Credit risk analysis method, device, equipment, storage medium and program product

Abstract

This application provides a credit risk analysis method applicable to the fields of big data and artificial intelligence. The method includes: responding to a credit risk analysis request for a target enterprise, acquiring multi-dimensional credit data of the target enterprise, and extracting a set of credit risk features for each dimension based on the credit data. These multiple dimensions include financial, cash flow, and financing dimensions. The method also involves inputting the set of credit risk features for each dimension into a corresponding sub-model to obtain analysis results for that dimension. These results include the initial credit risk level of the target enterprise under different performance periods and risk types in each dimension. Finally, the analysis results are input into a fusion model to obtain the final credit risk level of the target enterprise under different performance periods and risk types across multiple dimensions. This application also provides a credit risk analysis device, equipment, storage medium, and program product.

Description

Technical Field

[0001] This application relates to the fields of big data technology and artificial intelligence technology, and more specifically to a credit risk analysis method, apparatus, equipment, medium and program product. Background Technology

[0002] In the financial industry, companies often need to engage in large-scale debt transactions, and financial institutions such as banks typically analyze and predict a company's credit risk based on its credit data. Therefore, accurately analyzing and predicting a company's credit risk is crucial for both companies and financial institutions, as it can reduce potential financial losses.

[0003] Traditional credit risk analysis methods mainly rely on manual review, expert experience, and static rule models. These methods may have problems such as existing models often outputting a single risk score, failing to distinguish the severity of risks and time dimensions, and having low timeliness, relevance, and accuracy in risk analysis, which is not conducive to refined risk management. Summary of the Invention

[0004] In view of the above problems, embodiments of this application provide a credit risk analysis method, apparatus, device, medium, and program product.

[0005] According to a first aspect of this application, a credit risk analysis method is provided, comprising: in response to a credit risk analysis request for a target enterprise, acquiring credit data of the target enterprise in multiple dimensions, and extracting a set of credit risk features for each dimension based on the credit data, wherein the multiple dimensions include financial dimensions, cash flow dimensions, and financing dimensions; inputting the set of credit risk features for each dimension into a sub-model corresponding to each dimension to obtain analysis results for the corresponding dimension, wherein the analysis results include the initial credit risk level of the target enterprise under different performance periods and different risk types in each dimension; and inputting the analysis results into a fusion model to obtain the final credit risk level of the target enterprise under different performance periods and different risk types in the fusion of multiple dimensions.

[0006] According to embodiments of this application, obtaining credit data of a target enterprise in multiple dimensions includes: obtaining financial dimension credit data based on the target enterprise's financial statement information; obtaining cash flow dimension credit data based on the target enterprise's account cash flow information; and obtaining financing dimension credit data based on the target enterprise's financing business filing records, credit contract information, and guarantee registration information.

[0007] According to an embodiment of this application, the sub-model corresponding to each dimension is trained as follows: credit data of multiple enterprises are obtained, and a set of credit risk features for each enterprise is extracted based on the credit data; the credit risk feature set is labeled to obtain the first training sample for each dimension; a pre-built first model is trained based on the first training sample for each dimension to obtain the sub-model corresponding to each dimension, wherein the first model is constructed based on the gradient boosting tree algorithm.

[0008] According to an embodiment of this application, the fusion model is trained as follows: the credit risk feature set of each enterprise is input into the sub-model corresponding to each dimension to obtain the first analysis result of each enterprise, the first analysis result including the analysis results under multiple dimensions; the first analysis result is labeled to obtain the second training sample of the fusion model; the pre-constructed second model is trained based on the second training sample of the fusion model to obtain the fusion model, wherein the second model is constructed based on the logistic regression model.

[0009] According to embodiments of this application, the method further includes: matching a corresponding risk warning strategy from a preset risk control strategy library based on the final credit risk level and the corresponding preset threshold; generating a risk analysis report for the target enterprise based on credit data, the final credit risk level and the corresponding risk warning strategy, and storing it in the enterprise database for querying.

[0010] According to embodiments of this application, the method further includes: finding related companies that are associated with the target company and querying related risk analysis reports of the related companies in the company database; generating a risk analysis report of the target company based on credit data, the final credit risk level, and the corresponding risk warning strategy, including: generating a risk analysis report of the target company based on related risk analysis reports, credit data, credit risk level, and risk warning strategy.

[0011] According to embodiments of this application, different performance periods include a one-year performance period and a two-year performance period; different risk types include a first risk type and a second risk type, wherein the first risk type represents a higher degree of risk than the second risk type.

[0012] According to a second aspect of this application, a credit risk analysis apparatus is provided, comprising: an acquisition module, configured to, in response to a credit risk analysis request for a target enterprise, acquire credit data of the target enterprise across multiple dimensions, and extract a set of credit risk features for each dimension based on the credit data, wherein the multiple dimensions include a financial dimension, a cash flow dimension, and a financing dimension; a first analysis module, configured to input the set of credit risk features for each dimension into a sub-model corresponding to each dimension to obtain analysis results for the corresponding dimension, wherein the analysis results include the initial credit risk level of the target enterprise under different performance periods and different risk types in each dimension; and a second analysis module, configured to input the analysis results into a fusion model to obtain the final credit risk level of the target enterprise under different performance periods and different risk types in the fusion of multiple dimensions.

[0013] According to a third aspect of this application, an electronic device is provided, comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method described above.

[0014] According to a fourth aspect of this application, a computer-readable storage medium is also provided, on which a computer program or instructions are stored, wherein the computer program or instructions, when executed by a processor, implement the steps of the above-described method.

