WO2021159735A1

WO2021159735A1 - Credit risk assessment method and apparatus, and computer device and storage medium

Info

Publication number: WO2021159735A1
Application number: PCT/CN2020/124259
Authority: WO
Inventors: 苏雪琦; 王健宗; 程宁
Original assignee: 平安科技（深圳）有限公司
Priority date: 2020-09-18
Filing date: 2020-10-28
Publication date: 2021-08-19
Also published as: CN112116245A

Abstract

A credit risk assessment method and apparatus, and a computer device and a storage medium, which relate to the technical field of artificial intelligence, and can improve the accuracy of credit risk assessment and ensure the reliability of a credit risk assessment result. The method comprises: acquiring consumption behavior time series data and attribute data corresponding to a user to be assessed (101); inputting the consumption behavior time series data into a preset consumption behavior feature extraction model to carry out feature extraction, so as to obtain a consumption behavior feature vector corresponding to the consumption behavior time series data (102); inputting the attribute data into a preset attribute feature extraction model to carry out feature extraction, so as to obtain an attribute feature vector corresponding to the user to be assessed (103); and determining, according to the consumption behavior feature vector and the attribute feature vector, a credit risk assessment result corresponding to the user to be assessed (104). In the method, machine learning technology is used, and the method is applicable to credit risk assessment.

Description

Credit risk assessment method, device, computer equipment and storage medium

This application claims the priority of a Chinese patent application filed with the Chinese Patent Office on September 18, 2020, the application number is 202010984025.5, and the invention title is "Credit Risk Assessment Method, Apparatus, Computer Equipment and Storage Medium", the entire content of which is incorporated by reference Incorporated in this application.

Technical field

This application relates to the field of artificial intelligence technology, in particular to a credit risk assessment method, device, computer equipment and storage medium.

Background technique

Credit refers to the form of value movement that is conditional on repayment and interest payment. It usually includes credit activities such as bank deposits and loans. The occurrence of loans is bound to be accompanied by risks. Before the repayment period expires, major changes in the lender’s financial status are likely to affect Due to its ability to perform contracts, bad debts, bad debts and other risks occur, so it is of great significance to conduct credit risk assessments on lenders in advance.

The inventor discovered during the research process that at present, the lender’s credit risk is usually evaluated based on the lender’s personal information, value data, and credit data. Among them, personal information includes information such as the person’s gender and educational background. The data includes the monthly income and fixed assets of the lender, and the credit data includes the number of credit cards of the lender and the number of credit overdue times. However, the inventor realizes that the lender’s value data and credit information data are both static data. The characteristics of static data are stable and not easy to change or slow to change. It cannot reflect the future trend of the lender. Therefore, if you only rely on the lender Credit risk assessment based on static data cannot obtain accurate credit risk assessment results, and the accuracy of credit risk assessment is low.

Summary of the invention

This application provides a credit risk assessment method, device, computer equipment, and storage medium, which are mainly capable of improving the accuracy of credit risk assessment and ensuring the reliability of credit risk assessment results.

According to the first aspect of this application, a credit risk assessment method is provided, including:

Obtain the consumption behavior time series data and attribute data corresponding to the user to be evaluated;

Input the consumption behavior time series data into a preset consumption behavior feature extraction model for feature extraction, and obtain a consumption behavior feature vector corresponding to the consumption behavior time series data;

Inputting the attribute data into a preset attribute feature extraction model for feature extraction, and obtaining an attribute feature vector corresponding to the user to be assessed;

According to the consumption behavior feature vector and the attribute feature vector, a credit risk assessment result corresponding to the user to be assessed is determined.

According to the second aspect of this application, a credit risk assessment device is provided, including:

The acquiring unit is used to acquire the consumption behavior time series data and attribute data corresponding to the user to be evaluated;

The first extraction unit is configured to input the consumption behavior time series data into a preset consumption behavior feature extraction model for feature extraction to obtain a consumption behavior feature vector corresponding to the consumption behavior time series data;

The second extraction unit is configured to input the attribute data into a preset attribute feature extraction model for feature extraction, and obtain the attribute feature vector corresponding to the user to be evaluated;

The determining unit is configured to determine the credit risk assessment result corresponding to the user to be assessed according to the consumption behavior feature vector and the attribute feature vector.

According to a third aspect of the present application, there is provided a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the following steps are implemented:

Inputting the attribute data into a preset attribute feature extraction model for feature extraction, and obtaining the attribute feature vector corresponding to the user to be evaluated;

According to a fourth aspect of the present application, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, and the processor implements the following steps when the program is executed:

This application can simultaneously consider the impact of static data and dynamic data on credit risk assessment, thereby ensuring the reliability of credit risk assessment results and improving the prediction accuracy of credit risk assessment results.

Description of the drawings

The drawings described here are used to provide a further understanding of the application and constitute a part of the application. The exemplary embodiments and descriptions of the application are used to explain the application, and do not constitute an improper limitation of the application. In the attached picture:

Figure 1 shows a flowchart of a credit risk assessment method provided by an embodiment of the present application;

Figure 2 shows a flowchart of another credit risk assessment method provided by an embodiment of the present application;

Fig. 3 shows a schematic structural diagram of a credit risk assessment device provided by an embodiment of the present application;

Fig. 4 shows a schematic structural diagram of another credit risk assessment device provided by an embodiment of the present application;

Fig. 5 shows a schematic diagram of the physical structure of a computer device provided by an embodiment of the present application.

