CN114331735A

CN114331735A - Financial risk model training method and device and financial risk prediction method and device

Info

Publication number: CN114331735A
Application number: CN202111542825.2A
Authority: CN
Inventors: 陶雅楠; 张昊楠
Original assignee: Pross Technology Chongqing Co ltd
Current assignee: Pross Technology Chongqing Co ltd
Priority date: 2021-12-16
Filing date: 2021-12-16
Publication date: 2022-04-12

Abstract

The application provides a financial risk model training method, which comprises the following steps: acquiring a training sample set marked with a financial risk result; any training sample consists of preset financial characteristic indexes of multiple dimensions, and the value corresponding to the financial characteristic indexes is determined according to financial data and/or audit data of an enterprise; correcting the initial weight corresponding to the training sample according to a preset cost matrix; wherein the cost matrix is constructed based on losses incurred by each error classification; inputting the training sample set with the corrected weight into the financial risk model for training to obtain a trained financial risk model. Through the technical scheme, on one hand, the financial condition of an enterprise can be accurately reflected by determining financial characteristic indexes of multiple dimensions; on the other hand, the problem of sample data unbalance can be solved based on cost sensitive learning, and accuracy of model prediction is improved.

Description

Financial risk model training method and device and financial risk prediction method and device

Technical Field

The application relates to the technical field of computers, in particular to a financial risk model training method and device and a financial risk prediction method and device.

Background

With the further development of human economic activities and the perfection of related operating theories, in order to better expand the operating range and the operating scale of enterprises, increase the advantages in the same-industry competition and promote the maximization of enterprise benefits, modern enterprises mostly adopt the best praise and debt. Under the guidance of such an operation mode, enterprises raise funds and use the funds for production and operation activities by means of bank borrowing, bond issuing, leasing, commercial credit and the like.

Due to the fact that the enterprise has liabilities in operation, the phenomena that partial enterprises cannot pay for the liabilities due to poor operation, cash flow is reduced, finally the liabilities are not compensated, the yield is broken, and the like inevitably occur, and the situation is called enterprise financial risk.

For investors, there is a need to identify enterprises with financial risk in investment transactions to avoid irresponsible blood cost due to risk events. Meanwhile, for the society, maintaining the stability of the financial market is also a big matter of harmonious and stable relation society. Therefore, there is a need for early and effective identification and early warning of businesses with financial risk issues.

Disclosure of Invention

In view of the above, in order to solve the problem of enterprise financial risk prediction, the present application provides a financial risk model training method and apparatus, and a financial risk prediction method and apparatus.

Specifically, the method is realized through the following technical scheme:

in a first aspect, the present application provides a method for training a financial risk model based on cost-sensitive learning, including:

acquiring a training sample set marked with a financial risk result; any training sample consists of preset financial characteristic indexes of multiple dimensions, and the value corresponding to the financial characteristic indexes is determined according to financial data and/or audit data of an enterprise;

correcting the initial weight corresponding to the training sample according to a preset cost matrix; wherein the cost matrix is constructed based on losses incurred by each error classification;

inputting the training sample set with the corrected weight into the financial risk model for training to obtain a trained financial risk model.

In one embodiment of the present disclosure, the financial risk model includes a machine learning model.

In one embodiment of the present disclosure, the machine learning model comprises a decision tree model or a logistic regression model.

In an embodiment of the present disclosure, the financial characteristic index includes a secondary index obtained after processing a primary index in the financial statement of the enterprise and/or the audit data.

In one embodiment of the present disclosure, the financial risk outcome is at risk or no risk and the financial risk model is a binary model.

In one embodiment of the present disclosure, the financial risk result is a plurality of levels corresponding to financial risk, and the financial risk model is a multi-classification model.

In one embodiment of the present disclosure, the cost matrix is constructed based on the loss caused by each error classification, and includes:

determining loss caused when each classification is judged to be error classification based on a preset cost function; wherein, when correctly classified, the loss is zero;

and constructing the cost matrix according to the loss caused by the classification result corresponding to each classification.

In a second aspect, the present application provides a financial risk prediction method based on cost sensitive learning, including:

acquiring financial data and/or audit data of an enterprise, and determining numerical values corresponding to preset multi-dimensional financial characteristic indexes based on the financial data and/or the audit data;

inputting the numerical value corresponding to the financial characteristic index into the financial risk model trained according to the first aspect;

and outputting the financial risk result of the enterprise based on the trained financial risk model.

