A financial risk assessment method, device, equipment and storage medium

Abstract

The embodiments of the present application provide a financial risk assessment method, device, equipment and storage medium. The financial risk assessment method includes: obtaining financial data of the object to be assessed; analyzing the financial data, extracting financial dimension indicators of the object to be assessed, and screening the financial dimension indicators; inputting the screened financial dimension indicators into the financial risk assessment model, and outputting the financial risk assessment results of the object to be assessed; wherein the financial risk assessment model is obtained by integrating the prediction results of multiple risk prediction models. Through the embodiments of the present application, the financial risks of the object to be assessed are quantified to obtain more accurate financial risk assessment results.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of artificial intelligence technology, and in particular to a financial risk assessment method, apparatus, device and storage medium. Background Art

[0002] In the financial field, credit risk assessment is a key step for banks, financial institutions and other loan providers to decide whether to provide funds to potential borrowers. The risk of loan default is predicted and quantified by analyzing the borrower's repayment ability, income stability and other relevant factors. Traditional risk assessment methods rely on data such as financial statements and credit records. However, due to the small scale of some enterprises, opaque financial information, and lack of credit data, traditional risk assessment methods have certain limitations and need to be improved. Summary of the invention

[0003] In view of this, embodiments of the present application provide a financial risk assessment method, apparatus, device and storage medium to at least partially solve the above-mentioned problems.

[0004] According to a first aspect of an embodiment of the present application, a financial risk assessment method is provided, comprising:

[0005] Obtain financial data of the subject to be assessed;

[0006] Analyze the financial data, extract financial dimension indicators of the object to be evaluated, and screen the financial dimension indicators;

[0007] The screened financial dimension indicators are input into a financial risk assessment model, and a financial risk assessment result of the object to be assessed is output; wherein the financial risk assessment model is obtained by integrating the prediction results of multiple risk prediction models.

[0008] In an optional embodiment, the financial dimension indicators are screened.

[0009] include:

[0010] For each of the financial dimension indicators, calculating a first correlation coefficient between the financial dimension indicator and the target variable, and a second correlation coefficient between the financial dimension indicators;

[0011] Eliminate the financial dimension indicators whose first correlation coefficient is less than a first preset correlation coefficient, and process the financial dimension indicator pairs whose correlation coefficients among the financial dimension indicators are higher than a second preset correlation coefficient;

[0012] The importance of the financial dimension indicators is calculated, the financial dimension indicators whose importance is higher than a preset value are screened out, and the financial dimension indicators whose importance is higher than the preset value are combined.

[0013] In an optional embodiment, the target variables include: probability of default and prospects for corporate revenue growth.

[0014] In an optional embodiment, the financial risk assessment model is obtained by the following steps:

[0015] Obtain sample financial data for multiple samples;

[0016] Analyzing the sample financial data of the multiple samples, extracting sample financial dimension indicators of the multiple samples, and screening the sample financial dimension indicators;

[0017] Based on the screened sample financial dimension indicators, establish a variety of risk prediction models;

[0018] Selecting a target risk prediction model from the multiple risk prediction models;

[0019] The prediction results of the target risk prediction model are integrated to construct a financial risk assessment model.

[0020] In an optional embodiment, the target risk prediction model is selected from the multiple risk prediction models.

[0021] include:

[0022] Evaluating the prediction effect of each risk prediction model on the predicted target variable based on the sample financial data;

[0023] Based on the evaluation results, a target risk prediction model is screened out from the multiple risk prediction models.

[0024] In an optional embodiment, the prediction results of the target risk prediction model are integrated to construct a financial risk assessment model.

[0025] include:

[0026] For each sample, the prediction results of the target variable predicted by the target risk prediction model for each screened sample financial dimension indicator are integrated to obtain an initial integrated model;

[0027] The objective function of the initial fusion model is set based on the predicted target variable, and the parameters of the initial fusion model are iteratively updated to obtain the financial risk assessment model.

[0028] In an optional embodiment, the prediction results of the target variable predicted by the target risk prediction model for each of the screened sample financial dimension indicators are fused.

