A loan risk assessment method and system based on artificial intelligence

Abstract

The present invention relates to the field of credit risk technology, specifically to a loan risk assessment method and system based on artificial intelligence, including the following steps: collecting historical loan records, financial statements, bank flow and market behavior data of an enterprise, integrating the collected data, analyzing capital flow and debt repayment ability, and evaluating the financial status of the enterprise to generate a financial health index. In the present invention, the introduction of the financial health index improves the accuracy and efficiency of preliminary credit review, reduces the consumption of resources, and can instantly reflect the impact of market fluctuations on corporate credit by dynamically analyzing the correlation between lending behavior and market changes, providing a more forward-looking risk assessment. The application of product matching index information optimizes the loan product recommendation process, ensuring that financial products are highly consistent with corporate needs, thereby improving customer satisfaction and the effective delivery of loan products. The accuracy of risk level classification makes loan decisions more scientific.

Description

Technical Field

[0001] The present invention relates to the field of credit risk technology, and in particular to an artificial intelligence-based loan risk assessment method and system. Background Art

[0002] The field of credit risk technology involves the use of different methods and models to predict the likelihood of loan defaults and their financial consequences. In this field, financial institutions and other lending entities use various statistical, machine learning and artificial intelligence algorithms to assess the credit status of borrowers, including factors such as credit history, repayment ability and economic conditions, to help financial institutions make more informed judgments when approving loans and pricing loan terms, thereby reducing default rates and potential financial losses.

[0003] Among them, the loan risk assessment method is a systematic process for assessing the borrower's default risk. This process determines the borrower's credit rating by analyzing the borrower's credit record, financial status, repayment ability and other relevant factors. Its main purpose is to assess potential risks before the loan is issued to decide whether to approve the loan and determine appropriate loan terms and interest rates. This method enables financial institutions to effectively control and mitigate financial losses caused by loan defaults, thereby improving overall asset quality and stakeholder trust.

[0004] Traditional methods are inadequate in real-time data processing and dynamic risk prediction, and cannot effectively reflect immediate market changes, increasing the risk of misjudgment. The lack of a flexible loan product matching mechanism is also a major shortcoming of traditional methods. In the traditional evaluation process, product recommendations fail to fully consider the needs of the enterprise and its latest financial status, resulting in inefficient resource allocation, idle funds or excessive debt. Traditional methods are limited in their ability to optimize loan structures and dynamically adjust risk levels, limiting the flexibility and initiative of financial institutions in strategy adjustments and risk management, resulting in untimely responses to economic changes and increased risks of financial losses. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a loan risk assessment method and system based on artificial intelligence.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme: a loan risk assessment method based on artificial intelligence, comprising the following steps:

[0007] S1: Collect the company's historical loan records, financial statements, bank statements and market behavior data, integrate the collected data, analyze capital liquidity and debt repayment ability, and evaluate the company's financial status to generate a financial health index;

[0008] S2: Based on the financial health index, extract credit risk factors, quantitatively analyze the correlation between lending behavior and market changes, calculate the enterprise's default probability through risk factors, and obtain a default probability score;

[0009] S3: Based on the default probability score, using the long short-term memory network and the weighted Mahalanobis distance algorithm, identify the loan product with the best matching degree, analyze the corresponding relationship between product conditions and enterprise needs, and obtain product matching index information;

[0010] S4: Based on the product matching index information, a decision tree algorithm and a gated recurrent unit are used to perform a correlation analysis between the credit rating and the default risk, and a risk parameter is quantified for each credit level to obtain a risk level classification result;

[0011] S5: Based on the risk level classification results, evaluate the current loan structure, identify potential financing gaps and optimization points through industry comparative analysis, and generate loan structure optimization records;

[0012] S6: Based on the loan structure optimization record, using the historical data of financial performance, dynamic risk assessment is performed to predict the trend of asset and liability changes and obtain dynamic monitoring results of loan risks.

[0013] The present invention has the following improvements: the financial health index includes financial structure indicators, debt repayment ability indicators, operating ability indicators, profitability indicators, and development ability indicators; the default probability score includes credit score, financial status change information, and market behavior analysis results; the product matching index information includes loan type, guarantee measures, loan term, and interest rate options; the risk level classification results include default risk rating, expected loss rate, and credit recovery potential; the loan structure optimization record includes adjusted loan ratios, risk diversification strategies, and debt restructuring plans; the loan risk dynamic monitoring results include adjusted default probability forecasts, market volatility responses, and risk mitigation measures update results.

[0014] The present invention has the following improvements: collecting the historical loan records, financial statements, bank statements and market behavior data of the enterprise, integrating the collected data, analyzing the liquidity and debt repayment ability of the funds, and evaluating the financial status of the enterprise. The specific steps of generating the financial health index are:

[0015] S101: Collect historical loan records, financial statements, bank statements and market behavior data of enterprises, and perform data cleaning, including consistency checking, removing duplicates, correcting incorrect formats and filling missing values, to generate an integrated data set;

[0016] S102: Based on the integrated data set, analyzing the cash inflow and outflow patterns of the enterprise, evaluating the monthly and annual fluctuations in capital flows, and determining the stability and solvency of capital flows to obtain a capital liquidity assessment result;

[0017] S103: Based on the liquidity assessment results, the financial health of the enterprise is assessed, and industry standardized ratios are used to calculate key financial ratios, including debt-to-asset ratio and rate of return, to obtain a financial health index.