In a third aspect, the present application further provides a training device for a financial risk model based on cost-sensitive learning, the training device including:

the acquisition unit is used for acquiring the training sample set marked with the financial risk result; any training sample consists of preset financial characteristic indexes of multiple dimensions, and the value corresponding to the financial characteristic indexes is determined according to financial data and/or audit data of an enterprise;

the correcting unit is used for correcting the initial weight corresponding to the training sample according to a preset cost matrix; wherein the cost matrix is constructed based on losses incurred by each error classification;

and the training unit is used for inputting the training sample set with the corrected weight into the financial risk model for training to obtain the trained financial risk model.

In a fourth aspect, the present application further provides a financial risk prediction device based on cost sensitive learning, including:

the system comprises a determining unit, a processing unit and a processing unit, wherein the determining unit is used for acquiring financial data and/or audit data of an enterprise and determining numerical values corresponding to preset multi-dimensional financial characteristic indexes based on the financial data and/or the audit data;

the input unit is used for inputting the numerical value corresponding to the financial characteristic index into the financial risk model trained according to the first aspect;

and the output unit is used for outputting the financial risk result of the enterprise based on the trained financial risk model.

The technical scheme provided by the embodiment of the application can have the following beneficial effects:

the method comprises the steps of constructing a training sample through preset financial characteristic indexes of multiple dimensions, and constructing a cost matrix to correct the weight of the training sample by introducing a cost sensitive learning algorithm in a data preprocessing stage, so that a model meets the cost sensitive characteristic. Through the technical scheme, on one hand, the financial condition of an enterprise can be accurately reflected by determining financial characteristic indexes of multiple dimensions; on the other hand, the problem of sample data unbalance can be solved based on cost sensitive learning, and accuracy of model prediction is improved.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.

Drawings

FIG. 1 is a flow chart illustrating a method for training a cost-sensitive learning-based financial risk model according to an exemplary embodiment of the present application;

FIG. 2 is a flow chart illustrating a method for cost-sensitive learning based financial risk prediction according to an exemplary embodiment of the present application;

FIG. 3 is a diagram illustrating a hardware configuration of an electronic device according to an exemplary embodiment of the present application;

FIG. 4 is a block diagram of a training apparatus for a cost-sensitive learning based financial risk model according to an exemplary embodiment of the present application;

FIG. 5 is a block diagram of a financial risk prediction device based on cost-sensitive learning according to an exemplary embodiment of the present application.

Detailed Description

Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in this application and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

It is to be understood that although the terms first, second, third, etc. may be used herein to describe various information, such information should not be limited to these terms. These terms are only used to distinguish one type of information from another. For example, first information may also be referred to as second information, and similarly, second information may also be referred to as first information, without departing from the scope of the present application. The word "if" as used herein may be interpreted as "at … …" or "when … …" or "in response to a determination", depending on the context.

At present, the research aiming at the enterprise financial risk is mostly realized by a statistical-based method. For example, different evaluation index systems can be established for different types of enterprises, and the financial conditions of the enterprises are judged according to the scores of the indexes, but the identification accuracy and the identification efficiency are still to be improved.

In view of this, the present application provides a technical scheme for constructing a training sample through preset financial characteristic indexes of multiple dimensions, and constructing a cost matrix based on a cost-sensitive learning algorithm to modify the weight of the training sample, so that a trained model meets cost-sensitive characteristics.

In implementation, a set of training samples labeled with financial risk results may be obtained. Wherein, arbitrary training sample comprises the financial characteristic index of a plurality of predetermined dimensions, the numerical value that financial characteristic index corresponds is confirmed according to the financial data and/or the audit data of enterprise.

Then, the initial weight corresponding to the training sample can be corrected according to a preset cost matrix. Wherein the cost matrix is constructed based on the loss incurred by each misclassification.

Further, the training sample set after the weight correction can be input into the financial risk model for training to obtain a trained financial risk model.

In the technical scheme, the training sample is constructed through the preset financial characteristic indexes with multiple dimensions, and in the data preprocessing stage, the weight of the training sample is modified through constructing the cost matrix by introducing the cost sensitive learning algorithm, so that the model meets the cost sensitive characteristic. Through the technical scheme, on one hand, the financial condition of an enterprise can be accurately reflected by determining financial characteristic indexes of multiple dimensions; on the other hand, the problem of sample data unbalance can be solved based on cost sensitive learning, and accuracy of model prediction is improved.

