Training method and device of limit adjustment model, computer equipment and storage medium

Abstract

The invention relates to the technical field of data processing, and discloses a training method and device of a limit adjustment model, computer equipment and a storage medium. According to the method, a quota adjustment test request is obtained, and test element parameters, a test object account and a test quota adjustment rule are determined according to the quota adjustment test request; determining to-be-tested users and user feature information of the to-be-tested users according to the test object account; obtaining quota adjustment information and test index values of the to-be-tested users according to the test quota adjustment rules and the test element parameters; and generating a sample data set according to the user feature information of all the to-be-tested users, the quota adjustment information and the test index values, performing iterative training on the initial quota adjustment model according to the sample data set, and obtaining the quota adjustment model when a preset iteration stop condition is met. According to the method, the sample data acquisition efficiency is improved, meanwhile, the application universality and effectiveness of the limit adjustment model are improved, and the reasonability of model recommendation is ensured.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a training method, apparatus, computer equipment, and storage medium for a credit limit adjustment model. Background Technology

[0002] In the financial industry, adjusting user credit limits is a crucial task for financial institutions. By adjusting users' available credit or funds, institutions can balance risk and return, enhancing user loyalty. Currently, user credit limit management relies primarily on manual experience and rules, which fails to accurately identify user needs and scientifically assess the impact of different credit limit management strategies on risk and return. This easily leads to unreasonable credit limit allocation and adjustments, causing incalculable bad debt losses. To achieve reasonable credit limit allocation and adjustment, neural network-trained credit limit adjustment models can be used. However, the effectiveness of these models depends heavily on the quality of the training data. Most training data uses historical data, which is not only time-consuming and labor-intensive to collect but also lacks comprehensive coverage of application scenarios, thus limiting the effectiveness of the credit limit adjustment model. Therefore, a scientifically sound and reasonable training method for credit limit adjustment models is urgently needed. Summary of the Invention

[0003] Therefore, it is necessary to provide a training method, apparatus, computer equipment, and storage medium for a credit limit adjustment model to address the aforementioned technical problems, thereby solving the issues of poor training performance and limited applicability of existing credit limit adjustment models.

[0004] A training method for a credit limit adjustment model includes: Obtain a credit limit adjustment test request, and determine the test element parameters, test target account, and test credit limit adjustment rules based on the credit limit adjustment test request; Based on the test object account, determine the test users and the user characteristic information of each test user; The user feature information is grouped according to the test quota adjustment rules, and all the users to be tested are divided into at least two user groups, and the group quota adjustment rules for each user group are determined. According to the group quota adjustment rules and the test element parameters, the quota adjustment test is carried out on the users to be tested in each user group to obtain the quota adjustment information and test index value of each user to be tested. A sample dataset is generated based on the user characteristic information, credit limit adjustment information, and test indicator values ​​of all the users to be tested. The initial credit limit adjustment model is then iteratively trained based on the sample dataset, and the credit limit adjustment model is obtained when the preset iteration stopping condition is met.

[0005] A method for adjusting credit limits, comprising: Obtain a credit limit adjustment request, and determine the target feature information of the user to be analyzed based on the credit limit adjustment request; The target feature information is analyzed and processed according to the pre-trained quota adjustment model to obtain the initial adjustment quota corresponding to the operating indicators. The quota adjustment model is trained by the training method of the quota adjustment model described above. Based on the initial adjustment limit, a recommended risk indicator value is determined. When it is confirmed that the recommended risk indicator value is lower than a preset risk threshold, the limit adjustment recommendation information is determined based on the initial adjustment limit.

[0006] A training device for a credit limit adjustment model, comprising: The test request acquisition module is used to acquire the quota adjustment test request and determine the test element parameters, test object account and test quota adjustment rules based on the quota adjustment test request. The test user determination module is used to determine the test users and the user characteristic information of each test user based on the test object account; The user group segmentation module is used to group the user feature information according to the test quota adjustment rules, divide all the users to be tested into at least two user groups, and determine the group quota adjustment rules for each user group. The credit limit adjustment test module is used to perform credit limit adjustment tests on the users to be tested in each user group according to the group credit limit adjustment rules and the test element parameters, and obtain the credit limit adjustment information and test indicator values ​​of each user to be tested. The model training module is used to generate a sample dataset based on the user feature information, credit limit adjustment information and test index values ​​of all the users to be tested, and to iteratively train the initial credit limit adjustment model based on the sample dataset, and obtain the credit limit adjustment model when the preset iteration stopping condition is reached.

[0007] A computer device includes a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, wherein the processor, when executing the computer-readable instructions, implements the training method of the above-described credit limit adjustment model or the above-described credit limit adjustment method.

[0008] A computer-readable storage medium storing computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform a training method for the above-described quota adjustment model or the above-described quota adjustment method.

