Unhealthy asset evaluation system and method

Abstract

The invention discloses a system and a method for evaluating unhealthy assets, and relates to the technical field of asset evaluation, and the method comprises the following steps: firstly, collecting various financial data of borrowers and related parties through a big data analysis tool, and building a multi-dimensional data platform containing the comprehensive financial conditions of the borrowers; and according to the collected financial data, an asset and liability map of the borrower and the associated party thereof is constructed, and financial health conditions of the borrower and the associated party thereof are deeply known. Through combination of a big data analysis tool and a machine learning model, financial data of borrowers and related parties thereof can be comprehensively collected and analyzed, a clear asset and liability map is constructed, and potential hidden debt risks can be deeply mined. Through an intelligent risk prediction and guarantee supplement mechanism, a high-risk borrower can be automatically identified and guarantee requirements can be dynamically adjusted, so that potential loss of a loan institution is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of asset evaluation, and in particular to a system and method for evaluating non-performing assets. Background Art

[0002] The non-performing asset assessment system is a tool used to assess and manage the risks of non-performing assets (such as overdue loans, bad debts, low-value assets, etc.) in enterprises or financial institutions. The system combines data analysis, machine learning and financial models to comprehensively assess the value, risk level, and recovery possibility of non-performing assets. By collecting and processing multi-dimensional information such as historical data, market conditions, and economic environment, the non-performing asset assessment system can provide accurate decision-making support for management, help formulate asset disposal strategies, reduce losses, and optimize capital allocation. At the same time, the system can monitor and warn non-performing assets in real time to ensure timely detection of potential risks, thereby effectively avoiding possible financial crises in the future.

[0003] In the process of overdue loan assessment, although common assessment factors (such as borrower's credit status and collateral value) have been widely paid attention to, there are still some uncommon problems that may lead to serious consequences if they occur.

[0004] Specifically, the financial statements submitted by the borrower may show that its debt level is low, but in fact, due to its complex business structure (such as financing through affiliated companies, concealing debts or falsely reporting assets), its actual debt far exceeds that shown in the statements. Such behavior is often difficult to detect in a timely manner in conventional financial analysis, especially when the borrower involves a complex capital structure. If the assessment system fails to identify these concealed key information, the lender may overestimate the borrower's repayment ability, which in turn leads to a serious misjudgment of the risk of non-performing loans.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The purpose of the present invention is to provide a non-performing asset assessment system and method, which utilizes the combination of big data analysis tools and machine learning models to not only comprehensively collect and analyze the financial data of borrowers and their affiliates, build a clear asset-liability map, but also deeply explore potential hidden debt risks. Through intelligent risk prediction and collateral supplementation mechanisms, high-risk borrowers can be automatically identified and collateral requirements can be dynamically adjusted, thereby reducing potential losses for lending institutions. This method improves the accuracy and real-time nature of risk assessment, helps to achieve more scientific and comprehensive asset management, and reduces the risk of non-performing loans caused by hidden debts, so as to solve the problems in the above-mentioned background technology.

[0007] In order to achieve the above object, the present invention provides the following technical solution: a method for evaluating non-performing assets, comprising the following steps: First, we use big data analysis tools to collect various financial data of borrowers and their affiliates, and establish a multi-dimensional data platform that includes the borrowers' comprehensive financial status; Based on the collected financial data, we build an asset-liability map of the borrower and its affiliates to gain an in-depth understanding of the financial health of the borrower and its affiliates; Through in-depth mining of the asset-liability map, key data reflecting the existence of hidden debts of borrowers can be extracted, and the extracted data can be deeply analyzed to provide a decision-making basis for subsequent intelligent prediction of debt risks; Apply the trained machine learning model to new borrower data to conduct intelligent risk prediction, analyze the borrower’s financial status, and assess their hidden debt risk; When it is identified that the borrower has high-risk hidden debt, the intelligent collateral supplement mechanism is automatically activated. According to the prediction results, the borrower is required to provide additional assets as collateral, and the face value of the required collateral is calculated based on the risk level of the borrower's hidden debt.

[0008] Preferably, various financial data of the borrower and its affiliates are obtained through banks, tax authorities, credit rating agencies and third-party data providers.

[0009] Preferably, the specific steps of constructing the asset-liability map of the borrower and its related parties include: First, the collected financial data is preprocessed and cleaned to ensure the accuracy and consistency of the data; Then, based on the financial information of the borrower and its affiliates, the relationships between different businesses and legal entities are established; Next, the key financial indicators of each enterprise are integrated to build a multi-dimensional financial data model to show the overall liabilities, asset status and potential capital flow of the borrower and its related parties; Finally, by combining market information, industry data and macroeconomic factors, we can supplement the potential external environmental risks faced by borrowers and form a three-dimensional and comprehensive asset-liability map.

[0010] Preferably, key data reflecting the existence of hidden debts of the borrower are extracted, wherein the extracted key features include the degree of deviation between assets and liabilities in the balance sheet of the borrower and its affiliates and the degree of mismatch between cash inflows and outflows of the borrower and its affiliates. The extracted data are deeply analyzed to generate debt-asset deviation reference values ​​and cash flow mismatch reference values, respectively. The debt-asset deviation reference values ​​and cash flow mismatch reference values ​​provide a decision-making basis for subsequent intelligent prediction of debt risks.

