Credit risk assessment method and device, electronic device, and storage medium

Abstract

The present invention discloses a credit risk assessment method and apparatus, electronic device, and storage medium. The method comprises: selecting multiple financial indicators for credit assessment of a subject to be assessed to obtain an indicator system; ranking the indicator system using a random forest algorithm to obtain an indicator ranking result; converting multiple screening indicators ranked before a preset numerical threshold in the indicator ranking result using a principal component analysis strategy into multiple new comprehensive indicators; and, based on the multiple new comprehensive indicators, using a pre-constructed logistic regression model to measure the credit risk index of the subject to be assessed, wherein the credit risk index is used to determine whether the subject to be assessed is a credit risk subject. The present invention solves the technical problem of large errors in the assessment results of credit assessment strategies used by financial institutions in related technologies.

Description

Technical Field

[0001] The present invention relates to the field of financial technology, and in particular to a credit risk assessment method and device, electronic equipment, and storage medium. Background Art

[0002] The stable operation of financial institutions is of great significance to the financial market. Among the various risks in the operation of financial institutions, credit risk is one of the most important risks. Failure to effectively control credit risk will directly lead to an increase in the non-performing loans of financial institutions, and even cause the bankruptcy of financial institutions, causing panic in the financial market.

[0003] As for the current corporate credit risk measurement methods, the credit system construction is still not perfect, there are limitations in subjective evaluation of customer levels, and personnel costs are too high.

[0004] In related technologies, financial institutions have attempted to employ various credit risk measurement models to assess the credit risk of each lender. Currently, the most widely used credit risk measurement models include the KMV model and the credit metrics model. The KMV model primarily measures the probability of corporate default based on the lender's stock market performance. This model uses stock price and debt information to provide a predictive asset value estimate for a company's assets. However, this model is only applicable to listed companies, limiting its predictive capabilities for non-listed companies. It also restricts future model upgrades and may limit analysis of corporate default rates based solely on stock market information. The credit metrics model, on the other hand, calculates the credit risk VaR (Value at Risk) by analyzing data such as a company's credit default rate, loss rate, credit rating, and credit rating transitions over a period of time. This can be used to calculate the loss value of loans or loan portfolios under different credit profiles. However, this model also has a significant drawback: its assumptions about the rating transition matrix and default rate are overly ideal, inconsistent with actual market conditions, leading to prediction errors.

[0005] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0006] The embodiments of the present invention provide a credit risk assessment method and device, an electronic device, and a storage medium to at least solve the technical problem of large errors in assessment results of credit assessment strategies adopted by financial institutions in related technologies.

[0007] According to one aspect of an embodiment of the present invention, a credit risk assessment method is provided, comprising: selecting multiple financial indicators for credit assessment of a subject to be assessed to obtain an indicator system; sorting the indicator system using a random forest algorithm to obtain an indicator sorting result; converting multiple screening indicators ranked before a preset numerical threshold in the indicator sorting result into multiple new comprehensive indicators using a principal component analysis strategy; and measuring the credit risk index of the subject to be assessed using a pre-built logistic regression model based on the multiple new comprehensive indicators, wherein the credit risk index is used to determine whether the subject to be assessed is a credit risk subject.

[0008] Optionally, the step of selecting multiple financial indicators for credit assessment of the subject to be assessed to obtain an indicator system includes: obtaining financial data from the financial disclosure information of multiple assessed subjects; selecting a set of indicators that are crisis-related to the financial status of the subject to be assessed from the financial data; and screening out multiple financial indicators corresponding to preset financial indicator categories from the indicator set to obtain the indicator system, wherein the preset financial indicator categories include at least one of the following: debt repayment ability, subject development ability, subject operating ability, subject profitability, cash flow ability, and subject scale.

[0009] Optionally, the step of using a random forest algorithm to sort the indicator system to obtain an indicator sorting result includes: calculating the Gini index of the current node based on the number of subject categories and the proportion of each subject category in the current node to which the financial indicator to be evaluated belongs; obtaining the change in the Gini index of the previous node and the next node obtained after branching based on the current node; calculating the weight value of all the financial indicators to be evaluated at the current node based on the Gini index of the current node and the change in the Gini index of the previous node and the next node; and sorting all the financial indicators in the indicator system based on the weight value of each financial indicator to be evaluated to obtain the indicator sorting result.

[0010] Optionally, the step of using a principal component analysis strategy to convert multiple screening indicators ranked before a preset numerical threshold in the indicator ranking result into multiple new comprehensive indicators includes: screening the indicator ranking result according to the preset numerical threshold to obtain the multiple screening indicators; calculating the covariance matrix of all the screening indicators; standardizing the matrix data in the covariance matrix; and using the principal component analysis strategy to perform principal component analysis on the matrix data after standardization to obtain the multiple new comprehensive indicators.

[0011] Optionally, the step of calculating the covariance matrix of all the screening indicators includes: calculating the indicator average of all the screening indicators based on the indicator data corresponding to each of the screening indicators; calculating the covariance between every two of the screening indicators based on the indicator average of all the screening indicators and the total number of the screening indicators; and determining the covariance matrix corresponding to all the screening indicators by integrating the covariance between every two of the screening indicators.