[0015] According to a fifth aspect of this application, a computer program product is also provided, including a computer program or instructions that, when executed by a processor, implement the steps of the above-described method. Attached Figure Description

[0016] The above-mentioned contents, other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0017] Figure 1 The illustration shows an application scenario of the credit risk analysis method, apparatus, device, medium, and program product according to embodiments of this application;

[0018] Figure 2 A flowchart illustrating a credit risk analysis method according to an embodiment of this application is shown schematically.

[0019] Figure 3 This illustration schematically shows a diagram illustrating the final credit risk rating according to an embodiment of this application;

[0020] Figure 4 A schematic diagram of a training sub-model according to an embodiment of this application is shown.

[0021] Figure 5This illustration schematically shows a flowchart of generating a risk analysis report for a target enterprise according to an embodiment of this application;

[0022] Figure 6 This schematic diagram illustrates the structural block diagram of a credit risk analysis apparatus according to an embodiment of the present application;

[0023] Figure 7 A block diagram of an electronic device suitable for implementing a credit risk analysis method according to an embodiment of this application is shown schematically. Detailed Implementation

[0024] The embodiments of this application will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0025] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0026] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0027] In the technical solution of this application, the user information (including but not limited to user personal information, user image information, user device information, such as location information) and data (including but not limited to data used for analysis, stored data, and displayed data) involved are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of related data all comply with relevant laws, regulations, and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entry points for users to choose to authorize or refuse.

[0028] In scenarios where personal information is used for automated decision-making, the methods, devices, and systems provided in this application all provide users with corresponding operation entry points for users to choose to agree to or reject the automated decision results; if the user chooses to reject, the process enters the expert decision-making process.

[0029] As used in this paper, the term "model" refers to a model that learns the relationship between inputs and outputs from training data, enabling it to generate corresponding outputs for a given input after training. Model generation can be based on machine learning techniques. Deep learning is a machine learning algorithm that processes inputs and provides corresponding outputs using multiple layers of processing units. A neural network model is an example of a deep learning-based model. In this paper, "model" may also be referred to as a "machine learning model," "learning model," "machine learning network," or "learning network," and these terms are used interchangeably.

[0030] It should be understood that machine learning generally includes three phases: training, testing, and application (also known as inference). In the training phase, a given model is trained using a large amount of training data, iteratively updating parameter values ​​until the model can consistently generate inferences that meet the expected goals from the training data. Through training, the model can be considered to have learned the relationship between inputs and outputs (also known as the input-output mapping) from the training data. The parameter values ​​of the trained model are determined. In the testing phase, test inputs are applied to the trained model to test whether it can provide the correct output, thus determining the model's performance. In the application phase, the model can be used to process actual inputs based on the trained parameter values ​​to determine the corresponding output.

[0031] Figure 1 The illustration schematically depicts application scenarios of the credit risk analysis method, apparatus, device, storage medium, and program product according to embodiments of this application. For example... Figure 1 As shown, application scenario 100 according to an embodiment of this application may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired or wireless communication links or fiber optic cables. For example, a user can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send information, etc.

[0032] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be electronic devices such as smartphones, wearable devices, personal computers, intelligent voice interaction devices, smart home appliances, intelligent vehicles, in-vehicle terminals, aircraft, unmanned vending terminals, and extended reality devices. Extended reality devices can include virtual reality devices, augmented reality devices, and mixed reality devices. A client application for the target application can be installed and run on the terminal devices. This target application can include, but is not limited to, financial transaction applications, payment applications, shopping applications, web browser applications, search applications, instant messaging tools, email clients, and social media platform software (these are just examples). Furthermore, this application embodiment does not limit the form of the target application, and it can include, but is not limited to, applications, mini-programs, etc., installed on the terminal devices, and can also be in the form of web pages.

[0033] Server 105 can be a server providing various services, such as a backend management server supporting websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (this is just an example). The backend management server can analyze and process received user requests and other data, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services such as cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks, and basic cloud computing services such as big data. The server can be the backend server of the aforementioned target application, used to provide backend services to the clients of the target application.

[0034] It should be noted that the credit risk analysis method provided in this application embodiment can generally be executed by server 105 and / or terminal devices 101-103. Accordingly, the credit risk analysis device provided in this application embodiment can generally be set in server 105 and / or terminal devices 101-103.

[0035] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0036] Figure 2 A flowchart illustrating a credit risk analysis method according to an embodiment of this application is shown schematically. Figure 2 As shown, the credit risk analysis method 200 according to the embodiments of this application may include steps S210 to S230.

[0037] In step S210, in response to the credit risk analysis request for the target enterprise, the credit data of the target enterprise in multiple dimensions is obtained, and a set of credit risk features for each dimension is extracted based on the credit data. The multiple dimensions include financial dimension, cash flow dimension and financing dimension.

[0038] In the embodiments of this application, the target enterprise can refer to the object requiring credit risk analysis, such as a large or medium-sized general legal entity customer that has established a credit relationship with a financial institution. A credit risk analysis request can refer to a business requirement instruction initiated by a financial institution to conduct credit risk identification and analysis for the target enterprise. The multi-dimensional credit data covers three types of data: financial dimension credit data, cash flow dimension credit data, and financing dimension credit data.

[0039] In embodiments of this application, the consent or authorization of the enterprise user can be obtained before acquiring the enterprise user's information. For example, before step S210, a request to acquire the enterprise user's information can be sent to the enterprise user. If the user consents or authorizes the acquisition of user information, step S210 is executed.