[0009] In the training method, apparatus, computer equipment, and storage medium of the aforementioned credit limit adjustment model, the training method involves: acquiring a credit limit adjustment test request; determining test element parameters, test object accounts, and test credit limit adjustment rules based on the test request; determining the test users and their user characteristic information based on the test object accounts; grouping the user characteristic information according to the test credit limit adjustment rules, dividing all test users into at least two user groups, and determining the group credit limit adjustment rules for each user group; conducting credit limit adjustment tests on the test users in each user group according to the group credit limit adjustment rules and test element parameters, obtaining the credit limit adjustment information and test index values ​​for each test user; generating a sample dataset based on the user characteristic information, credit limit adjustment information, and test index values ​​of all test users; and iteratively training the initial credit limit adjustment model based on the sample dataset, obtaining the credit limit adjustment model when a preset iteration stopping condition is reached. This invention determines the test element parameters, test target accounts, and test credit limit adjustment rules based on the credit limit adjustment test request, establishing a credit limit adjustment stress test mechanism for the users to be tested. It can evaluate the impact of credit limit adjustment information on different indicators under different simulation scenarios as needed, providing diverse sample data for the training of the credit limit adjustment model. This invention avoids the hassle of collecting historical data, improves the efficiency of sample data acquisition, and enhances the applicability and effectiveness of the credit limit adjustment model, ensuring the rationality of model recommendations. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a flowchart illustrating a training method for a quota adjustment model in one embodiment of the present invention; Figure 2 This is a flowchart illustrating step S10 in the training method of the quota adjustment model in one embodiment of the present invention; Figure 3 This is a flowchart illustrating step S30 in the training method of the quota adjustment model in one embodiment of the present invention; Figure 4 This is a flowchart illustrating step S40 in the training method of the quota adjustment model in one embodiment of the present invention; Figure 5 This is a flowchart illustrating step S403 in the training method of the quota adjustment model in one embodiment of the present invention; Figure 6This is a flowchart illustrating step S50 in the training method of the quota adjustment model in one embodiment of the present invention; Figure 7 This is a schematic diagram of the fitting of an operating indicator for the training method of the quota adjustment model in one embodiment of the present invention; Figure 8 This is a schematic diagram of risk index fitting for the training method of the quota adjustment model in one embodiment of the present invention; Figure 9 This is a flowchart illustrating a method for adjusting credit limits according to an embodiment of the present invention; Figure 10 This is a schematic diagram of the structure of a training device for a quota adjustment model in one embodiment of the present invention; Figure 11 This is a schematic diagram of a computer device according to an embodiment of the present invention. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] The training method for the credit limit adjustment model provided in this embodiment can be applied to application scenarios where the server generates a sample dataset based on the credit limit adjustment test request sent by the operation and management personnel through the client, and iteratively trains the model based on the sample dataset. The client communicates with the server. The client includes, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0014] In one embodiment, the credit limit adjustment model is used to, upon receiving a credit limit adjustment request, determine the target feature information of the user to be analyzed based on the credit limit adjustment request, and analyze and process the target feature information according to the credit limit adjustment model to obtain and output credit limit adjustment recommendation information for the user to be analyzed, such as... Figure 1 As shown, a training method for a credit limit adjustment model is provided, including the following steps S10-S50.

[0015] S10. Obtain a credit limit adjustment test request, and determine the test element parameters, test target account, and test credit limit adjustment rules based on the credit limit adjustment test request.

[0016] In essence, a credit limit adjustment request refers to information used to request the server to adjust the financial service credit limit of a specified user. This request can be proactively generated by the application user based on their credit limit adjustment needs, or it can be automatically generated by the server according to a pre-set adjustment cycle. After receiving the credit limit adjustment request, the server identifies the user to be analyzed based on the request and obtains the target feature information of that user. The user to be analyzed refers to the application user specified in the credit limit adjustment request who requires a credit limit adjustment. Target feature information refers to user characteristic information used to characterize the user to be analyzed; user characteristic information refers to attribute characteristic information that distinguishes different users from other users.

[0017] A credit limit adjustment model is a pre-trained neural network model that recommends different credit limit adjustments based on the characteristics of different users. The server analyzes and processes the target feature information using the credit limit adjustment model to obtain credit limit adjustment recommendations for the user being analyzed, thus enabling the credit limit adjustment operation for that user to be performed based on these recommendations. The credit limit adjustment recommendation information refers to the recommended values ​​for adjusting the credit limit of the user being analyzed, output by the credit limit adjustment model. The credit limit adjustment model is trained using a large amount of credit limit adjustment test data as samples.

[0018] Financial service quota management personnel can create quota adjustment test requests by inputting configuration information through the client as needed. The client then sends these requests to the server. A quota adjustment test request is an information request to the server to divide users into different user groups based on different user characteristics for experimental quota adjustments. The server parses the received quota adjustment test request to obtain test element parameters, test target accounts, and test quota adjustment rules. Test element parameters refer to the statistical result indicators and statistical time ranges used to characterize the experimental quota adjustment. Test target accounts refer to the list of all user accounts designated to participate in the experimental quota adjustment. Test quota adjustment rules refer to the user group division rules and quota adjustment allocation rules during the experimental quota adjustment.

[0019] S20. Determine the users to be tested and the user characteristic information of each user to be tested based on the test object account.

[0020] Understandably, the server identifies the test users from all client users based on the test subject's account. Based on each test user's account, the server can obtain their user characteristic information. Test users refer to those participating in the experimental credit limit adjustment. User characteristic information refers to the attribute characteristics that distinguish different users from others. Each user has unique user characteristic information, such as age, gender, spending habits, credit score, and historical transaction records.

[0021] S30. The user feature information is grouped according to the test quota adjustment rules, and all the users to be tested are divided into at least two user groups, and the group quota adjustment rules for each user group are determined.

[0022] Understandably, the test credit limit adjustment rules include rules for grouping users based on user characteristic information. Different test users may share some or all of their user characteristic information. Therefore, the server processes the user characteristic information into groups according to the test credit limit adjustment rules, dividing all test users into at least two user groups. A user group refers to a set of users divided according to user characteristic information during experimental credit limit adjustments, that is, a set of users with at least one common user characteristic information. Test users within the same user group share common user characteristic information, while test users in different user groups share different common user characteristic information. In addition, each user group corresponds to its own group credit limit adjustment rules, which refer to the rules for allocating credit limit adjustments among users within a user group during experimental credit limit adjustments.

[0023] S40. Based on the group limit adjustment rules and the test element parameters, perform limit adjustment tests on the users to be tested in each user group to obtain the limit adjustment information and test index values ​​of each user to be tested.

