Training, application methods, devices, equipment, and media for user credit prediction models

Abstract

This application provides a training and application method, apparatus, device, and medium for a user credit prediction model, relating to the field of big data technology. The training method includes: acquiring training samples; grouping user data according to data type to obtain multiple data groups; selecting multiple target pre-trained models corresponding to a target business type from multiple trained pre-trained models; inputting the multiple target data groups corresponding to the target business type from the multiple data groups into the multiple target pre-trained models corresponding to the target data groups, respectively, to obtain the first credit prediction result of each target pre-trained model for each sample user; using the multiple first credit prediction results of each sample user and the user tags of the sample user as training samples to train the user credit prediction model, thereby obtaining a trained user credit prediction model. According to this application embodiment, accurate prediction of user credit can be achieved.

Description

Technical Field

[0001] This application belongs to the field of big data technology, and in particular relates to a training and application method, device, equipment and medium for a user credit prediction model. Background Technology

[0002] In recent years, internet finance, represented by consumer finance, has developed rapidly, and people's consumption concepts and levels have also undergone significant changes and improvements. Against this backdrop, in order to enhance the competitiveness of consumer finance products and reduce the default risk for users in consumer finance businesses, it is necessary to conduct access judgments for different consumer finance products, that is, to assess the default risk of users under different products.

[0003] Existing technologies often rely on expert experience to specify rules or use expert scoring card models to assess user default risk. However, user data has many dimensions and high sparsity, exceeding the data range that experts can handle. Furthermore, consumer finance products are diverse and updated rapidly, while expert scoring card models often cannot be updated and adjusted in a timely manner, resulting in low accuracy in predicting user credit. Summary of the Invention

[0004] This application provides a training and application method, apparatus, device, and medium for a user credit prediction model, which can accurately predict user credit.

[0005] In a first aspect, embodiments of this application provide a method for training a user credit prediction model. The method includes: acquiring training samples, which include user data and user tags of sample users, and the user data includes various types of data related to a target business type; grouping the user data according to data type to obtain multiple data groups; selecting multiple target pre-trained models corresponding to the target business type from multiple trained pre-trained models, wherein one pre-trained model corresponds to one type of data group; inputting the multiple target data groups corresponding to the target business type from the multiple data groups into the multiple target pre-trained models corresponding to the target data groups, respectively, to obtain a first credit prediction result for each sample user by each target pre-trained model; and training the user credit prediction model using the multiple first credit prediction results of each sample user and the user tags of the sample user as training samples to obtain a trained user credit prediction model.

[0006] According to the implementation of the first aspect of this application, after grouping user data according to data type to obtain multiple data groups, the training method of the user credit prediction model further includes: performing data processing on the multiple data groups respectively to obtain data features derived from user data in the multiple data groups; using the data features derived from user data in the multiple data groups and the user tags of sample users as training samples for the pre-trained model, training the pre-trained model to obtain the trained pre-trained model.

[0007] According to any of the foregoing embodiments of the first aspect of this application, data processing is performed on multiple data groups to obtain data features derived from user data in the multiple data groups, including: performing the following steps on each data group: deleting user data with duplicate values, abnormal values, and missing values ​​in the data group; for the deleted user data, if the value type of the user data is a fixed value, using the value of the user data as a data feature derived from the user data; if the value type of the user data is a variable value, calculating the value variable of the user data within a preset time period as a data feature derived from the user data.

[0008] According to any of the foregoing embodiments of the first aspect of this application, the pre-trained model is trained using data features derived from user data in multiple data groups and user tags of sample users as training samples to obtain a trained pre-trained model, including: matching data features derived from user data in each data group with user tags of sample users to determine the user tags corresponding to the data features; using data features belonging to the same data group and the user tags corresponding to the data features as training samples for a pre-trained model to train the pre-trained model, thereby obtaining multiple trained pre-trained models.

[0009] According to any of the foregoing embodiments of the first aspect of this application, training a pre-trained model to obtain multiple trained pre-trained models includes: inputting multiple data sets into multiple pre-trained models corresponding to the data sets to obtain a second credit prediction result for each sample user by each pre-trained model; calculating the accuracy index value and stability index value of each pre-trained model based on the second credit prediction result; determining whether the accuracy index value of each pre-trained model is greater than or equal to a first preset threshold, and whether the stability index value of each pre-trained model is less than or equal to a second preset threshold; for pre-trained models whose accuracy index value is less than the first preset threshold and / or whose stability index value is greater than the second preset threshold, adjusting the model parameters of the pre-trained model and re-acquiring training samples to train the adjusted pre-trained model until the accuracy index value of the pre-trained model is greater than or equal to the first preset threshold and the stability index value is less than or equal to the second preset threshold, thereby obtaining multiple trained pre-trained models.

[0010] According to any of the foregoing embodiments of the first aspect of this application, the sample user's data related to the target business type includes the sample user's basic information, credit information, transaction information, asset information, and income information.