[0040] For example, upon receiving a credit risk analysis request for a target company, three types of credit data for that company are obtained. Then, credit risk-related feature variables are extracted from these three types of credit data to form a credit risk feature set for each dimension. A credit risk feature set can refer to a combination of feature variables related to the company's credit risk extracted from credit data across various dimensions. For instance, the credit risk feature set for the financial dimension includes features such as solvency, size, profitability, operational capacity, growth potential, cash flow status, capital structure, and compliance. The credit risk feature set for the cash flow dimension includes features such as cash flow status, compliance of cash flow, basic information on cash inflows and outflows, cash turnover efficiency, counterparty concentration, account activity, cash accumulation level, abnormal transaction characteristics, and cross-border cash flow. The credit risk feature set for the financing dimension includes features such as debt history, basic company information, related repayment responsibilities, distribution of financing channels, financing cost level, and collateral / guarantee status.

[0041] In step S220, the credit risk feature set of each dimension is input into the sub-model corresponding to each dimension to obtain the analysis results under the corresponding dimension. The analysis results include the initial credit risk level of the target enterprise under different performance periods and different risk types in each dimension.

[0042] In the embodiments of this application, a sub-model can refer to a single-dimensional risk analysis and prediction model constructed for each of the three dimensions: finance, cash flow, and financing. Each sub-model is trained based on the Light Gradient Boosting Machine (LightGBM) algorithm. The performance period can refer to the time window for risk analysis and prediction, such as a 1-year performance period, a 2-year performance period, a 3-year performance period, and / or a 5-year performance period, etc., used to analyze and predict the credit risk of an enterprise in different future time periods. The risk type can refer to the category divided according to the severity of the risk, such as strong risk type, medium risk type, and / or weak risk type, which represent the risk level decreasing in the order of strong, medium, and weak.

[0043] For example, credit risk characteristics across three dimensions—financial, cash flow, and financing—are input into corresponding single-dimensional sub-models. Each sub-model, combined with a preset performance period and risk type, calculates the risk of the target company and ultimately outputs the analysis results for that dimension. This represents the initial credit risk level of the target company under each dimension, different performance periods, and risk type combinations. The analysis results can refer to the output data generated after inputting the target company's single-dimensional credit risk characteristics into the corresponding dimension's sub-model. This includes the probability of risk occurrence and its corresponding credit risk level for the target company under different performance periods and risk types within the corresponding dimension. The initial credit risk level corresponds to a quantified single-dimensional probability of risk occurrence. The probability of risk occurrence (risk rate) refers to the statistical probability of an insured event occurring within a certain period, expressed as a percentage or per thousand. In the insurance field, this indicator is used to measure the frequency of risk occurrence. The initial credit risk level can be obtained based on the probability of risk occurrence and its corresponding threshold; no specific limitations are imposed here, and it can be set according to actual needs.

[0044] For example, the analysis results can include the output of the financial dimension sub-model, specifically: a 18% probability of occurrence for high-risk events (1 year), corresponding to a medium-risk level; a 7% probability of occurrence for low-risk events (1 year), corresponding to a low-risk level; a 22% probability of occurrence for high-risk events (2 years), corresponding to a high-risk level; and a 9% probability of occurrence for low-risk events (2 years), corresponding to a low-risk level. The cash flow dimension sub-model output specifically includes: an 8% probability of occurrence for high-risk events (1 year), corresponding to a low-risk level; a 4% probability of occurrence for low-risk events (1 year), corresponding to a low-risk level; a 12% probability of occurrence for high-risk events (2 years), corresponding to a medium-risk level; and a 6% probability of occurrence for low-risk events (2 years), corresponding to a low-risk level. The financing dimension sub-model output shows a 25% probability of occurrence for high-risk events (1 year), corresponding to a high-risk level; an 11% probability of occurrence for low-risk events (1 year), corresponding to a medium-risk level; a 28% probability of occurrence for high-risk events (2 years), corresponding to a high-risk level; and a 13% probability of occurrence for low-risk events (2 years), corresponding to a medium-risk level. This example is for illustrative purposes only. Performance period, risk type, risk level, etc. are not limited to those in the example and are not specifically limited here. They can be set according to actual needs.

[0045] In step S230, the analysis results are input into the fusion model to obtain the final credit risk level of the target enterprise under different performance periods and different risk types in multiple dimensions.

[0046] In the embodiments of this application, the fusion model can refer to a risk prediction model that integrates multi-dimensional credit data, trained using the initial credit risk levels output by each single-dimensional sub-model as input and based on the Logistic Regression (LR) algorithm. The final credit risk level can refer to the risk level of the target enterprise under different performance periods and different risk types across multiple dimensions, as output by the fusion model.

[0047] Figure 3 This diagram illustrates the process of obtaining the final credit risk rating according to an embodiment of this application. Figure 3As shown, the sets of credit risk features from the financial dimension, the cash flow dimension, and the financing dimension are input into the financial dimension sub-model, respectively. Each sub-model, combined with a preset performance period and risk type, calculates the risk of the target company and outputs the analysis results for the corresponding dimension. This represents the initial credit risk level of the target company under each dimension and for different combinations of performance periods and risk types. All initial credit risk levels output from the three sub-models are then input into the fusion model. The fusion model weights and integrates these, outputting the final credit risk level of the target company after multi-dimensional information fusion, corresponding to different combinations of performance periods and risk types. For example, the fusion model might output the target company's 1-year high-risk level, 1-year low-risk level, 2-year high-risk level, and 2-year low-risk level. This example is for illustrative purposes only; the performance period and risk type are not limited to those shown in the example and are not specifically restricted. They can be set according to actual needs.