[0024] Understandably, each user group corresponds to its own group credit limit adjustment rules, but all user groups share common test element parameters. That is, the statistical result indicators and statistical time ranges for all test users are the same during the experimental credit limit adjustment process, facilitating unified data analysis and processing later. The server performs credit limit adjustment tests on the test users within each user group according to their respective group credit limit adjustment rules. Simultaneously, it collects the credit limit adjustment test results for each test user based on the test element parameters, obtaining the credit limit adjustment information and test indicator values ​​for each test user. Credit limit adjustment information refers to the specific values ​​conforming to the group credit limit adjustment rules used when adjusting the experimental credit limit for the test users. Test indicator values ​​refer to the quantified data of the statistical results during the experimental credit limit adjustment period for the test users.

[0025] S50. Generate a sample dataset based on the user characteristic information, credit limit adjustment information and test index values ​​of all the users to be tested, and iteratively train the initial credit limit adjustment model based on the sample dataset. When the preset iteration stopping condition is reached, the credit limit adjustment model is obtained.

[0026] Understandably, each user to be tested has their own corresponding user characteristic information, credit limit adjustment information, and test indicator values. A sample dataset can be obtained based on the user characteristic information, credit limit adjustment information, and test indicator values ​​of all users to be tested. The sample dataset is a collection of data generated after experimental credit limit adjustments to the users to be tested, used to train the credit limit adjustment model. The server trains the initial credit limit adjustment model based on the sample dataset to obtain the credit limit adjustment model. The initial credit limit adjustment model refers to a pre-built neural network model used to learn how to recommend different credit limit adjustments based on the characteristic information of different users; that is, the initial credit limit adjustment model is a credit limit adjustment model that has not yet been trained. The preset iteration stopping condition is a pre-set judgment rule or threshold used to stop the iterative training of the model, such as a judgment rule based on the convergence of the loss function, or an iteration number threshold based on reaching the maximum number of iterations.

[0027] In one specific embodiment, firstly, a sample dataset is acquired, an initial credit limit adjustment model is constructed, and a loss function is determined. Then, the model is iteratively trained using binary cross-entropy as the loss function, calculating the loss function and backpropagating to update the model parameters until the loss function converges or the maximum number of iterations is reached, at which point iteration stops, ultimately yielding the trained credit limit adjustment model.

[0028] This embodiment obtains a credit limit adjustment test request, determines the test element parameters, test object accounts, and test credit limit adjustment rules based on the request; identifies the test users and their user characteristic information based on the test object accounts; groups the user characteristic information according to the test credit limit adjustment rules, dividing all test users into at least two user groups, and determines the group credit limit adjustment rules for each user group; performs credit limit adjustment tests on the test users in each user group according to the group credit limit adjustment rules and test element parameters, obtaining the credit limit adjustment information and test indicator values ​​for each test user; generates a sample dataset based on the user characteristic information, credit limit adjustment information, and test indicator values ​​of all test users, and iteratively trains the initial credit limit adjustment model based on the sample dataset, obtaining the credit limit adjustment model when a preset iteration stopping condition is reached. This embodiment establishes a credit limit adjustment stress test mechanism for test users by determining the test element parameters, test object accounts, and test credit limit adjustment rules based on the credit limit adjustment test request. It can evaluate the impact of credit limit adjustment information on different indicators under different simulation scenarios as needed, providing diverse sample data for the training of the credit limit adjustment model. This embodiment avoids the hassle of collecting historical data, improves the efficiency of sample data acquisition, and enhances the applicability and effectiveness of the credit limit adjustment model, ensuring the rationality of the model's recommendations.

[0029] In one embodiment, such as Figure 2As shown, step S10, that is, before obtaining the quota adjustment test request, includes: S101. Obtain the test element configuration information, test account selection information, and at least two sets of group feature configuration information and group rule configuration information entered by the client. S102. Determine the test element parameters according to the test element configuration information, determine the test object account according to the test account selection information, and determine the test quota adjustment rule according to all the group feature configuration information and group rule configuration information; S103. Generate a quota adjustment test request based on the test element parameters, the test object account, and the test quota adjustment rules.

[0030] Understandably, quota operation and management personnel can input a series of configuration information, including test element configuration information, test account selection information, and at least two sets of group characteristic configuration information and group rule configuration information, through the quota adjustment test configuration page on the client side as needed. The server then creates a quota adjustment test request based on the configuration information. First, the server obtains the test element configuration information, test account selection information, and at least two sets of group characteristic configuration information and group rule configuration information entered by the client. The test element configuration information specifies the statistical result indicators and statistical time range for the experimental quota adjustment. The test account selection information specifies the account selection list for all users participating in the experimental quota adjustment. The configuration information includes at least two sets of group characteristic configuration information and group rule configuration information; each set of group characteristic configuration information and group rule configuration information can determine a user group during the experimental quota adjustment. The group characteristic configuration information specifies the characteristic configuration information that test users divided into the same user group need to possess during the experimental quota adjustment. The group rule configuration information specifies the quota adjustment allocation rules for each specific test user within the user group during the experimental quota adjustment. Next, the server determines the test element parameters based on the test element configuration information, determines the test target account based on the test account selection information, and determines the test credit limit adjustment rules based on all group feature configuration information and group rule configuration information. Finally, the server generates a credit limit adjustment test request based on the test element parameters, the test target account, and the test credit limit adjustment rules.

[0031] This embodiment obtains the necessary test element parameters, test object accounts, and test quota adjustment rules in the test request based on a series of configuration information, and realizes the automatic generation of quota adjustment test requests. This facilitates the subsequent division of users with different characteristics into different user groups for different quota allocations and adjustments, thereby improving the targeting and effectiveness of quota adjustment tests.

[0032] In one embodiment, the test quota adjustment rule includes at least two group characteristic conditions and a group quota adjustment rule associated with each of the group characteristic conditions; such as Figure 3 As shown, in step S30, namely, grouping the user feature information according to the test quota adjustment rules, dividing all the users to be tested into at least two user groups, and determining the group quota adjustment rules for each user group, includes: S301. Compare the user feature information of each user to be tested with all the group feature conditions to obtain the condition comparison results of each user to be tested. S302. Find the characteristic consistency result from all condition comparison results of each user to be tested, and classify the user to be tested into a user group with the group characteristic condition corresponding to the characteristic consistency result, and determine the group quota adjustment rule associated with the group characteristic condition as the group quota adjustment rule of the user group, until all users to be tested are classified.