[0018] The present invention is improved in that, based on the financial health index, credit risk factors are extracted, the correlation between lending behavior and market changes is quantitatively analyzed, and the default probability of the enterprise is calculated through the risk factors. The specific steps of obtaining the default probability score are:

[0019] S201: Based on the financial health index, identify risk factors associated with the enterprise's default probability, including debt level and cash flow fluctuations, collect factors and construct a risk analysis framework to generate a risk analysis infrastructure;

[0020] S202: Based on the risk analysis infrastructure, analyze the correlation between lending behavior and market dynamics, extract market sensitivity indicators by identifying the impact of changes in market conditions on lending behavior, and obtain market impact assessment results;

[0021] S203: Based on the market impact assessment results, a weight allocation method is used to integrate risk factors and market dynamic data, calculate the enterprise's probability of default under current market conditions, and obtain a default probability score.

[0022] The present invention is improved in that, based on the default probability score, a long short-term memory network and a weighted Mahalanobis distance algorithm are used to identify the loan product with the best matching degree, and the corresponding relationship between product conditions and enterprise needs is analyzed. The steps of obtaining product matching index information are specifically as follows:

[0023] S301: Based on the default probability score, a long short-term memory network is used to screen financial indicators, including profit margin and cash flow, and compare them with the characteristics of existing loan products to identify the matching degree between the loan products and the financial status of the enterprise, and generate product matching efficiency analysis information;

[0024] S302: Based on the product matching efficiency analysis information, a weighted Mahalanobis distance algorithm is used to compare the loan conditions of each product, including the repayment period and interest rate, with the capital demand and repayment capacity of the enterprise, evaluate the matching effectiveness, and obtain the loan suitability verification result;

[0025] S303: Based on the loan suitability verification results, the financial stability score and growth potential assessment of the enterprise are integrated, the suitability and potential risks of each recommended product are listed, and product matching index information is generated.

[0026] The present invention is improved in that the weighted Mahalanobis distance algorithm is according to the formula:

[0027]

[0028] Calculate the weighted Mahalanobis distance D between loan products and corporate financial needs WM , where x is the target enterprise financial demand vector, y is the loan product feature vector, S is the covariance matrix of the overall enterprise financial demand, and W is the weight matrix.

[0029] The present invention is improved in that, based on the product matching index information, a decision tree algorithm and a gated recurrent unit are used to perform a correlation analysis between credit rating and default risk, and a risk parameter is quantified for each credit level. The steps of obtaining a risk level classification result are specifically as follows:

[0030] S401: Based on the product matching index information, using a decision tree algorithm, collect historical default data of enterprises of multiple credit ratings, analyze the risk patterns of the data, identify risk factors that are sensitive to credit ratings, and generate a credit risk factor set;

[0031] S402: Based on the credit risk factor set, a gated recurrent unit is applied to compare and analyze the risk exposure under multiple credit levels, set a risk threshold for each level, and evaluate potential credit changes according to the threshold to obtain a credit risk quantification model;

[0032] S403: Based on the credit risk quantification model, the enterprise is classified into risk levels, including low risk, medium risk and high risk, and matching loan conditions and interest rates are specified for each level to generate risk level classification results.

[0033] The present invention is improved in that, based on the risk level classification result, the current loan structure is evaluated, and potential financing gaps and optimization points are identified through industry comparative analysis. The steps of generating a loan structure optimization record are specifically as follows:

[0034] S501: Based on the risk level classification result, compare the consistency of multiple loan conditions of the enterprise with the risk level, check the coverage and restrictions of loan products, identify areas with excessive risk concentration, and generate a preliminary analysis record of the loan structure;

[0035] S502: Based on the preliminary analysis record of the loan structure, using the industry average data as a benchmark, assess the degree of deviation between the enterprise's loan structure and the industry average, determine the financing gap and structural deficiencies, and obtain the financing gap and optimization plan;

[0036] S503: Based on the financing gap and optimization plan, formulate optimization measures, including loan condition adjustment, introduction of new loan products or modification of existing product conditions, match the needs and risk tolerance of the enterprise, and generate loan structure optimization records.

[0037] The present invention is improved in that, based on the loan structure optimization record, using the historical data of financial performance, dynamic risk assessment is performed to predict the trend of asset-liability changes, and the steps of obtaining the dynamic monitoring results of loan risks are specifically as follows:

[0038] S601: Based on the loan structure optimization record, collect the enterprise's historical financial data and current market dynamic information, analyze the impact of the data on the enterprise's asset-liability ratio and cash flow, identify the trend of changes in financial performance, and generate market and financial dynamic analysis information;

[0039] S602: Based on the market and financial dynamic analysis information, evaluate the accuracy of the asset-liability forecasting model, adjust the risk parameters according to market dynamics, optimize the responsiveness of the model, and obtain a risk assessment model update record;

[0040] S603: Update records based on the risk assessment model, conduct risk monitoring, reflect changes in market and financial conditions in real time, optimize the real-time nature of risk warning and management measures, and generate dynamic monitoring results for loan risks.

