User management and control method and device, electronic equipment, storage medium and program product

Abstract

The invention relates to the technical field of data analysis and artificial intelligence, in particular to a user management and control method and device, electronic equipment, a storage medium and a program product. The user management and control method comprises the following steps: acquiring multi-dimensional feature data information of a target user on a platform; verifying the multi-dimensional feature data information by using a target super user identification SIM card bound with the target user to obtain credible multi-dimensional feature data information; obtaining a target risk state of the target user and target risk management and control corresponding to the target risk state based on the credible multi-dimensional feature data information and a pre-trained arbitrage risk model; and executing target risk management and control on the target user. According to the method and the device, real-time locking, monitoring and management of potential risk accounts are realized, and counterfeit data can be eliminated by verifying the multi-dimensional feature data information of the target user through the super SIM card, so that the data authenticity is improved to reduce model misjudgment, and the operation risk is reduced.

Description

Technical Field

[0001] This disclosure relates to the fields of data analysis and artificial intelligence technology, and in particular to a user management method, device, electronic device, storage medium, and program product. Background Technology

[0002] With the development of mobile internet, preferential policies have become a key means of attracting new users and retaining existing ones, but they are also easily exploited by users, causing economic losses to enterprises. Therefore, it is extremely important to manage risky users. Currently, in the communications and internet field, due to the special nature of business and systems, the industry often uses "time-series user behavior logs + historical behavior recall prediction" to prevent arbitrage and thus identify risky users.

[0003] However, the aforementioned existing technologies are mostly focused on financial scenarios, and they still have significant shortcomings in meeting the demand for "real-time identification of risky arbitrage users" in the telecommunications and internet sector: First, risky arbitrage users can create false behavioral data by changing devices or falsifying device information, weakening the authenticity of logs and leading to model misjudgments; second, the technology has low compatibility with the scenarios for preventing and controlling preferential arbitrage in the telecommunications and internet sector, making it difficult to match its business risk characteristics; third, due to limitations in data collection cycles and log analysis efficiency, real-time performance is insufficient, making it impossible to quickly respond to arbitrage behavior. Therefore, how to effectively and in real-time identify risky arbitrage users in the telecommunications and internet sector and reduce operational risks has become an urgent technical problem to be solved. Summary of the Invention

[0004] In view of the above-mentioned technical problems, this disclosure provides a user management method, device, electronic device, storage medium and program product.

[0005] Firstly, this disclosure provides a user management method, including: Obtain multi-dimensional characteristic data of target users on the platform; The target user's super user identification SIM card is used to verify multi-dimensional feature data information and obtain credible multi-dimensional feature data information. Based on reliable multi-dimensional feature data and a pre-trained arbitrage risk model, the target risk status of the target user and the target risk management corresponding to the target risk status are obtained. Implement target risk management for target users.

[0006] Furthermore, according to the method of the first aspect of this disclosure, the multi-dimensional feature data information includes identity feature information, behavioral feature information, consumption feature information, customer service feature information, channel feature information, and device feature information.

[0007] Furthermore, according to the method of the first aspect of this disclosure, multi-dimensional feature data information is verified using the target super SIM card bound to the target user to obtain credible multi-dimensional feature data information, including: Use the target super SIM card to obtain the target hardware identifier of the target terminal device corresponding to the target user; Using the target super SIM card, the target hardware identifier is compared with the baseline hardware identifier corresponding to the target user stored in the target super SIM card to determine whether the target terminal device is a legitimate bound device; If the device is legally bound, the target hardware identifier is compared with the hardware identifier contained in the device feature information to verify the authenticity of the device feature information; When the requirement of authenticity is met, multi-dimensional feature data information is determined to be credible multi-dimensional feature data information.

[0008] Furthermore, according to the method of the first aspect of this disclosure, the arbitrage risk model includes a risk classification model and a Q-value table; Based on reliable multi-dimensional feature data and a pre-trained arbitrage risk model, the target user's target risk status and the corresponding target risk management are obtained, including: Preprocess reliable multi-dimensional feature data; The preprocessed credible multi-dimensional feature data is input into the risk classification model to obtain the target risk status of the target user. Based on the target risk status and the Q-value table, the target risk management corresponding to the target risk status is obtained. The Q-value table is used to describe the mapping relationship between the risk status, risk management, and predicted management benefits. The predicted management benefits are used to describe the predicted benefits after implementing risk management on the target user.

[0009] Furthermore, according to the method of the first aspect of this disclosure, the preprocessing includes at least one or more of the following: null value imputation, outlier handling, data standardization, formatting, binning coding, and similarity analysis.

[0010] Furthermore, according to the method of the first aspect of this disclosure, target risk management is performed on target users based on the target risk status and Q-value table, including: Based on the target risk status, find the maximum predictive control benefit matching the target risk status and the target risk control matching the maximum predictive control benefit from the Q value table. Implement target risk management for target users.

[0011] Furthermore, according to the method of the first aspect of this disclosure, the arbitrage risk model is trained in the following manner: Obtain historical multi-dimensional feature data of each user on the platform; By using the super SIM card corresponding to each user, historical multi-dimensional feature data information is verified to obtain credible historical multi-dimensional feature data information for each user; Based on the different risk statuses of each user and the actual risk management corresponding to the different risk statuses, an initial Q value table is constructed, in which the Q value is 0. The trusted historical multi-dimensional feature data of each user is divided into a training set and a test set; A deep neural network (DNN) model is trained using the training set to obtain the predicted risk status of each user. Based on each user's predicted risk status and initial Q-value table, update the parameters and initial Q-value table of the DNN model; The updated DNN model and the updated initial Q-value table were tested using a test set. Based on the test results, the trained DNN model and Q-value table were determined, and the trained DNN model was designated as a risk classification model.

[0012] Furthermore, according to the method of the first aspect of this disclosure, the historical multi-dimensional feature data information includes at least: Identity characteristics, behavioral characteristics, consumption characteristics, customer service characteristics, channel characteristics, and device characteristics.