[0033] Understandably, the test credit limit adjustment rules include at least two group characteristic conditions and group credit limit adjustment rules associated with each of these group characteristic conditions. Each user group involved in an experimental credit limit adjustment corresponds to a set of interrelated group characteristic conditions and group credit limit adjustment rules. Group characteristic conditions refer to the combination of characteristic information that test users divided into the same user group must possess during an experimental credit limit adjustment. Group credit limit adjustment rules specify the rules for allocating credit limit adjustments to each specific test user within the user group during an experimental credit limit adjustment.

[0034] In one embodiment, the test quota adjustment rule includes group characteristic conditions for three user groups. The first group characteristic condition, "premium user, medium-high demand, R2-new, unstable," corresponds to the first user group. The second group characteristic condition, "ordinary user, medium-high demand, R2-new, unstable," corresponds to the second user group. The third group characteristic condition, "ordinary user, medium-low demand, R1-new, stable," corresponds to the third user group. The group quota adjustment rule refers to the quota adjustment allocation rule for each user to be tested within the user group during experimental quota adjustments. The server compares the user characteristic information of each user to be tested with the three group characteristic conditions to obtain the condition comparison results for each user. The condition comparison results are used to characterize whether the user characteristic information of the user to be tested is consistent with the characteristic conditions of each group, including consistent and inconsistent results. A consistent result indicates that the user characteristic information of the user to be tested includes all the same attribute features specified in the group characteristic conditions. In this case, the condition comparison result for each user to be tested includes the results corresponding to the three group characteristic conditions, with one consistent result and two inconsistent results. Therefore, the server searches for consistent characteristics among all the condition comparison results for each user under test, and assigns the user to the user group corresponding to the group characteristic conditions of the consistent characteristic result. It also determines the group quota adjustment rule associated with that group characteristic condition as the group quota adjustment rule for the user group. For example, if the consistent characteristic result among the three condition comparison results for a user under test is the first group characteristic condition "Excellent User, Medium-High Demand, R2-new, Unstable," while the other two group characteristic conditions show inconsistent results, then the user under test is assigned to the first user group. The server iterates through all users under test until all users are assigned.

[0035] This embodiment describes a method for grouping user characteristic information based on test credit limit adjustment rules. This method can reasonably divide the users to be tested into different user groups and determine corresponding group credit limit adjustment rules for each user group. This facilitates more effective identification of potential risks through experimental credit limit adjustments for different groups and helps to take more reasonable credit limit adjustment controls.

[0036] In one embodiment, such as Figure 4 As shown, in step S40, which involves conducting credit limit adjustment tests on the users to be tested in each user group according to the group credit limit adjustment rules and the test element parameters, and obtaining the credit limit adjustment information and test indicator values ​​for each user to be tested, the following steps are included: S401. Determine the credit limit adjustment information for each user to be tested in each user group according to the group credit limit adjustment rules; S402. Determine the test indicators and test cycle based on the test element parameters; S403. Perform quota adjustment tests on each of the users to be tested according to the quota adjustment information, and obtain the test indicator value of each user to be tested based on all test result data corresponding to the test indicator within the test period.

[0037] Understandably, during the experimental credit limit adjustment process for users to be tested, the server first determines the credit limit adjustment information for each user in each user group according to the group credit limit adjustment rules. For example, within the same user group, 50% of the users to be tested receive a credit limit increase of 1000, and 50% receive a credit limit increase of 2000. Next, the server determines the test indicators and test period based on the test element parameters. The test indicators are quantitative metrics used to evaluate the effectiveness of the experimental credit limit adjustment, and the test period is the duration of the experimental credit limit adjustment. Finally, the server conducts credit limit adjustment tests on each user to be tested based on the credit limit adjustment information and statistically analyzes all test result data corresponding to the test indicators within the test period to obtain the test indicator value for each user to be tested. The test result data refers to the actual indicator results of the users to be tested during the experimental credit limit adjustment period. The test indicator value can be the average value of the test result data corresponding to the test indicator within the test period. For example, when the test indicator is an operational indicator and the test period is one month, the test result data refers to the actual operational result data of the user under test on each day during the experimental credit limit adjustment period, and the test indicator value refers to the average daily statistical operational result data of the user under test during the one-month period of the experimental credit limit adjustment.

[0038] This embodiment uses group-based credit limit adjustment rules to personalize credit limit adjustments based on the characteristics of different user groups. At the same time, by determining the test indicators and periods through test element parameters, a large amount of user behavior data can be collected during the experimental credit limit adjustment process. This helps to more accurately evaluate the user behavior patterns of each user group, optimize the credit limit adjustment strategy, and formulate a more precise credit limit adjustment strategy.

[0039] In one embodiment, such as Figure 5 As shown, step S403, namely, performing credit limit adjustment tests on each of the users to be tested based on the credit limit adjustment information, includes: S4031. Determine the credit limit type, credit limit adjustment method, and control experiment allocation information for each user group according to the group credit limit adjustment rules; S4032. Based on the control experiment allocation information, the users to be tested in each user group are divided into experimental users and control users, and the experimental quota adjustment information and the control quota adjustment information of each experimental user and the control quota adjustment information of each control user are determined according to the quota type, quota adjustment method and control experiment allocation information. S4033. Based on the experimental credit limit adjustment information, conduct credit limit adjustment tests on the experimental user to obtain the test result data of the experimental user; and based on the control credit limit adjustment information, conduct credit limit adjustment tests on the control user to obtain the test result data of the control user.