[0011] According to any of the foregoing embodiments of the first aspect of this application, the sample users are users who meet the preset screening rules and have preset type tags.

[0012] Secondly, embodiments of this application provide an application method for a user credit prediction model. The application method includes: acquiring user data of the user to be predicted; grouping the user data according to data type to obtain multiple data groups; inputting the multiple data groups into multiple pre-trained models corresponding to the data groups to obtain a third credit prediction result for each user to be predicted by each pre-trained model; inputting the multiple third credit prediction results of each user to be predicted into the user credit prediction model to obtain the final credit prediction result for the user to be predicted; the user credit prediction model is the user credit prediction model provided in the first aspect.

[0013] Thirdly, embodiments of this application provide a training apparatus for a user credit prediction model. The training apparatus includes: a first acquisition module for acquiring training samples, the training samples including user data and user tags of sample users, the user data including various types of data related to a target business type; a first grouping module for grouping the user data according to data type to obtain multiple data groups; a first selection module for selecting multiple target pre-trained models corresponding to the target business type from multiple trained pre-trained models, wherein one pre-trained model corresponds to one type of data group; a first input module for inputting the multiple target data groups corresponding to the target business type from the multiple data groups into the multiple target pre-trained models corresponding to the target data groups, respectively, to obtain a first credit prediction result for each sample user by each target pre-trained model; and a first training module for training the user credit prediction model using the multiple first credit prediction results of each sample user and the user tags of the sample user as training samples, to obtain a trained user credit prediction model.

[0014] Fourthly, embodiments of this application provide an application apparatus for a user credit prediction model. The application apparatus for the user credit prediction model includes: a second acquisition module for acquiring user data of the user to be predicted; a second grouping module for grouping the user data according to data type to obtain multiple data groups; a second input module for inputting the multiple data groups into multiple pre-trained models corresponding to the data groups to obtain a third credit prediction result of each pre-trained model for each user to be predicted; and a third input module for inputting multiple third credit prediction results of each user to be predicted into the user credit prediction model to obtain the final credit prediction result of the user to be predicted.

[0015] Fifthly, embodiments of this application provide an electronic device, which includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the training method for the user credit prediction model provided in the first aspect, and the steps of the application method for the user credit prediction model provided in the second aspect.

[0016] In a sixth aspect, embodiments of this application provide a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the steps of the training method for the user credit prediction model as provided in the first aspect, and the steps of the application method for the user credit prediction model as provided in the second aspect.

[0017] The user credit prediction model training and application method, apparatus, device, and medium of this application embodiment, after acquiring training samples, group the multi-dimensional user data in the training samples according to data type, so that each data group corresponds to a pre-trained model. Thus, when multiple target pre-trained models are selected according to the target business type, and the multiple first credit prediction results of the multiple target pre-trained models for sample users are used as training samples to train the user credit prediction model, the accuracy and robustness of the trained user credit prediction model will be higher when multiple target pre-trained models combine multi-dimensional user data, resulting in rich data dimensions and high utilization. Therefore, when using the multiple first credit prediction results of multiple target pre-trained models for model training, the model prediction result will also be more accurate. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating a training method for a user credit prediction model provided in an embodiment of this application.

[0020] Figure 2 This is a flowchart illustrating another training method for a user credit prediction model provided in an embodiment of this application.

[0021] Figure 3 This is a flowchart illustrating an application method for a user credit prediction model provided in an embodiment of this application.

[0022] Figure 4 This is a schematic diagram of the structure of a training device for a user credit prediction model provided in an embodiment of this application;

[0023] Figure 5 This is a schematic diagram of the structure of an application device for a user credit prediction model provided in an embodiment of this application;

[0024] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0025] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0026] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.

[0027] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0028] Various modifications and variations can be made to this application without departing from its spirit or scope, which will be apparent to those skilled in the art. Therefore, this application is intended to cover modifications and variations falling within the scope of the corresponding claims (the claimed technical solutions) and their equivalents. It should be noted that the embodiments provided in this application can be combined with each other without contradiction.

[0029] It should be noted that the acquisition, storage, use, and processing of data in this application embodiment all comply with the relevant provisions of national laws and regulations.

[0030] Before describing the technical solutions provided in the embodiments of this application, in order to facilitate understanding of the embodiments of this application, this application first specifically explains the problems existing in the prior art:

[0031] Because consumer finance businesses are characterized by numerous business scenarios, abundant data sources, diverse product types, and rapid iteration, they require predictive models applicable to different business scenarios to assess user default risk. Current technologies often rely on expert experience to specify rules or use expert scoring card models to assess user default risk. However, experts can only handle a limited range of data, and building expert scoring card models typically requires time for data collection and model development. The modeling process also necessitates coordination among experts from various fields, resulting in low model building efficiency, long iteration cycles, and an inability to make timely updates and adjustments. Therefore, in the era of big data, relying solely on expert experience or fixed scoring card models to predict user credit across different products cannot yield accurate prediction results.