[0013] Furthermore, according to the method of the first aspect of this disclosure, historical multi-dimensional feature data information is verified using the Super SIM card corresponding to each user, including: For each user, the corresponding Super SIM card performs the following operations: Obtain the hardware identifier of the user's currently used terminal device; The hardware identifier is compared with the base hardware identifier corresponding to the user stored in the Super SIM card to determine whether the terminal device is a legitimate bound device.

[0014] Furthermore, according to the method of the first aspect of this disclosure, verifying historical multi-dimensional feature data information using the super SIM card corresponding to each user also includes: If the device is legally bound, the hardware identifier is compared with the hardware identifier contained in the device feature information to verify the authenticity of the device feature information.

[0015] Furthermore, according to the method of the first aspect of this disclosure, reliable historical multi-dimensional feature data information is obtained, including: When the device feature information corresponding to a user meets the requirement of authenticity, the historical multi-dimensional feature data information of each user is determined to be reliable historical multi-dimensional feature data information.

[0016] Furthermore, according to the method of the first aspect of this disclosure, the method also includes: If the device is illegally bound, the user will be marked as a risk user, and the user's historical multi-dimensional feature data will be removed.

[0017] Furthermore, according to the method of the first aspect of this disclosure, the credible historical multi-dimensional feature data information is divided into a training set and a test set, including: Preprocess the credible historical multi-dimensional feature data of each user; The preprocessed credible historical multi-dimensional feature data of each user is divided into training set and test set.

[0018] Furthermore, according to the method of the first aspect of this disclosure, the preprocessing includes at least one or more of the following: null value imputation, outlier handling, data standardization, formatting, binning coding, and similarity analysis.

[0019] Furthermore, according to the method of the first aspect of this disclosure, the method also includes: Based on the time dimension, the credible historical multi-dimensional feature data information of each user is divided into new data and old data; Based on the logical relationship between the new and old data, the consistency of each user's behavior is confirmed. When the behavior of each user is consistent, the preprocessed credible historical multi-dimensional feature data of each user is divided into training set and test set.

[0020] Furthermore, according to the method of the first aspect of this disclosure, the method also includes: Based on the predicted management benefits and the actual management benefits corresponding to the predicted management benefits, update the risk classification model and Q-value table.

[0021] Secondly, this disclosure provides a user control device, including: The multi-dimensional feature data acquisition module is used to acquire multi-dimensional feature data information of the target user; The trusted multi-dimensional feature data information acquisition module is used to verify multi-dimensional feature data information by using the target super user identification SIM card bound to the target user, and to acquire trusted multi-dimensional feature data information. The risk status and risk management determination module is used to obtain the target risk status of the target user and the target risk management corresponding to the target risk status based on reliable multi-dimensional feature data and pre-trained arbitrage risk model. The risk management execution module is used to perform target risk management on target users.

[0022] Thirdly, this disclosure provides an electronic device, including: a memory for storing computer-readable instructions; and a processor for executing the computer-readable instructions, causing the electronic device to perform the method as described in any embodiment of the first aspect.

[0023] Fourthly, this disclosure provides a non-transitory computer-readable storage medium for storing computer-readable instructions that, when executed by a processor, cause the processor to perform the method as described in any embodiment of the first aspect.

[0024] Fifthly, this disclosure provides a computer program product, including a computer program that, when executed by a processor, implements the method as described in any embodiment of the first aspect.

[0025] This disclosure provides a user management method, device, electronic device, storage medium, and program product. This disclosure obtains multi-dimensional feature data of the target user, verifies the multi-dimensional feature data through the SIM card bound to the target user to ensure its credibility, and then combines it with an arbitrage risk model to determine the target user's target risk status and corresponding target risk management, and executes the target risk management on the target user. This achieves real-time locking and monitoring of potential risky accounts, and the verification of the target user's multi-dimensional feature data information through the super SIM card can eliminate forged data, improve data authenticity, reduce model misjudgment, and reduce operational risks.

[0026] It should be understood that both the foregoing general description and the following detailed description are exemplary and intended to provide further illustration of the claimed technology. Attached Figure Description

[0027] The above and other objects, features, and advantages of this disclosure will become more apparent from the more detailed description of the embodiments thereof in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this disclosure and form part of the specification. They are used together with the embodiments of this disclosure to explain the disclosure and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0028] Figure 1 A flowchart illustrating a user management method provided in this embodiment of the disclosure; Figure 2 A flowchart illustrating yet another user management method provided in this disclosure embodiment; Figure 3 A flowchart illustrating yet another user management method provided in this disclosure embodiment; Figure 4 A flowchart illustrating yet another user management method provided in this disclosure embodiment; Figure 5 A flowchart illustrating yet another user management method provided in this disclosure embodiment; Figure 6A structural block diagram of a user control device provided in this disclosure embodiment; Figure 7 A hardware block diagram of an electronic device provided in an embodiment of this disclosure; Figure 8 This is a schematic diagram of a computer-readable storage medium provided in an embodiment of this disclosure. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of this disclosure more apparent, exemplary embodiments according to this disclosure will now be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this disclosure, and not all embodiments of this disclosure. It should be understood that this disclosure is not limited to the exemplary embodiments described herein.

[0030] Figure 1 This is a flowchart illustrating a user management method provided in an embodiment of this disclosure. Figure 1 As shown, the method specifically includes the following steps: Step 101: Obtain multi-dimensional feature data information of target users on the platform.

[0031] In one embodiment of this disclosure, a platform refers to a functional carrier that connects users with each other and with content or services through specific products or services such as apps, web pages, and software, and carries various activities such as interaction, consumption, and storage. Examples include Douyin (TikTok), cloud storage, and cloud phones. Target users are specific user groups on the platform that require key monitoring and may exhibit risky behavior. Multi-dimensional feature information refers to various data and information that comprehensively characterize the target user from multiple different levels. Multi-dimensional feature data information of the target user is collected on a designated platform. The method for collecting multi-dimensional feature data information of the target user can employ existing probe technologies; the specific method is not limited herein.

[0032] For example, on the app, existing probe technology can be used to collect multi-dimensional feature data of target users on the platform. For instance, data such as "22-year-old student, registered through campus channels, and frequently logged in on common devices" can be collected for cloud phone user Xiao Zhang to prepare for subsequent verification.