[0012] Optionally, the principal component analysis strategy is adopted to perform principal component analysis on the matrix data after standardization, including: calculating the data variance based on the indicator data of all the screening indicators; calculating the correlation coefficient matrix that has an association relationship with the matrix data based on the covariance between each two screening indicators and the variance of the indicator data corresponding to multiple screening indicators; calculating the eigenvalues ​​and corresponding eigenvectors of the correlation coefficient matrix based on the correlation coefficient matrix and a pre-configured unit matrix to complete the principal component analysis.

[0013] Optionally, based on the correlation coefficient matrix and a preconfigured unit matrix, the eigenvalues ​​and corresponding eigenvectors of the correlation coefficient matrix are calculated to complete the step of principal component analysis, including: based on the correlation coefficient matrix and the preconfigured unit matrix, the eigenvalues ​​of the correlation coefficient matrix are calculated; substituting the correlation coefficient matrix and the corresponding eigenvalues ​​into a preset characteristic equation group, parsing the preset characteristic equation group, and obtaining a plurality of the eigenvectors, wherein each eigenvector is used to characterize the new comprehensive indicator.

[0014] Optionally, the step of using a pre-built logistic regression model to measure the credit risk index of the subject to be evaluated based on the multiple new comprehensive indicators includes: using the multiple new comprehensive indicators as variables of the logistic regression, substituting them into the selected sample data, and obtaining logistic regression coefficients; substituting the logistic regression coefficients into the logistic regression model to measure the violation parameters of the subject to be evaluated; and evaluating the credit risk index of the subject to be evaluated based on the violation parameters of the subject to be evaluated.

[0015] According to another aspect of an embodiment of the present invention, a credit risk assessment device is also provided, including: a selection unit, used to select multiple financial indicators for credit assessment of a subject to be assessed to obtain an indicator system; a sorting unit, used to sort the indicator system using a random forest algorithm to obtain an indicator sorting result; a conversion unit, used to convert multiple screening indicators ranked before a preset numerical threshold in the indicator sorting result into multiple new comprehensive indicators using a principal component analysis strategy; a measurement unit, used to measure the credit risk index of the subject to be assessed based on the multiple new comprehensive indicators using a pre-built logistic regression model, wherein the credit risk index is used to determine whether the subject to be assessed is a credit risk subject.

[0016] Optionally, the selection unit includes: a first acquisition module for acquiring financial data from the financial disclosure information of multiple evaluated entities; a first selection module for selecting a set of indicators that are crisis-related to the financial status of the entity to be evaluated from the financial data; a first screening module for screening out multiple financial indicators corresponding to preset financial indicator categories from the indicator set to obtain the indicator system, wherein the preset financial indicator categories include at least one of the following: debt repayment ability, entity development ability, entity operating ability, entity profitability, cash flow ability, and entity scale.

[0017] Optionally, the sorting unit includes: a first calculation module, used to calculate the Gini index of the current node based on the number of subject categories and the proportion of each subject category in the current node to which the financial indicator to be evaluated belongs; a second acquisition module, used to obtain the change in the Gini index of the previous node and the next node obtained after branching based on the current node; a second calculation module, used to calculate the weight values ​​of all the financial indicators to be evaluated at the current node based on the Gini index of the current node, the change in the Gini index of the previous node and the next node; a sorting module, used to sort all the financial indicators in the indicator system based on the weight value of each financial indicator to be evaluated, and obtain the indicator sorting result.

[0018] Optionally, the conversion unit includes: a second screening module, used to screen the indicator sorting results according to the preset numerical threshold to obtain the multiple screening indicators; a third calculation module, used to calculate the covariance matrix of all the screening indicators; a standardization processing module, used to standardize the matrix data in the covariance matrix; an analysis module, used to adopt the principal component analysis strategy to perform principal component analysis on the matrix data after standardization to obtain the multiple new comprehensive indicators.

[0019] Optionally, the third calculation module includes: a first calculation submodule, used to calculate the indicator average value of all the screening indicators based on the indicator data corresponding to each of the screening indicators; a second calculation submodule, used to calculate the covariance between every two of the screening indicators based on the indicator average value of all the screening indicators and the total number of the screening indicators; a first determination submodule, used to integrate the covariance between every two of the screening indicators and determine the covariance matrix corresponding to all the screening indicators.

[0020] Optionally, the analysis module includes: a third calculation submodule, used to calculate the data variance based on the indicator data of all the screening indicators; a fourth calculation submodule, used to calculate the correlation coefficient matrix that has an association relationship with the matrix data based on the covariance between each two of the screening indicators and the variance of the indicator data corresponding to multiple screening indicators; a fifth calculation submodule, used to calculate the eigenvalues ​​and corresponding eigenvectors of the correlation coefficient matrix based on the correlation coefficient matrix and a pre-configured unit matrix to complete principal component analysis.

[0021] Optionally, the fifth calculation submodule includes: a sixth calculation submodule, used to calculate the eigenvalues ​​of the correlation coefficient matrix based on the correlation coefficient matrix and a preconfigured unit matrix; an analysis submodule, used to substitute the correlation coefficient matrix and the corresponding eigenvalues ​​into a preset characteristic equation group, analyze the preset characteristic equation group, and obtain a plurality of eigenvectors, wherein each eigenvector is used to characterize the new comprehensive indicator.

[0022] Optionally, the measurement unit includes: a substitution module, used to substitute the multiple new comprehensive indicators as variables of logistic regression into the selected sample data to obtain logistic regression coefficients; a measurement module, used to substitute the logistic regression coefficients into the logistic regression model to measure the violation parameters of the subject to be evaluated; and an evaluation module, used to evaluate the credit risk index of the subject to be evaluated based on the violation parameters of the subject to be evaluated.