[0041] A loan risk assessment system based on artificial intelligence, the system comprising:

[0042] The data integration module integrates the company's historical loan records, financial statements, bank statements and market behavior data, analyzes capital flow and debt repayment ability, evaluates the company's financial status, and generates a financial health index;

[0043] The credit risk assessment module extracts credit risk factors based on the financial health index, quantitatively analyzes the correlation between lending behavior and market changes, and calculates the enterprise default probability to obtain a default probability score;

[0044] The product matching analysis module evaluates the financial stability and growth potential of the enterprise based on the default probability score, identifies and matches the optimal loan product, and analyzes the correspondence between product conditions and enterprise needs to obtain product matching index information;

[0045] The risk level classification module uses a decision tree algorithm and a gated recurrent unit to perform a correlation analysis between the credit level and the default risk based on the product matching index information, quantify the risk parameters of each credit level, divide the enterprise risk level, and obtain a risk level classification result;

[0046] The loan structure optimization module evaluates the current loan structure based on the risk level classification results, identifies potential financing gaps and optimization points in industry comparisons, quantifies deficiencies in the loan structure, updates the risk assessment model and matches market changes, and generates loan structure optimization records.

[0047] Compared with the prior art, the advantages and positive effects of the present invention are:

[0048] In the present invention, the introduction of the financial health index improves the accuracy and efficiency of preliminary credit review and reduces the consumption of resources. By dynamically analyzing the correlation between lending behavior and market changes, it can immediately reflect the impact of market fluctuations on corporate credit and provide a more forward-looking risk assessment. The application of product matching index information optimizes the loan product recommendation process and ensures that financial products are highly consistent with corporate needs, thereby improving customer satisfaction and the effective placement of loan products. The accuracy of risk level classification makes loan decisions more scientific. The optimization of loan structure further enhances the efficiency of capital allocation and the initiative of risk management, and enhances the adaptability and risk prevention and control capabilities of financial institutions in a constantly changing market. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 A flowchart of a loan risk assessment method based on artificial intelligence is proposed for the present invention;

[0050] Figure 2 The present invention proposes a detailed flow chart of step S1 in a loan risk assessment method based on artificial intelligence;

[0051] Figure 3 The present invention proposes a detailed flow chart of step S2 in a loan risk assessment method based on artificial intelligence;

[0052] Figure 4 The present invention proposes a detailed flow chart of step S3 in a loan risk assessment method based on artificial intelligence;

[0053] Figure 5 The present invention proposes a detailed flow chart of step S4 in a loan risk assessment method based on artificial intelligence;

[0054] Figure 6 The present invention proposes a detailed flow chart of step S5 in a loan risk assessment method based on artificial intelligence;

[0055] Figure 7 The present invention proposes a detailed flow chart of step S6 in a loan risk assessment method based on artificial intelligence;

[0056] Figure 8 A module diagram of a loan risk assessment system based on artificial intelligence is proposed for the present invention. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0058] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating positions or positional relationships, are based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0059] Example

[0060] See also Figure 1 The present invention provides a technical solution: a loan risk assessment method based on artificial intelligence, comprising the following steps:

[0061] S1: Collect the company's historical loan records, financial statements, bank statements and market behavior data, integrate the collected data, analyze capital liquidity and debt repayment ability, and evaluate the company's financial status to generate a financial health index;

[0062] S2: Based on the financial health index, extract credit risk factors, quantitatively analyze the correlation between lending behavior and market changes, calculate the company's default probability through risk factors, and obtain a default probability score;

[0063] S3: Based on the default probability score, using the long short-term memory network and weighted Mahalanobis distance algorithm, the loan product with the best matching degree is identified through the financial stability and growth potential of the enterprise, and the corresponding relationship between product conditions and enterprise needs is analyzed to obtain product matching index information;

[0064] S4: Based on the product matching index information, the decision tree algorithm and gated recurrent unit are used to analyze the correlation between credit rating and default risk, quantify the risk parameters for each credit level, and divide the enterprise risk level to obtain the risk level classification result;

[0065] S5: Based on the risk level classification results, evaluate the current loan structure, identify potential financing gaps and optimization points through industry comparative analysis, quantify deficiencies and potential improvement points in the loan structure, and generate loan structure optimization records;

[0066] S6: Based on the loan structure optimization records and using the historical data of financial performance, dynamic risk assessment is conducted to predict the trend of asset and liability changes, update the risk assessment model and match the market changes to obtain dynamic monitoring results of loan risks.

[0067] The financial health index includes financial structure indicators, debt-paying ability indicators, operating ability indicators, profitability indicators, and development ability indicators. The default probability score includes credit rating, financial status change information, and market behavior analysis results. Product matching index information includes loan type, loan term, and interest rate options. Risk level classification results include default risk rating, expected loss rate, and credit recovery potential. Loan structure optimization records include adjusted loan ratios, risk diversification strategies, and debt restructuring plans. The dynamic monitoring results of loan risks include adjusted default probability forecasts, market volatility responses, and risk mitigation measures update results.