Next, examples of the present application will be described in detail.

Referring to fig. 1, fig. 1 is a flowchart illustrating a method for training a financial risk model based on cost sensitive learning according to an exemplary embodiment of the present application, and as shown in fig. 1, the method includes the following steps:

step 101: acquiring a training sample set marked with a financial risk result; wherein, arbitrary training sample comprises the financial characteristic index of a plurality of predetermined dimensions, the numerical value that financial characteristic index corresponds is confirmed according to the financial data and/or the audit data of enterprise.

Step 102: correcting the initial weight corresponding to the training sample according to a preset cost matrix; wherein the cost matrix is constructed based on the loss incurred by each misclassification.

Step 103: inputting the training sample set with the corrected weight into the financial risk model for training to obtain a trained financial risk model.

In this embodiment, a training sample set labeled with financial risk results may be obtained.

Wherein, arbitrary training sample comprises the financial characteristic index of a plurality of predetermined dimensions, the numerical value that financial characteristic index corresponds is confirmed according to the financial data and/or the audit data of enterprise.

It should be noted that the financial data of the enterprise may include financial statements of the enterprise, and may also include analysis data provided by a third-party analysis organization.

For example, in enterprises appearing in the market in China, due to regulatory requirements, the enterprises appearing in the market need to regularly disclose corporate financial reports every year, the financial reports disclosed by the enterprises appearing in the market at present can be inquired and downloaded at official websites of Shanghai stock exchanges and Shenzhen stock exchanges, and the financial data of the enterprises required by research can be obtained by obtaining and analyzing the financial reports of the enterprises appearing in the market.

For another example, the analysis data provided by the third-party analysis organization for the financial data of the enterprise can be obtained from financial terminals providing data and information of financial markets at home and abroad.

In addition, as the enterprises which go to the market are required to disclose the financial reports of the previous year 4 months and 30 days before the second year at the latest in China, the financial reports of the enterprises which can be inquired in the current year are incomplete, and in the sample data selection, in order to consider the comprehensiveness of data acquisition and the timeliness of samples, the financial data of the enterprises of the latest complete financial years can be selected for research.

For example, the financial data of all-listed-a-shares in 2018, 2019 and 2020 can be selected to construct training samples. The 2020 enterprise financial data can be used as a test set to optimize the parameters of the model so as to improve the accuracy of prediction.

The audit data can be an audit report issued by a registered accountant for the enterprise.

When the financial risk result is labeled for the training sample, the training sample is historical data, so the financial condition of the real enterprise of the year in which the training sample is located can be labeled.

In one embodiment shown, the financial risk outcome is at risk or no risk and the financial risk model is a binary model.

For example, the financial risk results may be classified into two categories, risky and non-risky, depending on whether the financial status is abnormal, and the corresponding financial risk model is a binary model.

In another embodiment shown, the financial risk results are a plurality of levels of financial risk, and the financial risk model is a multi-classification model.

For example, the financial risk results may be classified as high risk, medium risk, low risk, and no risk according to the extent of enterprise loss, while the corresponding financial risk model is a multi-classification model.

It should be noted that, since the enterprise usually has credit problems such as debt problems or default problems when facing financial risks, the financial risk result can also be used to indicate the credit risk of the enterprise, and the trained financial risk model can also be used to predict the credit risk of the enterprise.

Further, the obtained financial data of the enterprise can be subjected to data processing to obtain a sample which can be used for training a financial risk model.

Specifically, numerical values corresponding to preset financial characteristic indexes of multiple dimensions can be calculated according to the obtained financial data of the enterprise, and training samples are constructed based on the financial characteristic indexes.

When selecting the index, the following key principles can be mainly followed:

firstly, the method is comprehensive in index, and when an enterprise financial risk index system is designed, not only the financial condition of the enterprise itself but also the influence of the environment where the enterprise is located need to be considered; the current financial index condition of the enterprise is considered, and the future predicted financial condition is also analyzed.

Secondly, the method is scientific in index selection, indexes are real and reliable, the method accords with the fact basis, and the phenomena of logic disorder and information stacking are avoided.

Then, the operability of the selection of the index, a data sample of which can be obtained, can be understood by the intended user.

Finally, the pertinence of the selected index is selected, different index systems need to be constructed according to different applicable objects and environments, and the established index systems can fully reflect the industrial characteristics and are adaptive to the local credit environment and the economic environment.