[0048] According to the embodiments of this application, by acquiring credit data from three dimensions—financial, cash flow, and financing—and combining the corresponding sub-models for each dimension with a multi-dimensional fusion model, a comprehensive identification and detailed analysis of credit risks for target enterprises at different performance stages and with different risk types can be achieved. This facilitates the differentiation of the severity and time dimension of risks according to actual needs, improves the timeliness, pertinence, and accuracy of risk analysis and prediction, and enhances the pertinence and decision-making efficiency of credit risk management for financial institutions.

[0049] In the embodiments of this application, obtaining credit data of the target enterprise in multiple dimensions includes: obtaining financial dimension credit data based on the target enterprise's financial statement information; obtaining cash flow dimension credit data based on the target enterprise's account cash flow information; and obtaining financing dimension credit data based on the target enterprise's financing business filing records, credit contract information, and guarantee registration information.

[0050] In the embodiments of this application, financial statement information can refer to written documents prepared by the target company reflecting its financial condition, operating results, and cash flows, including balance sheets, income statements, and cash flow statements. Account fund flow information can refer to detailed records of fund inflows and outflows in the accounts opened by the target company at financial institutions, including fund transfer amounts, counterparties, and transaction times. Financing business filing record information can refer to the filing and registration materials and retained records submitted by the target company to financial regulatory authorities or cooperating financial institutions when conducting financing business. Credit contract information can refer to various credit-related contracts signed between the target company and financial institutions, including terms such as credit limits, interest rates, and repayment methods. Guarantee registration information can refer to the security interest registration records processed by the target company with relevant registration agencies when providing collateral to obtain financing, as well as information such as guarantee liability agreements.

[0051] For example, data related to companies that have established credit relationships with financial institutions can be obtained, including their financial information, cash flow records, and financing data. Companies can also be grouped and modeled according to 12 industry categories to ensure the applicability and stability of the model across different business scenarios. Based on the financial statements such as balance sheets and profit and loss statements provided by the target companies, financial dimension credit data for analysis is extracted. Through the financial institutions' account management systems, the target companies' account cash flow information is retrieved to form cash flow dimension credit data. The target companies' financing business filing records, credit contract information signed with financial institutions, and guarantee registration information processed with relevant institutions are integrated to obtain financing dimension credit data.

[0052] According to the embodiments of this application, by acquiring credit data from the financial dimension, credit data from the cash flow dimension, and credit data from the financing dimension, it is beneficial to achieve multi-angle identification of credit risk, avoid the limitations of information from a single dimension, avoid information omissions, and improve the comprehensiveness and accuracy of credit risk analysis.

[0053] In the embodiments of this application, the sub-model corresponding to each dimension is trained as follows: credit data of multiple enterprises are obtained, and credit risk feature set of each enterprise is extracted based on the credit data; the credit risk feature set is labeled to obtain the first training sample of each dimension; the pre-built first model is trained based on the first training sample of each dimension to obtain the sub-model corresponding to each dimension, wherein the first model is constructed based on the gradient boosting tree algorithm.

[0054] In the embodiments of this application, the first training sample may refer to sample data formed by labeling the credit risk features of multiple enterprises across various dimensions with corresponding risk tags, and is used to train a single-dimensional sub-model. The first model may refer to a pre-built basic model framework, whose algorithm is the LightGBM algorithm. The LightGBM algorithm may refer to a multi-step decision tree algorithm based on ensemble learning, which can effectively handle high-dimensional nonlinear data and improve the prediction accuracy and stability of the model.

[0055] Figure 4 A flowchart illustrating the training of sub-models corresponding to each dimension according to an embodiment of this application is shown. Figure 4 As shown, credit data from multiple enterprises with credit relationships with financial institutions across different dimensions are obtained. Credit risk feature sets for these enterprises in three dimensions—financial risk, cash flow risk, and financing risk—are extracted and used as training input. Based on these enterprises' past risk event records, risk labels are applied to the credit risk feature sets for each dimension. These labels include combinations of different performance periods and risk types, serving as training labels and forming the first training sample for each dimension. The first training samples for each dimension are then input into a pre-built first model based on the gradient boosting tree algorithm for training. Through iterative optimization of the first model parameters, sub-models corresponding to each dimension are finally obtained: the financial dimension sub-model, the cash flow dimension sub-model, and the financing dimension sub-model. For example, the label indicating whether a enterprise has experienced a credit risk event is a primary indicator, including secondary indicators such as the occurrence of a strong risk event in the next 1 / 2 years or a weak risk event in the next 1 / 2 years.

[0056] In the embodiments of this application, the parameter configuration for constructing the first model based on the LightGBM algorithm can be as follows:

[0057] The tree structure parameters are as follows:

[0058] The number of decision trees (ntrees) was decreased by five levels: 30, 25, 20, 15, and 10, with 20 being the final number.

[0059] The maximum depth of the decision tree (max_depth) decreases in six increments: 15, 12, 10, 8, 6, and 4, ultimately set to 10.

[0060] The maximum number of leaf nodes in the decision tree (max_leaves) decreases from 15, 12, 10, 8, to 6, ultimately set to 12. This, along with max_depth, forms a "depth-leaf" dual constraint to prevent overfitting.

[0061] Minimum number of samples per leaf node (min_rows): 10

[0062] Minimum Hessian sum in leaf node (min_sum_hessian_in_leaf): 100

[0063] The learning rate and sampling are as follows:

[0064] The learning rate (learn_rate) increases in four increments: 0.05, 0.1, 0.2, and 0.5, ultimately settling at 0.1.

[0065] The bagging fraction is fixed at 0.5, with 50% of the training samples randomly sampled in each iteration.

[0066] Feature sampling rate (col_sample_rate_per_tree, corresponding to feature_fraction): fixed at 0.5, with 50% of features randomly sampled from each tree.