[0033] In one embodiment of this disclosure, the multi-dimensional feature data information includes at least identity feature information, behavioral feature information, consumption feature information, customer service feature information, channel feature information, and device feature information.

[0034] in: Identity characteristics information: including but not limited to the target user's natural attributes (such as age, gender), social relationships (such as family members, social networks), and occupational information.

[0035] Behavioral characteristics information: including but not limited to recording the target user's interactive behavior on the platform, such as page browsing path, click preferences, dwell time, etc., in order to capture the user's operating habits and interests.

[0036] Consumer characteristics information: including but not limited to the target user's purchase history, payment method preferences, return frequency and other economic activities. By analyzing the target user's purchase history, payment method preferences, return frequency and other economic activities, the spending power and credit risk of the target user can be assessed.

[0037] Customer service characteristic information: including but not limited to communication records between the target user and the customer service team, including inquiry questions, complaint handling results, etc., which helps to understand the target user's service needs and service satisfaction.

[0038] Channel characteristics information: including but not limited to different entry points accessed by target users, such as websites, mobile applications or offline stores, to understand the impact of different channels on the target user experience.

[0039] Device characteristic information: including but not limited to information about the terminal device used by the target user, such as operating system version, screen resolution, geographical location, hardware identifier, etc., to ensure that abnormal login behavior can be identified.

[0040] Based on the multi-dimensional characteristic data of the target users mentioned above, a detailed and accurate profile of the target users can be constructed.

[0041] Step 102: Verify the multi-dimensional feature data information using the target super user identification SIM card bound to the target user, and obtain credible multi-dimensional feature data information.

[0042] In one embodiment of this disclosure, the previously collected multi-dimensional feature data is verified through the target user's bound Subscriber Identity Module (SIM) [i.e., the main SIM card registered under the target user's real name] to obtain authentic and reliable user data (i.e., reliable multi-dimensional feature data information).

[0043] Step 103: Based on credible multi-dimensional feature data and a pre-trained arbitrage risk model, obtain the target risk status of the target user and the target risk management corresponding to the target risk status.

[0044] In one embodiment of this disclosure, the target risk status is the final judgment result of the arbitrage risk model on the potential risk level of the target user, and the target risk management is the specific countermeasures given based on this judgment result. Trustworthy multi-dimensional data is input into a pre-trained arbitrage risk model, allowing the arbitrage risk model to automatically determine the current target risk status of the target user and provide specific management measures corresponding to that target risk status (target risk management).

[0045] For example, target risk status can be categorized into low risk, medium risk, and high risk. Low risk corresponds to target risk control where no additional restrictions are imposed on the target user, ensuring smooth usage, and basic data retention is sufficient. Medium risk corresponds to target risk control where only light restrictions are imposed on the target user's "sensitive operations" or "core functions," which can be restored to normal after verification without affecting basic usage. High risk corresponds to target risk control where strict restrictions are imposed on the target user's account functions, or even account freezing, to prevent losses to the platform or other users. The above target risk statuses are merely illustrative; specific level classifications and corresponding target risk control measures can be set according to actual applications. This disclosure does not impose any limitations.

[0046] Step 104: Implement target risk management for target users.

[0047] In one embodiment of this disclosure, the target risk control is performed on the target user based on the obtained target risk control, thereby directly preventing or reducing the user's arbitrage risk.

[0048] In summary, according to the technical solution provided in the embodiments of this disclosure, this disclosure obtains multi-dimensional feature data of the target user, verifies the multi-dimensional feature data through the SIM card bound to the target user to ensure credibility, then combines the arbitrage risk model to determine the target risk status of the target user and the corresponding target risk control, and executes the target risk control on the target user, thereby realizing real-time locking and monitoring of potential risk accounts. Moreover, verifying the multi-dimensional feature data information of the target user through the super SIM card can eliminate forged data, improve data authenticity, reduce model misjudgment, and reduce operational risks.

[0049] Figure 2 This is a flowchart illustrating a user management method provided in an embodiment of this disclosure. Figure 2 As shown, the multi-dimensional feature data information is verified using the target super SIM card bound to the target user to obtain credible multi-dimensional feature data information. The specific steps include the following: Step 201: Use the target super SIM card to obtain the target hardware identifier of the target terminal device corresponding to the target user.

[0050] In one embodiment of this disclosure, the target super SIM card can capture and upload the target hardware unique identifier (i.e., the unique physical identifier of the target terminal device, such as IMEI number, MAC address, etc.) of the target terminal device. The target hardware unique identifier of the target terminal device (such as a mobile phone, tablet, etc.) currently inserted into the super SIM card is obtained through the target super SIM card bound to the target user.

[0051] Step 202: Using the target super SIM card, compare the target hardware identifier with the base hardware identifier corresponding to the target user stored in the target super SIM card to determine whether the target terminal device is a legitimate bound device.

[0052] In one embodiment of this disclosure, the target super SIM card has a built-in base hardware identifier (i.e., the hardware identifier of the target user's commonly used / authenticated terminal device, such as the IMEI of a commonly used mobile phone) pre-registered by the target user, and also has independent computing capabilities. After obtaining the target hardware identifier, the super SIM card will directly compare the target hardware identifier with the base hardware identifier to determine whether the target terminal device is a legitimate bound device.

[0053] Step 203: If the device is legally bound, compare the target hardware identifier with the hardware identifier contained in the device feature information to verify the authenticity of the device feature information.

[0054] In one embodiment of this disclosure, if the target hardware identifier matches the base hardware identifier, the target terminal device is determined to be a legitimate bound device (i.e., the currently used target device is a commonly used device that the target user has authenticated, and the risk is low). Further, to verify the authenticity of the device feature information, the hardware identifier (i.e., the device identifier of the terminal device used by the target user, such as the IMEI number recorded in the login log) contained in the acquired multi-dimensional feature data is extracted from the device feature information and compared with the target hardware identifier for consistency.

