Risk assessment methods, devices, electronic equipment, and readable storage media for objects

Abstract

This application provides a method, apparatus, electronic device, and readable storage medium for risk assessment of an object, relating to the field of computer technology. The method includes: acquiring relevant data for a target object across multiple dimensions, including first object data of the target object and transaction data between the target object and associated objects, wherein associated objects include objects that have transactions with the target object; for each dimension of the relevant data, determining a risk assessment result for the target object corresponding to that dimension based on a first risk prediction model corresponding to the data of that dimension; and determining a comprehensive risk assessment result for the target object based on the risk assessment results for each dimension using a second risk prediction model, thereby performing risk management on the target object based on the risk assessment results for each dimension and the comprehensive risk assessment result. This application improves the accuracy of the comprehensive risk assessment result and effectively reduces risk.

Description

Technical Field

[0001] This application relates to the field of computer technology, and more specifically, to a method, apparatus, electronic device, and readable storage medium for risk assessment of an object. Background Technology

[0002] To ensure transaction security, merchants need risk management during transactions. Currently, the main risk control technology involves disabling a merchant's ability to conduct transactions if they are identified as malicious before or during a transaction and subsequently penalized. However, malicious merchants (those suspected of operating in non-compliant industries, i.e., the targets of risk control) may have already transferred significant funds, causing economic losses. Furthermore, for some new transaction methods (such as corporate payments), there is a lack of risk control experience and overall risk control for these methods (e.g., when funds need to flow from a merchant to an external entity, the overall risk control for the merchant is poorly managed). In other words, the risk level of merchants cannot be accurately measured, and the risk management methods employed are relatively outdated, leaving many potential unknown risks. Summary of the Invention

[0003] This application provides a method, apparatus, electronic device, and readable storage medium for risk assessment of an object, which can perform risk assessment on a target object.

[0004] On the one hand, embodiments of this application provide a risk assessment method for an object, the method comprising:

[0005] Obtain relevant data for the target object from multiple dimensions. This data includes the target object's primary object data and transaction data between the target object and related objects. Related objects include objects that have transactions with the target object.

[0006] For each dimension of the relevant data, based on the first risk prediction model corresponding to the data of that dimension, the risk assessment result of the target object corresponding to that dimension is determined;

[0007] Based on the risk assessment results of each dimension corresponding to the target object, the comprehensive risk assessment result of the target object is determined through the second risk prediction model, and the target object is managed and controlled based on the risk assessment results of each dimension and the comprehensive risk assessment result.

[0008] On the other hand, embodiments of this application provide a risk assessment apparatus for an object, the apparatus comprising:

[0009] The data acquisition module is used to acquire relevant data of the target object in multiple dimensions. The relevant data in multiple dimensions includes the first object data of the target object and the transaction data between the target object and related objects. Among them, related objects include objects that have transactions with the target object.

[0010] The risk assessment result determination module is used to determine the risk assessment result of the target object corresponding to that dimension based on the first risk prediction model corresponding to the data of each dimension in the relevant data.

[0011] The risk management and control module is used to determine the comprehensive risk assessment result of the target object based on the risk assessment results of each dimension corresponding to the target object through the second risk prediction model, and to carry out risk management and control processing on the target object based on the risk assessment results of each dimension and the comprehensive risk assessment result.

[0012] In another aspect, embodiments of this application provide an electronic device, including a processor and a memory: the memory is configured to store a computer program, which, when executed by the processor, causes the processor to perform the risk assessment method for the aforementioned object.

[0013] In another aspect, embodiments of this application provide a computer-readable storage medium for storing a computer program that, when run on a computer, enables the computer to execute the risk assessment method for the aforementioned object.

[0014] The beneficial effects of the technical solutions provided in this application are:

[0015] In this embodiment of the application, when performing risk management on a target object, relevant data of the target object can be obtained, and the risk assessment result corresponding to that dimension of the target object can be determined based on the relevant data of the target object. That is, the risk assessment result of the target object can be measured based on multiple dimensions, and the comprehensive risk assessment result of the target object can be determined based on the risk assessment results of each dimension corresponding to the target object. At this time, the determined comprehensive risk assessment result integrates the measurement results of multiple dimensions, which can further improve the accuracy of the determined comprehensive risk assessment result. Accordingly, when performing risk management on the target object based on the risk assessment results of each dimension and the comprehensive risk assessment result, the risk can be effectively reduced. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below.

[0017] Figure 1aA schematic diagram of the structure of a system to which the risk assessment method for the object provided in the embodiments of this application is applicable;

[0018] Figure 1b A flowchart illustrating a risk assessment method for an object provided in an embodiment of this application;

[0019] Figure 2 A schematic diagram illustrating the partitioning of training samples provided in an embodiment of this application;

[0020] Figure 3 A data illustration provided for an embodiment of this application;

[0021] Figure 4 This is a schematic diagram of merchant-related data provided in an embodiment of this application;

[0022] Figure 5 A schematic diagram illustrating the construction of each first risk prediction model provided in an embodiment of this application;

[0023] Figure 6 A scenario diagram provided for an embodiment of this application;

[0024] Figure 7 This is a schematic diagram of the structure of a risk assessment device for an object provided in an embodiment of this application;

[0025] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0026] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0027] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this application means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0028] In this application embodiment, artificial intelligence technology can be used to determine the risk assessment result of the target object, and risk management and control measures can be carried out on the target object based on the determined comprehensive risk assessment result, thereby reducing some potential unknown risks.