[0067] The grid search and evaluation are as follows:

[0068] Three-fold hierarchical cross-validation was used, with the area under the receiver operating characteristic (AUC) as the evaluation metric. The ranking direction was to maximize this metric. The final optimal combination was: number of decision trees = 20, maximum depth of decision trees = 10, learning rate = 0.1, and maximum number of leaf nodes in decision trees = 12.

[0069] The validation set AUC was 0.874, an improvement of 3.8% over the baseline. The difference between the AUC of the training set and the test set (ΔAUC_train-test) was less than 0.005, meeting the regulatory stability requirements.

[0070] In the embodiments of this application, the financial dimension model takes a feature set constructed from financial statement data and derived financial indicators as input, trains the model using the LightGBM algorithm, and outputs the initial credit risk level corresponding to the probability of risk under the financial dimension; the cash flow dimension model takes a feature set constructed from the customer's account cash flow as input, trains the LightGBM model, and outputs the initial credit risk level corresponding to the probability of risk under the cash flow dimension; the financing dimension model takes a feature set constructed from the customer's financing data as input, trains the LightGBM model, and outputs the initial credit risk level corresponding to the probability of risk under the financing dimension.

[0071] In another embodiment of this application, each single-dimensional model can establish sub-models for different performance periods and different risk types of events, for a total of 12 sub-models. For example, each single-dimensional model establishes sub-models for four types of risk events, such as 1-year strong risk, 1-year weak risk, 2-year strong risk, and 2-year weak risk, for a total of 12 sub-models (3 dimensions multiplied by 4 types of events).

[0072] According to the embodiments of this application, by employing the gradient boosting tree algorithm to construct the first model and training the corresponding sub-models dimensionally, both the prediction accuracy of the model and its strong interpretability are ensured. Risk features are extracted from multi-dimensional credit data (financial, cash flow, financing, etc.) of multiple enterprises and labeled to form the first training sample for each dimension. Sub-models are trained separately for different dimensions, which facilitates the comprehensiveness and standardization of the training data for the sub-models, enhances the sub-models' ability to identify and analyze risks in specific dimensions, and enables multi-faceted analysis of corporate credit risk, providing reliable support for the subsequent fusion model to output accurate credit risk levels.

[0073] In the embodiments of this application, the fusion model is trained as follows: the credit risk feature set of each enterprise is input into the sub-model corresponding to each dimension to obtain the first analysis result of each enterprise, the first analysis result including the analysis results under multiple dimensions; the first analysis result is labeled to obtain the second training sample of the fusion model; the pre-constructed second model is trained based on the second training sample of the fusion model to obtain the fusion model, wherein the second model is constructed based on the logistic regression model.

[0074] The first analysis result can refer to the initial risk level corresponding to different performance periods and risk types for each enterprise across three dimensions: finance, cash flow, and financing. The second training sample can refer to the sample data formed by labeling the actual risk events of multiple enterprises based on the first analysis results, used to train the fusion model. The second model can refer to the pre-built basic model framework, whose algorithm is logistic regression (LR), meaning the second model can be a logistic regression model. Logistic regression is a generalized linear model primarily used for classification problems. It predicts the probability of a sample belonging to a certain category by linearly combining features and mapping the results to the (0,1) interval using a function. Logistic regression can not only predict classification results but also provide a probability estimate of belonging to a specific category.

[0075] In the embodiments of this application, a fusion model is constructed based on the output results of each single-dimensional sub-model. Different LR models are trained using varying performance periods and risk levels as divisions. The input to each model is the probability of occurrence output by the sub-model for each dimension, used to determine the corresponding final credit risk level. For example, the analysis results of strong risk over a one-year performance period under the financial, financing, and cash flow dimensions are used as training samples for the LR model. The output is the final probability of a customer experiencing a strong risk event within the next year, used to determine the corresponding final credit risk level. The fusion model includes four predictive LR sub-models. The fusion model facilitates the integration of multi-dimensional information to predict the final credit risk level under different performance periods and risk types. Each sub-model and the fusion model can be deployed within financial institutions such as banks, supporting monthly automatic updates and branch-level deployment, ensuring that the analysis and prediction results of the fusion model remain consistent with the latest status of the target enterprise, achieving dynamic identification and rapid response to credit risk.

[0076] According to the embodiments of this application, by constructing a second model using the LR algorithm, the advantages of the algorithm's high stability and strong interpretability are leveraged, balancing prediction accuracy and interpretability to meet both regulatory and business needs. Second training samples are obtained by labeling the first analysis results output by each dimension's sub-model, and a fusion model is trained. This breaks through the limitations of single-dimensional information, fully explores the correlations between risk characteristics across dimensions, improves the comprehensiveness and accuracy of credit risk prediction, and provides reliable decision support for financial institutions to conduct refined credit risk management.

[0077] In the embodiments of this application, a corresponding risk warning strategy is matched from a preset risk control strategy library based on the final credit risk level and the corresponding preset threshold; a risk analysis report of the target enterprise is generated based on the credit data, the final credit risk level and the corresponding risk warning strategy, and stored in the enterprise database for querying.

[0078] In the embodiments of this application, the preset threshold can refer to the critical value set by financial institutions based on business experience and risk control requirements for the final credit risk level of different performance periods and different risk types, used as a criterion for classifying the level of credit risk. The risk control strategy library can refer to a set of strategies pre-established by financial institutions, covering corresponding response measures for various risk levels, such as different levels of risk warning strategies like rigid control of strong risks, tracking and monitoring of medium risks, and routine management of weak risks. Risk warning strategies can refer to the response measures corresponding to various risk levels. A risk analysis report can refer to a comprehensive document integrating target enterprise credit data, final credit risk level, risk warning strategies, and other information, used to intuitively present the target enterprise's risk status and response suggestions. An enterprise database can refer to a database used by financial institutions to store basic data and risk analysis reports for multiple enterprises.