[0023] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is further provided, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute any one of the credit risk assessment methods described above.

[0024] According to another aspect of an embodiment of the present invention, an electronic device is also provided, comprising one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement any one of the credit risk assessment methods described above.

[0025] In the present invention, a plurality of financial indicators for credit assessment of the subject to be assessed are selected to obtain an indicator system, a random forest algorithm is used to sort the indicator system to obtain an indicator sorting result, a principal component analysis strategy is used to convert a plurality of screening indicators ranked before a preset numerical threshold in the indicator sorting result into a plurality of new comprehensive indicators, and based on the plurality of new comprehensive indicators, a pre-built logistic regression model is used to measure the credit risk index of the subject to be assessed, wherein the credit risk index is used to determine whether the subject to be assessed is a credit risk subject. In the present invention, based on the pre-selected financial data indicators, the principal component analysis method is used to convert the screened multiple indicators into several comprehensive indicators, and finally logistic regression is used to measure the credit risk, so that financial institutions can effectively obtain the credit risk and enterprise classification of the enterprise through the existing financial information of the subject to be assessed, making the credit assessment results more objective and reliable, thereby predicting the enterprise category and credit risk level, reducing the non-performing loans caused by the operating problems of the subject to be assessed, thereby solving the technical problem of large errors in the credit assessment strategies adopted by financial institutions in the related art. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0027] Figure 1 is a flow chart of an optional credit risk assessment method according to an embodiment of the present invention;

[0028] Figure 2 is a schematic diagram of an optional credit risk assessment device according to an embodiment of the present invention;

[0029] Figure 3 The figure is a hardware structure block diagram of an electronic device (or mobile device) according to a credit risk assessment method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0030] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0031] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0032] To facilitate those skilled in the art to understand the present invention, some of the terms or nouns involved in the embodiments of the present invention are explained below:

[0033] Random forest: A classifier containing multiple decision trees that can be used to evaluate the importance of indicators. The idea is to evaluate the contribution of each feature (indicator) on each tree in the random forest, and then take the average to compare the contribution of different indicators. The Gini coefficient is used here to measure the indicator. The formula is Wherein, k is the number of categories, and in the present invention, k is 2, and pmk represents the proportion of category k in node m.

[0034] Covariance: the overall error of two variables, in is the mean value of X, is the mean value of Y, and n is the number of samples.

[0035] Correlation coefficient: refers to the degree of linear correlation between variables.

[0036] It should be noted that the credit risk assessment method and apparatus thereof disclosed herein can be used in the field of financial technology when conducting credit risk assessment on various market entities (such as enterprises), loan entities, etc., and can also be used in any field other than the field of financial technology when conducting credit risk assessment on various market entities (such as enterprises), loan entities, etc. The application field of the credit risk assessment method and apparatus thereof disclosed herein is not limited.

[0037] It should be noted that the relevant information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display and analysis, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties. For example, an interface is set up between this system and the relevant user or organization. Before obtaining relevant information, it is necessary to send an acquisition request to the aforementioned user or organization through the interface, and obtain the relevant information after receiving the consent information fed back by the aforementioned user or organization.

[0038] The following embodiments of the present invention can be applied to various credit assessment systems, credit investigation software, and credit risk assessment software / interfaces in loan issuance systems. The relationship between enterprise classification and its financial information is found through a logistic regression model, and the decision boundary of the enterprise classification is found, so that the judgment result is more objective and reliable, thereby predicting the enterprise category and credit risk level. Through the credit risk assessment strategy of the present invention, it can help various financial institutions (for example, banking institutions) to screen out entities with high credit risks (for example, enterprises), and help financial institutions / fund issuance APP / fund issuance interfaces to conduct multi-faceted analysis of the enterprise's repayment ability in various credit usage scenarios (for example, before the loan), and avoid non-performing loans caused by business problems of the enterprise.

[0039] The present invention will be described in detail below with reference to various embodiments.

[0040] Example 1

[0041] According to an embodiment of the present invention, an embodiment of a credit risk assessment method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0042] Figure 1 is a flow chart of an optional credit risk assessment method according to an embodiment of the present invention. Figure 1 As shown, the method includes the following steps:

[0043] Step S101, selecting multiple financial indicators for credit evaluation of the subject to be evaluated to obtain an indicator system;

[0044] Step S102, using a random forest algorithm to sort the indicator system to obtain an indicator sorting result;

[0045] Step S103: using a principal component analysis strategy to convert multiple screening indicators ranked before a preset numerical threshold in the indicator ranking result into multiple new comprehensive indicators;

[0046] Step S104: Based on the multiple new comprehensive indicators, a pre-built logistic regression model is used to measure the credit risk index of the subject to be assessed, wherein the credit risk index is used to determine whether the subject to be assessed is a credit risk subject.