Detailed ways

Hereinafter, the present application will be described in detail with reference to the drawings and in conjunction with the embodiments. It should be noted that the embodiments in the application and the features in the embodiments can be combined with each other if there is no conflict.

The technical solution of the present application can be applied to the fields of artificial intelligence, blockchain and/or big data technology, such as neural network technology. Optionally, the data involved in this application, such as time series data, attribute data, and/or evaluation results, can be stored in a database, or can be stored in a blockchain, which is not limited in this application.

At present, the lender’s credit risk is usually evaluated based on the lender’s personal information, value data and credit information data. Among them, the personal information includes the gender and educational background of the lender, and the value data includes the lender’s monthly income. , Fixed assets, etc. The credit data includes the number of credit cards of the lender, the number of credit overdues, etc. However, the lender’s value data and credit information data are both static data. The characteristics of static data are stable and difficult to change or slow to change. It cannot reflect the future trend of the lender. Therefore, if you only use the lender’s static data for credit Risk assessment cannot obtain accurate credit risk assessment results, and the accuracy of credit risk assessment is low.

In order to solve the above-mentioned problems, an embodiment of the present application provides a credit risk assessment method. As shown in FIG. 1, the method includes:

101. Acquire consumption behavior time series data and attribute data corresponding to the user to be assessed.

Among them, the users to be assessed are users who need to conduct credit risk assessment, and the consumption behavior time series data of the users to be assessed are established based on the consumption records of the users to be assessed over a period of time, specifically including the historical consumption frequency time series data and consumption of the users to be assessed. Time series data of the total amount, time series data of the product category to which the maximum consumption amount belongs, time series data of consumption and monthly income proportions, etc., for example, obtain the consumption record of the evaluated user within one year, and determine the equal time interval as one month, and construct the to-be-assessed The user’s historical consumption frequency time series data is (3,2,3,5,7,8,9,10,3,6,7,5), which represents the consumption of the user to be assessed from January to December of the year Frequency: 3 consumptions in the first month, 2 consumptions in the second month, 3 consumptions in the third month, 5 consumptions in the fourth month, etc. In addition, the attribute data of users to be assessed includes: users to be assessed The personal information, value data and credit data of the user to be evaluated, where personal information includes the gender, age, education, hell, marital status, etc. of the user to be evaluated; value data mainly refers to the personal financial status of the user to be evaluated, specifically including: monthly income , Fixed assets, account balances, percentage of monthly consumption income, repayment amount, etc.; credit data includes: the number of credit cards of users to be assessed, total credit, the number of successful borrowings and the number of overdue payments, etc.

For the embodiment of this application, in order to overcome the defect of inaccurate credit risk assessment results caused by only the static data (attribute data) of the user to be assessed in the prior art, the embodiment of this application adopts the artificial intelligence field Machine learning technology introduces time-series data on the consumption time of users to be evaluated, that is, introduces dynamic data that can reflect the future trends of users to be evaluated, and combines it with static data to evaluate the credit risk of users to be evaluated. Assess the credit risk assessment accuracy of users and ensure the reliability of credit risk assessment results. The embodiments of this application are mainly suitable for credit risk assessment. The executive subject of the embodiments of this application is a device or equipment that can perform credit risk assessments for users to be assessed. It can be set on the client or server side.

Specifically, when a user to be evaluated applies for a loan, the user to be evaluated will upload personal information and value data in advance, and the credit business staff will receive the personal information and value data of the user to be evaluated for verification and store it in the server. At the same time, the credit business staff It will also retrieve the credit data of the users to be evaluated and the consumption records for a period of time for storage. After that, the credit evaluation staff will conduct a credit risk assessment of the users to be evaluated, so as to decide whether to lend to them based on the results of the credit risk assessment. The specific credit business The personnel can trigger the credit risk evaluation instruction by clicking the credit risk evaluation button in the server interface. After the server receives the credit risk evaluation instruction of the user to be evaluated, it will retrieve the consumption record and attribute data of the user to be evaluated over a period of time. And according to the consumption record of the user to be evaluated in a period of time, the consumption behavior time series data corresponding to the user to be evaluated is generated, so as to jointly predict the credit risk assessment result of the user to be evaluated based on the consumption behavior time series data and attribute data.

102. Input the consumption behavior time series data into a preset consumption behavior feature extraction model for feature extraction, and obtain a consumption behavior feature vector corresponding to the consumption behavior time series data.

Wherein, the preset consumption behavior feature extraction model may specifically be an Xgboost consumption behavior feature extraction model. For the embodiment of the present application, the Xgboost consumption behavior feature extraction model can process time series data. The Xgboost consumption behavior feature extraction model mainly includes an input layer. , Hidden layer and output layer, input the consumption behavior time series data corresponding to the user to be evaluated through the input layer, and then use the hidden layer to perform feature extraction on the consumption behavior time series data to obtain the consumption behavior feature vector corresponding to the user to be evaluated. Further, The output layer can predict the consumption behavior feature vector of the user to be evaluated, and obtain the preliminary risk assessment result corresponding to the user to be evaluated, which can introduce the time series data of the consumption behavior of the user to be evaluated, that is, dynamic data, and can use Xgboost consumption behavior characteristics The extraction model extracts the consumer behavior feature vector of the user to be evaluated, so that it can be combined with the static data of the user to be evaluated for joint credit risk assessment.