[0032] To address the problems in the prior art, embodiments of this application provide a method, apparatus, device, and medium for training and applying a user credit prediction model.

[0033] The training method of the user credit prediction model provided in the embodiments of this application will be introduced first.

[0034] Figure 1 This is a flowchart illustrating a training method for a user credit prediction model provided in an embodiment of this application. Figure 1 As shown, the method may include the following steps S101 to S105.

[0035] S101. Obtain training samples.

[0036] S102. Group the user data according to the data type to obtain multiple data groups.

[0037] S103. Select multiple target pre-trained models that correspond to the target business type from multiple trained pre-trained models.

[0038] S104. Input the multiple target data groups corresponding to the target business type from the multiple data groups into the multiple target pre-trained models corresponding to the target data groups, and obtain the first credit prediction result of each target pre-trained model for each sample user.

[0039] S105. Using multiple first credit prediction results for each sample user and the user tags of the sample user as training samples for the user credit prediction model, train the user credit prediction model to obtain the trained user credit prediction model.

[0040] The specific implementation methods of the above steps will be described in detail below.

[0041] The training method for the user credit prediction model in this application involves grouping multi-dimensional user data in the training samples according to data type after obtaining the training samples, so that each data group corresponds to a pre-trained model. This allows for the selection of multiple target pre-trained models based on the target business type, and the use of multiple first credit prediction results of sample users from these target pre-trained models as training samples to train the user credit prediction model. Because the multiple target pre-trained models combine multi-dimensional user data, resulting in rich data dimensions and high utilization, the accuracy and robustness of the trained user credit prediction model are higher when trained using multiple first credit prediction results from these multiple target pre-trained models. Consequently, when using a highly accurate user credit prediction model to predict user credit, the accuracy of the model prediction results will also be higher.

[0042] The specific implementation methods for each of the above steps are described below.

[0043] In S101, users who meet preset screening rules and have preset type labels are pre-selected from all users of the consumer finance business as sample users. Then, various types of data related to the target business type of the sample users, as well as the user tags of the sample users, are acquired as training samples for the pre-trained model and the user credit prediction model. Among them, the various types of data related to the target business type of the sample users include the sample users' basic information, credit information, transaction information, asset information, and income information.

[0044] As an example of S101, sample users can be users whose consumer finance loans have matured and who are more than 90 days overdue after the loan matures, as well as users who have not defaulted after their loans matured. Preset type labels can include "good user" and "bad user" labels. Accordingly, sample users whose loan matures and are more than 90 days overdue can be labeled as "bad users," and users who have not defaulted after their loans mature can be labeled as "good users."

[0045] As another example of S101, the credit information of sample users can include their credit report information and credit rating information; the transaction information of sample users can include their loan information, loan application amount information, and transaction history information; and the income information of sample users can include their housing provident fund information and salary disbursement information. Among these, basic information, credit report information, credit rating information, loan information, and loan application amount information can be classified as general user data, while transaction history information, asset information, housing provident fund information, and salary disbursement information can be classified as scenario-based user data. Each type of information in scenario-based user data corresponds to a specific business scenario. When acquiring scenario-based user data for sample users, if a sample user has multiple types of scenario-based user data, only the scenario-based user data corresponding to the sample user's most recent business scenario is acquired. Furthermore, each type of information can be considered as a separate data type, and multiple types of information can correspond to the same target business type.

[0046] In S102, the user data is grouped according to the data types it includes, resulting in multiple data groups composed of different types of user data.

[0047] As an example of S102, if each type of information is considered as a separate type of data, then user data can be divided into basic information data group, credit information data group, credit rating information data group, loan information data group, loan application amount information data group, cash flow information data group, asset information data group, provident fund information data group, and payroll information data group.

[0048] In S103, pre-trained models corresponding to each type of data group are pre-trained based on multiple data groups to obtain multiple trained pre-trained models, and then multiple target pre-trained models corresponding to the target business type are selected from them.

[0049] As an example of S103, pre-trained models can be obtained in advance based on multiple data sets, including pre-trained models corresponding to basic information data sets, credit information data sets, credit rating information data sets, loan information data sets, loan application amount information data sets, cash flow information data sets, asset information data sets, housing provident fund information data sets, and payroll information data sets.

[0050] Taking payroll information as an example, if it is necessary to predict user credit in the business scenario corresponding to payroll information, then the multiple target pre-trained models corresponding to the target business type can be determined as five pre-trained models corresponding to the data groups of five general user data types, and one pre-trained model corresponding to the payroll information data group.

[0051] In S104, multiple target data groups corresponding to the target business type are selected from multiple data groups, and then the multiple target data groups are input into the target pre-trained model corresponding to each target data group to obtain the first credit prediction result of each target pre-trained model for each sample user.