[0079] For example, after obtaining the final credit risk level of the target company, this level is compared with the risk threshold preset by the financial institution. Based on the comparison result, a corresponding risk warning strategy is matched from a pre-built risk control strategy library. By integrating the target company's financial, cash flow, and financing data across three dimensions, along with the final credit risk level and the matched risk warning strategy, a complete and logically clear target company risk analysis report is generated. This report is stored in the company's database for subsequent risk review, customer management, and business decision-making. A target company risk analysis report can also be automatically generated monthly and distributed to branches to guide risk assessment and customer management.

[0080] According to the embodiments of this application, by matching the final credit risk level with the corresponding risk warning strategy, a high degree of adaptation between risk control measures and corporate risk status is achieved, improving the timeliness and pertinence of credit risk intervention; by generating risk analysis reports and storing them in the corporate database, it is beneficial to provide financial institutions with comprehensive and intuitive decision-making basis, realize the standardized retention and traceability of risk information, support subsequent risk review, customer segmentation management and business strategy optimization, and improve the long-term effectiveness and scientific nature of credit risk management.

[0081] In the embodiments of this application, the process of finding related companies that are associated with the target company and querying the related risk analysis reports of the related companies in the enterprise database; generating the risk analysis report of the target company based on credit data, the final credit risk level and the corresponding risk warning strategy includes: generating the risk analysis report of the target company based on the related risk analysis report, credit data, credit risk level and risk warning strategy.

[0082] In the embodiments of this application, related enterprises can refer to enterprises that have a relationship with the target enterprise, such as equity control, business dealings, or joint guarantees. The risk status of such enterprises may have a mutual impact with that of the target enterprise. The related enterprise risk analysis report can refer to a credit risk analysis report of related enterprises that is pre-stored in the enterprise database, which includes the related enterprise's own credit data, final credit risk level, and matching risk warning strategies.

[0083] Figure 5 A flowchart illustrating the generation of a risk analysis report for a target enterprise according to an embodiment of this application is shown. Figure 5As shown, in step S510, related companies with a relationship to the target company are identified, and related risk analysis reports of these companies are retrieved from the company database. In step S520, based on the final credit risk level and the corresponding preset threshold, a corresponding risk warning strategy is matched from a preset risk control strategy library. In step S530, a risk analysis report for the target company is generated based on the related risk analysis report, credit data, credit risk level, and risk warning strategy. Alternatively, step S520 can be executed first, followed by step S510.

[0084] For example, before generating a risk analysis report for a target company, companies with related interests can be identified based on information such as the target company's equity structure, business cooperation list, and guarantee relationship ledger. Then, archived related risk analysis reports for these related companies can be retrieved from the company database. When generating the target company's risk analysis report, the related risk analysis reports of these companies are integrated with the target company's own credit data, final credit risk level, and matching risk warning strategies to generate the target company's risk analysis report, thereby improving the completeness of the risk analysis. For example, multi-level risk warning strategies can be designed. For instance, a high-risk warning strategy can automatically trigger risk alerts for companies whose predicted risk level exceeds a set threshold, suggesting rigid risk control measures such as suspending credit granting and strengthening post-loan inspections. A complementary long-term and short-term warning strategy can combine predictions for 1-year and 2-year periods to identify high-risk customers in the short term and potential high-risk customers in the long term, and formulate differentiated management strategies.

[0085] According to the embodiments of this application, by finding the related enterprises of the target enterprise and retrieving their related risk analysis reports, and integrating them into the risk analysis report of the target enterprise, the related transmission analysis of the risk of the target enterprise can be realized, thereby improving the comprehensiveness and completeness of the risk analysis, allowing risk control decisions to fully consider the risk transmission impact of related enterprises, improving the accuracy and foresight of credit risk management, and providing support for financial institutions to build a more three-dimensional and comprehensive risk prevention and control system.

[0086] In the embodiments of this application, different performance periods include a one-year performance period and a two-year performance period; different risk types include a first risk type and a second risk type, wherein the first risk type represents a higher degree of risk than the second risk type.

[0087] In the embodiments of this application, the first risk type can refer to the strong risk type in credit risk analysis and prediction, which is a risk category that represents a high degree of corporate credit risk, covering risk situations such as major defaults, large amounts of overdue debt, and a sharp deterioration in operating conditions that may seriously damage the credit assets of financial institutions. The second risk type can refer to the weak risk type in credit risk assessment, which is a risk category that represents a low degree of corporate credit risk, covering risk situations such as small amounts of overdue debt and slight short-term cash flow problems that do not pose a substantial threat to the safety of credit assets. The first risk type represents a higher degree of risk than the second risk type.

[0088] For example, in the process of analyzing the credit risk of a target enterprise, the time window for risk prediction, i.e., the performance period, is determined to include two types: a 1-year performance period and a 2-year performance period, used to predict the risk evolution trend of the enterprise in the next 1 year and the next 2 years, respectively. Simultaneously, the risk level classification dimension, i.e., the risk type, is determined to include two types: a first risk type and a second risk type, with the first risk type having a significantly higher risk level than the second risk type. For instance, the observation period is the past three years, and the performance period is set to 1 year and 2 years, used to train various models to predict the risk evolution trend of the enterprise within different time windows. The observation period can refer to the time interval used to collect historical data of sample enterprises during the training process of each model.