[0068] See also Figure 2 , collect the company's historical loan records, financial statements, bank statements and market behavior data, integrate the collected data, analyze capital liquidity and debt repayment ability, and evaluate the company's financial status. The specific steps for generating a financial health index are:

[0069] S101: Collect historical loan records, financial statements, bank statements and market behavior data of enterprises, and perform data cleaning, including consistency checking, removing duplicates, correcting incorrect formats and filling missing values, to generate an integrated data set;

[0070] S101: First, classify and organize the company's historical loan records, financial statements, bank statements, and market behavior data, and conduct a detailed review of each type of data to ensure the integrity and accuracy of the data. For example: conduct a data integrity check on loan records to ensure that there are no missing internal or external debts; in financial statements, check the data consistency between the balance sheet and the income statement to ensure the authenticity and reliability of the report data; conduct a time continuity check on bank statements to ensure that there are no missing time periods; for market behavior data, check and verify the frequency and volume of transactions to ensure that the data reflects the actual market situation. After data cleaning is completed, merge the data into an integrated data set.

[0071] S102: Based on the integrated data set, analyze the cash inflow and outflow patterns of the enterprise, evaluate the monthly and annual fluctuations in cash flows, and determine the stability and solvency of cash flows to obtain the cash liquidity assessment results;

[0072] S102: Utilize the integrated data set to analyze the company's cash inflows and outflows in detail, monitor the cyclical changes in cash flows, such as seasonal and monthly fluctuations, and evaluate the stability of cash flows through monitoring, with special attention paid to the maximum peak and minimum trough of cash flows at specific time points (such as the end of a quarter or the end of a year). Based on the analysis, calculate and evaluate the company's debt repayment capacity, that is, the company's ability to pay current debts in various time periods. By simulating the changes in cash flows under different economic environments, predict the company's financial performance and risks under different circumstances.

[0073] S103: Based on the results of the liquidity assessment, evaluate the financial health of the enterprise, use industry standardized ratios, calculate key financial ratios, including debt-to-asset ratio and rate of return, and obtain the financial health index;

[0074] S103: Comprehensively use the commonly used financial ratio analysis methods in the industry, such as debt-to-asset ratio and rate of return analysis, to measure the financial health of the company, combining the company's profitability (i.e. the ratio of the company's total revenue to total assets) with its debt repayment ability (i.e. the ratio of the company's total assets to total liabilities) to form a comprehensive financial health index. The index can clearly show the company's financial status in the same industry and help the company understand its competitive position in the market and its long-term development potential.

[0075] See also Figure 3 Based on the financial health index, we extract credit risk factors, quantitatively analyze the correlation between lending behavior and market changes, and calculate the default probability of the enterprise through risk factors. The specific steps to obtain the default probability score are as follows:

[0076] S201: Based on the financial health index, identify risk factors associated with the probability of corporate default, including debt levels and income fluctuations, collect factors and build a risk analysis framework to generate a risk analysis infrastructure;

[0077] S201: Using the financial health index as a basis, key risk factors directly related to corporate default behavior are identified. Debt levels and income fluctuations are of particular concern because they directly affect the company's debt repayment ability and operating stability. To this end, relevant data is collected, such as historical debt records, seasonal and cyclical fluctuations in income, and combined with the results of the financial health index to construct a multi-dimensional risk analysis framework. This framework includes indicators such as debt ratios and cash flow volatility. Through this structure, the company's financial risks can be systematically analyzed and predicted.

[0078] S202: Based on the risk analysis infrastructure, analyze the correlation between lending behavior and market dynamics, identify the impact of changes in market conditions on lending behavior, extract market sensitivity indicators, and obtain market impact assessment results;

[0079] S202: Through the risk analysis infrastructure, we conduct in-depth research on the interaction between market dynamics and corporate lending behavior, pay attention to the direct impact of market economic conditions such as interest rate changes, market demand fluctuations and macroeconomic policy adjustments on corporate lending behavior, analyze how factors affect the financial status and borrowing needs of enterprises, and extract market-sensitive indicators, such as the impact of market fluctuations on corporate cash flow and debt repayment ability. In this process, data analysis is used to reveal the specific impact of changes in market conditions on corporate lending behavior, and a detailed report on market impact assessment is formed.

[0080] S203: Based on the market impact assessment results, the risk factors and market dynamics data are integrated using a weight allocation method to calculate the enterprise's probability of default under current market conditions and obtain a probability of default score;

[0081] S203: Comprehensively utilize market impact assessment results and risk factors, and measure the contribution of each factor to the company's default probability by setting the weight of each factor. The weight allocation takes into account the sensitivity and actual impact of the factors, such as the different influences of debt levels and market fluctuations. Using this weighted method, combined with the company's actual market performance and financial status, an accurate default probability is calculated. The probability score reflects the default risk faced by the company in a given market and economic environment, providing an important risk assessment basis for creditors and investors.