[0052] As an example of S104, let's take payroll information as an example. Five general user data sets and the payroll information data set are selected from multiple data sets. Then, the basic information data set, credit information data set, credit rating information data set, loan information data set, loan application amount information data set, and payroll information data set are input into their respective pre-trained models. This yields the probability that each sample user is a bad user, predicted by the six pre-trained models, representing the first credit prediction result for each sample user by the six target pre-trained models.

[0053] In S105, the first credit prediction result of each target pre-trained model for each sample user is used as a data feature and matched with the user label of each sample user to determine the user label corresponding to each first credit prediction result. Then, multiple first credit prediction results of the same sample user and the user label of that sample user are substituted into the logistic regression algorithm to train the user credit prediction model, thereby obtaining the trained user credit prediction model.

[0054] As an example of S105, let's take payroll information as an example. By substituting the first credit prediction results of the six pre-trained models for each sample user, along with the user tags of each sample user, into the logistic regression algorithm for model training, a trained user credit prediction model for the business scenario corresponding to payroll information can be obtained. Similarly, user credit prediction models for other types of user data can also be obtained using the same training method, which will not be elaborated further in this embodiment.

[0055] According to some embodiments of this application, optionally, for each new business scenario, it is only necessary to train the pre-trained model corresponding to the data group of the scenario-type user data separately based on the scenario-type user data corresponding to the new business scenario.

[0056] In general, when selecting multiple target pre-trained models corresponding to the target business type from multiple pre-trained models, these models typically include pre-trained models corresponding to data sets of five general user types and pre-trained models corresponding to data sets of scenario-specific user types. Therefore, when training a user credit prediction model for a new business scenario, the pre-trained models corresponding to the data sets of the five general user types can be reused. In this case, it is only necessary to combine the newly trained pre-trained models corresponding to the data sets of scenario-specific user types to predict the user credit of the sample users, and then use the new prediction results as new training samples to train the user credit prediction model.

[0057] This reduces the workload of iterative training of the user credit prediction model, shortens the development cycle of the user credit prediction model, and accelerates the online deployment efficiency of the user credit prediction model, thereby further ensuring the accuracy of the user credit prediction model.

[0058] As another implementation of this application, such as Figure 2 As shown, in order to improve the accuracy of the pre-trained model, after S102, the training method of the user credit prediction model may also include the following steps S201 to S202.

[0059] S201. Perform data processing on multiple data groups separately to obtain data features derived from user data in multiple data groups.

[0060] When training the pre-trained model corresponding to each type of data set, it is necessary to pre-process the user data in multiple data sets to remove impurities and redundancy from the original user data, thereby obtaining data features with higher accuracy and usability.

[0061] S202. Using data features derived from user data in multiple data sets and user tags of sample users as training samples for the pre-trained model, the pre-trained model is trained to obtain a well-trained pre-trained model.

[0062] By substituting data features derived from user data in multiple datasets and user labels of sample users into a machine learning algorithm, a pre-trained model is trained, resulting in a well-trained model. Since these pre-trained models are trained based on data features with higher accuracy and usability, they can accurately reflect the data features of different datasets, thus leading to higher accuracy in the trained models.

[0063] For example, machine learning algorithms may include gradient boosting trees, linear regression, Naive Bayes, random forests, and ensemble models, etc. Gradient boosting trees may include gradient boosting decision trees (GBDT) and lightweight gradient boosting machines (LGBM), etc., and this application embodiment does not limit them.

[0064] To improve the accuracy of pre-trained models, S201, as one implementation method, can specifically include:

[0065] For each data group, perform the following steps: delete user data with duplicate, abnormal, or missing values; for the deleted user data, if the value type of the user data is a fixed value, use the value of the user data as a derived data feature; if the value type of the user data is a variable value, calculate the value variable of the user data within a preset time period as a derived data feature.

[0066] Specifically, data processing of user data across multiple datasets can involve two steps. First, remove user data with duplicate, outlier, or missing values ​​to eliminate impurities and redundancy from the original user data. Second, derive data features needed for training subsequent pre-trained models from the removed user data. For user data with fixed values, these values ​​can be directly used as derived data features. Since data such as loan information or transaction records of sample users may vary in value over different time periods, it is typically necessary to divide this type of data into different time windows (i.e., preset time periods) to calculate the numerical variables within each time window, thereby using these as derived data features.

[0067] For example, for user data with non-fixed values, if the user data is a numerical field, such as loan details, the sum, mean, quantile, minimum, maximum or standard deviation of the user data in each time window can be calculated; if the user data is a categorical field, such as loan type, the number of occurrences of the main type or the types of occurrences in each time window can be calculated. This application embodiment does not limit this.

[0068] To improve the accuracy of pre-trained models, S202, as one implementation method, can specifically include:

[0069] The data features derived from the user data in each data group are matched with the user tags of the sample users to determine the user tags corresponding to the data features. The data features and the user tags corresponding to the data features belonging to the same data group are used as training samples for a pre-trained model. The pre-trained model is trained to obtain multiple trained pre-trained models.

