Method and system for measuring and calculating enterprise pre-credit line based on industrial chain multi-dimensional characteristics
By building a multidimensional data set in the industrial chain and using a risk control model that integrates multidimensional features, the problem of difficulty in quantifying credit risks and lack of dynamic adaptability in the traditional credit assessment method is solved, and efficient and accurate credit risk prediction and credit line calculation are achieved.
Patent Information
- Application Number
- CN202411728069.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-28
- Publication Date
- 2025-04-11
AI Technical Summary
Traditional corporate credit assessment methods rely on static single financial data, making it difficult to fully quantify credit risks, and credit line calculations lack dynamic adaptability.
The enterprise pre-credit credit limit calculation method based on multi-dimensional characteristics of the industrial chain is adopted. By constructing a multi-dimensional data set of the industrial chain, combining key indicator data and key risk factor data, a risk control model integrating multi-dimensional characteristics is used for risk assessment, and the enterprise's pre-credit credit limit is calculated based on the industrial chain score and risk score.
It realizes efficient and accurate credit risk prediction and credit line calculation, which can comprehensively quantify the hierarchical relationships and market dynamics of the industrial chain, improves the accuracy of risk prediction and the scientificity of credit line, and is dynamically adaptable.
Smart Images
Figure CN119809790A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of artificial intelligence credit risk management, and in particular to a method and system for calculating an enterprise's pre-credit limit based on multi-dimensional characteristics of an industrial chain. Background Art
[0002] In the modern financial system, the credit assessment of enterprises is an important part of banks' risk control and credit decision-making. Traditional credit methods usually focus on the financial data and historical credit records of enterprises, and quantify their credit risk levels by analyzing indicators such as profitability, debt repayment ability and liquidity. However, traditional methods have limitations in data mining and risk prediction. They only rely on static and monotonous financial data and are difficult to capture the complex and dynamic risk environment faced by enterprises.
[0003] In recent years, the risk assessment method combined with artificial intelligence technology has been gradually applied to the application scenario of corporate credit assessment. However, existing technologies usually focus on the individual data characteristics of enterprises. This single-dimensional credit model has gradually exposed the following shortcomings: credit risk is difficult to fully quantify, and credit limit calculation lacks dynamic adaptability. Summary of the invention
[0004] In view of the shortcomings of the existing technology, the present invention proposes a method and system for calculating the pre-credit limit of an enterprise based on the multi-dimensional characteristics of the industrial chain, so as to improve the technical problems existing in the existing technology: the credit risk is difficult to be fully quantified and the credit limit calculation lacks dynamic adaptability.
[0005] The technical solution adopted by the present invention is as follows:
[0006] In the first aspect, a method for calculating the pre-credit limit of an enterprise based on the multi-dimensional characteristics of the industrial chain is provided, comprising the following steps:
[0007] Construct an industrial chain multidimensional data set including key indicator data and key risk factor data. The key indicator data include output value scale, capital flow, number of chain owners and company financial performance; divide the industrial chain into four levels according to the financial performance data of companies in different industrial chains, and assign different weights to the industrial chain of each level; calculate the comprehensive financial performance score of the industrial chain enterprises at each level, analyze the output value scale, capital flow and number of chain owners of the enterprises in the industrial chain, and obtain the industrial chain score of each industrial chain at the same level; input the key risk factor data into the risk control model that integrates multi-dimensional features to calculate the risk score of the enterprise; calculate the pre-credit limit of the enterprise according to the industrial chain score and risk score.
[0008] Furthermore, based on the financial performance data of companies in different industrial chains, the industrial chain is divided into four levels, and different weights are assigned to each level of the industrial chain, including:
[0009] The company financial performance data of multiple industrial chains are obtained from the industrial chain multidimensional data set; the company financial performance data includes: revenue, net profit, return on net assets, compound annual growth rate, debt ratio, and current asset ratio; the industrial chain is divided into primary industrial chain, secondary industrial chain, tertiary industrial chain and quaternary industrial chain according to the company financial performance data.