In addition, the method can also be combined with indexes selected by scholars at home and abroad when researching the enterprise financial risk.

In one embodiment, the financial characteristic indexes comprise secondary indexes obtained after processing primary indexes in the enterprise financial statement and/or the audit opinion indexes.

In one example, please refer to table 1, where table 1 is a table illustrating a financial characteristic index.

TABLE 1

As shown in table 1, 4 primary indicators and 16 secondary indicators may be selected for evaluating financial risks of enterprises, and the meanings of the indicators are known to those skilled in the art and are not described in detail.

The first-level index can be an index directly obtained according to financial data of an enterprise, and the second-level index is an index obtained after calculation according to the first-level index, so that financial characteristics of the enterprise can be reflected more specifically.

In addition, audit opinion indicators determined based on audit data may also be added, which may generally include 5 audit opinion types as shown in table 2, and since audit opinions are not a numerical value, a value may be assigned to each opinion type as shown in table 2 below:

TABLE 2

For example, the opinion type of the corresponding audit opinion indicator may be determined based on audit data in an audit report of the enterprise.

It should be noted that, when acquiring financial data of an enterprise, data may be first screened to remove the enterprise with a large data missing value.

In one example, the following table 3 shows statistical results of the stock in the a stock of 2018 and 2019 after performing statistical analysis on ST (Special treatment) stock and non-ST stock and performing elimination treatment.

TABLE 3

Wherein ST stock refers to a stock that has an abnormal financial or other condition to alert the stock of an investment risk.

It should be noted that, if the number of training samples in different classes is not very different, the classification result is usually not greatly affected, but if the number of training samples in different classes is too great, the classification result will be affected to a considerable extent.

For example, assuming that there are 998 countercase samples, but only 2 forward case samples, although the accuracy of the trained model for predicting the input as a countercase can reach 99.8%, such a model is often of no value because it is likely not to predict the forward case.

That is, when enterprise financial risk prediction is performed, the sample has a problem of category imbalance.

It can be understood that if model training is performed based on the samples constructed from the corporate financial data in table 3, the trained model will have poor prediction effect on ST stocks, which deviates from the original purpose of prediction using the financial risk model.

More importantly, the loss caused by classifying a company with financial risk as a company without financial risk problem is far greater than the loss caused by classifying a company without financial risk problem as a company with financial risk.

Therefore, it is desirable to avoid classifying a financially risky business as one that does not present a financial risk.

Therefore, a cost-sensitive learning algorithm can be introduced, wherein the cost-sensitive learning algorithm is an optimization algorithm for setting different weights for different types of samples under the premise of considering the misclassification cost, and quantizing the weights into sample cost to be merged into a model for learning.

Through training the financial risk model based on the cost sensitive learning algorithm, higher weight can be given to enterprises with financial risks, so that the classification condition of the enterprises with financial risks is concentrated on by the algorithm, the recall ratio of the enterprises with financial risks is improved, and the precision ratio of the enterprises with financial risks is reduced.

In this embodiment, the initial weight corresponding to the training sample may be modified according to a preset cost matrix. Wherein the cost matrix is constructed based on the loss incurred by each misclassification.

It is worth noting that the core element of the cost sensitive learning is to construct a cost matrix, which is shown in table 4 below.

TABLE 4

Among them, Cost_ijThe cost of predicting the ith sample as the jth sample is shown, and when i equals j, the cost is 0. The larger the loss due to misclassification, the larger the value of the cost. Further, if there are n classes, then a cost matrix of n x n can be constructed.

And the initial weight of each type of sample can be set to 1, and the initial weight of the training sample can be modified according to the following formula based on the preset cost matrix.

Where K represents the number of samples in the training set, K_nThe number of samples of the class i is shown, and cost (n) shows the misclassification cost of the class n, namely the cost for misclassifying other classes into n classes. And cost (i, n) represents the cost of classifying the actual n-type samples into i-type samples in the class with the total number of classes L.

In an embodiment shown, the loss caused when each classification is determined to be an erroneous classification may be determined based on a preset cost function; wherein, when correctly classified, the loss is zero; and constructing the cost matrix according to the loss caused by the classification result corresponding to each classification.

In one example, assuming that the preset cost function is F (i, j), i.e., F (i, j) represents a misclassification loss caused by judging i-class users as j-class, when the prediction is correct, i is j, no cost is generated.