[0011] Preferably, when the trained machine learning model is applied to new borrower data, the model uses the debt-asset deviation reference value and the cash flow mismatch reference value as input features, generates a debt risk index through calculation, and comprehensively evaluates the borrower's financial status by analyzing the debt risk index to accurately identify its hidden debt risks.

[0012] Preferably, the specific steps of conducting an in-depth analysis of the deviation between assets and liabilities in the balance sheet of the borrower and its affiliates and generating a reference value for the debt-asset deviation are as follows: The generation logic of the debt-asset deviation reference value is based on the proportion of various assets and liabilities in the balance sheet and the capital structure and capital flow pattern of the enterprise. The specific expression is: , in: is the debt-asset deviation from the reference value, is the total assets, is the total liabilities, indicating the total amount in the balance sheet, and are the values ​​of each asset category and liability category, is the weight coefficient of the asset or liability category, The number of asset and liability categories involved in the balance sheet.

[0013] Preferably, the specific steps of conducting an in-depth analysis of the mismatch between the cash inflows and outflows of the borrower and its related parties to generate a cash flow mismatch reference value are as follows: First, the cash flow matching function is defined by comprehensively considering the relationship between the cash inflow and outflow of the borrower and its related parties, focusing on the timing, amount difference and fund use of the two, and generating a cash flow mismatch reference value. The generated expression is: , in: Indicates that the cash flow does not match the reference value, Indicates k The cash inflow during the period, Indicates k The cash outflow during the period, It is the capital utilization ratio coefficient, which is used to adjust the matching degree between cash outflow and cash inflow. is the mismatch penalty index, controlling the sensitivity of cash flow mismatch, is the smoothing coefficient that adjusts the outflow amount for the mismatch effect, is the cash outflow impact attenuation index, is a very small value, ensuring that the function can still be calculated stably when the cash inflow is zero.

[0014] Preferably, the debt risk index generated when the trained machine learning model evaluates the borrower's financial situation is compared and analyzed with a pre-set debt risk index threshold to evaluate the borrower's financial situation. Specifically, if the debt risk index is greater than the debt risk index threshold, it indicates that the borrower has high-risk hidden debt; if the debt risk index is less than or equal to the debt risk index threshold, it indicates that the borrower's hidden debt is normal.

[0015] Preferably, the borrower is required to provide additional assets as collateral based on the prediction results, and the face value of the required collateral is calculated based on the risk level of the borrower's hidden debt. The specific steps are as follows: When a borrower is identified to have high-risk hidden debts, the smart collateral supplement mechanism is automatically activated, requiring the borrower to provide additional collateral to cover potential risks. The required collateral face value is calculated using the following formula: , the calculation expression is: , in: It is the debt risk index of the borrower, reflecting the current implicit debt risk of the borrower. It is a preset debt risk index threshold used to distinguish normal and high-risk borrowers. and are adjustment coefficients, which respectively control the impact of the debt risk index and the borrower's asset value on the demand for additional collateral, is the estimated total value of the borrower's existing assets, It is a nonlinear parameter used to adjust the additional par value of collateral; After the system starts the collateral supplement mechanism, the face value of the collateral is further adjusted according to the risk level of the borrower's implicit debt. For extremely high-risk borrowers, a larger amount of collateral is required. The specific adjustment expression is: , in: It is an adjustment factor, which is used to amplify or reduce the impact of risk level on the adjustment of the face value of collateral. is the nonlinear adjustment coefficient of risk level, It is the adjusted face value of the collateral, reflecting the dynamic adjustment results of the borrower's implicit debt risk.

[0016] A non-performing asset evaluation system, including a data collection and platform construction unit, an asset-liability map construction unit, a hidden debt data mining and analysis unit, an intelligent risk prediction unit, and an intelligent collateral supplementation mechanism unit; The data collection and platform construction unit first uses big data analysis tools to collect various financial data of borrowers and their affiliates, and establishes a multi-dimensional data platform that includes the borrowers' comprehensive financial status; The asset-liability map construction unit constructs an asset-liability map of the borrower and its affiliates based on the collected financial data, and provides an in-depth understanding of the financial health of the borrower and its affiliates; The hidden debt data mining and analysis unit extracts key data reflecting the existence of hidden debts of borrowers through in-depth mining of the asset-liability map, and conducts in-depth analysis on the extracted data to provide a decision-making basis for subsequent intelligent prediction of debt risks; Intelligent risk prediction unit, which applies the trained machine learning model to new borrower data to perform intelligent risk prediction, analyze the borrower's financial status, and assess their hidden debt risk; The intelligent collateral supplement mechanism unit automatically activates the intelligent collateral supplement mechanism when it is identified that the borrower has high-risk hidden debts. It requires the borrower to provide additional assets as collateral based on the prediction results, and calculates the face value of the required collateral based on the risk level of the borrower's hidden debts.