[0010] Furthermore, when dividing the industrial chain, the first-level industrial chain is divided according to whether it is profitable and the business growth situation; when the first-level industrial chain is not met, the second-level industrial chain is divided according to whether the business is stable, whether there is repayment pressure, and whether the short-term debt repayment ability is sufficient; when the second-level industrial chain is not met, the third-level industrial chain is divided according to the status of the industrial chain; the remaining industrial chains are divided into fourth-level industrial chains.
[0011] Furthermore, comprehensive financial performance scores are calculated for enterprises at each level of the industrial chain, and the output value scale, capital flow, and number of chain leaders of enterprises within the industrial chain are analyzed to obtain the industrial chain scores of each industrial chain at the same level, including: calculating the comprehensive financial performance score of the industrial chain based on the net asset return rate, compound annual growth rate of revenue, and compound annual growth rate of net profit of enterprises within the industrial chain; calculating the output value scale score of the industrial chain based on the output value scale of enterprises within the industrial chain; calculating the industrial capital allocation efficiency score of the industrial chain based on the capital flow of enterprises within the industrial chain; calculating the chain leader coefficient score based on the number of chain leaders within the industrial chain; calculating the industrial chain score of each industrial chain at the same level based on the comprehensive financial performance score, output value scale score, industrial capital allocation efficiency score, and chain leader coefficient score of the industrial chain.
[0012] Furthermore, key risk factor data include output value scale, number of chain leaders, historical default rate and per capita output.
[0013] Furthermore, the risk control model integrating multi-dimensional features is constructed as follows: extract key risk factor data from the multi-dimensional data set of the industrial chain and perform data preprocessing; construct a generalized linear model, use the key risk factor data for training, and output a first score; construct a nonlinear feature interaction model, use the key risk factor data for training, and output a second score; perform low-order and high-order feature interaction training based on the semantic segmentation model, link the first score and the second score, and obtain a risk control model integrating multi-dimensional features.
[0014] Furthermore, the enterprise's pre-credit limit is calculated using a weighted method based on the industry chain score and risk score.
[0015] Furthermore, the method for calculating the enterprise pre-credit limit based on the multi-dimensional characteristics of the industrial chain also includes: collecting business feedback data through real-time monitoring of actual loan and credit performance, and iteratively optimizing the industrial chain multi-dimensional data set, industrial chain score and risk score.
[0016] Furthermore, by real-time monitoring of actual lending and credit performance and collecting business feedback data, the industry chain multidimensional data set, industry chain scoring and risk scoring are iteratively optimized, including: regularly and dynamically updating key indicators in the industry chain multidimensional data set; expanding data dimensions and introducing new key risk factors; using the latest data set to update the reference dimensions of grading and retrain the risk control model that integrates multidimensional features; and adjusting the credit strategy based on the latest model scoring results.
[0017] In the second aspect, a system for calculating the pre-credit limit of an enterprise based on the multi-dimensional characteristics of the industrial chain is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method for calculating the pre-credit limit of an enterprise based on the multi-dimensional characteristics of the industrial chain provided in the first aspect is implemented.
[0018] It can be seen from the above technical solution that the beneficial technical effects of the present invention are as follows:
[0019] 1. The enterprise pre-credit limit calculation method based on the multi-dimensional characteristics of the industrial chain realizes efficient and accurate credit risk prediction and credit limit calculation by integrating multi-dimensional industrial chain data and intelligent risk assessment technology. This method can comprehensively quantify the hierarchical relationship and market dynamics of the industrial chain, deeply analyze the financial data of the enterprise, combine the historical loan and default data of the bank, and explore the complex correlation hidden in the data, thereby improving the accuracy of risk prediction and the scientific nature of credit limit.
[0020] 2. In model training, generalized linear models and deep factorization machines are used respectively. Generalized linear models can quickly capture and quantify linear relationships and provide the model with easily interpretable basic risk scores, while deep factorization machines are used to explore high-order nonlinear feature interactions, such as the potential relationship between capital flows and the number of chain owners. A combined training strategy is used in subsequent model training. The generalized linear model optimizes parameters through the maximum likelihood estimation method and efficiently handles linear feature relationships; at the same time, the deep factorization machine captures more complex patterns through deep neural networks in nonlinear feature interaction modeling. The two are combined through a semantic segmentation model, which not only ensures the efficiency and interpretability of the model, but also significantly improves the prediction ability.
