Bank credit evaluation system and method based on deep learning

Abstract

The invention discloses a bank credit evaluation system and method based on deep learning, and relates to the technical field of data analysis and evaluation, and the system comprises a data acquisition module, a credit evaluation module, a credit analysis module, and a decision output module. According to the method, customer groups are divided in a layered and detailed manner through data acquisition, a dynamic updating and exception rechecking mechanism is applied, a data set of key information is constructed, potential risks of daily demand expenditure of customers on credit are evaluated, and meanwhile income stability analysis is performed according to work to judge the stability of the work development trend of the customers; considering the influence of work on customer credit conditions, establishing a skill equivalent replacement risk model, simulating equivalent work replacement execution, performing customer economic risk evaluation, predicting credit risks generated by work, and integrating a historical credit data set and the work stability and development trend reflected by a work trend change coefficient to obtain a credit risk evaluation result. And a comprehensive customer credit assessment value is generated according to the fund demand and the risk condition embodied by the daily demand prediction risk value.

Description

Technical Field

[0001] This invention relates to the field of data analysis and evaluation technology, and in particular to a bank credit evaluation system and method based on deep learning. Background Technology

[0002] Deep learning is a branch of machine learning. Its core idea is to construct multi-layered neural networks to simulate the hierarchical information processing mechanism of the human brain, automatically learning abstract feature representations from data and completing complex tasks such as classification, regression, and generation. Bank credit assessment, also known as bank credit rating, refers to a comprehensive evaluation conducted by a bank's specialized credit assessment agency or department using rigorous scientific analysis methods and indicator systems to assess the ability and willingness of specific entities (such as enterprises, individuals, and governments) to repay their debts on time and in full in the future. The assessment uses intuitive rating symbols to represent the degree of credit risk. The assessment targets include various corporate clients, individual clients, and sovereign credit, and also provides decision-making references for financial market participants. For example, when approving corporate loan applications, banks use credit assessment results to decide whether to grant a loan, determine the loan amount and interest rate, and the assessment indicators include financial and non-financial indicators.

[0003] Currently, when acquiring bank customer data, risk assessment mainly relies on traditional financial statements and limited transaction records provided by customers. Relying solely on financial data cannot fully reflect their true situation and risks, ignores customers' work capabilities, and leads to overestimation of customers' creditworthiness. This results in inaccurate credit assessments, risk exposure bias, and an inability to clearly identify changes in factors that affect customer credit development trends, thus reducing the accuracy of credit rating.

[0004] To address the aforementioned technical shortcomings, a solution is proposed. Summary of the Invention

[0005] The purpose of this invention is to generate a customer's bank credit assessment value by assessing the potential risk to credit from the customer's daily expenses and conducting income stability analysis based on their work to determine the stability of the customer's work development trend, and by combining this with an equivalent replacement risk model.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a bank credit assessment system and method based on deep learning, comprising a data acquisition module, a credit assessment module, a credit analysis module, and a decision output module;

[0007] The data acquisition module constructs customer datasets, transaction datasets, work datasets, and historical credit datasets.

[0008] The credit analysis module includes a demand analysis unit and a volatility analysis unit. The demand analysis unit performs daily demand analysis for different periods based on customer datasets and transaction datasets to obtain daily demand standard thresholds, and then converts these daily demand standard thresholds into daily demand prediction risk values.

[0009] The fluctuation analysis unit performs economic fluctuation trend analysis based on customer dataset and transaction dataset, generates economic fluctuation impact factors based on income economic value at turning points, and performs work trend risk fluctuation analysis based on work dataset and work trend change coefficient based on economic fluctuation impact factors.

[0010] The impact analysis module performs job skill risk analysis using job datasets and customer datasets, establishes a skill equivalent replacement risk model, outputs a job equivalent replacement estimate based on the skill equivalent replacement risk model, simulates equivalent job replacement execution, and generates economic risk judgment signals through a preset evaluation mechanism.

[0011] The credit assessment module acquires historical credit datasets, work trend change coefficients, and daily demand prediction risk values ​​to generate customer credit assessment values. Based on the customer credit assessment values ​​and the work equivalent replacement estimate, it generates a decision suggestion information sheet.