[0040] Understandably, when determining the credit limit adjustment information for each test user based on the group credit limit adjustment rules, for each user group, all test users within the group need to be divided into an experimental group and a control group, and different credit limit adjustment strategies should be applied to the experimental group and the control group. The server parses the group credit limit adjustment rules for each user group to determine the credit limit type, credit limit adjustment method, and control experiment allocation information. The credit limit type, credit limit adjustment method, and control experiment allocation information may differ between different user groups. The credit limit type refers to the type used to distinguish different credit limit validity periods, including temporary credit limits and fixed credit limits. The credit limit adjustment method refers to the form used to distinguish different credit limit changes, including fixed amount methods and percentage amount methods. The control experiment allocation information refers to the proportion of test users allocated to the experimental group and the control group within the user group, and the credit limit adjustment value for each group.

[0041] The server randomly assigns test users in each user group to experimental and control groups based on the control experiment allocation information. Experimental users are those actually assigned to the experimental group, and control users are those actually assigned to the control group. For example, when the ratio of experimental to control groups in a user group is 50% each, then 50% of the test users are experimental users, and 50% are control users. Next, the server determines the experimental credit limit adjustment information for each experimental user and the control credit limit adjustment information for each control user based on the credit limit type, credit limit adjustment method, and control experiment allocation information. The experimental credit limit adjustment information refers to the specific values ​​used by experimental users in accordance with the credit limit type and adjustment method when making experimental credit limit adjustments. The control credit limit adjustment information refers to the specific values ​​used by control users in accordance with the credit limit type and adjustment method when making experimental credit limit adjustments. Finally, the server conducts credit limit adjustment tests on experimental users based on the experimental credit limit adjustment information to obtain the test results data for experimental users. Simultaneously, it conducts credit limit adjustment tests on control users based on the control credit limit adjustment information to obtain the test results data for control users.

[0042] In one embodiment, all users to be tested are divided into two user groups: a first user group and a second user group. The first user group has a temporary credit limit with a fixed adjustment method. In the control experiment allocation information, the experimental group has 50% of users and a credit limit adjustment of 100 (a 100% increase), while the control group has 50% of users and a credit limit adjustment of 110 (a 110% increase). The second user group has a fixed credit limit with a percentage adjustment method. In the control experiment allocation information, the experimental group has 50% of users and a credit limit adjustment of 10% (a 10% increase), while the control group has 50% of users and a credit limit adjustment of 20% (a 20% increase).

[0043] This embodiment divides all test users within each user group into an experimental group and a control group, and applies different credit limit adjustment strategies to the experimental and control groups to evaluate the impact of experimental credit limit adjustments on user behavior. This embodiment provides data support and decision-making basis for subsequent credit limit adjustment strategy optimization, ensuring the scientific nature and effectiveness of the credit limit adjustment test, and helping to formulate more reasonable credit limit adjustment strategies.

[0044] In one embodiment, the test indicator values ​​include operational indicator test values ​​corresponding to operating indicators and risk indicator test values ​​corresponding to risk indicators; such as Figure 6 As shown, step S50, namely generating a sample dataset based on the user characteristic information, quota adjustment information, and test indicator values ​​of each user to be tested in all the user groups, includes: S501. Fit the test values ​​of the operating indicators of all users to be tested in each user group to obtain the fitted values ​​of the operating indicators of each user to be tested. S502. Fit the risk index test values ​​of all users to be tested in each user group to obtain the risk index fitted values ​​of each user to be tested. S503. Generate a sample data set based on the user characteristic information, credit limit adjustment information, operational indicator fitting value and risk indicator fitting value of each user to be tested. S504. Obtain the sample dataset based on the set of all the said sample data groups.

[0045] Understandably, test indicator values ​​include test values ​​for operating indicators (Gross Merchandise Volume, GMV) and test values ​​for risk indicators (Risk). Operating indicator fit values ​​refer to the fit values ​​of indicators associated with user characteristic information and credit limit adjustment information, used to characterize the degree of operational performance. Risk indicator fit values ​​refer to the fit values ​​of indicators associated with user characteristic information and credit limit adjustment information, used to characterize the degree of risk. Each test user corresponds to a sample data set generated from user characteristic information, credit limit adjustment information, operating indicator fit values, and risk indicator fit values. The sample data set is a structured data combination used to characterize the relationship between user characteristic information, credit limit adjustment information, operating indicator fit values, and risk indicator fit values; that is, what user characteristics a test user has, what credit limit adjustment is applied, and what indicator changes will occur accordingly. The collection of sample data sets from all test users constitutes the sample dataset.

[0046] In one specific embodiment, such as Figure 7 As shown, the vertical axis represents the Gross Merchandise Volume (GMV), the horizontal axis represents the credit limit increase amount, and `fit` represents the fitted curve. By fitting the GMV test values ​​of all test users in each user group, the fitted GMV value for each test user can be obtained. Specifically, when the credit limit increase value of a test user is within the range of 0 to 1000, the fitted GMV value is the intersection of the fitted curve and the line with an horizontal axis of 1000 parallel to the vertical axis. Figure 8 As shown, the vertical axis represents the risk indicator (Risk), and the horizontal axis represents the credit limit increase amount (Credit Limit Increase Amount). "fit" represents the fitted curve, "bad" represents the risk index, and "good" represents the normal index. The sum of the risk index and the normal index is 1. By fitting the risk indicator test values ​​of all test users in each user group, the fitted risk indicator value for each test user can be obtained. Specifically, when the credit limit increase value of a test user is within the range of 0 to 1000, the fitted risk indicator value is the risk index of the bar chart with the horizontal axis at 1000. Combined with... Figure 7 and Figure 8 It can be seen that as the credit limit increases, the operating indicators first increase and then decrease, while the risk indicators do indeed become increasingly larger.

[0047] This embodiment reduces interference from data fluctuations or outliers through fitting, resulting in more stable and representative operating and risk indicators. This improves data quality and provides a data foundation for subsequent data analysis, model training, and strategy evaluation. It also helps to understand user behavior more accurately and optimize credit limit adjustment strategies.