[0017] In the above technical solution, the technical effects and advantages provided by the present invention are: The present invention uses a combination of big data analysis tools and machine learning models to not only comprehensively collect and analyze the financial data of borrowers and their affiliates, build a clear asset-liability map, but also deeply explore potential hidden debt risks. Through intelligent risk prediction and collateral supplementation mechanisms, high-risk borrowers can be automatically identified and collateral requirements can be dynamically adjusted, thereby reducing potential losses for lending institutions. This method improves the accuracy and real-time nature of risk assessment, helps to achieve more scientific and comprehensive asset management, and reduces the risk of non-performing loans caused by hidden debts. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0019] Figure 1 The present invention is a method flow chart of a non-performing asset evaluation method.

[0020] Figure 2 The module diagram of a non-performing asset evaluation system of the present invention is shown in FIG. DETAILED DESCRIPTION

[0021] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0022] The present invention provides Figure 1 A method for evaluating non-performing assets is shown, comprising the following steps: First, we use big data analysis tools to collect various financial data of borrowers and their affiliates, and establish a multi-dimensional data platform that includes the borrowers' comprehensive financial status; The various financial data of the borrower and its related parties include balance sheets, income statements, cash flow statements, etc. These data can come from banks, tax authorities, credit rating agencies and other third-party data providers. In addition, it is also necessary to collect external data related to the borrower, such as market information, industry data and changes in the macroeconomic environment. These data can be connected to the internal database of the financial institution through interfaces or APIs.

[0023] By collecting a large amount of financial data and external environment data, we lay the foundation for subsequent analysis. The key to this step is to ensure the comprehensiveness and integrity of the data, provide an accurate data source for the subsequent construction of the asset-liability map, and avoid missing key factors that may affect the assessment.

[0024] Third-party data providers usually include a variety of institutions and platforms that provide financial institutions with external data on borrowers' financial status, credit records, asset information, etc. These providers may include commercial credit rating companies (such as Dun & Bradstreet, Standard & Poor's, Moody's, etc.), financial information service providers (such as Bloomberg, Thomson Reuters, etc.), government regulatory agencies (such as tax return data provided by tax authorities), courts and legal institutions (providing bankruptcy records, litigation records, etc.), and professional market research companies (providing industry, market trends and company background analysis, etc.). These third-party data provide banks and financial institutions with a broader perspective of information, which helps to fully understand the financial health of borrowers and their related parties.

[0025] Based on the collected financial data, we build an asset-liability map of the borrower and its affiliates to gain an in-depth understanding of the financial health of the borrower and its affiliates; The map not only shows the basic financial indicators of the borrower, such as the total debt and asset size, but also reflects the details of the capital flow, loan structure, capital structure, etc. between different enterprises. This map can help financial institutions gain an in-depth understanding of the financial health of the borrower and its related parties and identify possible hidden debts. The asset-liability map is a comprehensive visualization of the financial status of the borrower and its related companies, which provides strong support for the assessment of hidden debts. In this step, the system will integrate data from multiple dimensions and clearly present information such as hidden debts, related transactions, and cross-company capital flows, thereby providing clear clues for subsequent analysis.

[0026] The specific steps to construct the asset-liability map of the borrower and its affiliates include: first, pre-processing and cleaning the collected financial data to ensure the accuracy and consistency of the data. Then, based on the financial information of the borrower and its affiliates, establish the relationship between different enterprises and legal entities, and identify their parent-subsidiary, joint venture, investment relationship and other structures. Next, integrate the key financial indicators of each enterprise, such as the balance sheet, income statement, cash flow statement, etc., to build a multi-dimensional financial data model to show the overall liabilities, asset status and potential capital flow of the borrower and its affiliates. In addition, it is necessary to combine market information, industry data and macroeconomic factors to supplement the external environmental risks that the borrower may face, so as to form a three-dimensional and comprehensive asset-liability map. This map will help assess the financial health and potential risks of the borrower and its affiliates, and provide a basis for subsequent risk assessment and decision-making.

[0027] Through in-depth mining of the asset-liability map, key data reflecting the existence of hidden debts of borrowers can be extracted, and the extracted data can be deeply analyzed to provide a decision-making basis for subsequent intelligent prediction of debt risks; Extract key data that reflects the existence of hidden debts of the borrower, among which the extracted key features include the degree of deviation between assets and liabilities in the balance sheet of the borrower and its affiliates and the degree of mismatch between cash inflows and outflows of the borrower and its affiliates. Conduct in-depth analysis on the extracted data to generate debt-asset deviation reference values ​​and cash flow mismatch reference values ​​respectively. The debt-asset deviation reference values ​​and cash flow mismatch reference values ​​provide a decision-making basis for subsequent intelligent prediction of debt risks.

[0028] A large deviation between assets and liabilities in the balance sheet of a borrower and its affiliates indicates that the borrower has hidden debt risk. Generally, the debt level of an enterprise should match its asset size, and liabilities should be supported by assets that the enterprise can reasonably repay. However, if the balance sheet of an enterprise shows that the asset size is large but the liabilities are low, or the liabilities increase significantly while the asset growth lags, this may mean that the enterprise has not fully disclosed its actual liabilities in the financial statements. For example, the enterprise may transfer part of its liabilities through affiliated companies or off-book financing, or use complex financial operations to hide part of the debt. This mismatch between assets and liabilities often indicates risks in the financial structure, which may cause the enterprise to face difficulties in repaying debts, thereby increasing the risk of non-performing loans. Therefore, a large deviation in the balance sheet can serve as an important signal for identifying hidden debt risks.