[0029] include:

[0030] The prediction results of different target risk prediction models are weighted and fused by multiplication fusion and / or exponential fusion.

[0031] According to a second aspect of an embodiment of the present application, a financial risk assessment device is provided, including:

[0032] An acquisition unit, used for acquiring financial data of the object to be evaluated;

[0033] An extraction unit, configured to analyze the financial data, extract the financial dimension indicators of the object to be evaluated, and screen the financial dimension indicators;

[0034] An evaluation unit is used to input the screened financial dimension indicators into a financial risk evaluation model and output the financial risk evaluation results of the object to be evaluated; wherein the financial risk evaluation model is obtained by integrating the prediction results of multiple risk prediction models.

[0035] According to a third aspect of an embodiment of the present application, an electronic device is provided, comprising a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, a method as described in any one of the first aspects is implemented.

[0036] According to a fourth aspect of an embodiment of the present application, a computer storage medium is provided, characterized in that it is configured to store computer program instructions, and the computer program instructions are used to execute any one of the methods in the first aspect.

[0037] According to the solution provided in the embodiment of the present application, firstly, the financial data of the object to be evaluated is obtained; the financial data is analyzed to extract the financial dimension indicators of the object to be evaluated, reflecting the overall business performance and development of the object to be evaluated; the financial dimension indicators are screened; the screened financial dimension indicators are input into the financial risk assessment model, and the financial risk assessment results of the object to be evaluated are output; wherein, the financial risk assessment model is obtained by integrating the prediction results of multiple risk prediction models, thereby combining the advantages of each risk prediction model to make the prediction results of the final financial risk assessment model more accurate, giving full play to the advantages of model fusion, quantifying the financial risks of the object to be evaluated, and obtaining more accurate financial risk assessment results. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of the present application. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0039] Figure 1 A flowchart of a method for assessing financial risks according to an embodiment of the present application;

[0040] Figure 2 A flowchart of steps for obtaining a financial risk assessment model according to an embodiment of the present application;

[0041] Figure 3 It is a structural diagram of a financial risk assessment device according to an embodiment of the present application. DETAILED DESCRIPTION

[0042] In order to enable those skilled in the art to better understand the technical solutions in the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the embodiments of the present application should fall within the scope of protection of the embodiments of the present application.

[0043] The specific implementation of the embodiment of the present application is further explained below in conjunction with the accompanying drawings of the embodiment of the present application.

[0044] Figure 1 A flowchart showing a method for assessing financial risks according to an embodiment of the present application is shown. Figure 1 As shown, a financial risk assessment method of the present application includes the following steps:

[0045] Step S102: Obtain the financial data of the object to be evaluated.

[0046] Before banks, financial institutions and other loan providers provide funds to borrowers, they need to conduct a comprehensive assessment of the borrowers based on a variety of data to determine whether the borrowers can repay the loans within the term.

[0047] In one feasible method, financial data is obtained from multiple perspectives around the object to be evaluated, for example, real-time acquisition of sales invoice data of the object to be evaluated, basic information of the object to be evaluated, industry data of the industry in which the object to be evaluated is located, and macroeconomic data.

[0048] Furthermore, the sales invoice data of the object to be evaluated includes detailed information such as invoice number, invoice date, name of the purchaser, name of goods / services, quantity, amount, tax amount, tax rate, etc., which reflects the detailed information of each transaction of the object to be evaluated. The basic information of the object to be evaluated includes: enterprise registration information, business scope, registered capital, establishment time, industrial and commercial judicial dishonesty record, etc., which reflects the legitimacy and business direction of the object to be evaluated. The industry data of the industry in which the object to be evaluated is located includes: market demand, market supply, industry scale, industry risk, etc., which reflects the industry situation of the industry in which the object to be evaluated is located, and is used to comprehensively evaluate the competitive advantages of the object to be evaluated. Macroeconomic data include: consumer price index, gross domestic product, etc., which can reflect the operation status of the entire national economy. The financial data of the object to be evaluated in multiple dimensions are obtained for further comprehensive analysis, so as to reflect the financial situation of the object to be evaluated from multiple angles, and provide a data basis for financial risk assessment of the object to be evaluated.