[0047] Through the above steps, multiple financial indicators for credit assessment of the subject to be assessed can be selected to obtain an indicator system, and the indicator system can be sorted using a random forest algorithm to obtain an indicator sorting result. The principal component analysis strategy is used to convert multiple screening indicators ranked before a preset numerical threshold in the indicator sorting result into multiple new comprehensive indicators. Based on the multiple new comprehensive indicators, a pre-built logistic regression model is used to measure the credit risk index of the subject to be assessed, wherein the credit risk index is used to determine whether the subject to be assessed is a credit risk subject. In this embodiment, based on the pre-selected financial data indicators, the principal component analysis method is used to convert the multiple screened indicators into several comprehensive indicators, and finally logistic regression is used to measure the credit risk, so that financial institutions can effectively obtain the credit risk and enterprise classification of the enterprise through the existing financial information of the subject to be assessed, making the credit assessment results more objective and reliable, thereby predicting the enterprise category and credit risk level, reducing non-performing loans caused by the operating problems of the subject to be assessed, and thus solving the technical problem of large errors in the credit assessment strategies adopted by financial institutions in related technologies.

[0048] The embodiments of the present invention are described in detail below in conjunction with the above implementation steps.

[0049] Step S101: Select multiple financial indicators for credit evaluation of the subject to be evaluated to obtain an indicator system.

[0050] It should be noted that the subject to be assessed in this embodiment includes various types, such as market entities (companies / enterprises), natural persons, loan entities, etc. In this embodiment, enterprises are used as the subject to be assessed for illustration. Before an enterprise is granted a loan or obtains funds, the credit rating and credit risk of the enterprise are assessed to help financial institutions screen out high-credit-risk enterprises and conduct a multi-faceted analysis of the repayment capacity of an enterprise before lending, thereby avoiding non-performing loans caused by business problems.

[0051] In an embodiment of the present invention, before conducting a credit assessment, financial indicators that have a greater impact on the company's financial crisis will be selected through the company's publicly available financial information to form an indicator system. Currently, many small and micro enterprises or non-listed companies do not disclose their financial statements. Therefore, the main data sources of the present invention are the financial statements of publicly available enterprises or listed companies, as well as financial indicator analysis.

[0052] Optionally, the step of selecting multiple financial indicators for credit assessment of the subject to be assessed to obtain an indicator system includes: obtaining financial data from the financial disclosure information of multiple assessed subjects; selecting a set of indicators that are crisis-related to the financial status of the subject to be assessed from the financial data; and screening out multiple financial indicators corresponding to preset financial indicator categories from the indicator set to obtain an indicator system, wherein the preset financial indicator categories include at least one of the following: debt repayment ability, subject development ability, subject operating ability, subject profitability, cash flow ability, and subject scale.

[0053] In an embodiment of the present invention, financial data in the financial disclosure information of multiple evaluated entities obtained in advance is analyzed to screen out indicators that have a greater impact on the financial crisis of the enterprise, covering various aspects such as the enterprise's scale, debt repayment ability, operating ability, profitability, cash flow ability, and development ability.

[0054] The following table 1 illustrates the structure of multiple indicators.

[0055] Table 1 Financial indicator structure and classification

[0056]

[0057]

[0058]

[0059] By analyzing the financial information / reports, a plurality of financial indicators as indicated in the above Table 1 can be obtained. After classification, the financial indicators form a corresponding indicator system.

[0060] Step S102: Random forest algorithm is used to sort the indicator system to obtain an indicator sorting result.

[0061] In this example, a random forest approach is used to rank the indicators in the financial indicator system. A random forest is a classifier consisting of multiple decision trees that can be used to evaluate the importance of indicators. The idea is to evaluate the contribution of each feature (presumably the evaluation indicator) on each tree in the random forest, then take the average to compare the contribution of different indicators.

[0062] In this embodiment, the random forest algorithm is used to sort the indicator system, and the steps of obtaining the indicator sorting results include: calculating the Gini index of the current node based on the number of subject categories and the proportion of each subject category in the current node to which the financial indicator to be evaluated belongs; obtaining the change in the Gini index of the previous node and the next node obtained after branching based on the current node; calculating the weight values ​​of all the financial indicators to be evaluated at the current node based on the Gini index of the current node and the change in the Gini index of the previous node and the next node; and sorting all the financial indicators in the indicator system based on the weight value of each financial indicator to be evaluated to obtain the indicator sorting result.

[0063] Optionally, this embodiment uses the Gini coefficient as a measurement indicator when ranking the indicators, and uses formula (1) to calculate the Gini coefficient. Formula (1) is as follows:

[0064]

[0065] Here, k is the number of categories. In this embodiment, k is 2. The category indicated by k is the enterprise / company category, which is divided into two categories, namely, credit-risk-free enterprises and credit-high-risk enterprises.

[0066] Among them, p mk It represents the proportion of category k in node m (associated indicator). Since there are multiple decision trees in the random forest, there can be multiple nodes on each decision tree. Different nodes represent different indicator contents. The contribution of the indicator on each tree in the random forest is determined through each node on the decision tree.

[0067] The feature importance scoring series formula (2) is as follows:

[0068]

[0069] in, Indicates the importance of the feature at node m, GI l and GI r They represent the Gini index of the two new nodes after branching, that is, the importance is the change in the Gini index before and after the m branch (that is, the change in the Gini index of the previous node and the next node mentioned above).

[0070] After obtaining the importance of all indicators (or features), the indicators can be sorted and some indicators can be retained.

[0071] The model with the addition of random forest has stronger predictive ability for the violating entities to be assessed, both in the training set and the validation set. That is, the screening of important indicators by random forest has a significant effect on improving the accuracy of the model, which can provide a reference for credit risk measurement modeling of financial institutions.