In another example, assuming that the financial risk result is risky and risk-free, the cost matrix constructed may be as shown in table 5 below, when i is 1, a financial inauguration enterprise; when i is 0, it is a normal finance business, and then F (1,1) is F (0,0) is 0.

TABLE 5

Cost value	Forecast as a normal financial enterprise	Forecasting as a financial inauguration enterprise
			Actually being a normal financial enterprise	0	F(0,1)
Actually financial risk enterprises	F(1,0)	0

In this embodiment, the training sample set after the weight correction may be input into the financial risk model for training, so as to obtain a trained financial risk model.

After the training samples are processed as described in step 102, the training sample set with the corrected weights is input into the financial risk model for training, so that the trained financial risk model with the cost sensitive characteristic can be obtained.

In one illustrated embodiment, the financial risk model includes a machine learning model.

In particular, the classification of financial risk results may be solved based on a machine learning model.

Preferably, in one illustrated embodiment, the machine learning model comprises a decision tree model or a logistic regression model.

The decision tree model is a classic classification algorithm in data mining, can recommend a classification rule displayed in a decision tree form from a group of unordered and irregular data, and performs classification and prediction analysis by using the rule. When the decision tree algorithm is selected, the person skilled in the art can determine the algorithm according to actual needs.

Logistic regression is also a classification model in machine learning, the output value of the model is always between 0 and 1, the output value represents the probability that data belongs to a certain classification, and the final result of the classification can be determined by setting a threshold value. When the threshold is set, the skilled person can determine the threshold according to actual needs.

After the model training is completed, the method for predicting the financial risk based on the cost sensitive learning is further provided.

Referring to fig. 2, fig. 2 is a flowchart illustrating a financial risk prediction method based on cost sensitive learning according to an exemplary embodiment of the present application, and as shown in fig. 2, the method includes the following steps:

step 201: acquiring financial data and/or audit data of an enterprise, and determining numerical values corresponding to preset multi-dimensional financial characteristic indexes based on the financial data and/or the audit data;

step 202: inputting the numerical value corresponding to the financial characteristic index into the trained financial risk model;

step 203: and outputting the financial risk result of the enterprise based on the trained financial risk model.

Since it basically corresponds to the foregoing method embodiment, reference may be made to the partial description of the foregoing method embodiment for relevant points, which is not described herein again.

Corresponding to the method embodiment, the application also provides an embodiment of the device.

Corresponding to the embodiment of the method, the application also provides an embodiment of a training device of the financial risk model based on cost-sensitive learning. The embodiment of the training device based on the financial risk model of the cost-sensitive learning can be applied to electronic equipment. The device embodiments may be implemented by software, or by hardware, or by a combination of hardware and software. Taking a software implementation as an example, as a logical device, the device is formed by reading, by a processor of the electronic device where the device is located, a corresponding computer program instruction in the nonvolatile memory into the memory for operation. From a hardware aspect, as shown in fig. 3, a hardware structure diagram of an electronic device shown in an exemplary embodiment of the present application is shown, except for the processor, the memory, the network interface, and the nonvolatile memory shown in fig. 3, the electronic device where the apparatus is located in the embodiment may also include other hardware according to an actual function of the electronic device, which is not described again.

Referring to fig. 4, fig. 4 is a block diagram of a training apparatus for a financial risk model based on cost-sensitive learning according to an exemplary embodiment of the present application, and as shown in fig. 4, the training apparatus 400 for a financial risk model based on cost-sensitive learning may be applied in the electronic device shown in fig. 3, and includes:

an obtaining unit 401, which obtains a training sample set labeled with a financial risk result; any training sample consists of preset financial characteristic indexes of multiple dimensions, and the value corresponding to the financial characteristic indexes is determined according to financial data and/or audit data of an enterprise;

a correcting unit 402, which corrects the initial weight corresponding to the training sample according to a preset cost matrix; wherein the cost matrix is constructed based on losses incurred by each error classification;

and a training unit 403, inputting the training sample set with the corrected weights into the financial risk model for training, and obtaining a trained financial risk model.

In an embodiment, the financial risk model comprises a machine learning model.

In an embodiment, the machine learning model comprises a decision tree model or a logistic regression model.

In an embodiment, the financial characteristic index includes a secondary index obtained after processing a primary index in the financial statement of the enterprise and/or the audit opinion index.