[0049] In order to ensure the quality of financial data and improve the interpretability of financial data, in a feasible way, after obtaining the financial data of the object to be evaluated, the financial data is preprocessed. Specifically, the financial data is first cleaned, including: processing missing values, for example, filling the mean, using interpolation or deleting financial data with missing values; processing outliers, for example, using box plots, using the 3σ principle and other methods to process outliers in financial data; unifying the data format, and adjusting the data fields of financial data to consistent units and formats. After the financial data is cleaned, data fusion is performed, including: according to the unique identification of the object to be evaluated, such as tax number, business registration number, etc., the financial data of different data sources of the object to be evaluated are associated to construct complete data of financial data; the financial data is standardized and normalized, for example, the continuous data is standardized by the Z-score normalization method, or the continuous data is normalized by the Min-Max normalization method.

[0050] Step S104: Analyze the financial data, extract the financial dimension indicators of the object to be evaluated, and screen the financial dimension indicators.

[0051] After obtaining the financial data of the subject to be evaluated, further analysis is performed on the financial data to extract financial dimension indicators that reflect the overall business performance and development of the subject to be evaluated from the financial data of each dimension. Specifically, by combining financial data of multiple dimensions, financial dimension indicators of multiple dimensions of the subject to be evaluated are further obtained, such as the dimension of amount, the dimension of time, the dimension of customer relationship, and the dimension of goods / services.

[0052] In one feasible approach, in the amount dimension, financial dimension indicators include: total sales amount, average invoice amount, and amount volatility, etc. Specifically, the total sales amount is the total sales amount of the subject to be evaluated over a period of time (such as the most recent year), reflecting the sales scale of the subject to be evaluated; the average invoice amount is the total sales amount divided by the number of invoices, reflecting the size of a single transaction amount of the subject to be evaluated; the amount volatility is the standard deviation and coefficient of variation of the sales amount, which is used to measure the stability of the sales volume of the subject to be evaluated.

[0053] In the time dimension, financial dimension indicators include: monthly / quarterly sales and sales growth rate, etc. Specifically, monthly / quarterly sales refers to the monthly / quarterly sales amount of the object to be evaluated, which is used to analyze the seasonal and cyclical sales of the object to be evaluated; sales growth rate refers to the month-on-month and year-on-year growth rate, which reflects the sales trend of the object to be evaluated.

[0054] In the customer relationship dimension, financial dimension indicators include: customer quantity, customer concentration, and customer loyalty, etc. Specifically, the customer quantity refers to the number of buyers of the object to be evaluated, reflecting the diversity of the customer group of the object to be evaluated; customer concentration refers to the sales proportion of the top five customers of the object to be evaluated, reflecting the degree of dependence of the object to be evaluated on major customers; customer loyalty refers to the proportion of repeated buyers of the object to be evaluated, reflecting the customer stability of the object to be evaluated.

[0055] In the commodity / service dimension, financial dimension indicators include: commodity category proportion, invoice issuance frequency, and invoice cancellation rate. Specifically, the commodity category proportion is the sales proportion of different commodities / services of the subject to be evaluated, reflecting the business structure of the subject to be evaluated; the invoice issuance frequency is the time interval between invoice issuance, reflecting whether the subject to be evaluated has abnormally frequent or sparse invoice issuance; the invoice cancellation rate is the proportion of the number of invoices cancelled by the subject to be evaluated, which is used to indicate whether there are any abnormalities in the finances of the subject to be evaluated or whether the business is stable.