[0029] The specific steps for conducting an in-depth analysis of the deviation between assets and liabilities in the balance sheet of the borrower and its related parties and generating a reference value for debt-asset deviation are as follows: The debt-asset deviation reference value measures the abnormality of the structure by analyzing the relationship between liabilities and assets in the balance sheet of the borrower and its affiliates. The logic for generating the debt-asset deviation reference value is based on the proportion of various assets and liabilities in the balance sheet and the capital structure and capital flow pattern of the enterprise. The specific expression is: , in: is the debt-asset deviation from the reference value, is the total assets, is the total liabilities, indicating the total amount in the balance sheet, and are the values ​​of each asset category (such as cash, accounts receivable, inventory, etc.) and liability category (such as long-term debt, short-term debt, etc.), The weight coefficient is the weight factor of the asset or liability category, which is set based on historical data, industry standards or macroeconomic environment. The weight coefficient reflects the risk weight of different asset and liability categories. Some high-risk categories (such as non-realizable assets or cross-border debts) will have higher weights. The number of asset and liability categories involved in the balance sheet; It directly reflects the degree of deviation between assets and liabilities. The higher it is, the heavier the asset burden of the enterprise is and the liabilities are not adequately supported by assets. By weighting the risk of each asset and liability category, we further reveal risk factors that may be hidden, such as hiding debts through affiliated companies or disguising true liabilities through long-term non-performing assets.

[0030] This formula can be used to more comprehensively assess the potential hidden debt risks in the balance sheets of borrowers and their affiliates. The generated debt-asset deviation reference value can help identify companies that appear to have a low debt ratio but actually have a higher debt risk.

[0031] It can be seen from the debt-asset deviation reference value that the larger the performance value of the debt-asset deviation reference value generated by in-depth analysis of the deviation between assets and liabilities in the balance sheet of the borrower and its affiliates, the greater the deviation between the assets and liabilities of the borrower and its affiliates, and the higher the hidden debt risk. Specifically, the debt-asset deviation reference value identifies the debt that the borrower may hide through incomplete disclosure or off-book operations by measuring the relationship between assets and liabilities and the structure of various types of assets and liabilities. When the debt-asset deviation reference value is high, it means that the debt level shown on the balance sheet of the borrower is much lower than the actual debt, and there may be a situation of concealing debt through means such as affiliated companies and false reports, which indicates a higher financial risk and debt repayment pressure, and increases the possibility of non-performing loans. On the contrary, a lower debt-asset deviation reference value indicates that the balance sheet is more standardized, the possibility of hidden debt is smaller, and the financial health of the borrower is more stable. Therefore, the debt-asset deviation reference value can be used as an effective indicator for assessing hidden debt risks.

[0032] A large mismatch between the cash inflows and outflows of a borrower and its related parties usually indicates that the borrower may have hidden debt risks. Under normal circumstances, a borrower should be able to pay its debts and operating expenses through cash flow generated from operating income. However, if a borrower shows a significant cash flow mismatch, such as stable or growing cash inflows but a significant increase in liabilities or abnormal cash outflows, this may mean that the borrower is hiding debts through opaque means, such as liabilities not listed in the financial statements, off-balance sheet financing, or loans made through related parties. This mismatch may be the result of the borrower using inflated income, deferred liabilities, or debt transfer through financial engineering, which may eventually lead to cash flow shortages and increase the risk of default. Therefore, cash flow mismatch is an important signal for identifying hidden debt risks, and financial institutions need to pay special attention to this indicator when conducting risk assessments.

[0033] The specific steps for conducting an in-depth analysis of the mismatch between the cash inflows and outflows of the borrower and its related parties to generate a cash flow mismatch reference value are as follows: First, the cash flow matching function is defined by comprehensively considering the relationship between the cash inflow and outflow of the borrower and its related parties, focusing on the timing, amount difference and fund use of the two, and generating a cash flow mismatch reference value. The generated expression is: , in: Indicates that the cash flow does not match the reference value, Indicates k Cash inflows during the period (such as sales revenue, loan income, etc.), Indicates k cash outflows during the period (e.g. operating costs, interest payments, debt repayments, etc.), It is the capital utilization ratio coefficient, which is used to adjust the matching degree between cash outflow and cash inflow (reflecting the matching degree between capital flow and capital use). is the mismatch penalty index, which controls the sensitivity of cash flow mismatch. Generally, a larger value will aggravate the impact of mismatch. It is a smoothing coefficient that adjusts the impact of outflow funds on mismatch. A larger value emphasizes the impact of cash outflow on mismatch. is the cash outflow impact attenuation index. As the outflow amount increases, the attenuation increases (used to control the nonlinear response of the cash flow mismatch reference value when the outflow amount is large). It is a very small value used to avoid division by zero errors and ensure that the function can still calculate stably when the cash inflow is zero; By calculating the difference between cash inflow and outflow in each period , which can measure the matching degree between the cash inflow and outflow of the borrower or its related parties. The larger the difference value, the more serious the mismatch of cash flow. Sensitivity adjustment: using the mismatch penalty index The cash flow mismatch metric can be weighted to prevent a small mismatch from affecting a large result. Small mismatches will have a larger impact, thereby accurately identifying small fund flow anomalies. Nonlinear impact of outflows: Through the smoothing coefficient and cash outflow impact decay index , further considering the impact of cash outflow amount, especially when large outflow may have a more profound impact on financial status and debt repayment ability, and adjusting the cash flow mismatch reference value to enhance the mismatch response in the case of large outflow. Finally, the summed cash flow mismatch reference value represents the overall cash flow mismatch of the borrower and its related parties. The higher the cash flow mismatch reference value, the more serious the mismatch is, and the greater the potential hidden debt risk of the borrower.