[0055] Step 204: When the authenticity requirement is met, the multi-dimensional feature data information is determined to be credible multi-dimensional feature data information.

[0056] In one embodiment of this disclosure, if the hardware identifier contained in the device feature information is consistent with the target hardware identifier, it indicates that the device feature information collected by the platform has not been tampered with (matches the actual device identifier), and the device feature information is determined to be authentic and valid. Therefore, the corresponding multi-dimensional feature data information can be determined to be reliable multi-dimensional feature data information.

[0057] Furthermore, if the two are inconsistent, it indicates that the device feature information may be forged or abnormal (e.g., the user has tampered with the device identifier recorded by the platform through technical means), and the device feature information is deemed unreliable. Accordingly, the target user is marked as a risk user, and the target user's multi-dimensional feature data information is removed.

[0058] In summary, the technical solution provided by this disclosure, through the target super SIM card, first obtains the real hardware identifier of the terminal device and determines the device's legitimacy, then verifies the consistency of the device feature information collected by the platform based on this real identifier; subsequently, using the authenticity of the device feature information as the core basis, it determines the overall authenticity and reliability of the multi-dimensional feature data information. This process can effectively eliminate risks such as illegal device use, forged device feature data, and inconsistencies between collected data and real devices, providing a reliable data foundation for subsequent arbitrage risk models to accurately assess the user's risk status.

[0059] Figure 3 This is a flowchart illustrating a user management method provided in an embodiment of this disclosure. Figure 3 As shown, the arbitrage risk model includes a risk classification model and a Q-value table.

[0060] Based on reliable multi-dimensional feature data and a pre-trained arbitrage risk model, the target risk status of the target user and the corresponding target risk management are obtained, specifically including the following steps: Step 301: Preprocess the credible multi-dimensional feature data information.

[0061] In one embodiment of this disclosure, although the verified credible multi-dimensional data is authentic, it may have problems such as disordered format, large numerical differences, and incomplete information. Preprocessing is to solve these problems, transforming the data into a unified format that the model can understand and calculate accurately. Preprocessing includes at least one or more of the following: null value imputation, outlier handling, data standardization, formatting, binning coding, and similarity analysis.

[0062] Among these methods, null value imputation addresses missing data; outlier handling addresses extremely unreasonable data; data standardization addresses large differences in numerical ranges; formatting addresses inconsistent formatting; binning coding addresses the inability to calculate values ​​for categorical data; and similarity analysis addresses data redundancy or correlation.

[0063] Step 302: Input the preprocessed credible multi-dimensional feature data information into the risk classification model to obtain the target risk status of the target user.

[0064] In one embodiment of this disclosure, the risk classification model is trained based on credible multi-dimensional feature data of a batch of historical user samples and their corresponding risk classification labels. The credible multi-dimensional feature data of the historical user samples includes multi-dimensional data such as identity, behavior, and device information generated by multiple users on the platform (all verified as authentic and reliable by the Super SIM card); each user sample corresponds to a unique risk classification label (e.g., low risk, medium risk, high risk). After inputting the preprocessed credible multi-dimensional feature data into the risk classification model, the risk classification model outputs the predicted target risk status of the target user.

[0065] Step 303: Based on the target risk status and the Q-value table, obtain the target risk management corresponding to the target risk status. The Q-value table is used to describe the mapping relationship between the risk status, risk management, and predicted management benefits. The predicted management benefits are used to describe the predicted benefits after implementing risk management on the target user.

[0066] In one embodiment of this disclosure, the Q-value table is a pre-defined two-dimensional or three-dimensional mapping table. Essentially, it is a decision-making experience base formed by the model through learning from historical data, recording the correspondence between risk status, risk management, and predicted management benefits (i.e., Q-value). Simply put, it pre-calculates how much benefit (or how much loss reduction) a certain management measure will bring to the platform for a user with a certain risk, and uses the "Q-value" (a quantitative score) to represent the level of benefit. Based on the target risk status and the Q-value table, the target risk management corresponding to the target risk status is obtained, including: Based on the target risk status, find the maximum predicted control benefit matching the target risk status and the target risk control matching the maximum predicted control benefit from the Q value table. Implement target risk management for target users.

[0067] Specifically, the Q-value table records all combinations of risk state → risk control → Q-value. First, all entries matching the risk state of the target user are selected from the Q-value table, forming a candidate set. Among the selected candidate entries, the predicted control benefits (Q-values) of different risk controls (i.e., control measures) are compared. The one with the largest Q-value is the optimal target risk control [because it means that for the platform, taking this measure will yield the greatest benefit (or reduce the greatest loss)]. Based on the determined target risk control, risk control is implemented for the target user.

[0068] For example, the Q-value table can be a three-dimensional data matrix of product, risk status, and risk management, representing the score corresponding to the implementation of risk management for a user under a certain risk status by a certain product.

[0069] In summary, based on the technical solutions provided in the embodiments of this disclosure, this disclosure, through the combination of risk classification models and Q-value tables, can quickly and effectively identify risky users and conduct risk management.

[0070] Furthermore, user management methods also include: Based on the predicted management benefits and the actual management benefits corresponding to the predicted management benefits, update the risk classification model and Q-value table.

[0071] Specifically, risk classification models and Q-value tables are trained based on historical data. However, in actual business operations, user behavior and risk patterns are constantly changing (for example, new fraud methods emerge), leading to a gap between the predicted control benefits (the Q-value calculated by the model) and the actual control benefits (the real benefits / losses after implementation). Therefore, it is necessary to use actual results to adjust the model and Q-value tables in reverse, making them more consistent with the current real business scenario and avoiding the use of old experience to judge new problems.

[0072] In summary, according to the technical solution provided in the embodiments of this disclosure, by using actual management benefits to modify the risk classification model and Q-value table, the risk classification model can more accurately identify new risks, and the Q-value table can better reflect current user feedback and business loss, ultimately achieving continuous optimization of risk management effectiveness.

[0073] Figure 4 This is a flowchart illustrating another user management method provided in an embodiment of the present disclosure.

[0074] like Figure 4 As shown, the arbitrage risk model is trained in the following way: Step 401: Obtain historical multi-dimensional feature data information of each user on the platform.