[0056] Further, the financial dimension indicators of the object to be evaluated are screened to find abnormal correlations or unreasonable patterns in the financial dimension indicators. By processing these financial dimension indicators, the quality of the data can be improved, making the data more reliable and more useful. Specifically, the financial dimension indicators are subjected to correlation analysis. In a feasible manner, the correlation analysis of the financial dimension indicators specifically includes: for each financial dimension indicator, calculating the first correlation coefficient between the financial dimension indicator and the target variable, and the second correlation coefficient between each financial dimension indicator. The financial dimension indicators whose first correlation coefficient is less than the first preset correlation coefficient are eliminated, thereby eliminating the financial dimension indicators whose correlation with the target variable is too low in the financial dimension indicators, thereby reducing the dependence of the financial risk assessment model on noise data and optimizing the prediction performance of the financial risk assessment model. In addition, the financial dimension indicator pairs whose second correlation coefficient is higher than the second preset correlation coefficient are processed. Specifically, the financial dimension indicator pairs whose second correlation coefficient is higher than the second preset correlation coefficient are eliminated, or the financial dimension indicator pairs whose second correlation coefficient is higher than the second preset correlation coefficient are subjected to dimensionality reduction, for example, the principal component analysis method is used for dimensionality reduction, thereby reducing the impact of multicollinearity and improving the prediction performance of the financial risk assessment model.

[0057] In one feasible approach, the target variables include: default probability and corporate revenue growth prospects. Default probability and corporate revenue growth prospects convert credit risk into digital indicators, which helps credit analysts to more intuitively understand the repayment ability of the subject to be assessed. Specifically, corporate revenue growth prospects refer to the company's future sales or revenue growth rate.

[0058] Screening the financial dimension indicators of the object to be evaluated also includes: calculating the importance of the financial dimension indicators, and screening out the financial dimension indicators whose importance is higher than the preset value. Specifically, a tree-based model, such as a random forest or a gradient boosting tree, is used to calculate the importance of the financial dimension indicators; or a linear model is used to calculate the importance of the financial dimension indicators, thereby selecting financial dimension indicators with high contribution and significant impact on the prediction results of the target variable, improving the prediction performance of the financial risk assessment model and reducing the calculation cost, and enhancing the transparency of the model, which is easy for business personnel to understand and apply.

[0059] Furthermore, the financial dimension indicators whose importance is higher than the preset value obtained by screening are combined. Specifically, interactive indicators and high-order indicators are constructed, such as indicator products or indicator squares, to capture nonlinear relationships and improve the fit and prediction accuracy of the financial risk assessment model.

[0060] Step S106: Input the screened financial dimension indicators into the financial risk assessment model, and output the financial risk assessment results of the object to be assessed.

[0061] In one feasible way, the financial risk assessment results of the subject to be assessed are divided into multiple grades for evaluating the financial risk of the subject to be assessed. For example, the financial risk assessment results of the subject to be assessed are divided into grades A, B, C, D and E. Among them, grade A represents an excellent evaluation, grade B represents a good evaluation, and grade C represents an acceptable evaluation. When the financial risk assessment results of the subject to be assessed are grades A, B and C, a loan can be directly issued to the subject to be assessed. Grade D represents an evaluation that requires review, and grade E represents an evaluation that requires the subject to be assessed to bring relevant materials to apply for a loan offline. When the financial risk assessment results of the subject to be assessed are grades D and E, a loan cannot be directly issued to the subject to be assessed.

[0062] In one feasible approach, the financial risk assessment model is obtained by integrating the prediction results of multiple risk prediction models. Figure 2 As shown, the following steps are included:

[0063] Step S202. Obtain sample financial data of multiple samples;

[0064] Step S204. Analyze the sample financial data, extract sample financial dimension indicators of multiple samples, and screen the sample financial dimension indicators;

[0065] Step S206. Establish multiple risk prediction models based on the screened sample financial dimension indicators;

[0066] A variety of risk prediction models are used to conduct multi-faceted risk prediction and improve the predictive performance of the final financial risk assessment model.

[0067] For example, a linear regression model is constructed to predict the growth prospects of corporate revenue. The model formula of the linear regression model is shown in Formula 1 below:

[0068]

[0069] Among them, Y 收入增长 is the enterprise revenue growth prospect, β0 is the intercept term, β i is the parameter to be estimated, x i is the i-th sample financial dimension indicator, n is the total number of sample financial dimension indicators, and ∈ is the error term.

[0070] The training method of the linear regression model is to train the linear regression model using the least squares method or the gradient descent method to minimize the mean square error.

[0071] For another example, a logistic regression model is used to predict whether an enterprise will default and output the default probability of the object to be evaluated. The model formula of the logistic regression model is as follows:

[0072]

[0073] Among them, α0 is the intercept term, α i is the parameter to be estimated.

