are defined and extracted as indicator features. During the extraction process, the indicator features can be dynamically expanded.

[0026] In the dimensionality reduction algorithm for qualitative indicators, an improved risk matrix method and a fuzzy comprehensive evaluation method are adopted. The improved risk matrix method adds the dimension of preventability to the classic risk matrix method, and sets three levels (high, medium, and low) for each dimension. The DeFiel method is used to evaluate each of the three dimensions one by one at the high, medium, and low levels. Then, the fuzzy comprehensive evaluation method is used for further evaluation to determine the key risks, specifically including:

[0027] Step 2-1 Let the set of evaluation dimensions be U = {probability of occurrence, severity of consequences, and possibility of prevention};

[0028] Step 2-2 Let the set of comments be V = {high, medium, low};

[0029] Step 2-3: Establish a fuzzy comprehensive risk assessment matrix;

[0030] Step 2-4: Use the Defield method to determine the weight coefficient vector of the evaluation indicators;

[0031] Steps 2-5 determine the fuzzy synthesis operator;

[0032] Steps 2-6 analyze the fuzzy comprehensive evaluation results to identify key risks;

[0033] The recursive feature elimination support vector machine model initializes the previous index features, trains the model using SVM, deletes the features with the smallest absolute weight, and then repeats the above process until the required number of features is reached.

[0034] The beneficial effects of this invention are:

[0035] (1) For the financing guarantee field, a complete risk control model is proposed, which scientifically and rationally achieves the effect of efficient risk control, greatly saves manpower and material resources, and reduces unnecessary operational process errors.

[0036] (2) In the model (industry evaluation index system table module, risk assessment model module, risk value module, penalty factor module, early warning mechanism module), the risk identification, risk analysis and risk measurement are decoupled to ensure the scalability of the risk control method. The algorithm module used has high analysis accuracy and fast response speed, realizing intelligent information monitoring.

[0037] (3) In application development, the alarm signal is used as the external interface of the model, which simplifies the complexity of system development.

[0038] (4) By dynamically binding or associating with relevant business objects and indicator systems, the needs of dynamic application evolution can be met, and the development cost of application systems can be reduced. Attached Figure Description

[0039] Figure 1 This is a partial schematic diagram of the indicator system table for the financing guarantee industry;

[0040] Figure 2 This is a schematic diagram of the improved risk matrix method for evaluation;

[0041] Figure 3 This is a schematic diagram of the algorithm flow of the fuzzy comprehensive evaluation method;

[0042] Figure 4 This is a flowchart illustrating the processing of the indicator system table;

[0043] Figure 5 This is a schematic diagram of the multi-layered de Felix method system;

[0044] Figure 6 This is a schematic diagram of a risk control model;

[0045] Figure 7 This is a diagram illustrating the early warning mechanism;

[0046] Figure 8 This is a schematic diagram of the technical route;

[0047] Figure 9 This is a business logic diagram. Detailed Implementation

[0048] The embodiments of the present invention are illustrated below with reference to the accompanying drawings and specific examples. Various details in this specification can be improved and applied to different scenarios, and can be modified or changed in various ways without departing from the spirit of the invention. It should be noted that the examples described below are merely for detailed explanation of the invention to facilitate the reader's understanding, and do not constitute any limitation or constraint on the invention.

[0049] It should be noted that this invention serves the financing guarantee industry, aiming to enhance the risk control capabilities of various financing guarantee companies to address industry challenges. This invention is specifically targeted at financing guarantee companies that primarily serve small and medium-sized enterprises (SMEs).

[0050] The present invention provides a risk control system for the financing guarantee industry, comprising an industry evaluation index system table module, a risk assessment model module, a risk value module, a penalty factor module, and an early warning mechanism module connected in sequence.

[0051] The industry evaluation index system module uses an improved risk matrix method module and a fuzzy comprehensive evaluation method module to reduce the dimensionality of qualitative indicators, and uses a support vector machine model with recursive feature elimination to reduce the dimensionality of quantitative data in order to screen out important indicators.

[0052] The risk assessment model module is used to determine the weight of each indicator based on the dimensionality-reduced indicator system table, using the multi-level DeFell method module, then setting alarm signals, and setting the risk threshold of each indicator according to industry standards.