[0070] Specifically, the user data in each data group belongs to different sample users, and these sample users have their own corresponding user tags. Therefore, based on the correspondence between user data and user tags, the correspondence between data features derived from the user data and user tags can be determined. Then, the data features derived from the user data in each data group are matched with the user tags of the sample users to determine the user tags corresponding to different data features. Finally, the data features derived from the user data belonging to the same data group and the corresponding user tags are used as training samples for a pre-trained model. This process trains the pre-trained model, resulting in multiple pre-trained models corresponding to different data groups.

[0071] As an example, data features derived from user data in the payroll information dataset can be matched with user tags of sample users with payroll information to determine the user tags corresponding to the data features in the payroll information dataset. Then, the data features and corresponding user tags of the payroll information dataset can be used as training samples for a pre-trained model corresponding to the payroll information dataset, and the model can be trained to obtain a pre-trained model for the payroll information dataset.

[0072] To improve the accuracy of the pre-trained model, the pre-trained model is trained to obtain multiple trained pre-trained models. Specifically, this may include: inputting multiple data sets into multiple pre-trained models corresponding to the data sets, obtaining a second credit prediction result for each sample user by each pre-trained model; calculating the accuracy index and stability index of each pre-trained model based on the second credit prediction result; determining whether the accuracy index of each pre-trained model is greater than or equal to a first preset threshold, and whether the stability index of each pre-trained model is less than or equal to a second preset threshold; for pre-trained models whose accuracy index is less than the first preset threshold and / or whose stability index is greater than the second preset threshold, adjusting the model parameters of the pre-trained model, and re-acquiring training samples to train the adjusted pre-trained model until the accuracy index of the pre-trained model is greater than or equal to the first preset threshold and the stability index is less than or equal to the second preset threshold, thus obtaining multiple trained pre-trained models.

[0073] Optionally, during the training process of the pre-trained model, it is necessary to evaluate whether the model performance meets the standards. Typically, the evaluation of model performance needs to consider both accuracy and stability. Specifically, multiple data sets can be input into multiple pre-trained models corresponding to those data sets to obtain the second credit prediction result for each sample user by each pre-trained model. Then, based on the second credit prediction results of each pre-trained model, the accuracy and stability metrics of each pre-trained model can be calculated.

[0074] For example, the accuracy metric of the model can be the Kolmogorov-Smirnov (KS) evaluation metric, and the stability metric can be the Population Stability Index (PSI). The KS evaluation metric can be understood as the maximum difference between the true positive rate (TPR) and the false positive rate (FPR) in the second credit prediction results of the pre-trained model.

[0075] Determining whether the accuracy and stability of each pre-trained model meet the standards can be done by determining whether the accuracy index value of each pre-trained model is greater than or equal to a first preset threshold, and whether the stability index value of each pre-trained model is less than or equal to a second preset threshold. The first preset threshold can be set to 0.20, and the second preset threshold can be set to 0.10. This application embodiment does not limit this.

[0076] For substandard pre-trained models with an accuracy metric value less than 0.20 and / or a stability metric value greater than 0.10, adjust their model parameters and reacquire the corresponding type of user data as new training samples. Iterate and train the adjusted pre-trained models until the accuracy metric values ​​of the substandard pre-trained models are all greater than or equal to 0.20 and the stability metric values ​​are all less than or equal to 0.10. This results in multiple pre-trained models that meet the performance requirements.

[0077] Figure 3 This is a flowchart illustrating an application method for a user credit prediction model provided in an embodiment of this application. Figure 3 As shown, the method may include the following steps S301 to S304.

[0078] S301. Obtain user data for the user to be predicted.

[0079] The user data acquired for the users to be predicted includes general user data such as basic information, credit information, credit rating information, loan information and loan application amount information, as well as scenario-based user data corresponding to the user's most recent business scenario, such as any one of the following: transaction history, asset information, housing provident fund information and payroll information.

[0080] S302. Group the user data according to the data type to obtain multiple data groups.

[0081] The user data of the users to be predicted is grouped according to the data type, resulting in multiple data groups composed of different types of user data.

[0082] As an example of S302, if the user data of the user to be predicted includes basic information, credit information, credit rating information, loan information, loan amount information, and payroll information, then after grouping the user data, we get the basic information data group, the credit information data group, the credit rating information data group, the loan information data group, the loan amount information data group, and the payroll information data group.

[0083] S303. Input multiple data sets into multiple pre-trained models corresponding to the data sets respectively, and obtain the third credit prediction result of each pre-trained model for each user to be predicted.

[0084] Multiple data sets are input into multiple pre-trained models corresponding to the data sets to obtain the third credit prediction result of each pre-trained model for each user to be predicted.

[0085] As an example of S303, the basic information data group, credit information data group, credit rating information data group, loan information data group, loan application amount information data group, and salary payment information data group are respectively input into the corresponding pre-trained models to obtain the third credit prediction result of each user to be predicted by the six pre-trained models.