[0072] Step S103 : using a principal component analysis strategy to convert multiple screening indicators ranked before a preset numerical threshold in the indicator ranking result into multiple new comprehensive indicators.

[0073] For example, Table 2 below illustrates the indicators screened by random forest.

[0074]

[0075]

[0076] Table 2 above shows that after sorting by importance, a total of 21 indicators with an importance greater than 0.01 were screened out.

[0077] Optionally, a principal component analysis strategy is used to convert multiple screening indicators ranked before a preset numerical threshold in the indicator sorting results into multiple new comprehensive indicators, including: screening the indicator sorting results according to the preset numerical threshold to obtain multiple screening indicators; calculating the covariance matrix of all screening indicators; standardizing the matrix data in the covariance matrix; and using the principal component analysis strategy to perform principal component analysis on the standardized matrix data to obtain multiple new comprehensive indicators.

[0078] The preset numerical threshold for going up the tree can be set by yourself. For example, for Table 2, the preset numerical threshold is set to 0.01. The importance exceeding the preset numerical threshold can be determined as a screening indicator.

[0079] Due to the high correlation and multicollinearity of the data, as well as the large number of indicators, directly substituting these indicators into the model will result in data duplication, affecting the accuracy of the model results. Therefore, after obtaining multiple screening indicators, it is necessary to conduct principal component analysis on the existing indicators to construct new predictive indicators. While preserving the original information as much as possible, the information in the original indicators is reflected through a smaller number of new indicators, and then the new indicators are further substituted into the model analysis. This can effectively reduce the correlation between indicators.

[0080] In this embodiment, the step of calculating the covariance matrix of all screening indicators includes: calculating the indicator average of all screening indicators based on the indicator data corresponding to each screening indicator; calculating the covariance between every two screening indicators based on the indicator average of all screening indicators and the total number of screening indicators; and determining the covariance matrix corresponding to all screening indicators by integrating the covariance between every two screening indicators.

[0081] The principal components of the data are equivalent to all eigenvalues ​​of its covariance matrix and their corresponding orthogonal normalized eigenvectors.

[0082] The covariance formula (3) between every two screening indicators (or two variables) is:

[0083]

[0084] in is x i (i is the index, x i is the average value of the data corresponding to indicator i), n is the number of samples, cov(x i ,x j ) is the covariance.

[0085] The covariance matrix of the data set is the set of covariances between the two screening indicators. It is a symmetric matrix, and its display formula (4) is as follows:

[0086]

[0087] By integrating the covariance between every two screening indicators, the covariance matrix corresponding to all screening indicators is obtained.

[0088] An optional step of performing principal component analysis on the standardized matrix data using a principal component analysis strategy includes: calculating the data variance based on the indicator data of all screening indicators; calculating the correlation coefficient matrix that has an association relationship with the matrix data based on the covariance between each two screening indicators and the variance of the indicator data corresponding to multiple screening indicators; and calculating the eigenvalues ​​and corresponding eigenvectors of the correlation coefficient matrix based on the correlation coefficient matrix and a pre-configured unit matrix to complete the principal component analysis.

[0089] Since the sample data has different dimensions, the numerical dispersion between different screening indicators will be large. Directly solving with the covariance matrix may lead to a decrease in the accuracy of the output results. Therefore, it is necessary to standardize the data. Performing principal component analysis on the standardized data is equivalent to finding the eigenvalues ​​and corresponding eigenvectors of the correlation coefficient matrix.

[0090] Among them, the correlation coefficient matrix is ​​as follows (5):

[0091]

[0092] in, To filter the data variance of the index data X corresponding to the index i, combined with the covariance cov(x i ,x j ), the correlation coefficient matrix can be obtained.

[0093] Optionally, based on the correlation coefficient matrix and a preconfigured unit matrix, the eigenvalues ​​and corresponding eigenvectors of the correlation coefficient matrix are calculated to complete the steps of principal component analysis, including: based on the correlation coefficient matrix and the preconfigured unit matrix, the eigenvalues ​​of the correlation coefficient matrix are calculated; substituting the correlation coefficient matrix and the corresponding eigenvalues ​​into a preset characteristic equation group, parsing the preset characteristic equation group, and obtaining multiple eigenvectors, wherein each eigenvector is used to represent a new comprehensive indicator.

[0094] For example, by solving the linear equation system shown in the following formula (6), the eigenvalue can be obtained.

[0095] det(λ k ER)=0 formula (6);

[0096] Among them, E is the unit matrix (a matrix with all 1s on the diagonal), and the contribution rate is m is the number of eigenvalues.

[0097] Table 3 below schematically illustrates the characteristic values ​​of the indicators obtained after principal component analysis.

[0098] Table 3 Principal component analysis indicators

[0099] K (Indicator) <![CDATA[λ k (Eigenvalue)]]> Contribution rate Cumulative contribution rate 1 4.673948744 0.222568988 0.222568988 2 3.093740731 0.147320987 0.369889975 3 2.543673015 0.121127286 0.491017261 4 1.973181385 0.093961018 0.58497828 5 1.453557191 0.069217009 0.654195289 6 1.11972553 0.053320263 0.707515552 7 1.050371395 0.050017685 0.757533238 8 0.953399817 0.045399991 0.802933229

[0100] The above embodiment of the present invention can use the principle of random forest to screen indicators for establishing a logistic regression model. When establishing a logistic regression model for credit risk measurement, the most important step is to select appropriate indicators with strong predictive ability for classification results.