[0053] The risk value module is used to obtain the values ​​of various indicators through user input and to calculate the risk value of each business process based on the risk assessment model module.

[0054] The penalty factor module is used to input indicator data into the random forest model module to predict the credit rating and determine the penalty factor bound to the credit rating.

[0055] The early warning mechanism module is used to input the result of multiplying the risk value module and the penalty factor module into the early warning mechanism, monitor the system through early warning tasks, and trigger early warning monitoring according to the early warning rules after obtaining the risk result, and display the early warning.

[0056] This invention discloses a risk control system for the financing guarantee industry, addressing issues such as poor risk control capabilities, strong subjectivity in risk assessment, and inefficient business processing within the industry. Specific implementation methods include:

[0057] A module for constructing an industry evaluation index system table is used. An improved risk matrix method and fuzzy comprehensive evaluation method are employed to reduce the dimensionality of qualitative indicators, and a support vector machine model with recursive feature elimination is used to reduce the dimensionality of quantitative data in order to select important indicators, including:

[0058] like Figure 1 As shown, an industry evaluation index system table is constructed based on industry standards and the requirements of guarantee companies, exhibiting different characteristics across different dimensions. The construction of the index system table must be comprehensive.

[0059] An improved risk matrix method and fuzzy comprehensive evaluation method are used to reduce the dimensionality of qualitative indicators. In the classic risk matrix method, risk has two dimensions: first, the probability of risk occurrence, which is considered in terms of probability; second, the severity of consequences, which is considered in terms of the consequences after the risk occurs. Considering that two dimensions of risk are insufficient for effective risk assessment, and taking into account practical application scenarios, this invention adds a dimension of preventability. Within each of these dimensions, three levels—high, medium, and low—are set, and the DeFiel method is used to evaluate each level individually. The risk assessment table below can be obtained based on the results of the DeFiel method.

[0060] Table 1

[0061]

[0062] Based on the evaluation results in the table above, fuzzy comprehensive evaluation is then used to further assess which risks are critical among all the risks. Figure 3 Here is the flowchart of the fuzzy comprehensive evaluation method algorithm:

[0063] Step 1: Assume the evaluation dimension set U = {probability of occurrence, severity of consequences, and possibility of prevention};

[0064] Step 2: Assume the set of comments V = {high, medium, low};

[0065] Step 3: As Figure 2 As shown, a fuzzy comprehensive evaluation matrix is ​​established for nine risks, including management team risk and financing project risk.

[0066]

[0067]

[0068]

[0069] Step 4: Use the Defield method to determine the weight coefficient vector of the evaluation index A = (0.3, 0.4, 0.3);

[0070] Step 5: Determine the fuzzy synthesis operator B i =AR i (i = 1, 2, ..., 9), we find:

[0071] B1=(0.67, 0.16, 0.17), B2=(0.60, 0.29, 0.11), B3=(0.50, 0.44, 0.06)

[0072] B4=(0, 29, 0.65, 0.06), B5= (0.51, 0.32, 0.17), B6= (0.26, 0.47, 0.27)

[0073] B7=(0.14, 0.43, 0.43), B8=(0.41, 0.50, 0.09), B9=(0.28, 0.60, 0.12)

[0074] Step 6: Analyze the fuzzy comprehensive evaluation results to identify key risks. This invention is based on the "maximum membership principle," where a risk is defined as "high" or higher than 50%. Four key risks were ultimately identified: management team risk, financing project risk, production and operation risk, and legal risk.

[0075] The fuzzy comprehensive evaluation module further assessed and identified key risks, which include: Management team risk dimension: changes in legal representative, key personnel, major shareholders, suspected corruption, frequent management personnel changes; Financing project risk: deteriorating prospects of financing projects, failure to complete financing projects on time, failure to achieve financing returns, financing projects not proceeding as planned, and financing funds not being used in financing projects; Production and operation risk dimension: imbalance between production and sales, excessively high product costs, lack of target customers, serious loss of major customers, and a significant decrease in demand for main products; Legal risk dimension: judgments on construction project contract disputes, contract disputes, labor disputes, sales contract disputes, copyright ownership and infringement disputes, online shopping contract disputes, bid rigging and unfair competition disputes, transportation contract disputes, computer software development contract disputes, and patent infringement disputes.