[0075] In one embodiment of this disclosure, the historical multi-dimensional feature data information includes at least: The information includes identity characteristics, behavioral characteristics, consumption characteristics, customer service characteristics, channel characteristics, and device characteristics. Historical multi-dimensional characteristic data can refer to data from two or three months ago. The specific timeframe can be selected based on actual needs, and this disclosure does not impose any limitations.

[0076] Specifically, a platform refers to a functional carrier that connects users with each other and with content or services through specific products or services such as apps, web pages, and software, and carries various activities such as interaction, consumption, and storage. Examples include Douyin (TikTok), cloud storage, and cloud phones. Each user refers to all registered users on the platform. Historical multi-dimensional feature data information refers to various data and information from multiple different levels in historical data that comprehensively characterize the target user. Historical multi-dimensional feature data information of each user is collected on the designated platform. The method for collecting historical multi-dimensional feature data information of each user can employ existing probe technologies; the specific method is not limited herein.

[0077] Step 402: Verify the historical multi-dimensional feature data information using the Super SIM card corresponding to each user, and obtain the credible historical multi-dimensional feature data information of each user.

[0078] In one embodiment of this disclosure, historical multi-dimensional feature data information is verified using the Super SIM card corresponding to each user, including: For each user, the corresponding Super SIM card performs the following operations: Obtain the hardware identifier of the user's currently used terminal device; The hardware identifier is compared with the base hardware identifier corresponding to the user stored in the Super SIM card to determine whether the terminal device is a legitimate bound device.

[0079] Specifically, the Super SIM card can capture and upload the unique hardware identifier (i.e., the unique physical identifier of the terminal device, such as IMEI number, MAC address, etc.) of the user's currently used terminal device. Through the Super SIM card bound to the user, it obtains the unique hardware identifier of the terminal device (such as a mobile phone, tablet, etc.) currently inserted into the Super SIM card. The Super SIM card directly compares this hardware identifier with the user-pre-registered baseline hardware identifier built into the Super SIM card, and determines whether the terminal device is a legitimate bound device based on the comparison result.

[0080] In one embodiment of this disclosure, the verification of historical multi-dimensional feature data information using the super SIM card corresponding to each user further includes: If the device is legally bound, the hardware identifier is compared with the hardware identifier contained in the device feature information to verify the authenticity of the device feature information.

[0081] Specifically, if the hardware identifier matches the baseline hardware identifier, the terminal device is determined to be a legitimate bound device (i.e., the device currently used by the user is a commonly used device that the user has authenticated, with low risk), thus proving that the device feature information is authentic and reliable. Further, to verify the authenticity of the device feature information, the hardware identifier (i.e., the device identifier of the terminal device used by the user, such as the IMEI number recorded in the login log) contained in the acquired historical multi-dimensional feature data is extracted from the device feature information and compared with the hardware identifier to verify the authenticity of the device feature information.

[0082] Furthermore, if the device is illegally bound, the user is marked as a risk user, and the user's historical multi-dimensional characteristic data information is removed.

[0083] Specifically, when the comparison results are inconsistent, the terminal device is determined to be an illegally bound device, the user is marked as a risk user, and the user's historical multi-dimensional feature data information is removed.

[0084] In one embodiment of this disclosure, obtaining reliable historical multi-dimensional feature data information includes: When the device feature information corresponding to a user meets the requirement of authenticity, the historical multi-dimensional feature data information of each user is determined to be reliable historical multi-dimensional feature data information.

[0085] Specifically, if the hardware identifier of the user's terminal device matches the hardware identifier contained in the device feature information, then the device feature information corresponding to the user satisfies the requirement of authenticity. Therefore, the user's historical multi-dimensional feature data is also authentic and reliable historical multi-dimensional information.

[0086] Step 403: Based on the different risk statuses of each user and the actual risk management corresponding to the different risk statuses, construct an initial Q value table, in which the Q value is 0.

[0087] In one embodiment of this disclosure, an initial Q index is constructed by taking the different risk states (such as low risk, medium risk, and high risk) that each user has been historically determined to be, and the control measures actually implemented for the user according to the different risk states (i.e., actual risk control). The Q values ​​in the table are all 0.

[0088] In practical applications, the initial Q-value table is based on the two-dimensional risk status → risk management, with the product added as a third dimension to form a three-dimensional combination framework of product-risk status-risk management, and then the initial Q value of all combinations is set to 0.

[0089] Step 404: Divide the trusted historical multi-dimensional feature data of each user into a training set and a test set.

[0090] In one embodiment of this disclosure, trusted historical multi-dimensional feature data information is divided into a training set and a test set, including: Preprocess the credible historical multi-dimensional feature data of each user; The preprocessed credible historical multi-dimensional feature data of each user is divided into training set and test set.

[0091] The preprocessing includes at least one or more of the following: null padding, outlier handling, data standardization, formatting, binning coding, and similarity analysis.

[0092] Specifically, null value imputation is used to address missing data issues. Outlier handling addresses extremely unreasonable data. Data standardization addresses large differences in numerical ranges. Formatting addresses inconsistent formatting. Binning coding addresses the inability to calculate categorical data. Similarity analysis addresses data redundancy or correlation issues. After preprocessing the credible historical multi-dimensional feature data of each user, a sample set containing user ID, credible historical multi-dimensional feature data, risk control, and control benefits is output. The sample set label represents the control benefit corresponding to performing risk control on a specific user, and risk control is based on the user's historical risk status. The above sample set is divided into a training set and a test set. The control benefit measures the return after risk control, calculated in conjunction with new user acquisition rewards and retention benefits, used to distinguish between genuine users and arbitrage users. Control benefit = new user acquisition reward + retention benefits - risk control cost. Genuine users have positive rewards and retention benefits, resulting in higher control benefit values, while arbitrage users have the opposite.

[0093] Step 405: Train a deep neural network (DNN) model using the training set to obtain the predicted risk status of each user.