[0048] In one embodiment, such as Figure 9 As shown, a method for adjusting credit limits is provided, including the following steps S60-S80: S60. Obtain a credit limit adjustment request, and determine the target feature information of the user to be analyzed based on the credit limit adjustment request; S70. The target feature information is analyzed and processed according to the pre-trained quota adjustment model to obtain the initial adjustment quota corresponding to the operating indicators. The quota adjustment model is trained by the training method of the quota adjustment model described above. S80. Determine a recommended risk indicator value based on the initial adjustment limit. When it is confirmed that the recommended risk indicator value is lower than a preset risk threshold, determine the limit adjustment recommendation information based on the initial adjustment limit.

[0049] Understandably, during the server-side analysis and processing of target feature information using the credit limit adjustment model, it needs to comprehensively consider operational and risk indicators to obtain recommended credit limit adjustments for the user being analyzed. The server first analyzes and processes the target feature information using the credit limit adjustment model to obtain an initial adjustment credit limit corresponding to the operational indicators. The initial adjustment credit limit refers to the credit limit adjustment value predicted by the credit limit adjustment model based on the target feature information, which maximizes the operational indicators. Then, based on the initial adjustment credit limit, a recommended risk indicator value is determined, and it is judged whether the recommended risk indicator value is lower than a preset risk threshold. The recommended risk indicator value refers to the risk indicator value corresponding to the initial adjustment credit limit and the target feature information. The preset risk threshold is a pre-set maximum critical value for the risk indicator used to determine whether the recommended risk indicator value meets the risk requirements. When the recommended risk indicator value is lower than the preset risk threshold, the credit limit adjustment recommendation information is determined based on the initial adjustment credit limit. When the recommended risk indicator value is higher than or equal to the preset risk threshold, the initial adjustment credit limit needs to be adjusted, and the adjusted operational and risk indicators are evaluated until the final credit limit adjustment recommendation information is determined.

[0050] The server receives credit limit adjustment requests, determines the target feature information of the user to be analyzed based on the requests, and analyzes and processes the target feature information according to the credit limit adjustment model to obtain credit limit adjustment recommendation information for the user to be analyzed. This enables reasonable credit limit adjustments for users with different characteristics, such as low-limit trial for new users, credit limit increase incentives for active users, and customized credit limit adjustments for high-value users.

[0051] This embodiment optimizes credit limit adjustment information based on a credit limit adjustment model, which can improve the rationality of credit limit allocation and adjustment, meet the needs of different users, and effectively achieve a balance between risk control and improved operational efficiency. By using the credit limit adjustment model for forecasting and comprehensively considering both operational and risk indicators, a more comprehensive assessment of users' operational capabilities and risk levels can be made, leading to more scientific and objective decision-making and improving the rationality of credit limit adjustments.

[0052] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0053] In one embodiment, a training device for a credit limit adjustment model is provided, which corresponds one-to-one with the training method for the credit limit adjustment model in the above embodiments. For example... Figure 10 As shown, the training device for the credit limit adjustment model includes a test request acquisition module 10, a test user determination module 20, a user group segmentation module 30, a credit limit adjustment testing module 40, and a model training module 50. Detailed descriptions of each functional module are as follows: The test request acquisition module 10 is used to acquire a quota adjustment test request and determine the test element parameters, the test object account, and the test quota adjustment rules based on the quota adjustment test request. The test user determination module 20 is used to determine the test users and the user characteristic information of each test user based on the test object account; User group segmentation module 30 is used to group the user feature information according to the test quota adjustment rules, divide all the users to be tested into at least two user groups, and determine the group quota adjustment rules for each user group. The quota adjustment test module 40 is used to perform quota adjustment tests on the users to be tested in each user group according to the group quota adjustment rules and the test element parameters, and obtain the quota adjustment information and test index values ​​of each user to be tested. The model training module 50 is used to generate a sample dataset based on the user feature information, credit limit adjustment information and test index values ​​of all the users to be tested, and to train the initial credit limit adjustment model based on the sample dataset to obtain the credit limit adjustment model.

[0054] In one embodiment, the test request acquisition module 10 includes: The configuration information acquisition unit is used to acquire the test element configuration information, test account selection information, and at least two sets of group feature configuration information and group rule configuration information entered by the client. The test content determination unit is used to determine test element parameters based on the test element configuration information, determine test object accounts based on the test account selection information, and determine the test quota adjustment rules based on all the group feature configuration information and group rule configuration information. The test request generation unit is used to generate a quota adjustment test request based on the test element parameters, the test object account, and the test quota adjustment rules.

[0055] In one embodiment, the user group segmentation module 30 includes: The user feature comparison unit is used to compare the user feature information of each user to be tested with all the group feature conditions to obtain the condition comparison results of each user to be tested. The user group determination unit is used to find characteristic consistency results from all condition comparison results of each user to be tested, and to classify the user to be tested into a user group with the group characteristic conditions corresponding to the characteristic consistency results, and to determine the group quota adjustment rule associated with the group characteristic conditions as the group quota adjustment rule of the user group, until all the users to be tested are classified.

[0056] In one embodiment, the quota adjustment test module 40 includes: The quota adjustment information determination unit is used to determine the quota adjustment information of each user to be tested in each user group according to the group quota adjustment rules; The test element determination unit is used to determine the test indicators and test cycle based on the test element parameters. The test indicator value acquisition unit is used to perform quota adjustment tests on each of the users to be tested according to the quota adjustment information, and to obtain the test indicator value of each user to be tested based on all test result data corresponding to the test indicator within the test period.

[0057] In one embodiment, the quota adjustment test module 40 further includes: The group quota adjustment rule parsing unit is used to determine the quota type, quota adjustment method and control experiment allocation information of each user group according to the group quota adjustment rule. The control experiment allocation unit is used to divide the users to be tested in each user group into experimental users and control users according to the control experiment allocation information, and to determine the experimental quota adjustment information of each experimental user and the control quota adjustment information of each control user according to the quota type, quota adjustment method and control experiment allocation information. The test result data acquisition unit is used to perform a credit limit adjustment test on the experimental user based on the experimental credit limit adjustment information to obtain the test result data of the experimental user, and to perform a credit limit adjustment test on the control user based on the control credit limit adjustment information to obtain the test result data of the control user.