[0086] S304. Input multiple third-party credit prediction results for each user to be predicted into the user credit prediction model to obtain the final credit prediction result for the user to be predicted.

[0087] As an example of S304, the six third-party credit prediction results of each user to be predicted can be input into the trained user credit prediction model to obtain the final credit prediction result of the user to be predicted.

[0088] The application method of the user credit prediction model in this application embodiment is applied to the user credit prediction model provided in any of the above embodiments. After obtaining training samples, the multi-dimensional user data in the training samples is grouped according to data type, so that each data group corresponds to a pre-trained model. Thus, when multiple target pre-trained models are selected according to the target business type, and the multiple first credit prediction results of the multiple target pre-trained models for sample users are used as training samples to train the user credit prediction model, the accuracy and robustness of the trained user credit prediction model will be higher when multiple target pre-trained models combine multi-dimensional user data, resulting in rich data dimensions and high utilization. Therefore, when using multiple first credit prediction results from multiple target pre-trained models for model training, the accuracy of the model prediction result will also be higher.

[0089] Based on the training method for the user credit prediction model provided in the above embodiments, this application also provides specific implementation methods for the training apparatus of the user credit prediction model. Please refer to the following embodiments.

[0090] like Figure 4 As shown, the training device 400 for the user credit prediction model provided in this application embodiment includes the following modules:

[0091] The first acquisition module 401 is used to acquire training samples. The training samples include user data and user tags of sample users. The user data includes various types of data related to the target business type of the sample users.

[0092] The first grouping module 402 is used to group user data according to data type to obtain multiple data groups;

[0093] The first selection module 403 is used to select multiple target pre-trained models corresponding to the target business type from multiple trained pre-trained models, wherein one pre-trained model corresponds to a data group of one type;

[0094] The first input module 404 is used to input multiple target data groups corresponding to the target business type from multiple data groups into multiple target pre-trained models corresponding to the target data groups, and obtain the first credit prediction result of each target pre-trained model for each sample user;

[0095] The first training module 405 is used to train the user credit prediction model by using multiple first credit prediction results of each sample user and the user tags of the sample user as training samples, and obtain the trained user credit prediction model.

[0096] The training apparatus for the user credit prediction model in this application, after acquiring training samples, groups the multi-dimensional user data in the training samples according to data type, so that each data group corresponds to a pre-trained model. Thus, when multiple target pre-trained models are selected based on the target business type, and the multiple first credit prediction results of the multiple target pre-trained models for sample users are used as training samples to train the user credit prediction model, the accuracy and robustness of the trained user credit prediction model will be higher because the multiple target pre-trained models combine multi-dimensional user data, resulting in rich data dimensions and high utilization. Therefore, when using the multiple first credit prediction results of the multiple target pre-trained models for model training, the accuracy of the model prediction result will also be higher.

[0097] In some embodiments, in order to improve the accuracy of the pre-trained model, the training device 400 of the user credit prediction model may further include: a second training module, used to process multiple data groups separately to obtain data features derived from user data in multiple data groups; and to train the pre-trained model using the data features derived from user data in multiple data groups and the user tags of sample users as training samples of the pre-trained model to obtain a trained pre-trained model.

[0098] In some embodiments, in order to improve the accuracy of the pre-trained model, the second training module is specifically used to delete user data with duplicate values, abnormal values, and missing values ​​in the data set; for the deleted user data, if the value type of the user data is a fixed value, the value of the user data is used as a data feature derived from the user data; if the value type of the user data is a variable value, the numerical variable of the user data within a preset time period is calculated as a data feature derived from the user data.

[0099] In some embodiments, in order to improve the accuracy of the pre-trained model, the second training module is specifically used to match the data features derived from the user data in each data group with the user tags of the sample users to determine the user tags corresponding to the data features; and to use the data features belonging to the same data group and the user tags corresponding to the data features as training samples for the pre-trained model to train the pre-trained model and obtain multiple trained pre-trained models.

[0100] In some embodiments, to improve the accuracy of the pre-trained model, the second training module is specifically used to input multiple data sets into multiple pre-trained models corresponding to the data sets, to obtain a second credit prediction result for each sample user by each pre-trained model; calculate the accuracy index value and stability index value of each pre-trained model based on the second credit prediction result; determine whether the accuracy index value of each pre-trained model is greater than or equal to a first preset threshold, and whether the stability index value of each pre-trained model is less than or equal to a second preset threshold; for pre-trained models whose accuracy index value is less than the first preset threshold and / or whose stability index value is greater than the second preset threshold, adjust the model parameters of the pre-trained model and re-acquire training samples to train the adjusted pre-trained model until the accuracy index value of the pre-trained model is greater than or equal to the first preset threshold and the stability index value is less than or equal to the second preset threshold, thereby obtaining multiple trained pre-trained models.

[0101] In some embodiments, the sample user data related to the target business type includes various types of data such as the sample user's basic information, credit information, transaction information, asset information, and income information.