[0074] The training method of the logistic regression model is to use the gradient ascent method or optimization algorithm to estimate the parameters and maximize the log-likelihood function.

[0075] For another example, a random forest model is used to predict the probability of default and the prospects for corporate revenue growth. Multiple decision trees are constructed, and each tree is trained on randomly sampled data and indicators. When predicting the probability of default, the output result is the voting result of the prediction results of each tree. When predicting the prospects for corporate revenue growth, the output result is the average of the prediction results of each tree. The training method of the random forest model is based on bagging (bootstrap method) and random feature selection to prevent overfitting.

[0076] For another example, we construct a gradient boosting decision tree (GBDT) model, combining multiple weak learners (decision trees) into a prediction model. By iteratively training a series of weak learners (decision trees), we gradually improve the model performance. Each iteration fits the previous residual and updates the model parameters. The final model is the weighted sum of all weak learners.

[0077] For another example, a deep neural network (DNN) model is constructed to capture complex nonlinear relationships and high-order interaction effects. The deep neural network model consists of an input layer, multiple hidden layers, and an output layer. It uses a back-propagation algorithm, uses an optimizer such as Adam to update parameters, and minimizes the loss function. The activation function of the deep neural network model is the Sigmoid function, the hyperbolic tangent function, or the ReLU (Rectified Linear Unit).

[0078] Step S208. Selecting a target risk prediction model from a plurality of risk prediction models;

[0079] The sample financial data is divided into training set, validation set and test set in a ratio of 6:2:2. The prediction effect of each risk prediction model on the predicted target variable is evaluated based on the sample financial data.

[0080] For example, for the classification problem of predicting the probability of default, the accuracy, recall, precision, F1-score, AUC (Area Under the Curve) value and KS (Kolmogorov-Smirnov, discrimination index) value of the risk prediction model are calculated to evaluate the prediction effect of the risk prediction model. Specifically, the accuracy is the ratio of the number of samples correctly predicted by the risk prediction model to the total number of samples; the recall is the ratio of the number of correctly predicted positive samples to the actual number of positive samples; the precision is the ratio of the number of correctly predicted positive samples to the number of samples predicted as positive; the F1-score is the harmonic mean of the precision and recall; the AUC value is the area under the ROC (Receiver Operating Characteristic Curve) curve, which measures the classification performance of the risk prediction model; the KS value is used to measure the ability of the risk prediction model to distinguish between positive and negative samples.

[0081] For another example, for the regression problem of predicting the growth prospects of corporate revenue, the mean square error, mean absolute error and determination coefficient of the risk prediction model are calculated to evaluate the prediction effect of the risk prediction model. Specifically, the mean square error is the average of the squares of the difference between the predicted value and the true value; the mean absolute error is the average of the absolute values ​​of the difference between the predicted value and the true value; and the determination coefficient reflects the degree of fit of the risk prediction model to the data.

[0082] For each prediction problem, the target risk prediction model is selected from multiple risk prediction models based on multiple evaluation criteria. In combination with multiple evaluation criteria, for example, a risk prediction model with an AUC value higher than a preset value and a mean square error lower than a preset value is selected as the target risk prediction model. In this way, the advantages of each target risk prediction model are combined to make the prediction results of the final financial risk assessment model more accurate, and the interpretability of the single model is retained in the fusion model.

[0083] Step S210: Integrate the prediction results of the target risk prediction model to construct a financial risk assessment model.

[0084] In one feasible approach, for each sample, the prediction results of the target risk prediction model for the target variable predicted by each screened sample financial dimension indicator are fused to obtain an initial fusion model.

[0085] Specifically, the prediction results of different target risk prediction models are weighted and fused by multiplication fusion and / or exponential fusion. Multiplication fusion enhances the interaction between models, fully utilizes the advantages of each model, and reduces the possible errors of a single model. Exponential fusion emphasizes the differences between models and improves the sensitivity of the model to data. The two fusion methods give full play to the advantages of model fusion and improve the prediction performance of the fusion model.