[0021] 3. Through real-time detection, it is possible to dynamically adjust risk assessment strategies and generate accurate pre-credit limits that meet the actual situation of the enterprise, making credit limit calculation dynamically adaptable, thereby reducing lending risks and improving the security of credit management. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for the specific embodiments or the prior art description. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn according to the actual scale.
[0023] Figure 1 This is a flow chart of a method for calculating a pre-credit limit according to Embodiment 1 of the present invention;
[0024] Figure 2 This is a hierarchical link diagram of the industrial chain according to an embodiment of the present invention. DETAILED DESCRIPTION
[0025] The following embodiments of the technical solution of the present invention are described in detail in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and are therefore only used as examples, and cannot be used to limit the protection scope of the present invention.
[0026] It should be noted that, unless otherwise specified, the technical terms or scientific terms used in this application should have the common meanings understood by those skilled in the art to which the present invention belongs.
[0027] Example 1
[0028] Regarding the technical problem that credit risk is difficult to fully quantify, the inventor found through research that the credit risk of an enterprise is not only affected by its own financial situation, but also closely related to its position in the industrial chain. The upstream and downstream relationships of an enterprise in the industrial chain directly affect the stability and risk resistance of its business. In addition, the driving effect of the chain leader enterprise will have an impact on the entire industrial chain. If the industrial chain information of an enterprise is ignored, it will lead to an underestimation or overestimation of the enterprise's true risk level.
[0029] As for the lack of dynamic adaptability in credit line calculation, the inventors found through research that under economic fluctuations or policy adjustments, the industrial chain in which the enterprise is located may undergo major changes, such as fluctuations in output value, adjustments in capital flows, etc. If only static historical data is used as a basis, it is difficult to timely reflect the dynamic changes in the industrial chain and the market, resulting in a disconnect between the credit line and the actual risk situation, thereby affecting the effective allocation of credit resources.
[0030] In view of the above research results, this embodiment provides a method for calculating the pre-credit limit of an enterprise based on the multi-dimensional characteristics of the industrial chain, comprising the following steps:
[0031] Step S1: Constructing a multidimensional data set of industrial chains including key indicator data and key risk factor data Step S11: Collecting key indicator data and key risk factor data related to the industrial chain
[0032] In a specific implementation, data collection of key indicators and key risk factors can be performed based on the public data of the industry chain and the data of the financial terminal Ifind. Key indicators include output value scale, capital flow, number of chain owners and company financial performance, among which company financial performance data include: revenue, net profit, return on net assets, compound annual growth rate, debt ratio, current asset ratio, etc.
[0033] Key risk factors include output value scale, number of chain leaders, historical default rate, per capita output, etc.
[0034] In this embodiment, some data are both key indicators and key risk factors.
[0035] Step S12: Use key indicator data to construct a multidimensional data set of the industrial chain
[0036] Optionally, when constructing a multidimensional data set for the industrial chain, you can first perform feature extraction and conversion on the original data of key indicators to generate structured data that reflects the characteristics of the enterprise and the industrial chain, and then construct the data set with the structured data.
[0037] In this step, different industrial chain multi-dimensional data can be constructed for different industries. Optionally, the constructed industrial chain multi-dimensional data set is stored in a MySQL database for use in subsequent model training.
[0038] As shown in Table 1, the industrial chain multidimensional data set collected in the embodiment of the present invention includes but is not limited to the data items shown in Table 1.
[0039] Table 1 Example of industrial chain multidimensional dataset
[0040]
[0041]
[0042] As shown in Table 2, the table shows the key indicator data of different industrial chains and their application scenarios in model training. In the subsequent steps, the generalized linear model (GLM) is combined with the deep factorization machine (DeepFM) to analyze the linear and nonlinear interactions between different features, providing a reliable basis for the accurate calculation of the enterprise's pre-credit limit.
[0043] Table 2 Example of enterprise data
[0044]
[0045]
[0046] Step S2: Divide the industrial chain into four levels according to the financial performance data of companies in different industrial chains, and assign different weights to each level of industrial chain.