[0034] It can be seen from the cash flow mismatch reference value that the larger the performance value of the cash flow mismatch reference value generated by in-depth analysis of the mismatch between the cash inflows and outflows of the borrower and its affiliates, the higher the mismatch between the cash inflows and outflows of the borrower and its affiliates, which usually means that the borrower has a higher hidden debt risk. Specifically, if the borrower shows obvious asymmetry or timing mismatch between cash inflows and outflows, it may indicate that it hides debts through off-book financing, false income or deferred liabilities, resulting in insufficient cash flow to cover the actual level of liabilities. This situation is often the manifestation of hidden debt risk. Therefore, a higher cash flow mismatch reference value indicates potential financial instability or risk, indicating that the borrower may face greater debt repayment difficulties. Conversely, it means that the borrower's financial situation is relatively healthy and the hidden debt risk is low.

[0035] Apply the trained machine learning model to new borrower data to conduct intelligent risk prediction, analyze the borrower’s financial status, and assess their hidden debt risk; When the trained machine learning model is applied to new borrower data, the model uses the debt-asset deviation reference value and the cash flow mismatch reference value as input features, and generates a debt risk index after calculation. By analyzing the debt risk index, the borrower's financial situation is comprehensively assessed and its hidden debt risks are accurately identified.

[0036] In the context of machine learning, a "trained machine learning model" means that after a series of data training processes, the model has the ability to extract patterns from the input data and make accurate predictions or classifications. The core of training is to help the machine learning algorithm learn the inherent patterns of the data through the relationship between the features in the historical data and the target variables. In this case, the trained machine learning model refers to the process of supervised learning through the existing borrower data (including financial indicators, historical debt records, etc.) and the annotation results of these data (such as the borrower's implicit debt risk level). Specifically, the machine learning algorithm uses a large amount of annotated historical data in the training stage. The algorithm completes the training process by continuously adjusting its own parameters to optimize its prediction ability on the given data.

[0037] For example, a machine learning model may use some common algorithms such as decision trees, support vector machines (SVMs), random forests, or deep neural networks (DNNs). During the training process, the algorithm continuously predicts the target variable (in this case, whether the borrower has hidden debt risk) through input features (such as debt-asset deviation from reference values, cash flow mismatch reference values, etc.). By training on a large amount of historical borrower data, the model can find out which features (such as different financial ratios, cash flow inflow and outflow mismatch, etc.) are associated with hidden debt risk, and the predictive power of these features for hidden debt risk. After training, the parameters of the model have been optimized and can predict new borrower data to analyze its debt risk. In other words, "training is complete" means that the model has learned from past data and can automatically identify hidden debt risks in similar new data. The training of the model is a process of continuously adjusting parameters and optimizing prediction accuracy until it reaches an acceptable prediction accuracy on a given data set. The trained model will become mature and able to automatically process new data, so as to conduct real-time hidden debt risk assessment in practical applications.

[0038] When the trained machine learning model is applied to new borrower data, the model first accepts new input features, in this case, the reference values ​​of debt-asset deviation and cash flow mismatch. These input features are calculated by the model to generate a comprehensive debt risk index. The debt risk index reflects the current implicit debt risk of the borrower and provides a basis for financial institutions to make further decisions. By using the optimized parameters obtained in the training phase, the model can accurately predict new data and determine whether the borrower has implicit debt risk.

[0039] Specifically, the debt-asset deviation reference value is used as one of the inputs to measure the degree of abnormality between the borrower's liabilities and assets. By calculating the deviation between assets and liabilities, the model can determine whether the borrower may conceal its actual debt situation by hiding debts or related transactions. Similarly, the cash flow mismatch reference value reflects the imbalance between the borrower's cash inflows and outflows. When the borrower raises funds through informal means or has a large amount of off-balance sheet liabilities, the difference between cash inflows and outflows will be very prominent. This mismatch may mean that the borrower has a high risk of hidden debt.

[0040] After taking the debt-asset deviation reference value and cash flow mismatch reference value as input features, the trained machine learning model will calculate the comprehensive score of the borrower's implicit debt risk, namely the debt risk index, through its learned parameters. The higher the value of the debt risk index, the more unstable the borrower's financial situation is and the greater the risk of implicit debt. Conversely, it means that the financial situation is relatively healthy. Ultimately, based on the debt risk index, financial institutions can make decisions such as whether to approve the loan, whether to adjust the loan amount, and whether to require additional guarantees.