[0012] Furthermore, the data acquisition module also includes the construction of customer datasets, transaction datasets, working datasets, and historical credit datasets, specifically including the following:

[0013] The customer group is pre-stratified into N dimensions based on the customer group, forming M customer categories;

[0014] By connecting with the bank's CRS system to the customer's authorized tax declaration data and social security payment records, annualized income and monthly average income are collected, and an age verification component is installed to establish a dynamic age verification mechanism in combination with biometric technology.

[0015] Customers are categorized into corresponding customer categories based on their annual income and age, thus forming a customer dataset.

[0016] Customer spending data is obtained through transaction records, and the transaction amount and timestamp of each transaction are extracted and analyzed to construct a customer consumption cycle map, which includes the monthly average expenditure, peak expenditure days, and time series features of consumption frequency distribution, forming a bank transaction dataset.

[0017] Obtain the client's job title, salary, and length of employment, and use intelligent technology analysis software to analyze the client's job skills and identify implicit skills.

[0018] Establish a knowledge graph-based professional relationship network, including hierarchical relationships between industry, position, and skill, and construct a job skills dataset;

[0019] By acquiring historical loan records and default counts, a historical credit dataset is constructed to generate a credit history timeline, marking the date of initial credit granting and credit change events.

[0020] Set up an automatic update strategy on a quarterly basis, and trigger manual review for customers whose income fluctuates abnormally by 30%.

[0021] Furthermore, based on the standard threshold of daily demand, it is converted into a work demand prediction risk value, specifically including the following:

[0022] S100. Based on the customer dataset and transaction dataset, obtain customer expenditures and annual customer income, and perform data integration and analysis on customers' daily needs and expenses.

[0023] S101. Obtain the monthly average expenditure for customers' daily needs within the period, and perform mean-based processing to obtain the standard threshold for demand. Then, organize the total monthly expenditure for customers within the period to obtain the median value of the total monthly expenditure for each month. Finally, calculate the expenditure difference between the highest monthly average expenditure and the monthly average expenditure to generate the prediction threshold.

[0024] S102. Integrate the actual average proportion with the standard demand threshold to obtain the daily demand forecast risk value.

[0025] Furthermore, the coefficient for change in work trends is generated, specifically including the following:

[0026] S200. Based on the customer dataset and transaction dataset, obtain customer revenue values ​​and establish an economic value change curve. Plot customer revenue values ​​on the vertical axis and time points on the horizontal axis, and perform curve change rate analysis to obtain the economic fluctuation impact factor X. In the above formula, I T and I T-1 These are the economic values ​​at the time point of the previous cycle and the corresponding economic values ​​at the time point of the current cycle, respectively.

[0027] S201. Obtain the nature of the customer's work through the work dataset, and assign a value to the customer's work industry sensitivity coefficient M by combining external economic data analysis.

[0028] S202. Based on the factors influencing customer economic fluctuations and industry sensitivity coefficients, conduct a job trend risk fluctuation analysis to generate a job trend change coefficient B.

[0029] Furthermore, a risk model for equivalent skill replacement is established, specifically including the following:

[0030] The skill equivalent replacement risk model is constructed from an input layer, a processing layer, and an output layer.

[0031] When constructing the processing layer, physical constraints are set for the initial network model based on a fully connected neural network architecture. The equivalence rule is positioned to obtain the job dataset and customer dataset, perform extended related job searches, extract relevant job datasets for each position based on job skill requirement analysis, and match customer skills with job requirement skills based on the similarity of job relevance. Based on job skill matching, direct and indirect equivalence matching are set, direct and indirect equivalence rules are defined and assigned dynamic weights, and related job matching value analysis is performed based on semantic similarity to obtain a dataset of directly and indirectly equivalent job matching values.

[0032] An initial function for job risk is constructed based on job position and job skills, and an estimated value of equivalent job replacement is predicted based on job skill matching.

[0033] The output layer outputs the equivalent replacement estimate for the forecast.

[0034] Furthermore, through a pre-set assessment mechanism, economic risk assessment signals are generated, specifically including the following:

[0035] S300, acquire equivalent replacement work generated by the skill equivalent replacement risk model, perform simulated equivalent work replacement execution operations, and estimate the economic value of the equivalent replacement work;

[0036] S301. Conduct a partial market salary analysis based on the equivalent replacement work to obtain the integrated salary economic value. Compare and analyze the salary economic value with the customer's preset repayment amount to obtain the difference between the two.