[0047] Step S21, obtaining the company financial performance data of multiple industrial chains from the industrial chain multidimensional data set, the company financial performance data including revenue, net profit, return on equity, compound annual growth rate, debt ratio, and current asset ratio data.
[0048] Among them, the three-year revenue compound annual growth rate (CAGR) is calculated as follows:
[0049]
[0050] In the formula, the final value is the indicator value three years later, the initial value is the indicator value three years ago, and n is set to 3.
[0051] Step S22: Divide the industrial chain into primary industrial chain, secondary industrial chain, tertiary industrial chain and quaternary industrial chain according to the company's financial performance data
[0052] In a specific implementation method, the rules for dividing the industrial chain adopt the relevant regulations of the financial industry, as follows:
[0053] The first-level industrial chain is divided according to whether it is profitable and the business growth situation; specifically, the industrial chain with positive current revenue, current net profit, return on net assets, three-year revenue compound annual growth rate, and three-year net profit compound annual growth rate is selected as the first-level industrial chain;
[0054] When the first-level industrial chain is not met, the second-level industrial chain is divided according to whether the business is stable, whether there is repayment pressure, and whether the short-term debt repayment ability is sufficient; specifically, the industrial chain whose current revenue, current net profit, and return on net assets are all positive, and whose three-year revenue compound annual growth rate or three-year net profit compound annual growth rate has a negative value is selected as the second-level industrial chain.
[0055] When the secondary industrial chain is not met, the tertiary industrial chain is divided according to the status of the industrial chain; specifically, the current revenue, current net profit, and return on net assets are all positive, the compound annual growth rate of revenue and the three-year compound annual growth rate of net profit are stable and not significantly negative, the debt ratio is less than 50% and the proportion of current assets is higher than 50% as the tertiary industrial chain. The tertiary industrial chain still has a relatively high industrial chain status or strong operational capabilities.
[0056] The remaining industrial chains that do not meet the above conditions are divided into four-level industrial chains.
[0057] like Figure 2 As shown, Figure 2A hierarchical chain diagram of the industrial chain is shown.
[0058] Step S23: assign different weights to each level of the industrial chain
[0059] Based on the level of the industrial chain, the following proportions are used for the weighting: the weight of the first-level industrial chain level is w1 = 0.4; the weight of the second-level industrial chain level is w2 = 0.3; the weight of the third-level industrial chain level is w3 = 0.2; the weight of the fourth-level industrial chain level is w4 = 0.1. By initially grading and weighting the industrial chain, the market size and credit demand characteristics can be quantified, thereby more accurately reflecting the priority of the industrial chain in credit decision-making.
[0060] Step S3: Calculate the comprehensive financial performance scores of the industrial chain enterprises at each level, and analyze the output value scale, capital flow, and number of chain owners of the enterprises in the industrial chain to obtain the industrial chain scores of each industrial chain at the same level.
[0061] Step S31: Calculate the comprehensive financial performance score of the industry chain based on the return on net assets, compound annual growth rate of revenue and compound annual growth rate of net profit of the enterprises in the industry chain.
[0062] Select representative enterprises in each industrial chain from the industrial chain multidimensional data set, extract the data of return on net assets, compound annual growth rate of revenue and compound annual growth rate of net profit of these enterprises, and use the multivariate linear regression algorithm (Z-Score) to standardize the three data to obtain the return on net assets score Z ROE , Revenue compound annual growth rate score Z 营收CAGR , Net Profit Compound Annual Growth Rate Score Z 净利润CAGR , which is calculated as follows:
[0063]
[0064] In the formula, X represents the financial indicator value of a specific enterprise, μ represents the mean value of the financial indicator of all enterprises in the industrial chain, and σ represents the standard deviation of the financial indicator of all enterprises.
[0065] Then, the standardized indicator scores are used to calculate the comprehensive financial performance scores according to the weights as follows:
[0066] Scaled_Score 财务 =0.4×Z ROE +0.3×Z 营收CAGR +0.3×Z 净利润CAGR
[0067] The comprehensive financial performance score is an assessment of the profitability and financial health of enterprises in the industry chain, which can reflect the collective financial performance of enterprises. Optionally, the median of the evaluation results of representative enterprises is taken as the overall weight of the industry chain.