[0041] Through the application of machine learning models, the process of borrower financial assessment has become more scientific and automated. Traditional financial assessment methods usually rely on manual analysis and empirical judgment, which are easily interfered by subjective factors and are also inefficient when processing large-scale data. Through automated data analysis and pattern recognition, machine learning models can not only greatly improve analysis efficiency, but also detect potential risk signals from massive data and reduce human omissions. In addition, with the continuous input of new data, the model can also be gradually updated and optimized to improve the accuracy and stability of predictions.

[0042] The machine learning model is not specifically limited here, and can achieve the deviation of debt-asset from the reference value. and cash flow does not match the reference value Conduct comprehensive analysis to generate a debt risk index In order to realize the technical solution of the present invention, the present invention provides a specific implementation method; debt risk index The generated expression is: , where Debt-asset deviation reference value and cash flow does not match the reference value The preset scaling factor of are all greater than 0. "Preset proportional coefficient" refers to the constant coefficient ( and ) is used to adjust different input characteristics (such as debt-asset deviation from the reference value and cash flow does not match the reference value ) for debt risk index These preset proportional coefficients reflect the weight of each feature in the final debt risk index. By presetting these proportional coefficients, the model can adjust the importance of input features according to actual conditions, so that the model can better adapt to actual risk assessment tasks. In the training stage of machine learning, these preset proportional coefficients are usually obtained by automatic optimization of historical data, but in the initial stage, a preliminary value is usually set based on prior knowledge or experience as the starting point for model training.

[0043] It can be seen from the debt risk index that the greater the performance value of the debt-asset deviation reference value generated by in-depth analysis of the deviation between assets and liabilities in the balance sheet of the borrower and its affiliates, the greater the performance value of the cash flow mismatch reference value generated by in-depth analysis of the mismatch between the cash inflows and outflows of the borrower and its affiliates, that is, the greater the performance value of the debt risk index generated when the financial status of the borrower is evaluated by the trained machine learning model, the higher the implicit debt risk of the borrower, and vice versa.

[0044] The debt risk index generated when the trained machine learning model evaluates the borrower's financial situation is compared and analyzed with the pre-set debt risk index threshold to evaluate the borrower's financial situation. Specifically: if the debt risk index is greater than the debt risk index threshold, it indicates that the borrower has high-risk hidden debt; if the debt risk index is less than or equal to the debt risk index threshold, it indicates that the borrower's hidden debt is normal and the risk is low.

[0045] When a borrower is identified to have high-risk hidden debts, the smart collateral supplement mechanism is automatically activated. The borrower is required to provide additional assets as collateral based on the prediction results, and the face value of the required collateral is calculated based on the risk level of the borrower's hidden debts. Based on the prediction results, the borrower is required to provide additional assets as collateral, and the face value of the required collateral is calculated based on the risk level of the borrower's hidden debt. The specific steps are as follows: When a borrower is identified to have high-risk hidden debts, the smart collateral supplement mechanism is automatically activated, requiring the borrower to provide additional collateral to cover potential risks. The required collateral face value is calculated using the following formula: , the calculation expression is: , in: It is the debt risk index of the borrower, reflecting the current implicit debt risk of the borrower. It is a preset debt risk index threshold used to distinguish normal and high-risk borrowers. and are adjustment coefficients, which respectively control the impact of the debt risk index and the borrower's asset value on the demand for additional collateral, is the estimated total value of the borrower's existing assets, It is a nonlinear parameter used to adjust the face value of the collateral supplement to ensure that riskier borrowers provide more collateral; The purpose of this step is that when the debt risk index exceeds the debt risk index threshold, based on the borrower's implicit debt risk, the system automatically calculates the required additional collateral face value through a nonlinear model and reminds the borrower to provide corresponding assets as collateral to reduce the risk exposure of financial institutions.

[0046] After the system starts the collateral supplement mechanism, the face value of the collateral is further adjusted according to the risk level of the borrower's implicit debt (determined by the value of the debt risk index). For extremely high-risk borrowers (the debt risk index is significantly higher than the debt risk index threshold, or even several times the debt risk index, which is not specifically limited here and can be set according to actual conditions), a larger amount of collateral needs to be provided. The calculation formula is as follows: , in: It is an adjustment factor, which is used to amplify or reduce the impact of risk level on the adjustment of the face value of collateral. It is a nonlinear adjustment coefficient for risk level, ensuring that borrowers with higher risk levels provide more collateral. It is the adjusted face value of the collateral, reflecting the dynamic adjustment result of the borrower's implicit debt risk; The purpose of this step is to flexibly and dynamically adjust the required collateral face value according to the risk level of the borrower's hidden debt. If the borrower has a high hidden debt risk, the system will automatically increase the collateral requirement, thereby protecting the lending institution from losses caused by default of high-risk borrowers. At the same time, through a refined adjustment mechanism, it ensures that the collateral requirements match the actual risks of the borrower, thereby improving the accuracy and fairness of risk management.