[0076] A support vector machine (SVM) model with recursive feature elimination is used to reduce the dimensionality of quantitative data to filter out important indicators. This includes: recursive feature elimination using an SVM-trained model initializes the previous indicator features, trains the model, removes features with the smallest absolute weight, and repeats this process until the desired number of features is reached. Since the desired number of features is manually defined, the final result may not be optimal. Therefore, cross-validation is performed to find the optimal number of features.

[0077] Figure 4 This demonstrates our processing operations for the indicator system table.

[0078] Based on the reduced-dimensionality indicator system table, the weights of each indicator are determined using the multi-level DeFiel method. Then, warning signals are set, and risk thresholds for each indicator are established according to industry standards. A risk assessment model is then built, including:

[0079] like Figure 5 The diagram illustrates the multi-layered DeFiel method. The first layer determines the weights of each indicator, i.e., the target layer. The second layer categorizes all indicators into qualitative and quantitative indicators. Quantitative indicators have five dimensions: operational risk, financial risk, related-party transaction risk, risk of funds raised for investment projects, and risk of external investment. Qualitative indicators have four dimensions: management team risk, financing project risk, production and operation risk, and legal risk. The third layer defines the dimensions to which each indicator belongs. The fourth layer represents the specific indicator items. After the comparison matrix is ​​determined through expert evaluation, the weights of each indicator can be calculated using a MATLAB program.

[0080] The multi-layered DeFiel method module comprises four layers. The first layer determines the weight of each indicator, i.e., the target layer. The second layer categorizes all indicators into qualitative or quantitative indicators. The quantitative indicators have the following five dimensions: operational risk, capital risk, related party misappropriation risk, risk of raising funds for investment projects, and external investment risk. The qualitative indicators have the following four dimensions: management team risk, financing project risk, production and operation risk, and legal risk. The third layer specifies the dimensions to which each indicator belongs. The fourth layer contains the specific indicator items.

[0081] The weights of each indicator are as follows: Operational Risk Dimension: Main Business Revenue Profit Margin: 0.0310, Operating Profit Ratio: 0.0386, Main Business Revenue Growth Rate: 0.0281, Accounts Receivable Turnover: 0.0219, Return on Equity: 0.0403, Cash Flow: 0.0428; Funding Risk Dimension: Debt-to-Asset Ratio: 0.0642, Current Ratio: 0.0502, Quick Ratio: 0.0488, Inventory Turnover: 0.0153, Profit Cash Ratio: 0.0271, Mandatory Cash Payment Ratio: 0.0248, Related Party Occupation Risk... Risk warning indicators, related party asset occupancy rate: 0.0110, related business revenue ratio: 0.0275; Fundraising investment project dimension: input-output ratio: 0.0282, project investment progress completion rate: 0.0469; External investment risk warning indicators, long-term equity investment ratio: 0.0438, investment return rate: 0.1094; Management team risk dimension: change of legal representative: 0.0208, change of key personnel: 0.0080, change of major shareholder / controlling shareholder: 0.0181, suspected embezzlement, fraud, etc.: 0.0122, change of management personnel Frequent activity: 0.0058; Financing project risk dimension: Deteriorating prospects of financing projects: 0.0287, Financing projects are unlikely to be completed on time: 0.0134, Failure to achieve financing returns: 0.0337, Financing projects not proceeding as planned: 0.0118; Production and operation risk dimension: Imbalance between production and sales: 0.0237, Excessively high product costs: 0.0145, Lack of target customers: 0.0202, Severe loss of major customers: 0.0225, Significant decrease in demand for main products: 0.0243; Legal risk dimension: Existence of construction engineering construction risks. Judgments concerning labor contract disputes: 0.0032; Judgments concerning contract disputes: 0.0032; Judgments concerning labor disputes: 0.0032; Judgments concerning sales contract disputes; Judgments concerning copyright ownership and infringement disputes: 0.0032; Judgments concerning online shopping contract disputes: 0.0032; Judgments concerning bid rigging and unfair competition disputes: 0.0032; Judgments concerning transportation contract disputes: 0.0032; Judgments concerning computer software development contract disputes: 0.0032; Judgments concerning infringement of invention patent rights: 0.0032.