[0089] According to the embodiments of this application, by clarifying the 1-year and 2-year performance periods of risk analysis and prediction and the classification criteria for the first and second risk types, a refined and tiered analysis of the credit risk of the target enterprise can be achieved. This provides a clear basis for financial institutions to distinguish the severity of risks, facilitates the formulation of differentiated early warning and control strategies for different risk levels, enhances the foresight, timeliness and initiative of credit risk management, and meets the regulatory requirements and business decision-making needs for the standardization of risk assessment criteria.

[0090] Based on the above-described credit risk analysis method, embodiments of this application also provide a credit risk analysis device. The following will be combined with... Figure 6 The device is described in detail.

[0091] Figure 6 A schematic block diagram of a credit risk analysis apparatus according to an embodiment of this application is shown.

[0092] like Figure 6 As shown, the credit risk analysis device 600 of this embodiment includes an acquisition module 610, a first analysis module 620, and a second analysis module 630.

[0093] The acquisition module 610 is used to respond to a credit risk analysis request for a target enterprise, acquire credit data of the target enterprise across multiple dimensions, and extract a set of credit risk features for each dimension based on the credit data. These multiple dimensions include financial dimensions, cash flow dimensions, and financing dimensions. In one embodiment, the acquisition module 610 can be used to execute step S210 as described above, which will not be repeated here.

[0094] The first analysis module 620 is used to input the credit risk feature set of each dimension into the sub-model corresponding to each dimension to obtain the analysis results under the corresponding dimension. The analysis results include the initial credit risk level of the target enterprise under different performance periods and different risk types in each dimension. In one embodiment, the first analysis module 620 can be used to execute step S220 described above, which will not be repeated here.

[0095] The second analysis module 630 is used to input the analysis results into the fusion model to obtain the final credit risk level of the target enterprise under different performance periods and different risk types across multiple dimensions. In one embodiment, the second analysis module 630 can be used to perform step S230 described above, which will not be repeated here.

[0096] According to an embodiment of this application, the acquisition module 610 is further configured to acquire financial dimension credit data based on the target company's financial statement information; acquire cash flow dimension credit data based on the target company's account cash flow information; and acquire financing dimension credit data based on the target company's financing business filing record information, credit contract information, and guarantee registration information.

[0097] According to an embodiment of this application, the credit risk analysis device 600 further includes a sub-model training module, which is used to acquire credit data of multiple enterprises and extract a set of credit risk features for each enterprise based on the credit data; label the credit risk feature set to obtain a first training sample for each dimension; and train a pre-built first model based on the first training sample for each dimension to obtain a sub-model corresponding to each dimension, wherein the first model is constructed based on the gradient boosting tree algorithm.

[0098] According to an embodiment of this application, the credit risk analysis device 600 further includes a fusion model training module. The fusion model training module is used to input the credit risk feature set of each enterprise into the sub-model corresponding to each dimension to obtain the first analysis result of each enterprise. The first analysis result includes the analysis results under multiple dimensions. The first analysis result is labeled to obtain the second training sample of the fusion model. The second model is trained based on the second training sample of the fusion model to obtain the fusion model. The second model is constructed based on the logistic regression model.

[0099] According to an embodiment of this application, the credit risk analysis device 600 further includes a risk warning module. The risk warning module is used to match a corresponding risk warning strategy from a preset risk control strategy library based on the final credit risk level and the corresponding preset threshold; generate a risk analysis report for the target enterprise based on credit data, the final credit risk level and the corresponding risk warning strategy, and store it in the enterprise database for querying.

[0100] According to an embodiment of this application, the risk warning module is also used to find related companies that are related to the target company, and query the related risk analysis report of the related companies in the enterprise database; and generate a risk analysis report of the target company based on the related risk analysis report, credit data, credit risk level and risk warning strategy.

[0101] According to an embodiment of this application, the credit risk analysis device 600 further includes a performance period module and a risk type module. The performance period module is used to set different performance periods, including a one-year performance period and a two-year performance period; the risk type module is used to set different risk types, including a first risk type and a second risk type, wherein the first risk type represents a higher level of risk than the second risk type.

[0102] According to embodiments of this application, any plurality of modules among the acquisition module 610, the first analysis module 620, and the second analysis module 630 can be combined into one module, or any one of these modules can be split into multiple modules. Alternatively, at least part of the functionality of one or more of these modules can be combined with at least part of the functionality of other modules and implemented in one module. According to embodiments of this application, at least one of the acquisition module 610, the first analysis module 620, and the second analysis module 630 can be at least partially implemented as a hardware circuit, such as a field-programmable gate array, a programmable logic array, a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit, or any other reasonable means of integrating or packaging the circuit, or implemented in any one of the three implementation methods of software, hardware, and firmware, or in a suitable combination of any of these. Alternatively, at least one of the acquisition module 610, the first analysis module 620, and the second analysis module 630 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0103] Figure 7 A block diagram of an electronic device suitable for implementing a credit risk analysis method according to an embodiment of this application is shown schematically.

[0104] like Figure 7As shown, an electronic device 700 according to an embodiment of this application includes a processor 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory 702 or a program loaded from a storage portion 708 into a random access memory 703. The processor 701 may include, for example, a general-purpose microprocessor, an instruction set processor and / or an associated chipset and / or a dedicated microprocessor. The processor 701 may also include onboard memory for caching purposes. The processor 701 may include a single processing unit or multiple processing units for executing different steps of the method flow according to an embodiment of this application.