[0037] S301. When the magnitude difference is less than the prediction threshold, it indicates that the equivalent replacement work is risky, and an economic risk warning signal is generated.

[0038] Furthermore, based on the customer's credit assessment value and the estimated value of equivalent job replacement, a decision recommendation information sheet is generated, specifically including the following:

[0039] S400: Obtain historical credit datasets, work trend change coefficients, and daily demand prediction risk values ​​to conduct customer credit assessment and analysis.

[0040] S401. By combining the number of defaults (C) with the work trend change coefficient (B) and the daily demand forecast risk value (Y), the customer credit assessment value is obtained and set as P. J represents the number of times the agreement needs to be kept;

[0041] S402. Based on the customer's credit assessment value and the estimated value of equivalent work replacement, a decision recommendation is generated. When the customer's credit assessment value is lower than the standard credit assessment value set by the bank, and the estimated value of equivalent work replacement is lower than the set replacement standard threshold, an information sheet that does not meet the bank's credit decision recommendation is generated.

[0042] A deep learning-based bank credit assessment method includes the following steps:

[0043] Step 1: Data Acquisition Module, constructing customer datasets, transaction datasets, working datasets, and historical credit datasets;

[0044] Step 2: Based on the customer dataset and transaction dataset, conduct daily demand analysis for different periods to obtain the daily demand standard threshold, and convert the daily demand standard threshold into the daily demand prediction risk value.

[0045] Step 3: Based on the customer dataset and transaction dataset, conduct economic fluctuation trend analysis, generate economic fluctuation impact factors based on the inflection point and income economic value, and conduct work trend risk fluctuation analysis based on the economic fluctuation impact factors and the work dataset to generate work trend change coefficients.

[0046] Step 4: Conduct job skill risk analysis using job datasets and customer datasets, establish a skill equivalent replacement risk model, output a job equivalent replacement estimate based on the skill equivalent replacement risk model, simulate equivalent job replacement execution, and generate economic risk judgment signals through a preset evaluation mechanism.

[0047] Step 5: Obtain historical credit datasets, work trend change coefficients, and daily demand forecast risk values; generate customer credit assessment values; and generate decision recommendation information sheets based on customer credit assessment values ​​and work equivalent replacement estimates.

[0048] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0049] This deep learning-based bank credit assessment system and method meticulously segments customer groups through data acquisition and employs dynamic updates and anomaly review mechanisms to construct a dataset of key information. It assesses the potential credit risk posed by customers' daily expenses and conducts income stability analysis based on their work to determine the stability of their employment trends. Considering the impact of work on customer creditworthiness, it establishes a skills-equivalent replacement risk model, simulates equivalent job replacement execution, assesses customer economic risk, and rationally predicts credit risk arising from work. By integrating historical credit datasets, job stability and development trends reflected by job trend change coefficients, and funding needs and risk status reflected by predicted daily expenses, it generates a comprehensive and objective customer credit assessment value. Furthermore, based on the estimated value of equivalent job replacement, it generates detailed and targeted decision-making recommendations. Attached Figure Description

[0050] Figure 1 A schematic diagram of the overall system structure of the present invention is shown;

[0051] Figure 2 A schematic diagram of the method flow structure of the present invention is shown. Detailed Implementation

[0052] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] Example 1:

[0054] like Figure 1 As shown, a bank credit assessment system based on deep learning includes a data acquisition module, a credit assessment module, a credit analysis module, and a decision output module.

[0055] The data acquisition module constructs customer datasets, transaction datasets, work datasets, and historical credit datasets.

[0056] The credit analysis module includes a demand analysis unit and a volatility analysis unit. The demand analysis unit performs daily demand analysis for different periods based on customer datasets and transaction datasets to obtain daily demand standard thresholds, and then converts these daily demand standard thresholds into daily demand prediction risk values.

[0057] The fluctuation analysis unit performs economic fluctuation trend analysis based on customer dataset and transaction dataset, generates economic fluctuation impact factors based on income economic value at turning points, and performs work trend risk fluctuation analysis based on work dataset and work trend change coefficient based on economic fluctuation impact factors.

[0058] The impact analysis module performs job skill risk analysis using job datasets and customer datasets, establishes a skill equivalent replacement risk model, outputs a job equivalent replacement estimate based on the skill equivalent replacement risk model, simulates equivalent job replacement execution, and generates economic risk judgment signals through a preset evaluation mechanism.