[0029] Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that utilize digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive technology within computer science that attempts to understand the essence of intelligence and produce new intelligent machines that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to possess perception, reasoning, and decision-making capabilities.

[0030] Artificial intelligence (AI) is a comprehensive discipline encompassing a wide range of fields, including both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies primarily include computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0031] In this application embodiment, the determination of the risk assessment result of the object involves machine learning technology, a branch of artificial intelligence software technology. Machine learning (ML) is a multidisciplinary field involving probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory, and many other disciplines. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to endow computers with intelligence; its applications span all areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and instructional learning.

[0032] Optionally, in this application embodiment, the relevant data of the target object can be stored based on a blockchain. Blockchain is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include a blockchain underlying platform, a platform product service layer, and an application service layer.

[0033] The underlying blockchain platform can include processing modules such as user management, basic services, smart contracts, and operational monitoring. The user management module is responsible for managing the identity information of all blockchain participants, including maintaining public and private key generation (account management), key management, and maintaining the correspondence between user real identities and blockchain addresses (access management). Furthermore, under authorization, it monitors and audits transactions of certain real identities and provides risk control rule configuration (risk control audit). The basic services module is deployed on all blockchain node devices to verify the validity of business requests. After consensus is reached on valid requests, they are recorded in storage. For a new business request, the basic services first perform interface adaptation parsing and authentication (interface adaptation), and then encrypt the business information using a consensus algorithm (consensus management). After encryption, the data is transmitted completely and consistently to the shared ledger (network communication) and recorded and stored. The smart contract module is responsible for contract registration, issuance, triggering, and execution. Developers can define contract logic using a programming language and publish it to the blockchain (contract registration). According to the contract terms, the key or other events are invoked to trigger execution and complete the contract logic. It also provides functions for contract upgrades and cancellations. The operation monitoring module is mainly responsible for deployment, configuration modification, contract settings, cloud adaptation, and real-time status visualization during product launch, such as alarms, monitoring network conditions, and monitoring the health status of node devices.

[0034] The platform's product service layer provides the basic capabilities and implementation frameworks for typical applications. Developers can leverage these basic capabilities, along with the specific characteristics of their business needs, to implement blockchain-based business logic. The application service layer provides blockchain-based application services to business stakeholders.

[0035] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0036] To better understand and explain the methods provided in the optional embodiments of this application, several terms involved in this application will first be introduced and explained:

[0037] Commercial payments: Transactions made by merchant users using the application to make payments.

[0038] Enterprise payment: A mobile payment product for a specific transaction scenario. Through this product, merchant B can proactively provide funds to user C.

[0039] Malicious merchants: Merchants suspected of engaging in industries that do not meet the requirements, i.e., merchants targeted by risk control measures;

[0040] Withdrawal crackdowns include, but are not limited to, intercepting the current corporate payment transaction and penalizing the merchant's ability to make corporate payments.

[0041] Same entity: Merchants with the same name

[0042] Optionally, the method provided in this application embodiment can be executed by a server or a terminal device. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The terminal device can be a smartphone, tablet, laptop, desktop computer, etc., but is not limited to these. The terminal device and the server can be directly or indirectly connected via wired or wireless communication, which is not limited herein. When the method is executed by the server, after determining the risk assessment results and comprehensive risk assessment results corresponding to the target object in each dimension, the risk assessment results and comprehensive risk assessment results corresponding to the target object can be provided to the user through the terminal device corresponding to the server.

[0043] As an optional implementation, Figure 1a The diagram shows a structural schematic of a system to which the risk assessment method for the object provided in this application is applicable, such as... Figure 1aAs shown, the system may include a server 10, terminal devices 11, and a database 12 connected to the server. The server 10 and terminal devices 11 are connected via a network 20. Optionally, the terminal device 11 can be the terminal device of any target object to be evaluated. The server 10 deploys a first risk prediction model and a second risk prediction model corresponding to the data of each dimension. The database 12 stores relevant data for multiple dimensions. For example, in the application scenario of merchant enterprise payment, the target object is the merchant. When the merchant conducts enterprise payment business, a risk assessment can be performed on the merchant. At this time, the server 10 can obtain relevant data for multiple dimensions of the merchant from the database 12, and then, based on the first risk prediction model corresponding to the data of each dimension, determine the risk assessment result of the merchant for each dimension. Then, based on the risk assessment result of the merchant for each dimension, the second risk prediction model is used to determine the comprehensive risk assessment result of the merchant. Optionally, the data for each dimension of the merchant may include, for example, Figure 4 The data shown in the example may include the merchant's name, transaction behavior information, etc., and the first risk prediction model corresponding to each dimension of data may refer to, for example... Figure 5 The models shown are pre-trained subject-based investigation status models, registration graph models, abnormal transaction behavior recognition models, and abnormal withdrawal behavior recognition models.