[0082] See also Figure 4 Based on the default probability score, using the long short-term memory network and the weighted Mahalanobis distance algorithm, the loan product with the best matching degree is identified through the financial stability and growth potential of the enterprise, and the corresponding relationship between the product conditions and the needs of the enterprise is analyzed. The specific steps for obtaining product matching index information are as follows:

[0083] S301: Based on the default probability score, long short-term memory network is used to screen financial indicators, including profit margin and cash flow, and compare them with the characteristics of existing loan products to identify the matching degree between loan products and corporate financial status, and generate product matching efficiency analysis information;

[0084] S301: By using long short-term memory networks to deeply analyze the company's financial data, especially profit margins and cash flows, the indicators reflect the profitability and liquidity of the company, and compare the financial indicators with the characteristics of existing loan products on the market, such as the product's funding limit, risk level and target customer group. Through this comparison, it is possible to identify which loan products' financial requirements are most consistent with the company's actual situation, and then evaluate the degree of match between each product and the company's financial status. This process generates a comprehensive analysis report on the efficiency of loan product matching by collecting and analyzing a large number of data points.

[0085] S302: Based on the product matching efficiency analysis information, the weighted Mahalanobis distance algorithm is used to compare the loan conditions of each product, including the repayment period and interest rate, with the capital demand and repayment capacity of the enterprise, evaluate the matching effectiveness, and obtain the loan suitability verification result;

[0086] S302: Use the weighted Mahalanobis distance algorithm to compare the specific conditions of different loan products, such as repayment period and interest rate, and conduct matching analysis with the company's capital needs and repayment ability. By calculating the distance between the characteristics of different loan products and the company's financial needs, find the closest loan options to ensure that the recommended loan products not only meet the financial indicators, but also match the company's payment ability and funding arrangements in terms of repayment conditions. The final result is a detailed loan suitability assessment result to help companies clarify which loan conditions are most consistent with their actual situation.

[0087] Weighted Mahalanobis distance algorithm, according to the formula:

[0088]

[0089] Calculate the weighted Mahalanobis distance D between loan products and corporate financial needs WM , where x is the target enterprise financial demand vector, y is the loan product feature vector, S is the covariance matrix of the overall enterprise financial demand, and W is the weight matrix.

[0090] Execution process:

[0091] First, calculate the covariance matrix S and its inverse matrix S of the overall enterprise financial demand -1 Then, the importance of each loan condition and enterprise demand is determined. To this end, the weight matrix W of each feature is obtained through historical data analysis. Then, the difference (xy) and its transposition (xy) between the target enterprise financial demand vector x and the feature vector y of the loan product are calculated. T , again, transpose the difference between the target enterprise's financial demand vector x and the loan product's characteristic vector y (xy) T , weight matrix W, inverse matrix S of covariance matrix -1 , multiply the target enterprise's financial demand vector x and the characteristic vector y of the loan product, and finally, take the square root to get the weighted Mahalanobis distance, which is used to determine the loan product that is closest to the enterprise's needs.

[0092] S303: Based on the loan suitability check results, the financial stability score and growth potential assessment of the enterprise are integrated, the suitability and potential risks of each recommended product are listed, and product matching index information is generated;

[0093] S303: Integrate the enterprise's financial stability score and its growth potential assessment. The score reflects the enterprise's long-term development prospects and short-term financial health status. Based on the information, the matching degree and potential risks faced by each recommended loan product are listed in detail. The matching index information will show the advantages and disadvantages of each loan product in detail, allowing the enterprise to make wise loan decisions based on a comprehensive understanding of various options. This process ensures that the loan product selected by the enterprise best meets its financial needs, while taking into account its long-term growth potential and the risks brought about by market changes.

[0094] See also Figure 5 Based on the product matching index information, the decision tree algorithm and gated recurrent unit are used to analyze the correlation between credit rating and default risk, quantify the risk parameters for each credit level, and divide the enterprise risk level. The specific steps to obtain the risk level classification result are as follows:

[0095] S401: Based on the product matching index information, a decision tree algorithm is used to collect historical default data of enterprises with multiple credit ratings, analyze the risk patterns of the data, identify risk factors that are sensitive to credit ratings, and generate a credit risk factor set;

[0096] S401: First, collect historical default data of enterprises with different credit ratings, including the frequency of default events, the amounts involved, and the recovery after default. Use the decision tree algorithm to analyze the data to identify and verify the most influential credit risk factors, such as debt levels, insufficient cash flow or declining profitability. The decision tree reveals the relationship between credit ratings and risk factors by constructing decision paths under different conditions, thereby forming a detailed set of credit risk factors.

[0097] S402: Based on the credit risk factor set, a gated recurrent unit is applied to compare and analyze the risk exposure under multiple credit levels, set a risk threshold for each level, and evaluate potential credit changes based on the threshold to obtain a credit risk quantification model;

[0098] S402: Use the gated recurrent unit (GRU) algorithm to process time series data, especially the changes in risk factors monitored at multiple time points. GRU effectively captures the dynamic changes of risk factors in different credit levels through its memory and update gate mechanism, so that the analysis can accurately reflect the true credit status of the enterprise in the economic cycle. Based on the analysis, a specific risk threshold is set for each credit level. The threshold can predict potential credit changes in the future. Finally, the threshold and analysis results are integrated into a credit risk quantification model, which can be used to monitor and predict changes in the credit status of the enterprise in real time.