[0101] As shown in Table 3 above, the eigenvalues ​​of the correlation matrix of the data set are obtained by principal component analysis. It can be seen that the first 8 principal components cover 80% of the information of the original data.

[0102] Solve the following equations based on the eigenvalues ​​obtained from the principal component analysis, formula (7):

[0103] RV k =λ k V k Formula (7);

[0104] Get the corresponding eigenvector V k , substitute the new indicator (i.e. the new comprehensive indicator mentioned above) into the logistic regression model in the next step.

[0105] Step S104: Based on the multiple new comprehensive indicators, a pre-built logistic regression model is used to measure the credit risk index of the subject to be assessed, wherein the credit risk index is used to determine whether the subject to be assessed is a credit risk subject.

[0106] Optionally, the step of measuring the credit risk index of the subject to be evaluated based on multiple new comprehensive indicators using a pre-built logistic regression model includes: substituting the multiple new comprehensive indicators as logistic regression variables into the selected sample data to obtain logistic regression coefficients; substituting the logistic regression coefficients into the logistic regression model to measure the violation parameters of the subject to be evaluated; and evaluating the credit risk index of the subject to be evaluated based on the violation parameters of the subject to be evaluated.

[0107] The logistic regression model is used to measure the credit risk of the entity to be evaluated. The logistic regression model can analyze the distinguishing boundary between non-compliant enterprises and non-compliant enterprises through the financial report of the entity to be evaluated. At the same time, after fitting the decision boundary, the logistic regression model can also establish the connection between the boundary and the classification probability, thereby predicting the type of enterprise.

[0108] Optionally, the logistic regression model in this embodiment is shown in the following formula (8):

[0109]

[0110] Among them, Y=1 means the subject to be evaluated is a violating subject, Y=0 means a non-violating subject, and w and b are parameters to be determined.

[0111] The eight new indicators obtained from the principal component analysis are used as variables for logistic regression and substituted into the selected sample data to obtain the following logistic regression coefficients:

[0112] w T V+b=0.358V1-0.304V2+0.307V3-0.391V4-0.583V5+0.057V6-0.049V7-0.299V8-2.56

[0113] By substituting the financial indicators of the entity under assessment into the model, the probability of non-compliance and the resulting credit risk index can be determined. The logistic regression model is highly applicable to the current status of the entity under assessment and can help financial institutions (e.g., commercial banks) conduct multi-faceted analyses of the entity's repayment capacity and identify high-credit-risk enterprises. Its high accuracy and stability in predicting non-compliance rates can help prevent non-performing loans (NPLs) caused by operational issues within the entity under assessment, making it a reliable model.

[0114] In this embodiment of the present invention, key financial data from each company's annual report is first obtained. 31 financial indicators that significantly impact corporate financial crisis are selected to form an indicator system. Key indicators are then screened using random forest analysis. Principal component analysis is then used to transform these multiple indicators into several comprehensive indicators. Finally, logistic regression is used to measure credit risk. This logistic regression model allows financial institutions to effectively determine the credit risk and corporate classification of the entity being assessed based on its existing financial information.

[0115] The present invention is described below in conjunction with another optional embodiment.

[0116] Example 2

[0117] This embodiment provides a credit risk assessment device, and the multiple implementation units involved in the assessment device correspond to the various implementation steps in the above-mentioned embodiment 1.

[0118] Figure 2 is a schematic diagram of an optional credit risk assessment device according to an embodiment of the present invention, such as Figure 2 As shown, the evaluation device may include: a selection unit 21, a sorting unit 22, a conversion unit 23, and a measurement unit 24, wherein:

[0119] A selection unit 21 is used to select multiple financial indicators for credit evaluation of the subject to be evaluated to obtain an indicator system;

[0120] A sorting unit 22 is used to sort the indicator system using a random forest algorithm to obtain an indicator sorting result;

[0121] A conversion unit 23 is configured to convert the plurality of screening indicators ranked before a preset numerical threshold in the indicator ranking result into a plurality of new comprehensive indicators using a principal component analysis strategy;

[0122] The measurement unit 24 is used to measure the credit risk index of the subject to be evaluated based on multiple new comprehensive indicators using a pre-built logistic regression model, wherein the credit risk index is used to determine whether the subject to be evaluated is a credit risk subject.

[0123] The above-mentioned credit risk assessment device can select multiple financial indicators for credit assessment of the subject to be assessed by the selection unit 21 to obtain an indicator system, and then use the random forest algorithm to sort the indicator system by the sorting unit 22 to obtain an indicator sorting result. The conversion unit 23 uses the principal component analysis strategy to convert multiple screening indicators ranked before a preset numerical threshold in the indicator sorting result into multiple new comprehensive indicators. The measurement unit 24 uses a pre-built logistic regression model based on the multiple new comprehensive indicators to measure the credit risk index of the subject to be assessed, wherein the credit risk index is used to determine whether the subject to be assessed is a credit risk subject. In this embodiment, based on the pre-selected financial data indicators, the multiple screening indicators are converted into several comprehensive indicators using the principal component analysis method, and finally the credit risk is measured using logistic regression. This allows financial institutions to effectively obtain the credit risk and enterprise classification of the enterprise based on the existing financial information of the subject to be assessed, making the credit assessment results more objective and reliable, thereby predicting the enterprise category and credit risk level, reducing non-performing loans caused by the operating problems of the subject to be assessed, and thus solving the technical problem of large errors in the credit assessment results of financial institutions in related technologies.