[0059] The credit assessment module acquires historical credit datasets, work trend change coefficients, and daily demand prediction risk values ​​to generate customer credit assessment values. Based on the customer credit assessment values ​​and the work equivalent replacement estimate, it generates a decision suggestion information sheet.

[0060] The data acquisition module uses a deep learning framework to automatically integrate customer datasets, transaction datasets, and work datasets. It uses an autoencoder to extract unsupervised features from unstructured data, resume texts, and consumption records, solving the problem of low efficiency in manual feature engineering in traditional methods. Combined with an incremental learning mechanism, it achieves end-to-end synchronization between data updates and model training, avoiding the lag of traditional batch processing modes.

[0061] The demand analysis unit captures the cyclical patterns of customer consumption behavior, automatically learns the demand characteristics at different time scales, generates more accurate daily demand standard thresholds, builds a demand prediction model, quantifies the degree of demand deviation by reconstructing errors, and directly outputs the daily demand prediction risk value. Compared with the traditional threshold method, it can provide early warning of abnormal consumption behavior.

[0062] The volatility analysis unit processes macroeconomic indicators, extracts local volatility characteristics, models long-term dependencies, accurately locates economic cycle turning points, constructs customer-industry correlation graphs, combines occupational mobility data in the job dataset, quantifies the transmission effect of economic fluctuations on individual income, and generates confidence levels for job trend change coefficients.

[0063] The impact analysis module, based on knowledge graph embedding technology, maps occupational skills into a high-dimensional vector space. It calculates skill equivalence scores through graph convolutional networks to achieve a quantitative assessment of cross-industry occupational replacement risk. It also uses generative adversarial networks to simulate equivalent job replacement scenarios and assess economic risks.

[0064] The credit assessment module structures historical credit datasets, maps time-series work trend change coefficients and daily demand risk value space to a unified latent space, uses work equivalence to replace the estimated value as the state input, designs a reward function to optimize decision-making suggestions and information sheets, and realizes personalized credit granting strategies.

[0065] It also includes a data acquisition module for building customer datasets, transaction datasets, working datasets, and historical credit datasets, specifically including the following:

[0066] The customer group is pre-stratified into N dimensions based on the customer group, forming M customer categories;

[0067] By connecting with the bank's CRS system to the customer's authorized tax declaration data and social security payment records, annualized income and monthly average income are collected, and an age verification component is installed to establish a dynamic age verification mechanism in combination with biometric technology.

[0068] Customers are categorized into corresponding customer categories based on their annual income and age, thus forming a customer dataset.

[0069] Customer spending data is obtained through transaction records, and the transaction amount and timestamp of each transaction are extracted and analyzed to construct a customer consumption cycle map, which includes the monthly average expenditure, peak expenditure days, and time series features of consumption frequency distribution, forming a bank transaction dataset.

[0070] Obtain the client's job title, salary, and length of employment, and use intelligent technology analysis software to analyze the client's job skills and identify implicit skills.

[0071] Establish a knowledge graph-based professional relationship network, including hierarchical relationships between industry, position, and skill, and construct a job skills dataset;

[0072] By acquiring historical loan records and default counts, a historical credit dataset is constructed to generate a credit history timeline, marking the date of initial credit granting and credit change events.

[0073] Set up an automatic update strategy on a quarterly basis, and trigger manual review for customers whose income fluctuates abnormally by 30%.

[0074] Based on the standard threshold of daily demand, it is converted into a predicted risk value for work demand, specifically including the following:

[0075] S100. Based on the customer dataset and transaction dataset, obtain customer expenditures and annual customer income, and perform data integration and analysis on customers' daily needs and expenses.

[0076] S101. Obtain the monthly average expenditure for customers' daily needs within the period, and perform mean-based processing to obtain the standard threshold for demand. Then, organize the total monthly expenditure for customers within the period to obtain the median value of the total monthly expenditure for each month. Finally, calculate the expenditure difference between the highest monthly average expenditure and the monthly average expenditure to generate the prediction threshold.

[0077] S102. Integrate the actual average proportion with the standard demand threshold to obtain the daily demand forecast risk value.