[0086] Among them, the formula for multiplication fusion is as follows: Formula 3:

[0087]

[0088] Among them, Score j is the fusion result, K is the number of target risk prediction models, Model k,j is the prediction result of the k-th target risk prediction model for the j-th sample, a k 、b k 、c k is the parameter to be optimized.

[0089] The formula for index fusion is as follows: Formula 4:

[0090]

[0091] Among them, w k is the weight coefficient of the target risk prediction model.

[0092] Therefore, the prediction results of the target risk prediction model for the target variable predicted by each screened sample financial dimension indicator are fused to obtain the initial fusion model.

[0093] In one feasible approach, the initial weight coefficients of the target risk prediction model are evenly distributed, or set according to the performance indicators of a single model, for example, the model with better performance is given a higher weight, and the weight coefficient is subsequently adjusted through an optimization algorithm. Parameter a k 、b k 、c k The initial value is set to 1 or according to the output scale of the target risk prediction model, where the output scale of the model refers to the numerical range, unit or magnitude of the model prediction result or output.

[0094] Furthermore, the objective function of the initial fusion model is set based on the predicted target variable. In one feasible approach, for the classification problem of predicting the probability of default, the log-likelihood function is used as the objective function, as shown in the following formula 5:

[0095]

[0096] Among them, yj is the true label of the jth sample. For classification problems, the true label is the category label, indicating the category to which the sample belongs, and N is the total number of samples.

[0097] For the regression problem of predicting the growth prospects of corporate revenue, the mean square error is used as the objective function, as shown in Formula 6 below:

[0098]

[0099] The parameters of the initial fusion model are optimized. Specifically, the gradient descent method is used to calculate the partial derivative of the objective function E with respect to each parameter and update the parameters, as shown in Formula 7 below:

[0100]

[0101] Where θ is the parameter to be optimized, η is the learning rate, and t is the number of iterations. The parameters are updated iteratively until the objective function converges or the preset number of iterations is reached.

[0102] The Adam optimizer is used to accelerate convergence by combining momentum and adaptive learning rate. Furthermore, a regularization term is added to the objective function to prevent overfitting. The adjusted objective function is as follows:

[0103] E reg =E+λ∑ θ θ 2 Formula 8

[0104] Among them, λ is the regularization coefficient. Further, the learning rate η and the regularization coefficient λ hyperparameters are adjusted to improve the optimization effect.

[0105] The parameters in the initial fusion model are updated to the optimized parameters to obtain the financial risk assessment model, which takes advantage of the model fusion and improves the prediction performance of the fused financial risk assessment model. The performance of the financial risk assessment model is evaluated on the validation set to prevent overfitting.

[0106] By evaluating the performance of the financial risk assessment model on the test set, different evaluation criteria, such as AUC value, KS value, etc., are calculated, and the performance of the financial risk assessment model is compared with that of a single risk prediction model to verify the effectiveness of the fusion.

[0107] In one feasible approach, after obtaining the financial risk assessment model, the model is retrained regularly with new data to maintain the timeliness of the model, and the model's predictive performance and data distribution changes are monitored to issue warnings and adjust the model in a timely manner.

[0108] In practical applications, the financial risk assessment model is deployed to the credit scoring card system, and the financial data of the subject to be assessed is obtained in real time to achieve automated financial risk assessment of the subject to be assessed. Based on the results of the financial risk assessment, the credit limit, interest rate, approval rate, etc. of the subject to be assessed are adjusted.

[0109] Therefore, through the solution of the embodiment of the present application, the financial dimension indicators of the object to be evaluated are extracted from the financial data of the object to be evaluated, thereby reflecting the overall business performance and development of the object to be evaluated; the financial risk assessment model is obtained by integrating the prediction results of multiple risk prediction models, and the advantages of each risk prediction model are combined to make the prediction results of the final financial risk assessment model more accurate, giving full play to the advantages of model fusion, quantifying the financial risks of the object to be evaluated, and obtaining more accurate financial risk assessment results.

[0110] See also Figure 3 As shown, the embodiment of the present application also provides a financial risk assessment device, including:

[0111] The acquisition unit is used to acquire the financial data of the object to be evaluated.