[0058] In one embodiment, the model training module 50 includes: The business indicator fitting unit is used to perform fitting processing on the business indicator test values ​​of all users to be tested in each user group to obtain the business indicator fitting values ​​of each user to be tested. The risk index fitting unit is used to fit the risk index test values ​​of all users to be tested in each user group to obtain the risk index fitting values ​​of each user to be tested. The sample data set generation unit is used to generate a sample data set based on the user characteristic information, credit limit adjustment information, operational indicator fitting value and risk indicator fitting value of each of the test users. The sample dataset generation unit is used to obtain a sample dataset based on the set of all said sample data groups.

[0059] Specific limitations regarding the training device for the credit limit adjustment model can be found in the limitations on the training method for the credit limit adjustment model described above, and will not be repeated here. Each module in the aforementioned training device for the credit limit adjustment model can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0060] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 11 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computational and control capabilities. The memory includes a readable storage medium and internal memory. The readable storage medium stores an operating system, computer-readable instructions, and a database. The internal memory provides an environment for the operation of the operating system and computer-readable instructions in the readable storage medium. The database stores data related to the training method of the credit limit adjustment model. The network interface communicates with external terminals via a network connection. When the computer-readable instructions are executed by the processor, a training method for a credit limit adjustment model is implemented. The readable storage medium provided in this embodiment includes both non-volatile and volatile readable storage media.

[0061] In one embodiment, a computer device is provided, including a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, wherein the processor executes the computer-readable instructions to implement the steps of a method for training a credit limit adjustment model: Obtain a credit limit adjustment test request, and determine the test element parameters, test target account, and test credit limit adjustment rules based on the credit limit adjustment test request; Based on the test object account, determine the test users and the user characteristic information of each test user; The user feature information is grouped according to the test quota adjustment rules, and all the users to be tested are divided into at least two user groups, and the group quota adjustment rules for each user group are determined. According to the group quota adjustment rules and the test element parameters, the quota adjustment test is carried out on the users to be tested in each user group to obtain the quota adjustment information and test index value of each user to be tested. A sample dataset is generated based on the user characteristic information, credit limit adjustment information, and test indicator values ​​of all the users to be tested. The initial credit limit adjustment model is then iteratively trained based on the sample dataset, and the credit limit adjustment model is obtained when the preset iteration stopping condition is met.

[0062] In addition, the processor can also perform the following steps when executing computer-readable instructions: Obtain a credit limit adjustment request, and determine the target feature information of the user to be analyzed based on the credit limit adjustment request; The target feature information is analyzed and processed according to the pre-trained quota adjustment model to obtain the initial adjustment quota corresponding to the operating indicators. The quota adjustment model is trained by the training method of the quota adjustment model described above. Based on the initial adjustment limit, a recommended risk indicator value is determined. When it is confirmed that the recommended risk indicator value is lower than a preset risk threshold, the limit adjustment recommendation information is determined based on the initial adjustment limit.

[0063] In one embodiment, one or more computer-readable storage media storing computer-readable instructions are provided. The readable storage media provided in this embodiment include non-volatile readable storage media and volatile readable storage media. The readable storage media stores computer-readable instructions, which, when executed by one or more processors, implement the steps of the following method for training a quota adjustment model: Obtain a credit limit adjustment test request, and determine the test element parameters, test target account, and test credit limit adjustment rules based on the credit limit adjustment test request; Based on the test object account, determine the test users and the user characteristic information of each test user; The user feature information is grouped according to the test quota adjustment rules, and all the users to be tested are divided into at least two user groups, and the group quota adjustment rules for each user group are determined. According to the group quota adjustment rules and the test element parameters, the quota adjustment test is carried out on the users to be tested in each user group to obtain the quota adjustment information and test index value of each user to be tested. A sample dataset is generated based on the user characteristic information, credit limit adjustment information, and test indicator values ​​of all the users to be tested. The initial credit limit adjustment model is then iteratively trained based on the sample dataset, and the credit limit adjustment model is obtained when the preset iteration stopping condition is met.

[0064] In addition, the processor can also perform the following steps when executing computer-readable instructions: Obtain a credit limit adjustment request, and determine the target feature information of the user to be analyzed based on the credit limit adjustment request; The target feature information is analyzed and processed according to the pre-trained quota adjustment model to obtain the initial adjustment quota corresponding to the operating indicators. The quota adjustment model is trained by the training method of the quota adjustment model described above. Based on the initial adjustment limit, a recommended risk indicator value is determined. When it is confirmed that the recommended risk indicator value is lower than a preset risk threshold, the limit adjustment recommendation information is determined based on the initial adjustment limit.

[0065] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by instructing related hardware with computer-readable instructions. These computer-readable instructions can be stored in a non-volatile readable storage medium or a volatile readable storage medium. When executed, these computer-readable instructions can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0066] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0067] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A training method for a credit limit adjustment model, characterized in that, The credit limit adjustment model is used to determine the target feature information of the user to be analyzed based on the credit limit adjustment request after receiving the request, and to analyze and process the target feature information according to the credit limit adjustment model to obtain and output credit limit adjustment recommendation information for the user to be analyzed. The training method includes: Obtain a credit limit adjustment test request, and determine the test element parameters, test target account, and test credit limit adjustment rules based on the credit limit adjustment test request; Based on the test object account, determine the test users and the user characteristic information of each test user; The user feature information is grouped according to the test quota adjustment rules, and all the users to be tested are divided into at least two user groups, and the group quota adjustment rules for each user group are determined. According to the group quota adjustment rules and the test element parameters, the quota adjustment test is carried out on the users to be tested in each user group to obtain the quota adjustment information and test index value of each user to be tested. A sample dataset is generated based on the user characteristic information, credit limit adjustment information, and test indicator values ​​of all the users to be tested. The initial credit limit adjustment model is then iteratively trained based on the sample dataset, and the credit limit adjustment model is obtained when the preset iteration stopping condition is met.