[0094] In one embodiment of this disclosure, a deep neural network (DNN) model is trained using a training set. During training, the input of the DNN model is the credible historical multi-dimensional feature data information of each user, and the output is the predicted risk status (label) of each user.

[0095] Step 406: Based on the predicted risk status and initial Q-value table of each user, update the parameters and initial Q-value table of the DNN model.

[0096] In one embodiment of this disclosure, the Q-value table is a core tool for decision-making in reinforcement learning, recording the expected benefit of taking a certain action in a certain state (here, the state can be understood as the user's risk state, the action can be understood as the risk management measure, and the benefit is the management benefit). The predicted risk state of the DNN model provides the current state input for the Q-value table. For example, if the DNN predicts that user A is in a "high-risk state", the Q-value table will locate the position corresponding to the "high-risk state", check the expected management benefit (Q value) of different management measures (such as strict review, limit on amount, etc.), compare the expected management benefit with the actual management benefit in the sample, calculate the error, and use backpropagation to iteratively update the parameters of the DNN model and the initial Q-value table according to the formula: new Q(s,a) ← old Q(s,a) + α[r + β×MaxQ(s',a') - old Q(s,a)]. Where α is the learning rate and β is the discount factor, both of which are pre-defined hyperparameters; MaxQ(s',a'): refers to the Q value corresponding to the best action of the user in the next state s' after taking action A1. After repeating the above steps multiple times, the Q value stabilizes within a reasonable range, accurately predicting the control benefits.

[0097] Step 407: Test the updated DNN model and the updated initial Q-value table using the test set.

[0098] In one embodiment of this disclosure, a test set is used to test the updated DNN model and the updated initial Q-value table (e.g., to verify the accuracy of the model's risk assessment and the deviation rate between the Q-value table's predicted returns and actual returns), to verify whether both achieve the goal of accurate risk assessment and effective risk management.

[0099] Step 408: Based on the test results, determine the trained DNN model and Q-value table. The trained DNN model is a risk classification model.

[0100] In one embodiment of this disclosure, if the test is passed (e.g., the model risk judgment accuracy is ≥90%, and the Q-value table profit prediction deviation rate is ≤10%), it means that after multiple rounds of updates and optimizations, the DNN model and the initial Q-value table have met the goals of accurate risk state judgment and good control benefits. At this time, they are officially determined as the trained DNN model and Q-value table.

[0101] Furthermore, if the test fails (e.g., accuracy is only 75%, deviation is 30%), return to step 406 to readjust the model parameters and Q-value table, and repeat the "update → test" cycle until the test meets the requirements.

[0102] In summary, based on the technical solution provided in the embodiments of this disclosure, this disclosure achieves three major technical effects through the core logic of "super SIM card verification of data credibility → DNN model and Q-value table collaborative training optimization → full-process closed-loop management and control": First, relying on the super SIM card to verify historical multi-dimensional feature data information, the authenticity of the data is guaranteed from the source, providing a reliable foundation for risk status judgment; Second, through the division of training and test sets and reinforcement learning iteration (optimizing model parameters and Q-value table based on Q-value update formula), a high-accuracy risk classification model (trained DNN model) and an accurate Q-value table are constructed, which can accurately identify different risk statuses of users and match measures that maximize the benefits of management and control based on the risk status (i.e., risk management and control); Third, a full-process self-optimization mechanism of "data verification → model training → management and control execution → effect feedback → model update" is established, which continuously improves the accuracy of risk judgment and the rationality of management and control as actual management and control data accumulates, ultimately achieving accurate and efficient management and control of risky users, and achieving the optimal balance between ensuring platform revenue (reducing arbitrage / loss) and maintaining a normal user experience.

[0103] Figure 5 This is a flowchart illustrating yet another user management method provided in an embodiment of this disclosure. Figure 5 As shown, the above user management method also includes the following steps: Step 501: Based on the time dimension, divide the trusted historical multi-dimensional feature data information of each user into new data and old data.

[0104] In one embodiment of this disclosure, the trusted historical multi-dimensional feature data information is the user's historical data accumulated on the platform (such as identity, behavior, device data, etc. over the past 3 months). Dividing new and old data based on the time dimension involves setting a time node (such as "the second month" as the boundary) to divide the data into two parts: Old data: Historical data prior to a given time point (such as user behavior data from 3 months ago to 2 months ago), representing the user's characteristics and risk patterns in earlier periods; New data: Historical data after a certain time point (such as user behavior data in the 3rd month), representing the user's recent characteristics and risk patterns.

[0105] Step 502: Based on the logical relationship between the new and old data, confirm the consistency of each user's behavior.

[0106] In one embodiment of this disclosure, the logical relationship is the reasonable correlation that a user's behavior should have over time, such as login habits, consumption preferences, device usage, etc., which is usually stable and will not change drastically without reason; while the behavior of risky users (such as arbitrage accounts) often shows illogical breaks or sudden changes.

[0107] For example, common logical relationship judgments include: checking whether the user's current operation conflicts with their past behavior, such as suddenly changing devices, frequently logging in from different locations, changing linked bank cards, or performing many sensitive actions in a short period of time. If these behaviors differ significantly from their previous habits, they will be marked as abnormal, possibly indicating forged behavior. In this case, the user will be marked as a risk user, and their historical multi-dimensional feature data will be removed.

[0108] Step 503: When the behavior of each user is consistent, divide the preprocessed credible historical multi-dimensional feature data of each user into a training set and a test set.

[0109] Specifically, step 502 has already determined which users exhibit consistent behavior (e.g., stable frequently used devices, stable login habits, and no unexplained sudden changes) and which exhibit inconsistent behavior (e.g., sudden and frequent device switching, logins from different locations, which are marked as risky users and their data removed) by comparing the old and new data. The preprocessed, reliable historical multi-dimensional feature data of users who meet the "consistent behavior" criteria are then divided into training and test sets.