In one embodiment, the financial risk outcome is risky and risk-free and the financial risk model is a binary model.

In one embodiment, the financial risk result is a plurality of levels corresponding to financial risk, and the financial risk model is a multi-classification model.

In one embodiment, the cost matrix is constructed based on the loss caused by each error classification, and comprises:

Referring to fig. 5, fig. 5 is a block diagram of a financial risk prediction apparatus based on cost-sensitive learning according to an exemplary embodiment of the present application, and as shown in fig. 5, the financial risk prediction apparatus 500 based on cost-sensitive learning may be applied in the electronic device shown in fig. 3, and includes:

the determining unit 501 is configured to obtain financial data and/or audit data of an enterprise, and determine preset numerical values corresponding to financial characteristic indexes of multiple dimensions based on the financial data and/or the audit data;

an input unit 502, which inputs the value corresponding to the financial characteristic index into the trained financial risk model;

an output unit 503, configured to output the financial risk result of the enterprise based on the trained financial risk model.

The embodiments in the present application are described in a progressive manner, and the same/similar parts in the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. Especially, for the embodiments of the client device and the apparatus, since they are substantially similar to the embodiments of the method, the description is simple, and for the relevant points, refer to the partial description of the embodiments of the method.

For the device embodiments, since they substantially correspond to the method embodiments, reference may be made to the partial description of the method embodiments for relevant points. The above-described embodiments of the apparatus are merely illustrative, wherein the modules described as separate parts may or may not be physically separate, and the parts displayed as modules may or may not be physical modules, may be located in one place, or may be distributed on a plurality of network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of the application. One of ordinary skill in the art can understand and implement it without inventive effort.

The apparatuses, modules or modules illustrated in the above embodiments may be implemented by a computer chip or an entity, or by an article with certain functions. A typical implementation device is a computer, which may take the form of a personal computer, laptop computer, cellular telephone, camera phone, smart phone, personal digital assistant, media player, navigation device, email messaging device, game console, tablet computer, wearable device, or a combination of any of these devices.

Corresponding to the method embodiment, the present specification also provides an embodiment of an electronic device. The electronic device includes: a processor and a memory for storing machine executable instructions; wherein the processor and the memory are typically interconnected by an internal bus. In other possible implementations, the device may also include an external interface to enable communication with other devices or components.

In this embodiment, the processor is caused to:

The present specification also provides another embodiment of an electronic device, corresponding to the above-described method embodiment. The electronic device includes: a processor and a memory for storing machine executable instructions; wherein the processor and the memory are typically interconnected by an internal bus. In other possible implementations, the device may also include an external interface to enable communication with other devices or components.

In this embodiment, the processor is caused to:

inputting the numerical value corresponding to the financial characteristic index into the financial risk model finished according to the training;

Other embodiments of the present application will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the invention following, in general, the principles of the application and including such departures from the present disclosure as come within known or customary practice within the art to which the invention pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the application being indicated by the following claims.

It will be understood that the present application is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the application is limited only by the appended claims.

The above description is only exemplary of the present application and should not be taken as limiting the present application, as any modification, equivalent replacement, or improvement made within the spirit and principle of the present application should be included in the scope of protection of the present application.

Claims

1. A training method of a financial risk model based on cost-sensitive learning comprises the following steps:

2. The method of claim 1, the financial risk model comprising a machine learning model.

3. The method of claim 2, the machine learning model comprising a decision tree model or a logistic regression model.

4. The method of claim 1, wherein the financial characteristic indicators comprise secondary indicators and/or audit opinion indicators obtained after processing primary indicators in the enterprise's financial statement.

5. The method of claim 1, the financial risk outcome being at risk or no risk, the financial risk model being a binary model.

6. The method of claim 1, wherein the financial risk results are a plurality of levels of financial risk correspondence, and the financial risk model is a multi-classification model.

7. The method of claim 1, the cost matrix constructed based on losses incurred by each misclassification, comprising:

8. A financial risk prediction method based on cost-sensitive learning, comprising:

inputting a value corresponding to the financial characteristic indicator into the trained financial risk model according to any one of claims 1-5;

9. A training device for a cost-sensitive learning based financial risk model, comprising:

10. A financial risk prediction device based on cost sensitive learning, comprising:

an input unit, inputting the value corresponding to the financial characteristic index into the trained financial risk model according to any one of claims 1-5;