[0099] S403: Based on the credit risk quantification model, the enterprise is classified into risk levels, including low risk, medium risk and high risk, and matching loan conditions and interest rates are specified for each level to generate risk level classification results;

[0100] S403: Based on the output of the credit risk quantification model, enterprises are classified into three levels: low risk, medium risk or high risk. Each risk level corresponds to different loan conditions and interest rates. The conditions and interest rates are carefully designed based on the risk tolerance and debt repayment potential of the enterprise. Low-risk enterprises obtain more favorable interest rates and more flexible repayment conditions, while high-risk enterprises face higher interest rates and stricter loan conditions. The risk level classification results not only help lending institutions to allocate resources reasonably.

[0101] See also Figure 6 Based on the risk level classification results, the current loan structure is evaluated, potential financing gaps and optimization points are identified through industry comparative analysis, deficiencies and potential improvement points in the loan structure are quantified, and the specific steps for generating loan structure optimization records are as follows:

[0102] S501: Based on the risk level classification results, compare the consistency of the enterprise's multiple loan conditions and risk levels, check the coverage and restrictions of loan products, identify areas with excessive risk concentration, and generate preliminary analysis records of loan structure;

[0103] S501: First, analyze the matching degree between various loan products and enterprise risk levels, focus on examining whether the loan conditions are suitable for enterprises at all levels of risk, especially check whether high-risk loan products reasonably cover relevant enterprise groups. In addition, evaluate whether the distribution of loan products at different risk levels is balanced, identify risk concentration areas, such as certain high-risk loans concentrated in specific industries or enterprise sizes. Through analysis, generate a preliminary analysis record containing the current loan structure assessment and potential risk points.

[0104] S502: Based on the preliminary analysis records of the loan structure, using the industry average data as a benchmark, evaluate the degree of deviation between the enterprise's loan structure and the industry average, determine the financing gap and structural deficiencies, and obtain the financing gap and optimization plan;

[0105] S502: Compare the company's loan structure with the industry average and analyze the differences in loan types, conditions and risk tolerance, especially in terms of loan product diversity, the degree of relaxation of loan conditions and risk prevention measures. The comparison reveals the main gaps and shortcomings in the company's financing structure, such as capital shortages or over-reliance on a certain type of loan. Based on the analysis results, formulate targeted financing gap and structure optimization plans.

[0106] S503: Based on the financing gap and optimization plan, formulate optimization measures, including loan condition adjustment, introduction of new loan products or modification of existing product conditions, match the needs and risk tolerance of the enterprise, and generate loan structure optimization records;

[0107] S503: Based on the previous analysis results and optimization plan, implement specific optimization measures, including adjusting the interest rates and repayment terms of existing loan products, introducing new loan products that meet current market demand and corporate risk characteristics, or making necessary modifications to existing products to better meet the needs of companies with different risk levels. Through these measures, optimize the company's loan structure and improve its financing efficiency and risk management capabilities. The loan structure optimization record generated in this process records in detail each measure implemented and the expected results.

[0108] See also Figure 7 Based on the loan structure optimization records, using the historical data of financial performance, dynamic risk assessment is carried out to predict the trend of asset-liability changes, update the risk assessment model and match the market changes. The specific steps to obtain the dynamic monitoring results of loan risks are as follows:

[0109] S601: Based on the loan structure optimization records, collect the company's historical financial data and current market dynamics information, analyze the impact of the data on the company's asset-liability ratio and cash flow, identify the changing trends of financial performance, and generate market and financial dynamic analysis information;

[0110] S601: First, based on the optimized loan structure records, collect the company's historical financial data, including the balance sheet, cash flow statement and income statement of the past few years. At the same time, consider the current market dynamics, such as interest rate changes, economic growth rate, industry competition and other factors. By analyzing the data, evaluate the specific impact on the company's debt-to-asset ratio and cash flow, and identify the main changing trends in the company's financial performance. The information is integrated into a market and financial dynamics analysis report to provide real-time data support for risk assessment.

[0111] S602: Based on the market and financial dynamic analysis information, evaluate the accuracy of the asset-liability forecasting model, adjust the risk parameters according to market dynamics, optimize the responsiveness of the model, and obtain the risk assessment model update record;

[0112] S602: Use the collected market and financial dynamic analysis information to evaluate the existing asset-liability forecasting model, including verifying the model's forecasting accuracy and response speed to new data. According to current and predicted market conditions, adjust the key risk parameters in the model, such as adjusting the sensitivity to changes in the economic cycle and the impact of interest rate changes, to improve the model's responsiveness to market changes. The update records generated in this process explain the specific content of the model adjustment and the expected effect.

[0113] S603: Update records based on the risk assessment model, conduct risk monitoring, reflect changes in market and financial conditions in real time, optimize the real-time nature of risk warning and management measures, and generate dynamic monitoring results for loan risks;

[0114] S603: Use the updated risk assessment model to perform continuous risk monitoring, including real-time tracking of market changes and the company's financial status, such as changes in the debt-to-asset ratio, fluctuations in cash flow, etc., and automatically generate risk warnings based on set thresholds and alarm mechanisms to help companies identify and respond to potential financial risks in a timely manner. The final output of this process is the dynamic monitoring results of loan risks, which can effectively optimize the loan structure and alleviate potential financial pressure.