[0110] In summary, according to the technical solution provided in the embodiments of this disclosure, this disclosure first splits the credible historical data of each user into new and old data based on time nodes, providing a time reference for behavior tracking; then, by comparing the logical correlation of login habits, device usage, consumption preferences, and other dimensions in the new and old data, it accurately identifies risky users whose behavior changes without reason and removes their data, filtering out interference samples that are "real data but abnormal behavior" from the source; finally, it only preprocesses and divides the training and testing sets for the credible historical data of users with consistent behavior, ensuring that the samples for subsequent model training are all "real and stable" high-quality data, ultimately making the trained risk classification model and Q-value table more consistent with the normal user behavior patterns, reducing misjudgments of normal users, while more accurately capturing the characteristics of risky users, improving the effectiveness and reliability of overall risk control, and reducing the risk of data forgery or inconsistency.

[0111] This disclosure also provides a user control device. Figure 6 This is a structural block diagram of a user control device provided in an embodiment of the present disclosure, such as... Figure 6 As shown, the user's control device 600 includes: a multi-dimensional feature data information acquisition module 601, a reliable multi-dimensional feature data information acquisition module 602, a risk status and risk control determination module 603, and a risk control execution module 604.

[0112] Specifically, the multi-dimensional feature data information acquisition module 601 is used to acquire multi-dimensional feature data information of the target user; Specifically, the trusted multi-dimensional feature data information acquisition module 602 is used to verify the multi-dimensional feature data information by using the target super user identification SIM card bound to the target user, and to acquire trusted multi-dimensional feature data information. Specifically, the risk status and risk control determination module 603 is used to obtain the target risk status of the target user and the target risk control corresponding to the target risk status based on credible multi-dimensional feature data information and a pre-trained arbitrage risk model. Specifically, the risk control execution module 604 is used to perform target risk control on target users.

[0113] Figure 7 This is a hardware block diagram of an electronic device provided according to an embodiment of the present disclosure. The electronic device 700 according to an embodiment of the present disclosure includes at least a processor and a memory for storing computer-readable instructions. When the computer-readable instructions are loaded and executed by the processor, the processor performs the user management method described in any of the preceding embodiments of the present disclosure.

[0114] Figure 7 The illustrated electronic device 700 specifically includes a central processing unit (CPU) 701, a graphics processing unit (GPU) 702, and a memory 703. These units are interconnected via a bus 704. The CPU 701 and / or GPU 702 can function as the aforementioned processors, and the memory 703 can function as the aforementioned memory storing computer-readable instructions. Furthermore, the electronic device 700 may also include a communication unit 705, a storage unit 706, an output unit 707, an input unit 708, and an external device 709, all of which are also connected to the bus 704.

[0115] Figure 8 This is a schematic diagram of a computer-readable storage medium provided in an embodiment of this disclosure. (As shown...) Figure 8 As shown, a computer-readable storage medium 800 according to an embodiment of the present disclosure stores computer-readable instructions 801 thereon. When the computer-readable instructions 801 are executed by a processor, the user control method described with reference to the above figures according to any embodiment of the present disclosure is performed. The computer-readable storage medium includes, but is not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, optical disk, magnetic disk, etc.

[0116] This disclosure further provides a computer program product, including a computer program that, when executed by a processor, implements the user management method described in any of the preceding embodiments of this disclosure.

[0117] The above is subject to the attached document. Figure 1 To be continued Figure 8 This disclosure provides a detailed description of user management methods, apparatus, electronic devices, storage media, and program products according to embodiments of the present disclosure. The technical solution corresponding to the user management method provided in this disclosure has the following technical effects: 1. By combining advanced probe technology and super SIM card verification, real-time and accurate capture of user behavior patterns and identity information is achieved, thereby constructing a dynamically updated and highly reliable user feature database (i.e., the multi-dimensional feature data information of each user mentioned above).

[0118] 2. A risk arbitrage model based on user behavior and devices was implemented, enabling real-time updates from user characteristics to the model and control strategies.

[0119] 3. A three-dimensional matrix model of Q-value (i.e., the aforementioned Q-value table) is proposed to be constructed and integrated with the DNN model to obtain the optimal control action plan.

[0120] 4. Enables real-time locking of risky accounts.

[0121] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this disclosure.

[0122] The basic principles of this disclosure have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this disclosure are merely examples and not limitations, and should not be considered as essential features of each embodiment of this disclosure. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the scope of this disclosure to the necessity of employing the aforementioned specific details for implementation.

[0123] The block diagrams of devices, apparatuses, devices, and systems disclosed herein are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0124] Additionally, as used herein, the "or" used in a list of items beginning with "at least one" indicates a separate list, such that a list of, for example, "at least one of A, B, or C" means A or B or C, or AB or AC or BC, or ABC (i.e., A and B and C). Furthermore, the word "exemplary" does not imply that the described example is preferred or better than other examples.

[0125] It should also be noted that in the systems and methods of this disclosure, the components or steps can be decomposed and / or recombined. These decompositions and / or recombinations should be considered as equivalent solutions to this disclosure.

[0126] Various changes, substitutions, and modifications can be made to the technology described herein without departing from the teachings defined by the appended claims. Furthermore, the scope of the claims of this disclosure is not limited to the specific aspects of the processes, machines, manufactures, events, means, methods, and actions described above. Currently existing or later-developed processes, machines, manufactures, events, means, methods, or actions that perform substantially the same function or achieve substantially the same result as the corresponding aspects described herein can be utilized. Therefore, the appended claims include such processes, machines, manufactures, events, means, methods, or actions within their scope.

[0127] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of this disclosure. Therefore, this disclosure is not intended to be limited to the aspects shown herein, but rather to be carried out within the widest scope consistent with the principles and novel features disclosed herein.

[0128] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this disclosure to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations therein.

Claims

1. A user management method, characterized in that, include: Obtain multi-dimensional characteristic data of target users on the platform; The multi-dimensional feature data information is verified using the target super user identification SIM card bound to the target user to obtain credible multi-dimensional feature data information; Based on the credible multi-dimensional feature data and the pre-trained arbitrage risk model, the target risk status of the target user and the target risk management corresponding to the target risk status are obtained. Perform the target risk management on the target user.

2. The user management method according to claim 1, characterized in that, The multi-dimensional feature data information includes identity feature information, behavioral feature information, consumption feature information, customer service feature information, channel feature information, and device feature information.