103. Input the attribute data into a preset attribute feature extraction model for feature extraction, and obtain an attribute feature vector corresponding to the user to be evaluated.

Among them, the preset attribute feature extraction model may be a preset convolutional neural network model. For the embodiment of the present application, in order to combine the attribute data (static data) of the user to be evaluated and the time series data of consumption behavior (dynamic data), they are treated together. To evaluate the loan risk of the appraising user, the attribute data of the user to be evaluated needs to be extracted. Specifically, the convolutional neural network model can be used to extract the feature of the attribute data of the user to be evaluated. The convolutional neural network model mainly includes an input layer, a hidden layer, and The output layer inputs the attribute data corresponding to the user to be evaluated through the input layer, and then uses the hidden layer to perform feature extraction on the input attribute data to obtain the attribute feature vector corresponding to the user to be evaluated, and output the attribute feature vector through the output layer.

104. Determine a credit risk assessment result corresponding to the user to be assessed according to the consumption behavior feature vector and the attribute feature vector.

Among them, the credit risk evaluation result of the user under evaluation includes default and non-default. For example, if the output credit risk evaluation result is 0, it means that the user under evaluation will default. That is, if the user under evaluation lends, the user under evaluation is very It may default and cannot repay the loan on time; if the output credit risk assessment result is 1, it means that the user under evaluation will not default, that is, if the user under evaluation makes a loan, the user under evaluation will probably not default and will be able to repay the loan on time. For the embodiment of the present application, in order to improve the credit risk assessment accuracy of the user to be evaluated, it is necessary to conduct a risk assessment of the user to be evaluated based on both static data and dynamic data.

Specifically, according to the consumption behavior feature vector, the preliminary risk assessment result corresponding to the user to be assessed is determined. The preliminary risk assessment result is the preliminary risk assessment result obtained only based on the time series data of the consumption behavior of the user to be assessed, and the attribute data of the user needs to be considered. Therefore, according to the preliminary risk assessment results, consumer behavior feature vectors, and attribute feature vectors, the final credit risk assessment results corresponding to the users to be assessed can be determined. Specifically, the preliminary risk assessment results, consumer behavior feature vectors, and attribute features can be The vectors are collectively used as input vectors, which are input to the preset credit risk assessment model for credit risk assessment, and the credit risk assessment results corresponding to the users to be assessed are obtained. The preset credit risk assessment model may be a preset logistic regression model. The credit risk assessment result output by the preset logistic regression model is used to determine whether or not to lend to the user under evaluation. For example, if the credit risk assessment result output by the preset logistic regression model is 0, it is determined that the user under evaluation is likely to default, so it is not correct It lends; if the credit risk assessment result output by the preset logistic regression model is 1, it is determined that the user to be assessed is likely to not default, so lending to it.

The credit risk assessment method provided by the embodiments of the present application is compared with the current method of evaluating the credit risk of the user to be assessed based on the static data of the user to be assessed. This application can obtain the corresponding consumption behavior time series data and attributes of the user to be assessed. Data; and input the consumption behavior time series data into a preset consumption behavior feature extraction model for feature extraction to obtain the consumption behavior feature vector corresponding to the consumption behavior time series data; at the same time, input the attribute data into the preset The attribute feature extraction model performs feature extraction to obtain the attribute feature vector corresponding to the user to be assessed; finally, according to the consumption behavior feature vector and the attribute feature vector, the credit risk assessment result corresponding to the user to be assessed is determined, thereby By obtaining the time series data of the consumption behavior of the user to be evaluated, dynamic data reflecting the future change trend of the user to be evaluated can be introduced. When the user to be evaluated is subjected to risk assessment, the consumption behavior feature vector and attribute feature vector of the user to be evaluated can be extracted and based on Consumer behavior feature vectors and attribute feature vectors predict the credit risk assessment results of users to be assessed, and can consider the impact of static data and dynamic data on credit risk assessment at the same time, so as to ensure the reliability of credit risk assessment results and improve the credit risk assessment results Prediction accuracy.

Further, in order to better explain the above credit risk assessment process, as a refinement and extension of the above embodiment, an embodiment of the present application provides another credit risk assessment method, as shown in FIG. 2, the method includes :

201. Acquire consumption behavior time series data and attribute data corresponding to the user to be evaluated.

For the embodiment of this application, in order to obtain the consumption behavior time series data, step 201 specifically includes: obtaining the consumption record and the corresponding consumption time of the user to be assessed within a preset time period; according to the consumption record and the consumption time , Determine the consumption behavior time series data corresponding to the user to be evaluated. Among them, the preset time period can be set according to business needs. It should be noted that in order to ensure the accuracy of the credit risk assessment of the user to be evaluated, the preset time period should not be set too short, for example, to obtain the user to be evaluated last year All consumption records from January to December and their corresponding consumption time, and according to the consumption time corresponding to multiple consumption records of the user to be evaluated in a year, the historical consumption frequency time series data of the user to be evaluated and the time series data of customer consumption amount are respectively counted , The time series data of the commodity category to which the largest consumption amount belongs, the time series data of consumption and the proportion of monthly income, for example, the time series data of the consumption and the proportion of monthly income of the users to be assessed from January to December last year are (0.5, 0.3, 0.8 ,0.2,0.5,0.9,0.7,0.9,0.8,0.6,0.7,0.5), in addition, when the user to be evaluated applies for a loan, personal information and value data will be filled in, and the personal information and value data of the user to be evaluated by the credit business staff After verification, it will be stored on the server, and at the same time, the credit data related to the user under evaluation will be called and stored on the server, so that when the user under evaluation conducts credit risk evaluation, the risk assessment is carried out based on the time series data and attribute data of the user under evaluation. , To improve the accuracy of the credit risk assessment of users to be assessed.