[0082] Warning signals and risk thresholds need to be set by taking into account industry standards, requirements of financing guarantee companies, and other factors.

[0083] The values ​​of various indicators are obtained through user input, and the risk value of each business process is calculated based on the risk assessment model, including:

[0084] like Figure 6As shown, the risk assessment model comprises three main modules: risk identification, risk analysis, and risk measurement. Risk identification involves defining evaluation indicators and setting warning signals. Risk analysis primarily studies how to quantitatively or qualitatively evaluate warning signals and generate corresponding risk assessment results. Risk measurement studies how to comprehensively assess the risk evaluation results across various business dimensions and obtain the final evaluation result. Quantitative analysis results are directly incorporated into risk measurement. For qualitative risk measurement results, a multi-level analysis method is used to quantitatively analyze non-quantitative events and transform qualitative analysis into weights for final risk measurement. The risk assessment module is extracted from the indicator system and business objects and linked to guarantee projects, thus achieving a reasonable risk control effect.

[0085] The indicator data is input into a random forest model to predict credit ratings and determine the penalty factors tied to the credit ratings, including:

[0086] The random forest model needs to be trained before the indicator data is input. It uses the SME listing data in the Guotai An database as the dataset for training to predict credit ratings. The credit rating is divided into 18 levels (referencing industry standards) and corresponds to different penalty factors.

[0087] The result of multiplying the risk value by the penalty factor is input into the early warning mechanism. The early warning mechanism monitors the system through early warning tasks, and after obtaining the risk result, triggers early warning monitoring according to the early warning rules, and displays the early warning, including:

[0088] The risk value calculated by the risk assessment module is multiplied by the corresponding penalty factor to obtain the risk result, which is then input into the early warning mechanism. For example... Figure 7 As shown, the early warning mechanism comprises three parts: early warning task definition, early warning task monitoring, and early warning result display. The early warning mechanism monitors tasks by defining alarm signals associated with them, triggers early warning rules through data perception, pushes early warning messages to relevant business processes, and displays the early warning information or generates relevant early warning reports.

[0089] Based on a comprehensive analysis of the entire business process in the financing guarantee sector, this invention utilizes methodologies such as requirements engineering and software engineering, and technologies such as workflow engines and rule engines to research and design risk control solutions and systems for the financing guarantee industry. The specific technical approach of this invention is as follows: Figure 8 As shown, the scheme is divided into three parts: data collection, risk assessment, and risk warning. The research content of the three parts is integrated and mutually supportive.

[0090] In the data collection section, the focus is on the risk assessment indicator system. Workflow and data sensing technologies are used to collect relevant business data from the relevant process links of the business system. After data processing, quantitative and qualitative source data required for risk control assessment are formed.

[0091] In the risk assessment section, the focus is on completing the risk evaluation of business source data. The specific implementation steps are as follows: the risk identification module completes the definition of alarm signals, the risk analysis module completes the risk analysis of alarm signals, the risk assessment module completes the risk measurement, and the measurement results are pushed to the risk early warning and control.

[0092] The risk warning section focuses on warning rules. The warning definition clarifies the monitoring task, and by initiating task detection, it completes risk monitoring of relevant warning signal points. Based on message mechanisms and rule engine technology, it evaluates the risk assessment and measurement results of warning signals, judges the warning rules, and displays the risk assessment results and warning judgment results.

[0093] Figure 9 This invention demonstrates the business logic for the target field: from the initial customer application to the processing of related business, the account manager is responsible. Our risk assessment model and mechanism transform relevant business processing information and push it to the risk control manager's dashboard. Post-insurance information is pushed to the early warning monitoring dashboard. The risk control manager processes related business and provides feedback. The general manager supervises and holds the account manager and risk control manager accountable, and is open to arbitration requests. After the project successfully completes the post-insurance supervision phase, it enters the project closure phase and is archived.