[0115] See also Figure 8 , a loan risk assessment system based on artificial intelligence, the system includes:

[0116] The data integration module integrates the company's historical loan records, financial statements, bank statements and market behavior data, analyzes capital flow and debt repayment ability, evaluates the company's financial status, and generates a financial health index;

[0117] The credit risk assessment module extracts credit risk factors based on the financial health index, quantitatively analyzes the correlation between lending behavior and market changes, and calculates the enterprise's default probability to obtain a default probability score;

[0118] The product matching analysis module evaluates the financial stability and growth potential of enterprises based on the default probability score, identifies and matches the best loan products, analyzes the correspondence between product conditions and enterprise needs, and obtains product matching index information;

[0119] The risk level classification module uses the decision tree algorithm and gated recurrent unit to analyze the correlation between credit level and default risk based on product matching index information, quantify the risk parameters of each credit level, divide the enterprise risk level, and obtain the risk level classification result;

[0120] The loan structure optimization module evaluates the current loan structure based on the risk level classification results, identifies potential financing gaps and optimization points in industry comparisons, quantifies deficiencies in the loan structure, updates the risk assessment model and matches market changes, and generates loan structure optimization records.

[0121] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A loan risk assessment method based on artificial intelligence, characterized in that: The following steps are involved: Collect the company's historical loan records, financial statements, bank statements and market behavior data, integrate the collected data, analyze capital liquidity and debt repayment ability, and evaluate the company's financial status to generate a financial health index; Based on the financial health index, credit risk factors are extracted, the correlation between lending behavior and market changes is quantitatively analyzed, and the default probability of the enterprise is calculated through the risk factors to obtain a default probability score; Based on the default probability score, using the long short-term memory network and the weighted Mahalanobis distance algorithm, identify the loan product with the best matching degree, analyze the corresponding relationship between product conditions and enterprise needs, and obtain product matching index information; Based on the product matching index information, a decision tree algorithm and a gated recurrent unit are used to perform a correlation analysis between credit rating and default risk, quantify risk parameters for each credit level, and obtain a risk level classification result; Based on the risk level classification results, evaluate the current loan structure, identify potential financing gaps and optimization points through industry comparative analysis, and generate loan structure optimization records; Based on the loan structure optimization records, using historical data of financial performance, dynamic risk assessment is performed to predict asset-liability change trends and obtain dynamic monitoring results of loan risks; Based on the default probability score, the long short-term memory network and the weighted Mahalanobis distance algorithm are used to identify the loan product with the best matching degree, and the corresponding relationship between product conditions and enterprise needs is analyzed. The specific steps for obtaining product matching index information are as follows: Based on the default probability score, a long short-term memory network is used to screen financial indicators, including profit margin and cash flow, and compare them with the characteristics of existing loan products to identify the degree of match between loan products and corporate financial conditions, and generate product matching efficiency analysis information; Based on the product matching efficiency analysis information, the weighted Mahalanobis distance algorithm is used to compare the loan conditions of each product, including the repayment period and interest rate, with the capital demand and repayment capacity of the enterprise, evaluate the matching effectiveness, and obtain the loan suitability verification result; Based on the loan suitability check results, the financial stability score and growth potential assessment of the enterprise are integrated, the suitability and potential risks of each recommended product are listed, and product matching index information is generated; Based on the product matching index information, a decision tree algorithm and a gated recurrent unit are used to perform a correlation analysis between credit rating and default risk, and a risk parameter is quantified for each credit level. The specific steps for obtaining the risk level classification result are as follows: Based on the product matching index information, a decision tree algorithm is used to collect historical default data of enterprises with multiple credit ratings, analyze the risk patterns of the data, identify risk factors that are sensitive to credit ratings, and generate a credit risk factor set; Based on the credit risk factor set, a gated recurrent unit is applied to compare and analyze the risk exposure under multiple credit levels, set a risk threshold for each level, and evaluate potential credit changes according to the threshold to obtain a credit risk quantification model; Based on the credit risk quantification model, the risk levels of enterprises are divided into low risk, medium risk and high risk levels, and matching loan conditions and interest rates are specified for each level to generate risk level classification results.

2. The artificial intelligence-based loan risk assessment method according to claim 1, characterized in that: The financial health index includes financial structure indicators, debt repayment ability indicators, operating ability indicators, profitability indicators, and development ability indicators. The default probability score includes credit score, financial status change information, and market behavior analysis results. The product matching index information includes loan type, guarantee measures, loan term, and interest rate options. The risk level classification results include default risk rating, expected loss rate, and credit recovery potential. The loan structure optimization record includes adjusted loan ratios, risk diversification strategies, and debt restructuring plans. The loan risk dynamic monitoring results include adjusted default probability forecasts, market volatility responses, and risk mitigation measures update results.