Further, the acquired attribute data of the user to be evaluated may include multiple pieces of attribute data, and different data may have a strong correlation. In order to achieve the purpose of feature dimensionality reduction and prevent overfitting, it can be correlated. Only part of the attribute data in the attribute data is selected as input variables. Based on this, the method also includes: if there are multiple attribute data, verifying the correlation between each attribute data to obtain The correlation verification result; according to the correlation verification result, the target attribute data is filtered from the various attribute data, and further, the correlation between the various attribute data is verified to obtain one of the various attribute data The correlation verification result between the various attribute data includes: calculating the correlation coefficient between the various attribute data; determining the correlation verification result between the various attribute data according to the calculated correlation coefficients, and at the same time, the According to the correlation verification result, screening the target attribute data from the respective attribute data includes: determining, according to the correlation verification result, the attribute data that is relevant and the attributes that are not relevant in the respective attribute data Data; filter a preset number of attribute data from the relevant attribute data, and determine the preset number of attribute data and the non-relevant attribute data as target attribute data.

Specifically, a preset correlation verification algorithm can be used to calculate the correlation coefficient between each attribute data, and the specific calculation formula of the correlation coefficient is as follows:

Wherein, ρ _s is the correlation coefficient between the different attributes of _{_{data, d i = x i '-y}} i', x i ' is the attribute data X x component of rank _i, y _i' is the component attribute data Y y _i of rank, d _i is x _i 'and y _i' of rank deviation, n is the dimension attribute data X and Y, after calculating the attribute data X and the correlation coefficient Y, according to the correlation coefficient The size judges whether the attribute data X and Y are correlated, for example, if the calculated correlation coefficient is greater than or equal to 0.829, then the attribute data X and Y are considered to be correlated; if the calculated correlation coefficient is less than 0.829 , It is considered that there is no correlation between the attribute data X and Y, so that the correlation verification result between different attribute data can be determined. Further, according to the correlation verification result between different attribute data, multiple attribute data with relevance are determined, and then a preset number of attribute data is filtered from the multiple attribute data with relevance, for example, it is determined that there are 5 There are correlations between the two attribute data. In order to achieve the purpose of feature dimensionality reduction, two attribute data are filtered from five relevant attribute data, and the filtered attribute data and non-relevant attribute data are selected. Commonly used as target attribute data, so that the target attribute data can be input to the preset attribute feature extraction model for feature extraction, and the attribute feature vector corresponding to the user to be evaluated is obtained, thereby achieving the purpose of feature dimensionality reduction and prevention of overfitting .

202. Input the consumption behavior time series data into a preset consumption behavior feature extraction model for feature extraction, and obtain a consumption behavior feature vector corresponding to the consumption behavior time series data.

Among them, the preset consumption behavior feature extraction model may specifically be the Xgboost consumption behavior feature extraction model. The Xgboost consumption behavior feature extraction model can process time series data. For the embodiment of this application, the dynamic data of the user to be evaluated (consumption behavior Time series data), then input it into the Xgboost consumer behavior feature extraction model for feature extraction, and obtain the consumer behavior feature vector corresponding to the user to be evaluated. At the same time, according to the consumer behavior feature vector, the credit risk of the user to be evaluated can be preliminary Evaluate.

203. Input the attribute data into a preset attribute feature extraction model for feature extraction, and obtain an attribute feature vector corresponding to the user to be evaluated.

For the embodiment of the present application, after performing correlation verification on multiple attribute data of the user to be evaluated, and after determining the target attribute data in the multiple attribute data, the determined target attribute data is input into a preset attribute feature extraction model for feature extraction, Wherein, the preset attribute feature extraction model may specifically be a convolutional neural network model, and the attribute feature vector of the user to be evaluated is extracted through the convolutional neural network model.

204. Determine a preliminary risk assessment result corresponding to the user to be assessed according to the consumption behavior feature vector.

For this embodiment, after using the Xgboost consumption behavior feature extraction model to extract the consumption behavior feature vector of the user to be evaluated, the output layer of the Xgboost model can be used to make a preliminary assessment of the credit risk of the user to be evaluated based on the extracted consumption behavior feature vector. Obtain a preliminary risk assessment result of the user to be assessed, where the preliminary risk assessment result includes the default or non-default of the user to be assessed.

205. Determine a credit risk assessment result corresponding to the user to be assessed according to the preliminary risk assessment result, the consumption behavior feature vector, and the attribute feature vector.