[0078] Generate the coefficient for changes in work trends, specifically including the following:

[0079] S200. Based on the customer dataset and transaction dataset, obtain customer revenue values ​​and establish an economic value change curve. Plot customer revenue values ​​on the vertical axis and time points on the horizontal axis, and perform curve change rate analysis to obtain the economic fluctuation impact factor X. In the above formula, I T and I T-1 These are the economic values ​​at the time point of the previous cycle and the corresponding economic values ​​at the time point of the current cycle, respectively.

[0080] S201. Obtain the nature of the customer's work through the work dataset, and assign a value to the customer's work industry sensitivity coefficient M by combining external economic data analysis.

[0081] S202. Based on the factors influencing customer economic fluctuations and industry sensitivity coefficients, conduct a job trend risk fluctuation analysis to generate a job trend change coefficient B.

[0082] Establish a risk model for equivalent skill replacement, which includes the following:

[0083] The skill equivalent replacement risk model is constructed from an input layer, a processing layer, and an output layer.

[0084] When constructing the processing layer, physical constraints are set for the initial network model based on a fully connected neural network architecture. The equivalence rule is positioned to obtain the job dataset and customer dataset, perform extended related job searches, extract relevant job datasets for each position based on job skill requirement analysis, and match customer skills with job requirement skills based on the similarity of job relevance. Based on job skill matching, direct and indirect equivalence matching are set, direct and indirect equivalence rules are defined and assigned dynamic weights, and related job matching value analysis is performed based on semantic similarity to obtain a dataset of directly and indirectly equivalent job matching values.

[0085] An initial function for job risk is constructed based on job position and job skills, and an estimated value of equivalent job replacement is predicted based on job skill matching.

[0086] The output layer outputs the equivalent replacement estimate for the forecast.

[0087] An economic risk assessment signal is generated through a pre-set assessment mechanism, specifically including the following:

[0088] S300, acquire equivalent replacement work generated by the skill equivalent replacement risk model, perform simulated equivalent work replacement execution operations, and estimate the economic value of the equivalent replacement work;

[0089] S301. Conduct a partial market salary analysis based on the equivalent replacement work to obtain the integrated salary economic value. Compare and analyze the salary economic value with the customer's preset repayment amount to obtain the difference between the two.

[0090] S301. When the magnitude difference is less than the prediction threshold, it indicates that the equivalent replacement work is risky, and an economic risk warning signal is generated.

[0091] Based on the customer's credit assessment score and the estimated value of equivalent job replacement, a decision recommendation information sheet is generated, which includes the following:

[0092] S400: Obtain historical credit datasets, work trend change coefficients, and daily demand prediction risk values ​​to conduct customer credit assessment and analysis.

[0093] S401. By combining the number of defaults (C) with the work trend change coefficient (B) and the daily demand forecast risk value (Y), the customer credit assessment value is obtained and set as P. J represents the number of times the agreement needs to be kept;

[0094] S402. Based on the customer's credit assessment value and the estimated value of equivalent work replacement, a decision recommendation is generated. When the customer's credit assessment value is lower than the standard credit assessment value set by the bank, and the estimated value of equivalent work replacement is lower than the set replacement standard threshold, an information sheet that does not meet the bank's credit decision recommendation is generated.

[0095] Example 2:

[0096] like Figure 2 As shown, a deep learning-based bank credit assessment method includes the following steps:

[0097] Step 1: Data Acquisition Module, constructing customer datasets, transaction datasets, working datasets, and historical credit datasets;

[0098] Step 2: Based on the customer dataset and transaction dataset, conduct daily demand analysis for different periods to obtain the daily demand standard threshold, and convert the daily demand standard threshold into the daily demand prediction risk value.

[0099] Step 3: Based on the customer dataset and transaction dataset, conduct economic fluctuation trend analysis, generate economic fluctuation impact factors based on the inflection point and income economic value, and conduct work trend risk fluctuation analysis based on the economic fluctuation impact factors and the work dataset to generate work trend change coefficients.

[0100] Step 4: Conduct job skill risk analysis using job datasets and customer datasets, establish a skill equivalent replacement risk model, output a job equivalent replacement estimate based on the skill equivalent replacement risk model, simulate equivalent job replacement execution, and generate economic risk judgment signals through a preset evaluation mechanism.

[0101] Step 5: Obtain historical credit datasets, work trend change coefficients, and daily demand forecast risk values; generate customer credit assessment values; and generate decision recommendation information sheets based on customer credit assessment values ​​and work equivalent replacement estimates.