[0068] Step S32: Calculate the output value scale score of the industrial chain according to the output value scale of enterprises in the industrial chain
[0069] The output value scale score can reflect the market size of each industrial chain. Specifically, the output value scale score is calculated as follows:
[0070]
[0071] In the formula, optional, weight w 产值 Set to 0.3.
[0072] Step S33: Calculate the industry capital allocation efficiency score of the industry chain according to the capital flow of enterprises in the industry chain
[0073] The industrial capital allocation efficiency score can measure the capital flow and support strength. Specifically, the capital flow efficiency of the industrial chain is calculated based on the number of important projects in each industrial chain, the total annual investment and the average investment amount of a single project; specifically, the capital flow efficiency is calculated as follows:
[0074]
[0075] In the formula, N represents the number of important items, I total represents the total annual investment, w i Represents the weight of a single important item, P i represents the output of the i-th project.
[0076] The industrial capital allocation efficiency score of a single industrial chain is calculated as follows:
[0077]
[0078] In the formula, optional, weight w 资金 Set to 0.2.
[0079] Step S34: Calculate the chain leader coefficient score according to the number of chain leaders in the industrial chain
[0080] The chain leader in the industrial chain is the core enterprise, which has a driving effect on upstream and downstream enterprises. The chain leader coefficient score can reflect the influence of the core enterprise. Specifically, the number of chain leaders in each industrial chain is counted. The more chain leader enterprises there are, the more leverage they have in credit expansion. The chain leader coefficient score is calculated as follows:
[0081]
[0082] In the formula, optional, w 链主 Set to 0.2.
[0083] Step S35: Calculate the industry chain score of each industry chain at the same level based on the comprehensive financial performance score, output value scale score, industry capital allocation efficiency score, and chain leader coefficient score of the industry chain.
[0084] The calculation results of steps S31-S34 are weighted and integrated, and for each level of industrial chain, the industrial chain score of each industrial chain in the same level is calculated according to different weights. Specifically, the industrial chain score is calculated according to the following formula:
[0085]
[0086] Where, Scaled_Score Final Indicates the industry chain score, Scaled_Score i is the score of different dimensions, w i is the weight of each dimension, among which the weight of comprehensive financial performance score is 0.2; the weight of output value scale score is 0.4; the weight of industrial capital allocation efficiency score is 0.3; and the weight of chain leader coefficient score is 0.1.
[0087] In this step S3, the position and credit risk of the enterprise in the industrial chain are comprehensively evaluated, the analysis results of each dimension are weighted and integrated, and the scores of each industrial chain in the same level are refined.
[0088] Step S4: Input the key risk factor data into the risk control model integrating multi-dimensional features to calculate the risk score of the enterprise
[0089] The risk control model that integrates multi-dimensional features includes a generalized linear model and a nonlinear feature interaction model. The generalized linear model is linked through a semantic segmentation model. The construction method of the risk control model is as follows:
[0090] 1. Preprocess data for key risk factors
[0091] The key risk factors extracted from the multidimensional data set of the industrial chain are preprocessed. The key risk factors include output value scale, number of chain owners, historical default rate, per capita output, etc. Data preprocessing includes processing missing values, outliers, and standardizing features to ensure consistency and stability when data is input into the model.
[0092] 2. Build a generalized linear model, use key risk factor data for training, and output the first score
[0093] The generalized linear model (GLM) selects the logarithmic probability function (Logit) as the link function, models the risk score of the key risk factors on the probability of enterprise default, and estimates the model parameters through maximum likelihood estimation (MLE). Then, combined with the bank's historical loan and default data, the performance of the risk control model is evaluated using K-fold cross validation. The data is divided into K parts, and K-1 parts of data are used to train the model in each round, and the remaining 1 part is used to verify the model to evaluate the robustness and generalization ability of the model. In this embodiment, the generalized linear model risk score can be expressed as the following function:
[0094]
[0095] In the formula, S GLM represents the first score calculated by the generalized linear model, X j represents the risk factor, β j represents the regression coefficient of the corresponding risk factor, and β0 represents the intercept term.