For the embodiments of this application, in order to improve the accuracy of credit risk assessment of users to be assessed, it is necessary to simultaneously consider the consumption behavior time series data (dynamic data), attribute data (static data) of the users to be assessed, and the preliminary risk assessment results output by the Xgboost model. Based on this, step 205 specifically includes: integrating the preliminary risk assessment result, the consumer behavior feature vector, and the attribute feature vector to obtain an integrated feature vector; and inputting the integrated feature vector to a preset The credit risk assessment model performs risk assessment, and obtains the credit risk assessment result corresponding to the user to be assessed. Among them, the preset credit risk assessment model may be a preset logistic regression model.

For example, the consumer behavior feature vector of the user to be evaluated is (x1, x2, x3), the attribute feature vector is (x4, x5, x6, x7), the preliminary risk assessment result obtained by the Xgboost model is u, and the above vectors are integrated , The integrated feature vector is (x1,x2,x3,x4,x5,x6,x7,u), and the integrated feature vector (x1,x2,x3,y1,y2,y3,y4,u) As the input variable of the preset logistic regression model, the credit risk of the user to be evaluated is evaluated. When the credit risk of the user to be evaluated is specifically evaluated, the probability that the user to be evaluated will not default can be calculated. The specific calculation formula is as follows:

Among them, α, β are the estimated parameters of the preset logistic regression model, (x ₁ ,x ₂ ,...,x _n ,u) are the integrated variables, and E(y) is the probability that the user to be evaluated will not default, according to The calculated probability determines whether the user under evaluation will default.

For the embodiments of this application, when constructing the preset consumption behavior feature extraction model, the preset attribute feature extraction model, and the preset credit risk assessment model, they can be trained as a whole to collect the consumption behavior time series data and attributes of a large number of users Data, and the corresponding repayment situation, determine the sample data set, train the sample data set, construct the preset consumption behavior feature extraction model, the preset attribute feature extraction model and the preset credit wind evaluation model, when the credit risk is preset The evaluation model is specifically a logistic regression model, and the parameters of the logistic regression model are adjusted using a preset maximum likelihood algorithm to obtain the optimal evaluation parameters.

The embodiment of this application provides another credit risk evaluation method. Compared with the current method of evaluating the credit risk of the user to be evaluated based on the static data of the user to be evaluated, this application can obtain the time series data and corresponding consumption behavior of the user to be evaluated. Attribute data; and input the consumption behavior time series data into a preset consumption behavior feature extraction model for feature extraction to obtain the consumption behavior feature vector corresponding to the consumption behavior time series data; at the same time, input the attribute data into the preset Set the attribute feature extraction model to perform feature extraction to obtain the attribute feature vector corresponding to the user to be evaluated; finally, according to the consumption behavior feature vector and the attribute feature vector, determine the credit risk assessment result corresponding to the user to be evaluated, and By acquiring the time series data of the consumption behavior of the user to be evaluated, it is possible to introduce dynamic data reflecting the future change trend of the user to be evaluated. When the user to be evaluated is subjected to risk assessment, by extracting the consumption behavior feature vector and attribute feature vector of the user to be evaluated, and Predict the credit risk assessment results of users to be assessed based on the consumption behavior feature vector and attribute feature vector, and can consider the impact of static data and dynamic data on the credit risk assessment at the same time, so as to ensure the reliability of the credit risk assessment results and improve the credit risk assessment The prediction accuracy of the result.

Further, as a specific implementation of FIG. 1, an embodiment of the present application provides a credit risk assessment device. As shown in FIG. 3, the device includes: an acquiring unit 31, a first extracting unit 32, a second extracting unit 33, and Determine unit 34.

The acquiring unit 31 may be used to acquire the consumption behavior time series data and attribute data corresponding to the user to be evaluated. The acquiring unit 31 is the main functional module of the device for acquiring the consumption behavior time series data and attribute data corresponding to the user to be evaluated.

The first extraction unit 32 may be configured to input the consumption behavior time series data into a preset consumption behavior feature extraction model for feature extraction, and obtain a consumption behavior feature vector corresponding to the consumption behavior time series data. The first extraction unit 32 is the main functional module in this device that inputs the consumption behavior time series data into a preset consumption behavior feature extraction model for feature extraction, and obtains the consumption behavior feature vector corresponding to the consumption behavior time series data, and is also The core module.

The second extraction unit 33 may be configured to input the attribute data into a preset attribute feature extraction model for feature extraction, and obtain the attribute feature vector corresponding to the user to be evaluated. The second extraction unit 33 is a main functional module of the device that inputs the attribute data into a preset attribute feature extraction model for feature extraction, and obtains the attribute feature vector corresponding to the user to be evaluated.

The determining unit 34 may be configured to determine the credit risk assessment result corresponding to the user to be assessed based on the consumption behavior feature vector and the attribute feature vector. The determining unit 34 is a main functional module of the device that determines the credit risk assessment result corresponding to the user to be assessed based on the consumption behavior feature vector and the attribute feature vector, and is also a core module.

Further, in order to determine the credit risk assessment result corresponding to the user to be assessed, as shown in FIG. 4, the determining unit 34 includes a first determining module 341 and a second determining module 342.

The first determining module 341 may be used to determine a preliminary risk assessment result corresponding to the user to be assessed according to the consumption behavior feature vector.