3. The artificial intelligence-based loan risk assessment method according to claim 1, characterized in that: Collect the company's historical loan records, financial statements, bank statements and market behavior data, integrate the collected data, analyze the liquidity and debt repayment ability, and evaluate the company's financial status. The specific steps to generate the financial health index are: Collect historical loan records, financial statements, bank statements and market behavior data of enterprises, and perform data cleaning, including consistency checking, removing duplicates, correcting incorrect formats and filling in missing values, to generate integrated data sets; Based on the integrated data set, analyze the cash inflow and outflow patterns of the enterprise, evaluate the monthly and annual fluctuations in cash flows, and determine the stability and solvency of cash flows to obtain cash liquidity assessment results; Based on the liquidity assessment results, the financial health of the enterprise is assessed, and industry standardized ratios are used to calculate key financial ratios, including debt-to-asset ratio and rate of return, to obtain a financial health index.

4. The artificial intelligence-based loan risk assessment method according to claim 1, characterized in that: Based on the financial health index, the credit risk factor is extracted, the correlation between lending behavior and market changes is quantitatively analyzed, and the default probability of the enterprise is calculated through the risk factor. The specific steps for obtaining the default probability score are as follows: Based on the financial health index, identify risk factors associated with the probability of corporate default, including debt levels and cash flow fluctuations, collect factors and construct a risk analysis framework to generate a risk analysis infrastructure; Based on the risk analysis infrastructure, the correlation between lending behavior and market dynamics is analyzed, and by identifying the impact of changes in market conditions on lending behavior, market sensitivity indicators are extracted to obtain market impact assessment results; Based on the market impact assessment results, a weight allocation method is used to integrate risk factors and market dynamics data, calculate the company's probability of default under current market conditions, and obtain a default probability score.

5. The artificial intelligence-based loan risk assessment method according to claim 1, characterized in that: The weighted Mahalanobis distance algorithm is based on the formula: Calculate the weighted Mahalanobis distance between loan products and corporate financial needs ,in, is the target enterprise's financial demand vector, is the loan product feature vector, is the covariance matrix of the overall enterprise financial demand, is the weight matrix.

6. The artificial intelligence-based loan risk assessment method according to claim 1, characterized in that: Based on the risk level classification results, the current loan structure is evaluated, and potential financing gaps and optimization points are identified through industry comparative analysis. The specific steps for generating loan structure optimization records are as follows: Based on the risk level classification results, compare the consistency of the enterprise's multiple loan conditions with the risk level, check the coverage and restrictions of loan products, identify areas with excessive risk concentration, and generate preliminary analysis records of loan structure; Based on the preliminary analysis records of the loan structure, using the industry average data as a benchmark, evaluate the degree of deviation of the enterprise's loan structure from the industry average, determine the financing gap and structural deficiencies, and obtain the financing gap and optimization plan; Based on the financing gap and optimization plan, we formulate optimization measures, including loan condition adjustment, introduction of new loan products or modification of existing product conditions, to match the needs and risk tolerance of the enterprise and generate loan structure optimization records.

7. The artificial intelligence-based loan risk assessment method according to claim 1, characterized in that: Based on the loan structure optimization record, using the historical data of financial performance, dynamic risk assessment is performed to predict the trend of asset-liability changes, and the steps to obtain the dynamic monitoring results of loan risks are as follows: Based on the loan structure optimization records, collect the company's historical financial data and current market dynamics information, analyze the impact of the data on the company's asset-liability ratio and cash flow, identify the changing trends of financial performance, and generate market and financial dynamics analysis information; Based on the market and financial dynamic analysis information, evaluate the accuracy of the asset-liability forecasting model, adjust risk parameters according to market dynamics, optimize the responsiveness of the model, and obtain risk assessment model update records; Based on the risk assessment model, records are updated to conduct risk monitoring, reflect changes in market and financial conditions in real time, optimize the real-time nature of risk warning and management measures, and generate dynamic monitoring results for loan risks.

8. A loan risk assessment system based on artificial intelligence, characterized in that: According to any one of claims 1 to 7, the artificial intelligence-based loan risk assessment method is implemented, and the system comprises: The data integration module integrates the company's historical loan records, financial statements, bank statements and market behavior data, analyzes capital flow and debt repayment ability, evaluates the company's financial status, and generates a financial health index; The credit risk assessment module extracts credit risk factors based on the financial health index, quantitatively analyzes the correlation between lending behavior and market changes, and calculates the enterprise default probability to obtain a default probability score; The product matching analysis module evaluates the financial stability and growth potential of the enterprise based on the default probability score, identifies and matches the optimal loan product, analyzes the correspondence between product conditions and enterprise needs, and obtains product matching index information; The risk level classification module uses a decision tree algorithm and a gated recurrent unit to perform a correlation analysis between the credit level and the default risk based on the product matching index information, quantify the risk parameters of each credit level, divide the enterprise risk level, and obtain a risk level classification result; The loan structure optimization module evaluates the current loan structure based on the risk level classification results, identifies potential financing gaps and optimization points in industry comparisons, quantifies deficiencies in the loan structure, updates the risk assessment model and matches market changes, and generates loan structure optimization records.

Images