[0096] 3. Build a nonlinear feature interaction model, use key risk factor data for training, and output the second score
[0097] The nonlinear feature interaction model uses the deep factorization machine (DeepFM) to model the nonlinear relationship between risk factors, capture high-order interaction features such as output value scale and number of chain owners, historical default rate and liquidity, and explore the complex correlations hidden in the data. In this embodiment, the deep factorization machine model can be expressed as the following function:
[0098]
[0099] In the formula, S DeepFM represents the second score calculated by the nonlinear feature interaction model, w i represents the weight of the linear part, (V i ,V j ) represents the risk factor X i and X j The inner product between them is used to capture the second-order interaction; f deep (X) represents the deep network part, which is used to learn high-order feature interactions.
[0100] 4. Perform interactive training of low-order and high-order features based on the semantic segmentation model, link the first score and the second score, and obtain a risk control model that integrates multi-dimensional features.
[0101] Then, based on the semantic segmentation model (UNet), low-order and high-order feature interactions are constructed, and the model is linked through a multi-layer network. The nonlinear feature interaction model is combined with the calculation results of the generalized linear model. After training, a risk control model integrating multi-dimensional features is obtained to correct the credit risk score. In this embodiment, the risk control model can be expressed as:
[0102] S Final =αS GLM +(1―α)S DeepFM
[0103] In the formula, S Final Represents the risk score calculated by the risk control model that integrates multi-dimensional features; optionally, the weighting coefficient α is set to 0.8.
[0104] In this step, in order to better model the complex relationship between these features, a combination of generalized linear model (GLM) and deep factorization machine (DeepFM) is used in model training. GLM can quickly capture and quantify linear relationships, providing the model with an easily interpretable basic risk score, while DeepFM is used to explore high-order nonlinear feature interactions, such as the potential relationship between capital flow and the number of chain owners. A combined training strategy is used in subsequent model training. GLM optimizes parameters through maximum likelihood estimation and efficiently handles linear feature relationships; at the same time, DeepFM captures more complex patterns through deep neural networks in nonlinear feature interaction modeling. The combination of the two not only ensures the efficiency and interpretability of the model, but also significantly improves the predictive ability. When in use, the key risk factor data is input into the trained risk control model that integrates multidimensional features to obtain the risk score of the enterprise. Table 3 below is an example of application scenarios of different industrial chains in model training.
[0105] Table 3 Application scenarios of different industrial chains in model training
[0106]
[0107] S5. Calculate the pre-credit limit of the enterprise in a weighted manner based on the industry chain score and risk score
[0108] According to the industrial chain scores of each level of industrial chain obtained in step S3 and the risk score of the risk control model integrating multi-dimensional features obtained in step S4, the pre-credit limit of the enterprise is calculated. In this embodiment, the pre-credit limit is calculated according to the following formula:
[0109] Pre-credit limit
[0110] =(The weight of the industrial chain level to which the enterprise belongs × Scaled_Score Final ×γ+S Final ×δ) × basic credit limit of the enterprise
[0111] In the formula, optionally, the industrial chain scoring coefficient γ is set to 0.4, and the risk scoring coefficient δ is set to 0.6. The weight of the industrial chain level to which the enterprise belongs is obtained through step S23, and the basic credit limit of the enterprise is derived from the average of the historical credit records of the enterprise in various financial institutions.
[0112] This embodiment uses accuracy, AUC, F1-Score, Recall and other indicators to evaluate the calculation effect of the pre-credit limit calculation method, in which a risk control model integrating multi-dimensional features is used as a benchmark. The model needs to predict the probability of default and give a reasonable credit limit based on the multi-dimensional financial characteristics and industrial chain information of the enterprise. The experimental results are shown in Table 4.