[0094] In summary, this invention discloses a risk control solution and system for the financing guarantee industry, addressing issues such as poor risk control capabilities, strong subjectivity in risk assessment, and inefficient business processing. First, this invention constructs an industry evaluation index system table. An improved risk matrix method and fuzzy comprehensive evaluation method are used to reduce the dimensionality of qualitative indicators, and a support vector machine model with recursive feature elimination is used to reduce the dimensionality of quantitative data to select important indicators. Then, based on the dimensionality-reduced index system table, the weight of each indicator is determined using the multi-level DeFilh method, and alarm signals are set. Risk thresholds for each indicator are set according to industry standards to establish a risk assessment model. Next, the values ​​of each indicator are obtained through user input, thereby calculating the risk value of each business link. Subsequently, the index data is input into a random forest model for credit rating prediction, with the credit rating bound to a penalty factor. Finally, the result of multiplying the risk value by the penalty factor is input into an early warning mechanism. The early warning mechanism monitors the system through early warning tasks, and after obtaining the risk result, triggers early warning monitoring according to early warning rules and displays the early warning. This invention proposes a complete risk control model and system, which can effectively improve the risk control capabilities of guarantee companies; by decoupling risk identification, risk analysis, and risk assessment, it ensures the scalability of risk control methods; by setting alarm signals as interfaces for external models, it ensures the scalability of risk control methods; and by dynamically binding or associating with relevant business objects and indicator systems, it can meet the needs of dynamic application evolution and reduce application system development costs.

Claims

1. A risk control system for the financing guarantee industry, characterized in that, It includes the following modules connected in sequence: industry evaluation indicator system table, risk assessment model, risk value, penalty factor, and early warning mechanism. The industry evaluation index system module uses an improved risk matrix method module and a fuzzy comprehensive evaluation method module to reduce the dimensionality of qualitative indicators, and uses a support vector machine model with recursive feature elimination to reduce the dimensionality of quantitative data in order to screen out important indicators. The risk assessment model module is used to determine the weight of each indicator based on the dimensionality-reduced indicator system table, using the multi-level DeFell method module, then setting alarm signals, and setting the risk threshold of each indicator according to industry standards. The risk value module is used to obtain the values ​​of various indicators through user input and to calculate the risk value of each business process based on the risk assessment model module. The penalty factor module is used to input indicator data into the random forest model module to predict the credit rating and determine the penalty factor bound to the credit rating. The early warning mechanism module is used to input the result of multiplying the risk value module and the penalty factor module into the early warning mechanism, monitor the system through early warning tasks, and trigger early warning monitoring according to the early warning rules after obtaining the risk result, and display the early warning. The improved risk matrix method module adds a prevention probability dimension to the classic risk matrix method, and sets three levels of high, medium and low for each dimension. The DeFell method is used to evaluate the high, medium and low levels one by one for each of the three dimensions. The fuzzy comprehensive evaluation method module further evaluates and identifies the key risks. The identified key risks are: management team risk dimension, financing project risk dimension, production and operation risk dimension, and legal risk dimension.

2. The risk control system for the financing guarantee industry according to claim 1, characterized in that, The recursive feature elimination support vector machine model includes: initializing the previous index features using support vector machine as the training model, training the model with SVM, deleting the features with the smallest absolute value weight, and then repeating the above process until the required number of features is reached.

3. The risk control system for the financing guarantee industry according to claim 1, characterized in that, The multi-layered DeFiel method module comprises four layers. The first layer determines the weight of each indicator, i.e., the target layer. The second layer categorizes all indicators into qualitative or quantitative indicators. The quantitative indicators have the following five dimensions: operational risk, capital risk, related party misappropriation risk, risk of raising funds for investment projects, and external investment risk. The qualitative indicators have the following four dimensions: management team risk, financing project risk, production and operation risk, and legal risk. The third layer specifies the dimensions to which each indicator belongs. The fourth layer contains the specific indicator items.