[0047] Through the intelligent risk management mechanism, when the borrower is identified to have high-risk hidden debt, the borrower is automatically required to provide additional assets as collateral, thereby effectively reducing the potential loss risk of financial institutions. According to the risk level of the borrower's hidden debt, the system dynamically calculates and adjusts the face value of the required collateral to ensure that the value of the collateral is proportional to the borrower's financial risk. This not only enhances the risk prevention and control capabilities of lending institutions for high-risk borrowers, but also ensures that borrowers provide sufficient collateral, reduces the financial pressure that may be caused by default, and achieves more accurate and fair risk management.

[0048] The above-mentioned non-performing asset assessment method can effectively solve the problem of the difficulty in identifying the hidden debts of borrowers. By combining big data analysis tools and machine learning models, it is possible to not only comprehensively collect and analyze the financial data of borrowers and their affiliates, build a clear asset-liability map, but also deeply explore potential hidden debt risks. Through intelligent risk prediction and collateral supplementation mechanisms, the system can automatically identify high-risk borrowers and dynamically adjust collateral requirements, thereby reducing potential losses for lending institutions. This method improves the accuracy and real-time nature of risk assessment, helps to achieve more scientific and comprehensive asset management, and reduces the risk of non-performing loans caused by hidden debts.

[0049] The present invention provides Figure 2 A non-performing asset evaluation system shown includes a data collection and platform construction unit, an asset-liability map construction unit, a hidden debt data mining and analysis unit, an intelligent risk prediction unit, and an intelligent collateral supplement mechanism unit; The data collection and platform construction unit first uses big data analysis tools to collect various financial data of borrowers and their affiliates, and establishes a multi-dimensional data platform that includes the borrowers' comprehensive financial status; The asset-liability map construction unit constructs an asset-liability map of the borrower and its affiliates based on the collected financial data, and provides an in-depth understanding of the financial health of the borrower and its affiliates; The hidden debt data mining and analysis unit extracts key data reflecting the existence of hidden debts of borrowers through in-depth mining of the asset-liability map, and conducts in-depth analysis on the extracted data to provide a decision-making basis for subsequent intelligent prediction of debt risks; Intelligent risk prediction unit, which applies the trained machine learning model to new borrower data to perform intelligent risk prediction, analyze the borrower's financial status, and assess their hidden debt risk; The intelligent collateral supplement mechanism unit automatically activates the intelligent collateral supplement mechanism when it is identified that the borrower has high-risk hidden debts. It requires the borrower to provide additional assets as collateral based on the prediction results, and calculates the face value of the required collateral based on the risk level of the borrower's hidden debts.

[0050] A non-performing asset evaluation method provided in an embodiment of the present invention is implemented by the above-mentioned non-performing asset evaluation system. The specific method and process of a non-performing asset evaluation system are detailed in the embodiment of the above-mentioned non-performing asset evaluation method, which will not be repeated here.

[0051] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0052] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0053] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for evaluating non-performing assets, characterized in that: The following steps are involved: First, we use big data analysis tools to collect various financial data of borrowers and their affiliates, and establish a multi-dimensional data platform that includes the borrowers' comprehensive financial status; Based on the collected financial data, we build an asset-liability map of the borrower and its affiliates to gain an in-depth understanding of the financial health of the borrower and its affiliates; Through in-depth mining of the asset-liability map, key data reflecting the existence of hidden debts of borrowers can be extracted, and the extracted data can be deeply analyzed to provide a decision-making basis for subsequent intelligent prediction of debt risks; Apply the trained machine learning model to new borrower data to conduct intelligent risk prediction, analyze the borrower’s financial status, and assess their hidden debt risk; When it is identified that the borrower has high-risk hidden debt, the intelligent collateral supplement mechanism is automatically activated. According to the prediction results, the borrower is required to provide additional assets as collateral, and the face value of the required collateral is calculated based on the risk level of the borrower's hidden debt.

2. A method for evaluating non-performing assets according to claim 1, characterized in that: Various financial data of borrowers and their affiliates are obtained through banks, tax authorities, credit rating agencies and third-party data providers.

3. A method for evaluating non-performing assets according to claim 1, characterized in that: The specific steps to construct the asset-liability map of the borrower and its related parties include: First, the collected financial data is preprocessed and cleaned to ensure the accuracy and consistency of the data; Then, based on the financial information of the borrower and its affiliates, the relationships between different businesses and legal entities are established; Next, the key financial indicators of each enterprise are integrated to build a multi-dimensional financial data model to show the overall liabilities, asset status and potential capital flow of the borrower and its related parties; Finally, by combining market information, industry data and macroeconomic factors, we can supplement the potential external environmental risks faced by borrowers and form a three-dimensional and comprehensive asset-liability map.

4. A method for evaluating non-performing assets according to claim 1, characterized in that: Extract key data that reflects the existence of hidden debts of the borrower, among which the extracted key features include the degree of deviation between assets and liabilities in the balance sheet of the borrower and its affiliates and the degree of mismatch between cash inflows and outflows of the borrower and its affiliates. Conduct in-depth analysis on the extracted data to generate debt-asset deviation reference values ​​and cash flow mismatch reference values ​​respectively. The debt-asset deviation reference values ​​and cash flow mismatch reference values ​​provide a decision-making basis for subsequent intelligent prediction of debt risks.