[0113] Table 4
[0114] index The results of traditional generalized linear model calculation The results of the calculation in this example Validation set accuracy 0.88 0.92 Validation set AUC 0.86 0.90 F1-Score 0.80 0.89 Precision 0.85 0.90 Recall 0.88 0.91
[0115] In Table 4, the accuracy is used to measure the overall prediction accuracy of the model for defaulting and non-defaulting enterprises; AUC represents the prediction ability of the model at different thresholds, and F1-Score combines Precision and Recall to measure the comprehensive performance of the model in practical applications. Through the above multi-dimensional indicator evaluation, the performance of the pre-credit limit calculation method provided by this embodiment in enterprise credit risk prediction and credit limit evaluation is fully verified, especially in terms of accuracy and AUC, which is significantly better than the traditional generalized linear model, and effectively improves the accuracy of pre-credit calculation by financial institutions.
[0116] The enterprise pre-credit limit calculation method based on the multi-dimensional characteristics of the industrial chain provided in this embodiment realizes efficient and accurate credit risk prediction and credit limit calculation by integrating multi-dimensional industrial chain data and intelligent risk assessment technology. This method can comprehensively quantify the hierarchical relationship and market dynamics of the industrial chain, deeply analyze the financial data of the enterprise, combine the historical loan and default data of the bank, and mine the complex correlation hidden in the data, thereby improving the accuracy of risk prediction and the scientific nature of credit limit.
[0117] In some implementations, the enterprise pre-credit limit calculation method based on the multi-dimensional characteristics of the industrial chain also includes step S5, which monitors the actual loan and credit performance in real time, collects business feedback data, iteratively optimizes the industrial chain multi-dimensional data set, industrial chain score and risk score, and continuously improves the accuracy and applicability of the calculation method. Step S5 is as follows:
[0118] Step S51: regularly obtain the latest data from the financial terminal Ifind and relevant departments of the industrial chain, and dynamically update the key indicators in the industrial chain multidimensional data set. At the same time, the data dimension can also be expanded, and new key risk factors can be introduced to enhance the applicability of the model.
[0119] Step S52: Use the latest data set to update the reference dimension of the classification and retrain the risk control model that integrates multi-dimensional features. During the training process, use the time series segmentation method or rolling window technology to ensure the stability and generalization ability of the model in different time periods. At the same time, use the validation set to perform performance evaluation to measure the performance of the model in terms of AUC, F1, accuracy and other indicators.
[0120] Step S53: Adjust the credit strategy based on the latest model scoring results. Optimize the pre-credit limit, risk mitigation measures and credit approval process of each industry chain and enterprise to ensure that the credit strategy is consistent with the actual risk situation and market demand. For high-risk enterprises, dynamically set stricter risk control measures, such as increasing guarantees or reducing credit limits.
[0121] Step S5 can dynamically adjust the risk assessment strategy through real-time detection and generate an accurate pre-credit limit that meets the actual situation of the enterprise, so that the credit limit calculation has dynamic adaptability, thereby reducing lending risks and improving the security of credit management.
[0122] Example 2
[0123] This embodiment provides a system for calculating the pre-credit limit of an enterprise based on the multi-dimensional characteristics of the industrial chain, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the method for calculating the pre-credit limit of an enterprise based on the multi-dimensional characteristics of the industrial chain described in Example 1.
[0124] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.
Claims
1. A method for calculating the pre-credit limit of an enterprise based on the multi-dimensional characteristics of the industrial chain, characterized in that: The following steps are involved: Construct a multi-dimensional data set of the industry chain including key indicator data and key risk factor data. The key indicator data include output value scale, capital flow, number of chain owners and company financial performance; According to the financial performance data of companies in different industrial chains, the industrial chain is divided into four levels, and different weights are given to each level of industrial chain; Calculate the comprehensive financial performance scores of enterprises in each level of the industrial chain, analyze the output value scale, capital flow, and number of chain owners of enterprises in the industrial chain, and obtain the industrial chain score of each industrial chain in the same level; Input key risk factor data into the risk control model that integrates multi-dimensional features to calculate the enterprise's risk score; The enterprise's pre-credit limit is calculated based on the industry chain score and risk score.