[0124] Optionally, the selection unit includes: a first acquisition module, used to obtain financial data from the financial disclosure information of multiple evaluated entities; a first selection module, used to select a set of indicators that are crisis-related to the financial status of the entity to be evaluated from the financial data; a first screening module, used to screen out multiple financial indicators corresponding to preset financial indicator categories from the indicator set to obtain an indicator system, wherein the preset financial indicator categories include at least one of the following: debt repayment ability, entity development ability, entity operating ability, entity profitability, cash flow ability, and entity scale.

[0125] Optionally, the sorting unit includes: a first calculation module, used to calculate the Gini index of the current node based on the number of subject categories and the proportion of each subject category in the current node to which the financial indicator to be evaluated belongs; a second acquisition module, used to obtain the change in the Gini index of the previous node and the next node obtained after branching based on the current node; a second calculation module, used to calculate the weight values ​​of all financial indicators to be evaluated at the current node based on the Gini index of the current node, the change in the Gini index of the previous node and the next node; a sorting module, used to sort all financial indicators in the indicator system based on the weight value of each financial indicator to be evaluated, and obtain the indicator sorting result.

[0126] Optionally, the conversion unit includes: a second screening module, used to screen the index sorting results according to a preset numerical threshold to obtain multiple screening indicators; a third calculation module, used to calculate the covariance matrix of all screening indicators; a standardization processing module, used to standardize the matrix data in the covariance matrix; an analysis module, used to adopt a principal component analysis strategy to perform principal component analysis on the standardized matrix data to obtain multiple new comprehensive indicators.

[0127] Optionally, the third calculation module includes: a first calculation submodule, used to calculate the indicator average of all screening indicators based on the indicator data corresponding to each screening indicator; a second calculation submodule, used to calculate the covariance between every two screening indicators based on the indicator average of all screening indicators and the total number of screening indicators; a first determination submodule, used to integrate the covariance between every two screening indicators and determine the covariance matrix corresponding to all screening indicators.

[0128] Optionally, the analysis module includes: a third calculation submodule, used to calculate the data variance based on the indicator data of all screening indicators; a fourth calculation submodule, used to calculate the correlation coefficient matrix that has an association relationship with the matrix data based on the covariance between each two screening indicators and the variance of the indicator data corresponding to multiple screening indicators; a fifth calculation submodule, used to calculate the eigenvalues ​​and corresponding eigenvectors of the correlation coefficient matrix based on the correlation coefficient matrix and a pre-configured unit matrix to complete the principal component analysis.

[0129] Optionally, the fifth calculation submodule includes: a sixth calculation submodule, used to calculate the eigenvalues ​​of the correlation coefficient matrix based on the correlation coefficient matrix and a pre-configured unit matrix; an analysis submodule, used to substitute the correlation coefficient matrix and the corresponding eigenvalues ​​into a preset characteristic equation group, analyze the preset characteristic equation group, and obtain multiple eigenvectors, wherein each eigenvector is used to represent a new comprehensive indicator.

[0130] Optionally, the measurement unit includes: a substitution module, used to substitute multiple new comprehensive indicators as variables of logistic regression into the selected sample data to obtain logistic regression coefficients; a measurement module, used to substitute the logistic regression coefficients into the logistic regression model to measure the violation parameters of the subject to be evaluated; and an evaluation module, used to evaluate the credit risk index of the subject to be evaluated based on the violation parameters of the subject to be evaluated.

[0131] The above-mentioned credit risk assessment device may further include a processor and a memory. The above-mentioned selection unit 21, sorting unit 22, conversion unit 23, measurement unit 24, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to implement corresponding functions.

[0132] The processor includes a kernel that retrieves corresponding program units from memory. One or more kernels can be provided. By adjusting kernel parameters, a pre-built logistic regression model is used to measure the credit risk index of the entity to be assessed based on multiple new comprehensive indicators. The credit risk index is used to determine whether the entity to be assessed is a credit risk entity.

[0133] The above-mentioned memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0134] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing an initialization program having the following method steps: selecting multiple financial indicators for credit evaluation of a subject to be evaluated to obtain an indicator system; using a random forest algorithm to sort the indicator system to obtain an indicator sorting result; using a principal component analysis strategy to convert multiple screening indicators ranked before a preset numerical threshold in the indicator sorting result into multiple new comprehensive indicators; based on the multiple new comprehensive indicators, using a pre-built logistic regression model to measure the credit risk index of the subject to be evaluated, wherein the credit risk index is used to determine whether the subject to be evaluated is a credit risk subject.

[0135] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned credit risk assessment methods.

[0136] According to another aspect of an embodiment of the present invention, an electronic device is also provided, comprising one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by one or more processors, the one or more processors implement any one of the above-mentioned credit risk assessment methods.

[0137] Figure 3 FIG is a hardware structure block diagram of an electronic device (or mobile device) according to a credit risk assessment method of an embodiment of the present invention. Figure 3As shown, the electronic device may include one or more (illustrated as 102a, 102b, ..., 102n in the figure) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, a keyboard, a power supply and / or a camera. It will be understood by those skilled in the art that Figure 3 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 3 More or fewer components than shown, or with Figure 3 Different configurations shown.

[0138] In the above embodiments of the present invention, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0139] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of the units can be a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0140] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0141] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0142] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, etc. Various media that can store program codes.