[0105] Random access memory 703 stores various programs and data required for the operation of electronic device 700. Processor 701, read-only memory 702, and random access memory 703 are interconnected via bus 704. Processor 701 executes various steps of the method flow according to embodiments of this application by executing programs in read-only memory 702 and / or random access memory 703. It should be noted that the programs may also be stored in one or more memories other than read-only memory 702 and random access memory 703. Processor 701 may also execute various steps of the method flow according to embodiments of this application by executing programs stored in said one or more memories.

[0106] According to embodiments of this application, the electronic device 700 may further include an input / output interface 705, which is also connected to a bus 704. The electronic device 700 may also include one or more of the following components connected to the input / output interface 705: an input section 706 including a keyboard, mouse, etc.; an output section 707 including a cathode ray tube, liquid crystal display, etc., and a speaker, etc.; a storage section 708 including a hard disk, etc.; and a communication section 709 including a network interface card, such as a local area network card, modem, etc. The communication section 709 performs communication processing via a network such as the Internet. A drive 710 is also connected to the input / output interface 705 as needed. A removable medium 711, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 710 as needed so that computer programs read from it can be installed into the storage section 708 as needed.

[0107] Embodiments of this application also provide a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of this application.

[0108] According to embodiments of this application, the computer-readable storage medium can be a non-volatile computer-readable storage medium, such as including but not limited to: portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory, portable compact disk read-only memory, optical storage devices, magnetic storage devices, or any suitable combination thereof. In embodiments of this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this application, the computer-readable storage medium may include the read-only memory 702, and / or random access memory 703, and / or one or more memories other than read-only memory 702 and random access memory 703 described above.

[0109] Embodiments of this application also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code is used to cause the computer system to implement the methods provided in the embodiments of this application.

[0110] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 709, and / or installed from a removable medium 711. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0111] In embodiments of this application, the computer program can be downloaded and installed from a network via communication section 709, and / or installed from removable medium 711. When the computer program is executed by processor 701, it performs the functions defined in the system of embodiments of this application. According to embodiments of this application, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0112] According to embodiments of this application, program code for executing the computer programs provided in the embodiments of this application can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0113] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0114] Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application.

Claims

1. A credit risk analysis method, characterized in that, The method includes: In response to a credit risk analysis request for a target company, the system acquires credit data for the target company across multiple dimensions and extracts a set of credit risk features for each dimension based on the credit data. The multiple dimensions include financial dimensions, cash flow dimensions, and financing dimensions. The credit risk feature set of each dimension is input into the sub-model corresponding to each dimension to obtain the analysis results under the corresponding dimension. The analysis results include the initial credit risk level of the target enterprise under different performance periods and different risk types in each dimension. The analysis results are input into the fusion model to obtain the final credit risk level of the target enterprise under the different performance periods and different risk types of the multiple dimensions.

2. The method according to claim 1, characterized in that, The acquisition of multi-dimensional credit data of the target enterprise includes: Obtain financial dimension credit data based on the target company's financial statement information; Credit data based on the target company's account fund flow information is obtained from the fund flow dimension; and Credit data in the financing dimension is obtained based on the target company's financing business filing records, credit contract information, and guarantee registration information.

3. The method according to claim 1, characterized in that, The sub-model corresponding to each dimension is trained in the following manner: Obtain the credit data of multiple enterprises, and extract the credit risk feature set of each enterprise based on the credit data; The credit risk feature set is labeled to obtain the first training sample for each dimension; A pre-constructed first model is trained based on the first training samples of each dimension to obtain the sub-model corresponding to each dimension, wherein the first model is constructed based on the gradient boosting tree algorithm.

4. The method according to claim 3, characterized in that, The fusion model is trained as follows: The credit risk feature set of each enterprise is input into the sub-model corresponding to each dimension to obtain the first analysis result of each enterprise. The first analysis result includes the analysis results under the multiple dimensions. The first analysis results are labeled to obtain the second training samples of the fusion model; The second model is trained based on the second training samples of the fusion model to obtain the fusion model, wherein the second model is constructed based on the logistic regression model.

5. The method according to claim 1, characterized in that, Also includes: Based on the final credit risk level and the corresponding preset threshold, a corresponding risk warning strategy is matched from the preset risk control strategy library; A risk analysis report for the target company is generated based on the credit data, the final credit risk level, and the corresponding risk warning strategy, and stored in the company database for future reference.

6. The method according to claim 5, characterized in that, Also includes: Based on the target company, find related companies that are associated with the target company, and query the related risk analysis report of the related companies in the company database; The process of generating a risk analysis report for the target company based on the credit data, the final credit risk level, and the corresponding risk warning strategy includes: A risk analysis report for the target company is generated based on the associated risk analysis report, the credit data, the credit risk level, and the risk warning strategy.

7. The method according to claim 1, characterized in that, The different performance periods include a one-year performance period and a two-year performance period; the different risk types include a first risk type and a second risk type, wherein the first risk type represents a higher degree of risk than the second risk type.

8. A credit risk analysis device, characterized in that, The device includes: The acquisition module is used to respond to a credit risk analysis request for a target enterprise, acquire credit data of the target enterprise in multiple dimensions, and extract a set of credit risk features for each dimension based on the credit data, wherein the multiple dimensions include financial dimension, cash flow dimension and financing dimension; The first analysis module is used to input the credit risk feature set of each dimension into the sub-model corresponding to each dimension to obtain the analysis results under the corresponding dimension. The analysis results include the initial credit risk level of the target enterprise under different performance periods and different risk types in each dimension; and The second analysis module is used to input the analysis results into the fusion model to obtain the final credit risk level of the target enterprise under the different performance periods and different risk types of the multiple dimensions.

9. An electronic device, comprising: One or more processors; Memory, used to store one or more computer programs. The characteristic feature is that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 7.

11. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 7.

Images