5. A method for evaluating non-performing assets according to claim 4, characterized in that: When the trained machine learning model is applied to new borrower data, the model uses the debt-asset deviation reference value and the cash flow mismatch reference value as input features, and generates a debt risk index after calculation. By analyzing the debt risk index, the borrower's financial status is comprehensively assessed and its hidden debt risks are accurately identified.

6. A method for evaluating non-performing assets according to claim 4, characterized in that: The specific steps for conducting an in-depth analysis of the deviation between assets and liabilities in the balance sheet of the borrower and its affiliates and generating a reference value for debt-asset deviation are as follows: The generation logic of the debt-asset deviation reference value is based on the proportion of various assets and liabilities in the balance sheet and the capital structure and capital flow pattern of the enterprise. The specific expression is: ， in: is the debt-asset deviation from the reference value, is the total assets, is the total liabilities, indicating the total amount in the balance sheet, and are the values ​​of each asset category and liability category, is the weight coefficient of the asset or liability category, The number of asset and liability categories involved in the balance sheet.

7. A method for evaluating non-performing assets according to claim 4, characterized in that: The specific steps for conducting an in-depth analysis of the mismatch between the cash inflows and outflows of the borrower and its related parties and generating a cash flow mismatch reference value are as follows: First, the cash flow matching function is defined by comprehensively considering the relationship between the cash inflow and outflow of the borrower and its related parties, focusing on the timing, amount difference and fund use of the two, and generating a cash flow mismatch reference value. The generated expression is: ， in: Indicates that the cash flow does not match the reference value, Indicates k The cash inflow during the period, Indicates k The cash outflow during the period, It is the capital utilization ratio coefficient, which is used to adjust the matching degree between cash outflow and cash inflow. is the mismatch penalty index, controlling the sensitivity of cash flow mismatch, is the smoothing coefficient that adjusts the outflow amount for the mismatch effect, is the cash outflow impact attenuation index, is a very small value, ensuring that the function can still be calculated stably when the cash inflow is zero.

8. A method for evaluating non-performing assets according to claim 5, characterized in that: The debt risk index generated when evaluating the borrower's financial situation through the trained machine learning model is compared and analyzed with the pre-set debt risk index threshold to evaluate the borrower's financial situation. Specifically: if the debt risk index is greater than the debt risk index threshold, it indicates that the borrower has high-risk hidden debt; if the debt risk index is less than or equal to the debt risk index threshold, it indicates that the borrower's hidden debt is normal.

9. A method for evaluating non-performing assets according to claim 8, characterized in that: Based on the prediction results, the borrower is required to provide additional assets as collateral, and the face value of the required collateral is calculated based on the risk level of the borrower's hidden debt. The specific steps are as follows: When a borrower is identified to have high-risk hidden debts, the smart collateral supplement mechanism is automatically activated, requiring the borrower to provide additional collateral to cover potential risks. The required collateral face value is calculated using the following formula: , the calculation expression is: ， in: It is the debt risk index of the borrower, reflecting the current implicit debt risk of the borrower. It is a preset debt risk index threshold used to distinguish normal and high-risk borrowers. and are adjustment coefficients, which respectively control the impact of the debt risk index and the borrower's asset value on the demand for additional collateral, is the estimated total value of the borrower's existing assets, It is a nonlinear parameter used to adjust the additional par value of collateral; After the system starts the collateral supplement mechanism, the face value of the collateral is further adjusted according to the risk level of the borrower's implicit debt. For extremely high-risk borrowers, a larger amount of collateral is required. The specific adjustment expression is: ， in: It is an adjustment factor, which is used to amplify or reduce the impact of risk level on the adjustment of the face value of collateral. is the nonlinear adjustment coefficient of risk level, It is the adjusted face value of the collateral, reflecting the dynamic adjustment results of the borrower's implicit debt risk.

10. A non-performing asset evaluation system, used to implement the non-performing asset evaluation method according to any one of claims 1 to 9, characterized in that: It includes data collection and platform construction unit, asset-liability map construction unit, implicit debt data mining and analysis unit, intelligent risk prediction unit and intelligent collateral supplement mechanism unit; The data collection and platform construction unit first uses big data analysis tools to collect various financial data of borrowers and their affiliates, and establishes a multi-dimensional data platform that includes the borrowers' comprehensive financial status; The asset-liability map construction unit constructs an asset-liability map of the borrower and its affiliates based on the collected financial data, and provides an in-depth understanding of the financial health of the borrower and its affiliates; The hidden debt data mining and analysis unit extracts key data reflecting the existence of hidden debts of borrowers through in-depth mining of the asset-liability map, and conducts in-depth analysis on the extracted data to provide a decision-making basis for subsequent intelligent prediction of debt risks; Intelligent risk prediction unit, which applies the trained machine learning model to new borrower data to perform intelligent risk prediction, analyze the borrower's financial status, and assess their hidden debt risk; The intelligent collateral supplement mechanism unit automatically activates the intelligent collateral supplement mechanism when it is identified that the borrower has high-risk hidden debts. It requires the borrower to provide additional assets as collateral based on the prediction results, and calculates the face value of the required collateral based on the risk level of the borrower's hidden debts.

Images