[0143] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A credit risk assessment method, characterized in that: include: Select multiple financial indicators for credit assessment of the subject to be assessed to obtain an indicator system; Random forest algorithm is used to sort the indicator system to obtain indicator sorting results; The principal component analysis strategy is used to convert multiple screening indicators ranked before the preset numerical threshold in the indicator ranking results into multiple new comprehensive indicators, including: Filtering the indicator sorting results according to the preset numerical threshold to obtain a plurality of the filtering indicators; Calculating the covariance matrix of all the screening indicators; Performing standardization on the matrix data in the covariance matrix; Using the principal component analysis strategy, performing principal component analysis on the matrix data after standardization to obtain the multiple new comprehensive indicators; Based on the multiple new comprehensive indicators, a pre-built logistic regression model is used to measure the credit risk index of the subject to be evaluated, wherein the credit risk index is used to determine whether the subject to be evaluated is a credit risk subject.

2. The method according to claim 1, characterized in that The steps of selecting multiple financial indicators for credit assessment of the subject to be assessed and obtaining an indicator system include: Access financial data from the financial disclosures of multiple assessed entities; Selecting a set of indicators that are crisis-related to the financial status of the subject to be assessed from the financial data; A plurality of financial indicators corresponding to preset financial indicator categories are screened out from the indicator set to obtain the indicator system, wherein the preset financial indicator categories include at least one of the following: debt repayment ability, entity development ability, entity operating ability, entity profitability, cash flow ability, and entity scale.

3. The method according to claim 1, characterized in that The step of using a random forest algorithm to sort the indicator system to obtain an indicator sorting result includes: Calculate the Gini index of the current node based on the number of entity categories and the proportion of each entity category in the current node to which the financial indicator to be evaluated belongs; Obtaining a change in the Gini index of a previous node and a next node obtained after branching based on the current node; Calculating the weight values ​​of all the financial indicators to be evaluated at the current node by combining the Gini index of the current node and the changes in the Gini indexes of the previous node and the next node; Based on the weight value of each financial indicator to be evaluated, all financial indicators in the indicator system are sorted to obtain the indicator sorting result.

4. The method according to claim 1, wherein The step of calculating the covariance matrix of all the screening indicators includes: Calculate the average value of all the screening indicators based on the indicator data corresponding to each of the screening indicators; Calculating the covariance between every two screening indicators based on the indicator averages of all the screening indicators and the total number of the screening indicators; The covariance between every two screening indicators is integrated to determine the covariance matrix corresponding to all the screening indicators.

5. The method according to claim 1, wherein The step of performing principal component analysis on the matrix data after standardization using the principal component analysis strategy includes: Calculating data variance based on the indicator data of all the screening indicators; Calculating a correlation coefficient matrix associated with the matrix data based on the covariance between each two screening indicators and the variance of the indicator data corresponding to the plurality of screening indicators; Based on the correlation coefficient matrix and a pre-configured identity matrix, the eigenvalues ​​and corresponding eigenvectors of the correlation coefficient matrix are calculated to complete principal component analysis.

6. The method according to claim 5, characterized in that The step of calculating the eigenvalues ​​and corresponding eigenvectors of the correlation coefficient matrix based on the correlation coefficient matrix and a pre-configured identity matrix to complete principal component analysis includes: Calculating eigenvalues ​​of the correlation coefficient matrix based on the correlation coefficient matrix and a preconfigured identity matrix; Substituting the correlation coefficient matrix and the corresponding eigenvalues ​​into a preset characteristic equation group, the preset characteristic equation group is analyzed to obtain a plurality of eigenvectors, wherein each eigenvector is used to characterize the new comprehensive indicator.

7. The method according to claim 1, characterized in that The step of measuring the credit risk index of the subject to be assessed using a pre-built logistic regression model based on the multiple new comprehensive indicators includes: Substituting the multiple new comprehensive indicators as variables for logistic regression into the selected sample data to obtain logistic regression coefficients; Substituting the logistic regression coefficient into the logistic regression model to measure the violation parameter of the subject to be evaluated; Evaluate the credit risk index of the subject to be evaluated based on the violation parameters of the subject to be evaluated.

8. A credit risk assessment device, characterized in that: include: A selection unit is used to select multiple financial indicators for credit assessment of the subject to be assessed to obtain an indicator system; A sorting unit, configured to sort the indicator system using a random forest algorithm to obtain an indicator sorting result; A conversion unit is used to convert the multiple screening indicators ranked before the preset numerical threshold in the indicator ranking result into multiple new comprehensive indicators using a principal component analysis strategy, wherein the conversion unit includes: A second screening module is used to screen the index sorting results according to the preset numerical threshold to obtain the multiple screening indicators; A third calculation module is used to calculate the covariance matrix of all the screening indicators; A standardization processing module, used for performing standardization processing on the matrix data in the covariance matrix; An analysis module, configured to perform principal component analysis on the matrix data after standardization using the principal component analysis strategy to obtain the multiple new comprehensive indicators; A measurement unit is used to measure the credit risk index of the subject to be evaluated based on the multiple new comprehensive indicators using a pre-built logistic regression model, wherein the credit risk index is used to determine whether the subject to be evaluated is a credit risk subject.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the credit risk assessment method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: The system comprises one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the credit risk assessment method according to any one of claims 1 to 7.

Images