4. A risk control system for the financing guarantee industry according to claim 1, characterized in that, The risk assessment model module includes three main modules: risk identification, risk analysis, and risk measurement. The risk identification module is used to define evaluation indicators and set alarm signals. The risk analysis module is used to study how to quantitatively or qualitatively evaluate alarm signals and generate corresponding risk assessment results. The risk measurement module is used to study how to comprehensively evaluate the risk assessment results across various business dimensions and obtain the final evaluation results. Quantitative analysis results are directly incorporated into risk measurement. For qualitative risk assessment results, a multi-level analysis method is used to perform quantitative analysis on non-quantitative events and convert qualitative analysis into weights for final risk measurement.

5. A risk control system for the financing guarantee industry according to claim 1, characterized in that, The random forest model module needs to be trained before the indicator data is input. It uses the SME listing data in the Guotai An database as the dataset for training to achieve the prediction of credit rating. The credit rating is divided into 18 levels (referencing industry standards) and corresponds to different penalty factors.

6. A risk control system for the financing guarantee industry according to claim 1, characterized in that, The early warning mechanism module includes three parts: early warning task definition, early warning task monitoring, and early warning result display. The early warning mechanism monitors the task by defining the alarm signals associated with the early warning task, triggers early warning rules through data perception, pushes early warning messages to the corresponding business links, and displays the early warning information or generates relevant early warning reports.

7. A method for a risk control system in the financing guarantee industry, characterized in that, Includes the following steps: Step 1: Construct an industry evaluation index system table module. Use an improved risk matrix method module and a fuzzy comprehensive evaluation method module to reduce the dimensionality of qualitative indicators, and use a support vector machine model with recursive feature elimination to reduce the dimensionality of quantitative data in order to screen out important indicators. In the dimensionality reduction algorithm of qualitative indicators, an improved risk matrix method and a fuzzy comprehensive evaluation method are adopted. The improved risk matrix method adds the dimension of prevention probability to the classic risk matrix method. Three levels of high, medium and low are set under each dimension. The DeFell method is used to evaluate the high, medium and low levels one by one for the three dimensions. Then, the fuzzy comprehensive evaluation method is used for further evaluation to determine the key risks. Step 2: Based on the reduced-dimensionality indicator system table, the weight of each indicator is determined using the multi-level DeFilh method module. Then, alarm signals are set, and the risk thresholds of each indicator are set according to industry standards to establish a risk assessment model module. Step 3: Obtain the values ​​of each indicator from the user input method, calculate the risk value of each business link according to the risk assessment model module, and construct the risk value module; Step 4: Input the indicator data into the random forest model module to predict the credit rating and determine the credit rating binding penalty factor module; Step 5: Input the result of multiplying the risk value module and the penalty factor module into the early warning mechanism module. The early warning mechanism module monitors the system through early warning tasks, obtains the risk result, triggers early warning monitoring according to the early warning rules, and displays the early warning.

8. The method as described in claim 7, characterized in that, Step 1 specifically includes: The indicator system is established using a multi-dimensional mechanism. Through enterprise profiling, the basic situation of the enterprise and the business information of the enterprise in various dimensions are defined and extracted as indicator features. During the extraction process, the indicator features can be dynamically expanded. Improved risk matrix method and fuzzy comprehensive evaluation method, specifically including: Step 2-1 Let the set of evaluation dimensions be U = {probability of occurrence, severity of consequences, and possibility of prevention}; Step 2-2: Let the set of comments be V = {High, Medium, Low}; Step 2-3: Establish a fuzzy comprehensive risk assessment matrix; Step 2-4: Use the Defield method to determine the weight coefficient vector of the evaluation indicators; Steps 2-5 determine the fuzzy synthesis operator; Steps 2-6 analyze the fuzzy comprehensive evaluation results to identify key risks; The recursive feature elimination support vector machine model initializes the previous index features, trains the model using SVM, deletes the features with the smallest absolute weight, and then repeats the above process until the required number of features is reached.

9. The method as described in claim 8, characterized in that, In step 2-6, when analyzing the fuzzy comprehensive evaluation results, based on the "maximum membership principle", a "high" value greater than or equal to 50% can be identified as a significant risk. The four significant risks identified are: management team risk, financing project risk, production and operation risk, and legal risk.

Images