The second determining module may be configured to determine the credit risk assessment result corresponding to the user to be assessed based on the preliminary risk assessment result, the consumption behavior feature vector, and the attribute feature vector.

Further, in order to determine the credit risk evaluation result corresponding to the user to be evaluated, the second determination module 341 includes: an integration sub-module and an evaluation sub-module.

The integration sub-module may be used to integrate the preliminary risk assessment result, the consumer behavior feature vector, and the attribute feature vector to obtain an integrated feature vector.

The evaluation sub-module may be used to input the integrated feature vector into a preset credit risk evaluation model for risk evaluation, and obtain a credit risk evaluation result corresponding to the user to be evaluated.

Further, in order to perform correlation verification on the attribute data, the device further includes: a verification unit 35 and a screening unit 36.

The verification unit 35 may be used to verify the correlation between the various attribute data if there are multiple attribute data, and obtain the correlation verification result between the various attribute data.

The screening unit 36 may be used to screen target attribute data from the various attribute data according to the correlation verification result.

The second extraction unit 33 may be specifically configured to input the target attribute data into a preset attribute feature extraction model for feature extraction, and obtain the attribute feature vector corresponding to the user to be evaluated.

Further, in order to determine the correlation verification result between the various attribute data, the verification unit 35 includes: a calculation module 351 and a determination module 352.

The calculation module 351 may be used to calculate the correlation coefficient between the various attribute data.

The determining module 352 may be used to determine the correlation verification result between the respective attribute data according to the respective calculated correlation coefficients.

Further, in order to filter the target attribute data from the various attribute data, the screening unit 36 includes: a determination module 361 and a screening module 362.

The determining module 361 may be configured to determine, according to the correlation verification result, the attribute data with relevance and the attribute data with no relevance in the respective attribute data.

The screening module 362 may be used to filter a preset number of attribute data from the relevant attribute data, and determine that the preset number of attribute data and the non-relevant attribute data are Target attribute data.

Further, in order to obtain the consumption behavior time series data corresponding to the user to be evaluated, the obtaining unit 31 includes: an obtaining module 311 and a determining module 312.

The acquiring module 311 may be used to acquire the consumption record of the user to be assessed within a preset time period and the corresponding consumption time.

The determining module 312 may be used to determine the consumption behavior time series data corresponding to the user to be evaluated according to the consumption record and the consumption time.

It should be noted that, for other corresponding descriptions of the various functional modules involved in the credit risk assessment device provided in the embodiment of the present application, reference may be made to the corresponding description of the method shown in FIG. 1, which will not be repeated here.

Based on the above-mentioned method shown in Figure 1, correspondingly, an embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the following steps are implemented: The consumption behavior time series data and attribute data; input the consumption behavior time series data into a preset consumption behavior feature extraction model for feature extraction, and obtain the consumption behavior feature vector corresponding to the consumption behavior time series data; input the attribute data into The preset attribute feature extraction model performs feature extraction to obtain the attribute feature vector corresponding to the user to be assessed; and the credit risk assessment result corresponding to the user to be assessed is determined according to the consumption behavior feature vector and the attribute feature vector. Optionally, when the program is executed by the processor, other steps of the method in the foregoing embodiment may be implemented, which will not be repeated here. Further optionally, the storage medium involved in this application, such as a computer-readable storage medium, may be non-volatile or volatile.

Based on the above-mentioned method shown in FIG. 1 and the embodiment of the apparatus shown in FIG. 3, an embodiment of the present application also provides a physical structure diagram of a computer device. As shown in FIG. 5, the computer device includes: a processor 41, The memory 42 and the computer program that is stored on the memory 42 and can run on the processor, wherein the memory 42 and the processor 41 are both set on the bus 43, the processor 41 implements the following steps when the program is executed: The consumption behavior time series data and attribute data corresponding to the user; input the consumption behavior time series data into a preset consumption behavior feature extraction model for feature extraction, and obtain the consumption behavior feature vector corresponding to the consumption behavior time series data; convert the attribute data Input to a preset attribute feature extraction model for feature extraction to obtain the attribute feature vector corresponding to the user to be assessed; determine the credit risk assessment result corresponding to the user to be assessed according to the consumption behavior feature vector and the attribute feature vector . Optionally, the processor 41 may also execute the program to implement other steps of the method in the foregoing embodiment, which will not be repeated here.

Through the technical solution of this application, this application can obtain the consumption behavior time series data and attribute data corresponding to the user to be evaluated; and input the consumption behavior time series data into a preset consumption behavior feature extraction model for feature extraction to obtain the consumption behavior The consumption behavior feature vector corresponding to the time series data; at the same time, the attribute data is input into the preset attribute feature extraction model for feature extraction to obtain the attribute feature vector corresponding to the user to be evaluated; finally according to the consumption behavior feature vector And the attribute feature vector to determine the credit risk assessment result corresponding to the user to be evaluated, and thus by acquiring the time series data of the consumption behavior of the user to be evaluated, dynamic data reflecting the future change trend of the user to be evaluated can be introduced. When conducting risk assessment, by extracting the consumer behavior feature vector and attribute feature vector of the user to be assessed, and predicting the credit risk assessment result of the user to be assessed based on the consumer behavior feature vector and attribute feature vector, the credit risk assessment result of the user to be assessed can be considered at the same time as static data and dynamic data. The influence of risk assessment can ensure the reliability of credit risk assessment results and improve the prediction accuracy of credit risk assessment results.