[0102] The size of the interval and threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by those skilled in the art for each set of sample data; as long as it does not affect the ratio between the parameter and the quantized value.

[0103] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0104] In the two embodiments provided in this application, it should be understood that the disclosed apparatus and system can be implemented in other ways; for example, the apparatus embodiments described above are merely illustrative, and the division of modules is merely a logical functional division. In actual implementation, there may be other division methods, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed; furthermore, the coupling or direct coupling or communication connection between the shown or discussed mutuals can be through some interfaces, and the indirect coupling or communication connection between the apparatus or modules can be electrical, mechanical or other forms.

[0105] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A bank credit assessment system based on deep learning, characterized in that, It includes a data acquisition module, a credit assessment module, a credit analysis module, and a decision output module; The data acquisition module constructs customer datasets, transaction datasets, work datasets, and historical credit datasets. The credit analysis module includes a demand analysis unit and a volatility analysis unit. The demand analysis unit performs daily demand analysis for different periods based on customer datasets and transaction datasets to obtain daily demand standard thresholds, and then converts these daily demand standard thresholds into daily demand prediction risk values. The fluctuation analysis unit performs economic fluctuation trend analysis based on customer dataset and transaction dataset, generates economic fluctuation impact factors based on income economic value at turning points, and performs work trend risk fluctuation analysis based on work dataset and work trend change coefficient based on economic fluctuation impact factors. The impact analysis module performs job skill risk analysis using job datasets and customer datasets, establishes a skill equivalent replacement risk model, outputs a job equivalent replacement estimate based on the skill equivalent replacement risk model, simulates equivalent job replacement execution, and generates economic risk judgment signals through a preset evaluation mechanism. The credit assessment module acquires historical credit datasets, work trend change coefficients, and daily demand prediction risk values ​​to generate customer credit assessment values. Based on the customer credit assessment values ​​and the work equivalent replacement estimate, it generates a decision suggestion information sheet.

2. The deep learning-based bank credit assessment system according to claim 1, characterized in that, It also includes a data acquisition module for building customer datasets, transaction datasets, working datasets, and historical credit datasets, specifically including the following: The customer group is pre-stratified into N dimensions based on the customer group, forming M customer categories; By connecting with the bank's CRS system to the customer's authorized tax declaration data and social security payment records, annualized income and monthly average income are collected, and an age verification component is installed to establish a dynamic age verification mechanism in combination with biometric technology. Customers are categorized into corresponding customer categories based on their annual income and age, thus forming a customer dataset. Customer spending data is obtained through transaction records, and the transaction amount and timestamp of each transaction are extracted and analyzed to construct a customer consumption cycle map, which includes the monthly average expenditure, peak expenditure days, and time series features of consumption frequency distribution, forming a bank transaction dataset. Obtain the client's job title, salary, and length of employment, and use intelligent technology analysis software to analyze the client's job skills and identify implicit skills. Establish a knowledge graph-based professional relationship network, including hierarchical relationships between industry, position, and skill, and construct a job skills dataset; By acquiring historical loan records and default counts, a historical credit dataset is constructed to generate a credit history timeline, marking the date of initial credit granting and credit change events. Set up an automatic update strategy on a quarterly basis, and trigger manual review for customers whose income fluctuates abnormally by 30%.

3. The deep learning-based bank credit assessment system according to claim 1, characterized in that, Based on the standard threshold of daily demand, it is converted into a predicted risk value for work demand, specifically including the following: S100. Based on the customer dataset and transaction dataset, obtain customer expenditures and annual customer income, and perform data integration and analysis on customers' daily needs and expenses. S101. Obtain the monthly average expenditure for customers' daily needs within the period, and perform mean-based processing to obtain the standard threshold for demand. Then, organize the total monthly expenditure for customers within the period to obtain the median value of the total monthly expenditure for each month. Finally, calculate the expenditure difference between the highest monthly average expenditure and the monthly average expenditure to generate the prediction threshold. S102. Integrate the actual average proportion with the standard demand threshold to obtain the daily demand forecast risk value.