[0112] In one feasible method, financial data is obtained from multiple perspectives around the object to be evaluated, for example, real-time acquisition of sales invoice data of the object to be evaluated, basic information of the object to be evaluated, industry data of the industry in which the object to be evaluated is located, and macroeconomic data.

[0113] The extraction unit is used to analyze financial data, extract financial dimension indicators of the object to be evaluated, and screen the financial dimension indicators.

[0114] After obtaining the financial data of the subject to be evaluated, further analysis is performed on the financial data to extract financial dimension indicators that reflect the overall business performance and development of the subject to be evaluated from the financial data of each dimension. Specifically, by combining financial data of multiple dimensions, financial dimension indicators of multiple dimensions of the subject to be evaluated are further obtained, such as the dimension of amount, the dimension of time, the dimension of customer relationship, and the dimension of goods / services.

[0115] Furthermore, the financial dimension indicators of the object to be evaluated are screened to find abnormal correlations or unreasonable patterns in the financial dimension indicators. By processing these financial dimension indicators, the quality of the data can be improved, making the data more reliable and useful. Specifically, the financial dimension indicators are analyzed for correlation; the importance of the financial dimension indicators is calculated, and the financial dimension indicators with importance higher than the preset value are screened out; and the financial dimension indicators with importance higher than the preset value obtained by screening are combined.

[0116] The evaluation unit is used to input the screened financial dimension indicators into the financial risk evaluation model and output the financial risk evaluation results of the object to be evaluated; wherein the financial risk evaluation model is obtained by integrating the prediction results of multiple risk prediction models.

[0117] Specifically, obtain sample financial data of multiple samples; analyze the sample financial data, extract sample financial dimension indicators of multiple samples, and screen the sample financial dimension indicators; establish multiple risk prediction models based on the screened sample financial dimension indicators; select the target risk prediction model from the multiple risk prediction models, and for each prediction problem, screen the target risk prediction model from the multiple risk prediction models based on multiple evaluation criteria. Combine multiple evaluation criteria, for example, select a risk prediction model with an AUC value higher than a preset value and a mean square error lower than a preset value as the target risk prediction model. In this way, the advantages of each target risk prediction model are combined to make the prediction results of the final financial risk assessment model more accurate, and retain the interpretability of the single model in the fusion model.

[0118] The prediction results of the target risk prediction model are integrated to construct a financial risk assessment model. In one feasible method, for each sample, the prediction results of the target variable predicted by the target risk prediction model for each screened sample financial dimension indicator are integrated to obtain an initial fusion model.

[0119] Specifically, the prediction results of different target risk prediction models are weighted and fused by multiplication fusion and / or exponential fusion. Multiplication fusion enhances the interaction between models, fully utilizes the advantages of each model, and reduces the possible errors of a single model. Exponential fusion emphasizes the differences between models and improves the sensitivity of the model to data. The two fusion methods give full play to the advantages of model fusion and improve the prediction performance of the fusion model.

[0120] The information interaction, execution process, etc. between the units in the above-mentioned device are based on the same concept as the embodiment of the method of the present application. For specific contents, please refer to the description in the embodiment of the method of the present application, and will not be repeated here.

[0121] An embodiment of the present application also provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the financial risk assessment method described in the above embodiment is implemented.

[0122] An embodiment of the present application also provides a computer storage medium configured to store computer program instructions, and the computer program instructions are used to execute the financial risk assessment method described in the above embodiments.

[0123] It should be pointed out that, according to the needs of implementation, the various components / steps described in the embodiments of the present application can be split into more components / steps, or two or more components / steps or partial operations of components / steps can be combined into new components / steps to achieve the purpose of the embodiments of the present application.

[0124] The above-mentioned method according to the embodiment of the present application can be implemented in hardware, firmware, or implemented as software or computer code that can be stored in a recording medium (such as a CD-ROM, RAM, floppy disk, hard disk or magneto-optical disk), or implemented as a computer code originally stored in a remote recording medium or a non-temporary machine-readable medium downloaded through a network and to be stored in a local recording medium, so that the method described herein can be stored in such software processing on a recording medium using a general-purpose computer, a dedicated processor or programmable or dedicated hardware (such as an application-specific integrated circuit (ASIC) or a field programmable gate array (FPGA)). It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component (for example, a random access memory (RAM), a read-only memory (ROM), a flash memory, etc.) that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method described herein is implemented. In addition, when a general-purpose computer accesses

[0125] When code is used to implement the methods shown herein, execution of the code transforms a general-purpose computer into a special-purpose computer for performing the methods shown herein.