3. The user management method according to claim 2, characterized in that, The step of verifying the multi-dimensional feature data information using the target super SIM card bound to the target user to obtain credible multi-dimensional feature data information includes: Use the target super SIM card to obtain the target hardware identifier of the target terminal device corresponding to the target user; Using the target super SIM card, the target hardware identifier is compared with the base hardware identifier corresponding to the target user stored in the target super SIM card to determine whether the target terminal device is a legitimate binding device; If the device is legally bound, the target hardware identifier is compared with the hardware identifier contained in the device feature information to verify the authenticity of the device feature information; When the authenticity requirement is met, the multi-dimensional feature data information is determined to be credible multi-dimensional feature data information.

4. The user management method according to claim 1, characterized in that, The arbitrage risk model includes a risk classification model and a Q-value table; The process of obtaining the target risk status of the target user and the corresponding target risk management based on the credible multi-dimensional feature data information and the pre-trained arbitrage risk model includes: The credible multi-dimensional feature data information is preprocessed; The preprocessed credible multi-dimensional feature data information is input into the risk classification model to obtain the target risk status of the target user. Based on the target risk status and the Q-value table, the target risk management corresponding to the target risk status is obtained. The Q-value table is used to describe the mapping relationship between the risk status, the risk management, and the predicted management benefits. The predicted management benefits are used to describe the predicted benefits after the risk management is implemented on the target user.

5. The user management method according to claim 4, characterized in that, The preprocessing includes at least one or more of the following: null value imputation, outlier handling, data standardization, formatting, binning coding, and similarity analysis.

6. The user management method according to claim 4, characterized in that, The step of performing target risk management on the target user based on the target risk status and the Q-value table includes: Based on the target risk status, find the maximum predicted control benefit matching the target risk status and the target risk control matching the maximum predicted control benefit from the Q value table; Perform the target risk management on the target user.

7. The user management method according to claim 4, characterized in that, The arbitrage risk model is trained in the following way: Obtain historical multi-dimensional feature data of each user on the platform; The historical multi-dimensional feature data information is verified using the super SIM card corresponding to each user to obtain the credible historical multi-dimensional feature data information of each user; Based on the different risk statuses of each user and the actual risk management corresponding to the different risk statuses, an initial Q value table is constructed, in which the Q value is 0; The trusted historical multi-dimensional feature data information of each user is divided into a training set and a test set; The training set is used to train a deep neural network (DNN) model to obtain the predicted risk status of each user; Based on the predicted risk status of each user and the initial Q-value table, update the parameters of the DNN model and the initial Q-value table; The DNN model with updated parameters and the updated initial Q-value table are tested using the test set. Based on the test results, the trained DNN model and Q-value table were determined, and the trained DNN model was designated as a risk classification model.

8. The user management method according to claim 7, characterized in that, The historical multi-dimensional feature data information includes at least: Identity characteristics, behavioral characteristics, consumption characteristics, customer service characteristics, channel characteristics, and device characteristics.

9. The user management method according to claim 7, characterized in that, The verification of the historical multi-dimensional feature data information using the super SIM card corresponding to each user includes: For each user, the corresponding Super SIM card performs the following operations: Obtain the hardware identifier of the user's currently used terminal device; The hardware identifier is compared with the base hardware identifier corresponding to the user stored in the super SIM card to determine whether the terminal device is a legitimate bound device.

10. The user management method according to claim 9, characterized in that, The verification of the historical multi-dimensional feature data information using the super SIM card corresponding to each user also includes: If the device is legally bound, the hardware identifier is compared with the hardware identifier contained in the device feature information to verify the authenticity of the device feature information.

11. The user management method according to claim 10, characterized in that, The acquisition of reliable historical multi-dimensional feature data information includes: When the device feature information corresponding to the user meets the authenticity requirement, the historical multi-dimensional feature data information of each user is determined to be reliable historical multi-dimensional feature data information.

12. The user management method according to claim 10, characterized in that, The method further includes: If the device is illegally bound, the user is marked as a risk user, and the user's historical multi-dimensional feature data information is removed.

13. The user management method according to claim 8, characterized in that, The step of dividing the credible historical multi-dimensional feature data information into a training set and a test set includes: Preprocess the trusted historical multi-dimensional feature data information of each user; The preprocessed trusted historical multi-dimensional feature data of each user is divided into a training set and a test set.

14. The user management method according to claim 13, characterized in that, The preprocessing includes at least one or more of the following: null value imputation, outlier handling, data standardization, formatting, binning coding, and similarity analysis.

15. The user management method according to claim 13, characterized in that, The method further includes: Based on the time dimension, the trusted historical multi-dimensional feature data information of each user is divided into new data and old data; Based on the logical relationship between the new data and the old data, the consistency of each user's behavior is confirmed; When the behaviors of each user are consistent, the preprocessed trusted historical multi-dimensional feature data of each user is divided into a training set and a test set.

16. The user management method according to claim 4, characterized in that, The method further includes: Based on the predicted management benefits and the actual management benefits corresponding to the predicted management benefits, update the risk classification model and the Q-value table.

17. A user control device, characterized in that, include: The multi-dimensional feature data acquisition module is used to acquire multi-dimensional feature data information of the target user; A trusted multi-dimensional feature data information acquisition module is used to verify the multi-dimensional feature data information using the target super user identification SIM card bound to the target user, and to acquire trusted multi-dimensional feature data information. The risk status and risk management determination module is used to obtain the target risk status of the target user and the target risk management corresponding to the target risk status based on the credible multi-dimensional feature data information and the pre-trained arbitrage risk model. The risk control execution module is used to perform the target risk control on the target user.

18. An electronic device, characterized in that, include: Memory, used to store computer-readable instructions; as well as A processor for executing the computer-readable instructions, causing the electronic device to perform the user control method as described in any one of claims 1 to 16.

19. A readable storage medium for storing computer-readable instructions, characterized in that, When the computer-readable instructions are executed by the processor, the processor performs the user management method as described in any one of claims 1 to 16.

20. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the user management method as described in any one of claims 1 to 16.

Images