2. The training method for the quota adjustment model as described in claim 1, characterized in that, Before obtaining the quota adjustment test request, the following is included: Obtain the test element configuration information, test account selection information, and at least two sets of group feature configuration information and group rule configuration information entered by the client; The test element parameters are determined based on the test element configuration information, the test target account is determined based on the test account selection information, and the test quota adjustment rule is determined based on all the group feature configuration information and group rule configuration information. A limit adjustment test request is generated based on the test element parameters, the test target account, and the test limit adjustment rules.

3. The training method for the quota adjustment model as described in claim 1, characterized in that, The test quota adjustment rules include at least two group characteristic conditions and group quota adjustment rules associated with each of the group characteristic conditions; The step of grouping the user feature information according to the test quota adjustment rules, dividing all the users to be tested into at least two user groups, and determining the group quota adjustment rules for each user group includes: The user feature information of each user to be tested is compared with all the group feature conditions to obtain the condition comparison results of each user to be tested. Find the characteristic consistency result from all condition comparison results of each user to be tested, and classify the user to be tested into a user group with the group characteristic condition corresponding to the characteristic consistency result, and determine the group quota adjustment rule associated with the group characteristic condition as the group quota adjustment rule of the user group, until all users to be tested are classified.

4. The training method for the quota adjustment model as described in claim 1, characterized in that, The step of conducting credit limit adjustment tests on users in each user group according to the group credit limit adjustment rules and the test element parameters, and obtaining credit limit adjustment information and test indicator values ​​for each user, includes: The credit limit adjustment information for each user to be tested in each user group is determined according to the group credit limit adjustment rules. The test indicators and test cycle are determined based on the aforementioned test element parameters; Based on the credit limit adjustment information, a credit limit adjustment test is conducted on each of the users to be tested, and the test indicator value of each user to be tested is obtained based on all test result data corresponding to the test indicator within the test period.

5. The training method for the quota adjustment model as described in claim 4, characterized in that, The step of conducting credit limit adjustment tests on each of the users to be tested based on the credit limit adjustment information includes: The credit limit type, credit limit adjustment method, and control experiment allocation information for each user group are determined according to the group credit limit adjustment rules. Based on the control experiment allocation information, the users to be tested in each user group are divided into experimental users and control users, and the experimental quota adjustment information and the control quota adjustment information of each experimental user and the control quota adjustment information of each control user are determined according to the quota type, quota adjustment method and control experiment allocation information. Based on the experimental credit limit adjustment information, the experimental user is tested for credit limit adjustment to obtain the test result data of the experimental user. Similarly, based on the control credit limit adjustment information, the control user is tested for credit limit adjustment to obtain the test result data of the control user.

6. The training method for the credit limit adjustment model as described in claim 1, characterized in that, The test indicator values ​​include the test values ​​of operating indicators corresponding to operating indicators and the test values ​​of risk indicators corresponding to risk indicators; The step of generating a sample dataset based on the user characteristic information, credit limit adjustment information, and test indicator values ​​of each user to be tested in all the user groups includes: The operational indicator test values ​​of all users to be tested in each user group are fitted to obtain the operational indicator fitted values ​​of each user to be tested. The risk index test values ​​of all users to be tested in each user group are fitted to obtain the risk index fitted values ​​of each user to be tested. A sample data set is generated based on the user characteristic information, credit limit adjustment information, operational indicator fitting value, and risk indicator fitting value of each user to be tested. The sample dataset is obtained from the set of all the said sample data groups.

7. A method for adjusting credit limits, characterized in that, include: Obtain a credit limit adjustment request, and determine the target feature information of the user to be analyzed based on the credit limit adjustment request; The target feature information is analyzed and processed according to the pre-trained quota adjustment model to obtain the initial adjustment quota corresponding to the operating indicators. The quota adjustment model is trained by the training method of the quota adjustment model as described in any one of claims 1 to 6. Based on the initial adjustment limit, a recommended risk indicator value is determined. When it is confirmed that the recommended risk indicator value is lower than a preset risk threshold, the limit adjustment recommendation information is determined based on the initial adjustment limit.

8. A training device for a credit limit adjustment model, characterized in that, include: The test request acquisition module is used to acquire the quota adjustment test request and determine the test element parameters, test object account and test quota adjustment rules based on the quota adjustment test request. The test user determination module is used to determine the test users and the user characteristic information of each test user based on the test object account; The user group segmentation module is used to group the user feature information according to the test quota adjustment rules, divide all the users to be tested into at least two user groups, and determine the group quota adjustment rules for each user group. The credit limit adjustment test module is used to perform credit limit adjustment tests on the users to be tested in each user group according to the group credit limit adjustment rules and the test element parameters, and obtain the credit limit adjustment information and test indicator values ​​of each user to be tested. The model training module is used to generate a sample dataset based on the user feature information, credit limit adjustment information and test index values ​​of all the users to be tested, and to iteratively train the initial credit limit adjustment model based on the sample dataset, and obtain the credit limit adjustment model when the preset iteration stopping condition is reached.

9. A computer device comprising a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, characterized in that, When the processor executes the computer-readable instructions, it implements the training method of the quota adjustment model as described in any one of claims 1 to 6 or the quota adjustment method as described in claim 7.

10. A computer-readable storage medium storing computer-readable instructions, characterized in that, When the computer-readable instructions are executed by one or more processors, the one or more processors cause the one or more processors to perform the training method of the quota adjustment model as described in any one of claims 1 to 6 or the quota adjustment method as described in claim 7.

Images