[0126] Those of ordinary skill in the art will appreciate that the units and method steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for specific applications, but such implementation should not be considered to be beyond the scope of the embodiments of the present application.

[0127] The above implementation methods are only used to illustrate the embodiments of the present application, and are not limitations on the embodiments of the present application. Ordinary technicians in the relevant technical field can make various changes and modifications without departing from the spirit and scope of the embodiments of the present application. Therefore, all equivalent technical solutions also belong to the scope of the embodiments of the present application. The scope of patent protection of the embodiments of the present application should be limited by the claims.

Claims

1. A financial risk assessment method, characterized in that: include: Obtain financial data of the subject to be assessed; Analyze the financial data, extract financial dimension indicators of the object to be evaluated, and screen the financial dimension indicators; The screened financial dimension indicators are input into a financial risk assessment model, and a financial risk assessment result of the object to be assessed is output; wherein the financial risk assessment model is obtained by integrating the prediction results of multiple risk prediction models.

2. The method according to claim 1, characterized in that: Filter the financial dimension indicators. include: For each of the financial dimension indicators, calculating a first correlation coefficient between the financial dimension indicator and the target variable, and a second correlation coefficient between the financial dimension indicators; Eliminate the financial dimension indicators whose first correlation coefficient is less than a first preset correlation coefficient, and process the financial dimension indicator pairs whose correlation coefficients among the financial dimension indicators are higher than a second preset correlation coefficient; The importance of the financial dimension indicators is calculated, the financial dimension indicators whose importance is higher than a preset value are screened out, and the financial dimension indicators whose importance is higher than the preset value are combined.

3. The method according to claim 2, characterized in that The target variables include: probability of default and corporate revenue growth prospects.

4. The method according to claim 3, characterized in that The financial risk assessment model is obtained by the following steps: Obtain sample financial data for multiple samples; Analyzing the sample financial data of the multiple samples, extracting sample financial dimension indicators of the multiple samples, and screening the sample financial dimension indicators; Based on the screened sample financial dimension indicators, establish a variety of risk prediction models; Selecting a target risk prediction model from the multiple risk prediction models; The prediction results of the target risk prediction model are integrated to construct a financial risk assessment model.

5. The method according to claim 4, characterized in that The target risk prediction model is selected from the multiple risk prediction models. include: Evaluating the prediction effect of each risk prediction model on the predicted target variable based on the sample financial data; Based on the evaluation results, a target risk prediction model is screened out from the multiple risk prediction models.

6. The method according to claim 4, characterized in that The prediction results of the target risk prediction model are integrated to construct a financial risk assessment model. include: For each sample, the prediction results of the target variable predicted by the target risk prediction model for each screened sample financial dimension indicator are integrated to obtain an initial integrated model; The objective function of the initial fusion model is set based on the predicted target variable, and the parameters of the initial fusion model are iteratively updated to obtain the financial risk assessment model.

7. The method according to claim 6, characterized in that The prediction results of the target variable predicted by the target risk prediction model for each of the screened sample financial dimension indicators are integrated, include: The prediction results of different target risk prediction models are weighted and fused by multiplication fusion and / or exponential fusion.

8. A financial risk assessment device, characterized in that: include: An acquisition unit, used for acquiring financial data of the object to be evaluated; An extraction unit, configured to analyze the financial data, extract the financial dimension indicators of the object to be evaluated, and screen the financial dimension indicators; An evaluation unit is used to input the screened financial dimension indicators into a financial risk evaluation model and output the financial risk evaluation results of the object to be evaluated; wherein the financial risk evaluation model is obtained by integrating the prediction results of multiple risk prediction models.

9. An electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer storage medium, characterized in that: The device is configured to store computer program instructions for executing the method according to any one of claims 1 to 7.

Images