Obviously, those skilled in the art should understand that the above-mentioned modules or steps of this application can be implemented by a general computing device, and they can be concentrated on a single computing device or distributed in a network composed of multiple computing devices. Above, alternatively, they can be implemented with program codes executable by a computing device, so that they can be stored in a storage device for execution by the computing device, and in some cases, they can be executed in a different order than here. Perform the steps shown or described, or fabricate them into individual integrated circuit modules respectively, or fabricate multiple modules or steps of them into a single integrated circuit module for implementation. In this way, this application is not limited to any specific combination of hardware and software.

The above descriptions are only preferred embodiments of the application, and are not intended to limit the application. For those skilled in the art, the application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this application shall be included in the protection scope of this application.

Claims

A credit risk assessment method, which includes:

Obtain the consumption behavior time series data and attribute data corresponding to the user to be evaluated;

Input the consumption behavior time series data into a preset consumption behavior feature extraction model for feature extraction, and obtain a consumption behavior feature vector corresponding to the consumption behavior time series data;

Inputting the attribute data into a preset attribute feature extraction model for feature extraction, and obtaining an attribute feature vector corresponding to the user to be assessed;

According to the consumption behavior feature vector and the attribute feature vector, a credit risk assessment result corresponding to the user to be assessed is determined.
The method according to claim 1, wherein the determining a credit risk assessment result corresponding to the user to be assessed based on the consumption behavior feature vector and the attribute feature vector comprises:

Determine the preliminary risk assessment result corresponding to the user to be assessed according to the consumption behavior feature vector;

According to the preliminary risk assessment result, the consumption behavior feature vector and the attribute feature vector, the credit risk assessment result corresponding to the user to be assessed is determined.
The method according to claim 2, wherein the determining the credit risk assessment result corresponding to the user to be assessed based on the preliminary risk assessment result, the consumption behavior feature vector, and the attribute feature vector comprises:

Integrating the preliminary risk assessment result, the consumer behavior feature vector, and the attribute feature vector to obtain an integrated feature vector;

The integrated feature vector is input into a preset credit risk assessment model for risk assessment, and the credit risk assessment result corresponding to the user to be assessed is obtained.
The method according to claim 1, wherein, after said acquiring the consumption behavior time series data and attribute data corresponding to the user to be evaluated, the method further comprises:

If there are multiple attribute data, verify the correlation between the various attribute data, and obtain the correlation verification result between the various attribute data;

Screening target attribute data from the various attribute data according to the correlation verification result;

The inputting the attribute data into a preset attribute feature extraction model for feature extraction to obtain the attribute feature vector corresponding to the user to be assessed includes:

The target attribute data is input into a preset attribute feature extraction model for feature extraction, and the attribute feature vector corresponding to the user to be evaluated is obtained.
The method according to claim 4, wherein the verifying the correlation between the various attribute data to obtain the correlation verification result between the various attribute data comprises:

Calculating the correlation coefficient between the various attribute data;

According to the calculated correlation coefficients, the correlation verification result between the various attribute data is determined.
The method according to claim 4, wherein the screening target attribute data from the respective attribute data according to the correlation verification result comprises:

According to the correlation verification result, determine the attribute data that has correlation and the attribute data that does not have correlation among the various attribute data;

A preset number of attribute data is filtered from the attribute data with relevance, and the preset number of attribute data and the attribute data that have no relevance are determined as target attribute data.
The method according to any one of claims 1 to 6, wherein said acquiring the consumption behavior time series data corresponding to the user to be assessed comprises:

Acquiring the consumption record and the corresponding consumption time of the user to be assessed in a preset time period;

According to the consumption record and the consumption time, determine the consumption behavior time series data corresponding to the user to be assessed.
A credit risk assessment device, which includes:

The acquiring unit is used to acquire the consumption behavior time series data and attribute data corresponding to the user to be evaluated;

The first extraction unit is configured to input the consumption behavior time series data into a preset consumption behavior feature extraction model for feature extraction to obtain a consumption behavior feature vector corresponding to the consumption behavior time series data;

The second extraction unit is configured to input the attribute data into a preset attribute feature extraction model for feature extraction, and obtain the attribute feature vector corresponding to the user to be evaluated;

The determining unit is configured to determine the credit risk assessment result corresponding to the user to be assessed according to the consumption behavior feature vector and the attribute feature vector.
A computer-readable storage medium on which a computer program is stored, wherein the following steps are implemented when the computer program is executed by a processor:

Obtain the consumption behavior time series data and attribute data corresponding to the user to be evaluated;

Input the consumption behavior time series data into a preset consumption behavior feature extraction model for feature extraction, and obtain a consumption behavior feature vector corresponding to the consumption behavior time series data;

Inputting the attribute data into a preset attribute feature extraction model for feature extraction, and obtaining an attribute feature vector corresponding to the user to be assessed;

According to the consumption behavior feature vector and the attribute feature vector, a credit risk assessment result corresponding to the user to be assessed is determined.
8. The computer-readable storage medium according to claim 9, wherein when determining the credit risk assessment result corresponding to the user to be assessed according to the consumption behavior characteristic vector and the attribute characteristic vector, the specific realization is implemented:

Determine the preliminary risk assessment result corresponding to the user to be assessed according to the consumption behavior feature vector;

According to the preliminary risk assessment result, the consumption behavior feature vector and the attribute feature vector, the credit risk assessment result corresponding to the user to be assessed is determined.
10. The computer-readable storage medium according to claim 10, wherein the determination of the credit risk assessment result corresponding to the user to be assessed is based on the preliminary risk assessment result, the consumption behavior feature vector, and the attribute feature vector When, the specific realization:

Integrating the preliminary risk assessment result, the consumer behavior feature vector, and the attribute feature vector to obtain an integrated feature vector;

The integrated feature vector is input into a preset credit risk assessment model for risk assessment, and the credit risk assessment result corresponding to the user to be assessed is obtained.
8. The computer-readable storage medium according to claim 9, wherein after the acquisition of the consumption behavior time series data and attribute data corresponding to the user to be assessed, the computer program is further used to implement when the computer program is executed by the processor:

If there are multiple attribute data, verify the correlation between the various attribute data, and obtain the correlation verification result between the various attribute data;

Screening target attribute data from the various attribute data according to the correlation verification result;

When the attribute data is input into a preset attribute feature extraction model for feature extraction, and the attribute feature vector corresponding to the user to be evaluated is obtained, the specific implementation is as follows:

The target attribute data is input into a preset attribute feature extraction model for feature extraction, and the attribute feature vector corresponding to the user to be evaluated is obtained.
The computer-readable storage medium according to claim 12, wherein when the target attribute data is selected from the respective attribute data according to the correlation verification result, the specific realization is achieved:

According to the correlation verification result, determine the attribute data that has correlation and the attribute data that does not have correlation among the various attribute data;

A preset number of attribute data is filtered from the attribute data with relevance, and the preset number of attribute data and the attribute data that have no relevance are determined as target attribute data.
The computer-readable storage medium according to any one of claims 9-13, wherein when said acquiring the time series data of consumption behavior corresponding to the user to be evaluated, the specific implementation is as follows:

Acquiring the consumption record and the corresponding consumption time of the user to be assessed in a preset time period;

According to the consumption record and the consumption time, determine the consumption behavior time series data corresponding to the user to be assessed.
A computer device includes a memory, a processor, and a computer program stored on the memory and running on the processor, wherein the computer program is executed by the processor to implement the following steps:

Obtain the consumption behavior time series data and attribute data corresponding to the user to be evaluated;

Input the consumption behavior time series data into a preset consumption behavior feature extraction model for feature extraction, and obtain a consumption behavior feature vector corresponding to the consumption behavior time series data;

Inputting the attribute data into a preset attribute feature extraction model for feature extraction, and obtaining an attribute feature vector corresponding to the user to be assessed;

According to the consumption behavior feature vector and the attribute feature vector, a credit risk assessment result corresponding to the user to be assessed is determined.
15. The computer device according to claim 15, wherein, when determining the credit risk assessment result corresponding to the user to be assessed according to the consumption behavior feature vector and the attribute feature vector, the specific realization is implemented:

Determine the preliminary risk assessment result corresponding to the user to be assessed according to the consumption behavior feature vector;

According to the preliminary risk assessment result, the consumption behavior feature vector and the attribute feature vector, the credit risk assessment result corresponding to the user to be assessed is determined.
16. The computer device according to claim 16, wherein when determining the credit risk assessment result corresponding to the user to be assessed according to the preliminary risk assessment result, the consumption behavior feature vector, and the attribute feature vector, the specific accomplish:

Integrating the preliminary risk assessment result, the consumer behavior feature vector, and the attribute feature vector to obtain an integrated feature vector;

The integrated feature vector is input into a preset credit risk assessment model for risk assessment, and the credit risk assessment result corresponding to the user to be assessed is obtained.
The computer device according to claim 15, wherein, after the acquisition of the consumption behavior time series data and attribute data corresponding to the user to be evaluated, the computer program is further used to realize when the computer program is executed by the processor:

If there are multiple attribute data, verify the correlation between the various attribute data, and obtain the correlation verification result between the various attribute data;

Screening target attribute data from the various attribute data according to the correlation verification result;

When the attribute data is input into a preset attribute feature extraction model for feature extraction, and the attribute feature vector corresponding to the user to be evaluated is obtained, the specific implementation is as follows:

The target attribute data is input into a preset attribute feature extraction model for feature extraction, and the attribute feature vector corresponding to the user to be evaluated is obtained.
18. The computer device according to claim 18, wherein when the target attribute data is selected from the respective attribute data according to the correlation verification result, the specific realization is achieved:

According to the correlation verification result, determine the attribute data that has correlation and the attribute data that does not have correlation among the various attribute data;

A preset number of attribute data is filtered from the attribute data with relevance, and the preset number of attribute data and the attribute data that have no relevance are determined as target attribute data.
The computer device according to any one of claims 15-19, wherein when the acquisition of the consumption behavior time series data corresponding to the user to be evaluated is specifically implemented:

Acquiring the consumption record and the corresponding consumption time of the user to be assessed in a preset time period;

According to the consumption record and the consumption time, determine the consumption behavior time series data corresponding to the user to be assessed.