2. The method for calculating the enterprise pre-credit limit based on the multi-dimensional characteristics of the industrial chain according to claim 1 is characterized in that: According to the financial performance data of companies in different industrial chains, the industrial chain is divided into four levels, and different weights are given to each level of the industrial chain, including: Obtaining financial performance data of companies in multiple industrial chains from the industrial chain multidimensional data set; the financial performance data of the companies include: revenue, net profit, return on net assets, compound annual growth rate, debt ratio, and current asset ratio; According to the company's financial performance data, the industrial chain is divided into primary industrial chain, secondary industrial chain, tertiary industrial chain and quaternary industrial chain.
3. The method for calculating the enterprise pre-credit limit based on the multi-dimensional characteristics of the industrial chain according to claim 2 is characterized in that: When dividing the industrial chain, the first-level industrial chain is divided according to whether it is profitable and the business growth; When the first-level industrial chain is not satisfied, the second-level industrial chain is divided according to whether the business is stable, whether there is repayment pressure, and whether the short-term debt repayment capacity is sufficient; When the secondary industrial chain is not satisfied, the tertiary industrial chain is divided according to the status of the industrial chain; The remaining industrial chain is divided into four levels.
4. The method for calculating the enterprise pre-credit limit based on the multi-dimensional characteristics of the industrial chain according to claim 1 is characterized in that: The comprehensive financial performance scores of enterprises in each level of the industrial chain are calculated respectively, and the output value scale, capital flow, and number of chain owners of enterprises in the industrial chain are analyzed to obtain the industrial chain scores of each industrial chain in the same level, including: Calculate the comprehensive financial performance score of the industry chain based on the return on equity, compound annual growth rate of revenue and compound annual growth rate of net profit of enterprises in the industry chain; Calculate the output value scale score of the industrial chain based on the output value scale of enterprises in the industrial chain; According to the capital flow of enterprises in the industrial chain, the industrial capital allocation efficiency score of the industrial chain is calculated; Calculate the chain leader coefficient score based on the number of chain leaders in the industrial chain; The industrial chain score of each industrial chain at the same level is calculated based on the comprehensive financial performance score, output value scale score, industrial capital allocation efficiency score, and chain leader coefficient score of the industrial chain.
5. The method for calculating the enterprise pre-credit limit based on the multi-dimensional characteristics of the industrial chain according to claim 1 is characterized in that: The key risk factor data include output value scale, number of chain leaders, historical default rate and per capita output.
6. The method for calculating the enterprise pre-credit limit based on the multi-dimensional characteristics of the industrial chain according to claim 5 is characterized in that: The risk control model integrating multi-dimensional features is constructed as follows: Extract key risk factor data from the industry chain multidimensional data set and perform data preprocessing; Construct a generalized linear model, train it using key risk factor data, and output the first score; Construct a nonlinear feature interaction model, train it using key risk factor data, and output a second score; Based on the semantic segmentation model, low-order and high-order feature interactive training is performed to link the first score and the second score to obtain a risk control model that integrates multi-dimensional features.
7. The method for calculating the enterprise pre-credit limit based on the multi-dimensional characteristics of the industrial chain according to claim 1 is characterized in that: The enterprise's pre-credit limit is calculated using a weighted method based on the industry chain score and risk score.
8. The method for calculating the enterprise pre-credit limit based on the multi-dimensional characteristics of the industrial chain according to claim 1 is characterized in that: Also includes: By real-time monitoring of actual loan and credit performance and collecting business feedback data, we iteratively optimize the industry chain multidimensional data set, industry chain score and risk score.
9. The method for calculating the enterprise pre-credit limit based on the multi-dimensional characteristics of the industrial chain according to claim 8 is characterized in that: By real-time monitoring of actual lending and credit performance, collecting business feedback data, and iteratively optimizing the industry chain multidimensional data set, industry chain score, and risk score, including: Regularly and dynamically update key indicators in the industry chain multidimensional data set; expand data dimensions and introduce new key risk factors; Use the latest data sets to update the reference dimensions of classification and retrain the risk control model that integrates multi-dimensional features; Adjust the credit strategy based on the latest model scoring results.
10. A system for calculating enterprise pre-credit limit based on multi-dimensional characteristics of industrial chain, characterized in that: It includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the enterprise pre-credit limit calculation method based on the multi-dimensional characteristics of the industrial chain as described in any one of claims 1 to 9.