[0102] In some embodiments, the sample users are users who meet the preset filtering rules and have preset type labels.

[0103] Figure 4 Each module in the illustrated device has the ability to implement Figure 1 The functions of each step in the process and their corresponding technical effects are described briefly and will not be elaborated here.

[0104] Based on the application method of the user credit prediction model provided in the above embodiments, this application also provides specific implementation methods of the application apparatus for the user credit prediction model. Please refer to the following embodiments.

[0105] like Figure 5 As shown, the application device 500 for the user credit prediction model provided in this application embodiment includes the following modules:

[0106] The second acquisition module 501 is used to acquire user data of the user to be predicted.

[0107] The second grouping module 502 is used to group user data according to data type to obtain multiple data groups;

[0108] The second input module 503 is used to input multiple data sets into multiple pre-trained models corresponding to the data sets respectively, and obtain the third credit prediction result of each pre-trained model for each user to be predicted.

[0109] The third input module 504 is used to input multiple third-party credit prediction results for each user to be predicted into the user credit prediction model to obtain the final credit prediction result for the user. This demonstrates the device's unique advantages.

[0110] The application device for the user credit prediction model in this application embodiment is applied to the user credit prediction model provided in any of the above embodiments. After obtaining training samples, the multi-dimensional user data in the training samples is grouped according to data type, so that each data group corresponds to a pre-trained model. Thus, when multiple target pre-trained models are selected according to the target business type, and multiple first credit prediction results of the multiple target pre-trained models for sample users are used as training samples to train the user credit prediction model, the accuracy and robustness of the trained user credit prediction model will be higher when multiple target pre-trained models combine multi-dimensional user data, resulting in rich data dimensions and high utilization. Therefore, when using multiple first credit prediction results from multiple target pre-trained models for model training, the accuracy of the model prediction result will also be higher.

[0111] Figure 5 Each module in the illustrated device has the ability to implement Figure 3 The functions of each step in the process and their corresponding technical effects are described briefly and will not be elaborated here.

[0112] Based on the training method of the user credit prediction model provided in the above embodiments, this application also provides specific implementation methods of electronic devices. Please refer to the following embodiments.

[0113] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0114] The electronic device may include a processor 601 and a memory 602 storing computer program instructions.

[0115] Specifically, the processor 601 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0116] Memory 602 may include mass storage for data or instructions. For example, and not limitingly, memory 602 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. In one example, memory 602 may include removable or non-removable (or fixed) media, or memory 602 may be non-volatile solid-state memory. Memory 602 may be internal or external to the integrated gateway disaster recovery device.

[0117] In one example, memory 602 may be read-only memory (ROM). In one example, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0118] Memory 602 may include read-only memory (ROM), random access memory (RAM), disk storage media device, optical storage media device, flash memory device, electrical, optical, or other physical / tangible memory storage device. Therefore, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to one aspect of this application.

[0119] The processor 601 reads and executes computer program instructions stored in the memory 602 to achieve... Figure 1 The method / steps S101 to S105 in the illustrated embodiment achieve the following: Figure 1 The technical effects achieved by executing the methods / steps shown in the examples are not elaborated here for the sake of brevity.

[0120] In one example, the electronic device may also include a communication interface 603 and a bus 610. For example, Figure 6As shown, the processor 601, memory 602, and communication interface 603 are connected through bus 610 and complete communication with each other.

[0121] The communication interface 603 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0122] Bus 610 includes hardware, software, or both, that couples components of an electronic device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 610 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, this application contemplates any suitable bus or interconnect.

[0123] Furthermore, in conjunction with the training method of the user credit prediction model in the above embodiments, this application embodiment can provide a computer-readable storage medium for implementation. This computer-readable storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the user credit prediction model training methods in the above embodiments. Examples of computer-readable storage media include non-transitory computer-readable storage media, such as electronic circuits, semiconductor memory devices, ROM, random access memory, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, and hard disks.

[0124] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0125] The functional blocks shown in the above-described block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0126] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0127] The aspects of this application have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by dedicated hardware performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0128] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A training method for a user credit prediction model, characterized in that, The method comprises: obtaining training samples, the training samples comprising user data and user labels of sample users, the user data comprising multiple types of data related to a target business type of the sample users, the multiple types of data related to the target business type of the sample users comprising basic information, credit information, transaction information, asset information and income information of the sample users; grouping the user data according to data types to obtain multiple data groups; selecting multiple target pre-training models corresponding to the target business type from multiple trained pre-training models, wherein one pre-training model corresponds to one type of data group; inputting multiple target data groups corresponding to the target business type in multiple data groups into multiple target pre-training models corresponding to the target data groups respectively to obtain first credit prediction results of each sample user by each target pre-training model; training a user credit prediction model by taking multiple first credit prediction results of each sample user and user labels of the sample users as training samples of the user credit prediction model to obtain a trained user credit prediction model.