4. The deep learning-based bank credit assessment system according to claim 1, characterized in that, Generate the coefficient for changes in work trends, specifically including the following: S200. Based on the customer dataset and transaction dataset, obtain customer revenue values ​​and establish an economic value change curve. Plot customer revenue values ​​on the vertical axis and time points on the horizontal axis, and perform curve change rate analysis to obtain the economic fluctuation impact factor X. In the above formula, I T and I T-1 These are the economic values ​​at the time point of the previous cycle and the corresponding economic values ​​at the time point of the current cycle, respectively. S201. Obtain the nature of the customer's work through the work dataset, and assign a value to the customer's work industry sensitivity coefficient M by combining external economic data analysis. S202. Based on the factors influencing customer economic fluctuations and industry sensitivity coefficients, conduct a job trend risk fluctuation analysis to generate a job trend change coefficient B.

5. The deep learning-based bank credit assessment system according to claim 1, characterized in that, Establish a risk model for equivalent skill replacement, which includes the following: The skill equivalent replacement risk model is constructed from an input layer, a processing layer, and an output layer. When constructing the processing layer, physical constraints are set for the initial network model based on a fully connected neural network architecture. The equivalence rule is positioned to obtain the job dataset and customer dataset, perform extended related job searches, extract relevant job datasets for each position based on job skill requirement analysis, and match customer skills with job requirement skills based on the similarity of job relevance. Based on job skill matching, direct and indirect equivalence matching are set, direct and indirect equivalence rules are defined and assigned dynamic weights, and related job matching value analysis is performed based on semantic similarity to obtain a dataset of directly and indirectly equivalent job matching values. An initial function for job risk is constructed based on job position and job skills, and an estimated value of equivalent job replacement is predicted based on job skill matching. The output layer outputs the equivalent replacement estimate for the forecast.

6. The deep learning-based bank credit assessment system according to claim 1, characterized in that, An economic risk assessment signal is generated through a pre-set assessment mechanism, specifically including the following: S300, acquire equivalent replacement work generated by the skill equivalent replacement risk model, perform simulated equivalent work replacement execution operations, and estimate the economic value of the equivalent replacement work; S301. Conduct a partial market salary analysis based on the equivalent replacement work to obtain the integrated salary economic value. Compare and analyze the salary economic value with the customer's preset repayment amount to obtain the difference between the two. S301. When the magnitude difference is less than the prediction threshold, it indicates that the equivalent replacement work is risky, and an economic risk warning signal is generated.

7. The deep learning-based bank credit assessment system according to claim 1, characterized in that, Based on the customer's credit assessment score and the estimated value of equivalent job replacement, a decision recommendation information sheet is generated, which includes the following: S400: Obtain historical credit datasets, work trend change coefficients, and daily demand prediction risk values ​​to conduct customer credit assessment and analysis. S401. By combining the number of defaults (C) with the work trend change coefficient (B) and the daily demand forecast risk value (Y), the customer credit assessment value is obtained and set as P. J represents the number of times the agreement needs to be kept; S402. Based on the customer's credit assessment value and the estimated value of equivalent work replacement, a decision recommendation is generated. When the customer's credit assessment value is lower than the standard credit assessment value set by the bank, and the estimated value of equivalent work replacement is lower than the set replacement standard threshold, an information sheet that does not meet the bank's credit decision recommendation is generated.

8. A deep learning-based bank credit assessment method, characterized in that, Includes the following steps: Step 1: Data Acquisition Module, constructing customer datasets, transaction datasets, working datasets, and historical credit datasets; Step 2: Based on the customer dataset and transaction dataset, conduct daily demand analysis for different periods to obtain the daily demand standard threshold, and convert the daily demand standard threshold into the daily demand prediction risk value. Step 3: Based on the customer dataset and transaction dataset, conduct economic fluctuation trend analysis, generate economic fluctuation impact factors based on the inflection point and income economic value, and conduct work trend risk fluctuation analysis based on the economic fluctuation impact factors and the work dataset to generate work trend change coefficients. Step 4: Conduct job skill risk analysis using job datasets and customer datasets, establish a skill equivalent replacement risk model, output a job equivalent replacement estimate based on the skill equivalent replacement risk model, simulate equivalent job replacement execution, and generate economic risk judgment signals through a preset evaluation mechanism. Step 5: Obtain historical credit datasets, work trend change coefficients, and daily demand forecast risk values; generate customer credit assessment values; and generate decision recommendation information sheets based on customer credit assessment values ​​and work equivalent replacement estimates.

Images