2. The method of claim 1, wherein, After the grouping of the user data according to data types to obtain multiple data groups, the method further comprises: performing data processing on multiple data groups respectively to obtain data features derived from the user data in multiple data groups; training the pre-training model by taking data features derived from the user data in multiple data groups and user labels of the sample users as training samples of the pre-training model to obtain a trained pre-training model.

3. The method of claim 2, wherein, The data processing on multiple data groups respectively to obtain data features derived from the user data in multiple data groups comprises: performing the following steps on each data group respectively: deleting the user data with repeated values, abnormal values and missing values in the data group; for the deleted part of the user data, in the case that the value type of the user data is a fixed value, taking the value of the user data as a data feature derived from the user data; in the case that the value type of the user data is an unfixed value, calculating the value variable of the user data in a preset time period as a data feature derived from the user data.

4. The method according to claim 2 or 3, characterized in that, The training of the pre-training model by taking data features derived from the user data in multiple data groups and user labels of the sample users as training samples of the pre-training model to obtain a trained pre-training model comprises: matching the data features derived from the user data in each data group with the user labels of the sample users to determine the user labels corresponding to the data features; taking the data features and the user labels corresponding to the data features belonging to the same data group as a training sample of one pre-training model to train the pre-training model to obtain multiple trained pre-training models.

5. The method of claim 4, wherein, The training of the pre-training models comprises: inputting the data groups into the pre-training models corresponding to the data groups respectively to obtain second credit prediction results of each pre-training model for each sample user; calculating the accuracy index value and the stability index value of each pre-training model according to the second credit prediction results; determining whether the accuracy index value of each pre-training model is greater than or equal to a first preset threshold value and whether the stability index value of each pre-training model is less than or equal to a second preset threshold value; adjusting the model parameters of the pre-training model and reacquiring training samples for model training of the pre-training model until the accuracy index value of the pre-training model is greater than or equal to the first preset threshold value and the stability index value of the pre-training model is less than or equal to the second preset threshold value, thereby obtaining the trained pre-training models.

6. The method of claim 1, wherein, The sample user is a user meeting a preset screening rule and having a preset type label.

7. A method of applying a user credit prediction model, characterized by, The method comprises: acquiring user data of a to-be-predicted user; grouping the user data according to data types to obtain data groups; inputting the data groups into pre-training models corresponding to the data groups respectively to obtain third credit prediction results of each pre-training model for each to-be-predicted user; inputting the third credit prediction results of each to-be-predicted user into the user credit prediction model to obtain a final credit prediction result of the to-be-predicted user; The user credit prediction model is the user credit prediction model trained according to any one of claims 1-5. 8.A device for training a user credit prediction model, comprising: The device comprises: a first acquisition module configured to acquire training samples, wherein the training samples comprise user data and user labels of sample users, the user data comprises multiple types of data related to a target business type of the sample users, and the multiple types of data related to the target business type of the sample users comprise basic information, credit information, transaction information, asset information and income information of the sample users; a first grouping module configured to group the user data according to data types to obtain data groups; a first selection module configured to select target pre-training models corresponding to the target business type from trained pre-training models, wherein one pre-training model corresponds to one type of data group; a first input module configured to input target data groups corresponding to the target business type in the data groups into target pre-training models corresponding to the target data groups respectively to obtain first credit prediction results of each target pre-training model for each sample user; The first training module is configured to train a user credit prediction model by taking the first credit prediction results of each sample user and the user label of the sample user as training samples of the user credit prediction model, and obtain a trained user credit prediction model.

9. An application device for a user credit prediction model, characterized in that, The device comprises: The second obtaining module is configured to obtain user data of a to-be-predicted user. The second grouping module is configured to group the user data according to data types, and obtain a plurality of data groups. The second input module is configured to input the plurality of data groups into a plurality of pre-trained models corresponding to the data groups respectively, and obtain third credit prediction results of each to-be-predicted user by each pre-trained model. The third input module is configured to input the plurality of third credit prediction results of each to-be-predicted user into the user credit prediction model, and obtain a final credit prediction result of the to-be-predicted user. The user credit prediction model is the user credit prediction model trained by any one of claims 1-5.

10. An electronic device, comprising: The electronic device comprises a processor, a memory, and a computer program stored on the memory and executable on the processor, wherein the computer program is executed by the processor to implement the steps of the user credit prediction model training method according to any one of claims 1 to 6, and the steps of the user credit prediction model application method according to claim 7.

11. A computer readable storage medium, characterized in that, The computer program is stored on the computer readable storage medium, and when executed by the processor, the computer program implements the steps of the user credit prediction model training method according to any one of claims 1 to 6, and the steps of the user credit prediction model application method according to claim 7.

12. A computer program product, characterised in that, The instructions in the computer program product are executed by the processor of the electronic device, so that the electronic device performs the steps of the user credit prediction model training method according to any one of claims 1 to 6, and the steps of the user credit prediction model application method according to claim 7.

Images