[0044] Furthermore, after determining the merchant's comprehensive risk assessment result, server 10 can perform risk control processing on the merchant based on the risk assessment results of each dimension and the comprehensive risk assessment result. If the merchant is identified as an untrustworthy merchant, the enterprise payment transaction for that merchant will be blocked; if the merchant is identified as a trustworthy merchant, the enterprise payment transaction for that merchant will be executed. To facilitate further management of the merchant by the user, the risk control processing result, as well as the risk assessment results of each dimension and the comprehensive risk result corresponding to the merchant, can be sent to terminal device 11. Terminal device 11 will then provide the risk control processing result, as well as the risk assessment results of each dimension and the comprehensive risk result corresponding to the merchant, to the user (such as the enterprise's manager or relevant person in charge).

[0045] It is understood that the application scenarios applicable to the method provided in this application embodiment include, but are not limited to, risk assessment scenarios for merchants' enterprise payment business. Any other scenario used to assess the risk assessment results of a target object is also applicable. For example, the method provided in this application embodiment can be applied in a secure payment application scenario. In this case, when a user (i.e., the paying user) makes a payment, relevant data from multiple dimensions of the fund inflow user can be obtained, and the data from each dimension can be input into the corresponding first risk prediction model to determine the risk assessment results of the fund inflow user for each dimension. Then, based on the risk assessment results of the fund inflow user for each dimension, a second risk prediction model is used to determine the comprehensive risk assessment result of the fund inflow user. If the comprehensive risk assessment result of the fund inflow user is deemed unreliable, the transaction can be closed to ensure the safety of the paying user's funds.

[0046] Figure 1b A flowchart illustrating a risk assessment method for an object provided in an embodiment of this application is shown. Figure 1b As shown, the method includes:

[0047] Step S101: Obtain relevant data of the target object in multiple dimensions. The relevant data in multiple dimensions includes the first object data of the target object and the transaction data between the target object and related objects. The related objects include objects that have transactions with the target object.

[0048] Here, the target object refers to the object that requires risk assessment. By conducting a risk assessment on the target object, it can be determined whether the target object poses a risk, and thus, based on the assessment results, it can be determined whether risk management of the target object is necessary. The target object may vary depending on the application scenario. This application embodiment does not limit the specific type of the target object; it can be any object that requires risk assessment and monitoring, such as a merchant, a general user, or a user or merchant with specific / designated business capabilities, etc. For example, the target object could be a merchant with corporate payment capabilities.

[0049] The relevant data of the target object refers to data associated with the target object, such as the first object data of the target object, transaction data between the target object and related objects, and related objects can include objects that have transactions with the target object. Optionally, the relevant data of the target object may also include the second object data of related objects.

[0050] The object data of the target object (i.e., the first object data mentioned above) can refer to data related to the target object itself, and this application embodiment does not specifically limit it. This may include, but is not limited to, the name of the target object (e.g., if the target object is a merchant, the name of the target object can refer to the merchant's business name), complaint information of the target object (e.g., if the target object is a merchant, complaint information can refer to complaints involving that merchant), status information of the target object (e.g., when the target object is a merchant, the status information of the target object can refer to whether the target object has been marked as a malicious merchant), registration information of the target object (e.g., when the target object is a merchant, the status information of the target object can refer to the merchant's merchant number registration information), rating information of the target object, etc. The transaction data between the target object and related objects can refer to the data involved in the transaction between the target object and related objects, and this application embodiment does not specifically limit this. For example, when the target object is a merchant, the transaction data can refer to normal transaction behavior data, abnormal transaction behavior data, transaction complaint data, normal withdrawal behavior data, abnormal withdrawal behavior data, and text data during withdrawal, etc.

[0051] The second object data of the associated object can refer to data related to the associated object itself. The specific data that the second object data of the associated object may include can be similar to the data types that the first object data of the target object may include. For details, please refer to the explanation of the first object data above, which will not be repeated here.

[0052] In this embodiment of the application, since the associated object is an object that has a transaction with the target object, there will be some correlation between the associated object and the target object. Correspondingly, the second object data of the associated object can also reflect some attributes of the target object in essence. If the relevant data used to determine the risk assessment results and comprehensive risk assessment results of the target object in each dimension can also include the second object data of the associated object, the data on which it is based will be richer, and the risk assessment results and comprehensive risk assessment results of the target object in each dimension determined based on the relevant data will also be more accurate.

[0053] Step S102: For each dimension of the relevant data, based on the first risk prediction model corresponding to the dimension of the data, determine the risk assessment result of the target object corresponding to that dimension.

[0054] This application does not limit the method of dividing data dimensions; it can be a coarse-grained or fine-grained division, which can be configured according to actual needs, such as dividing according to data type. In the embodiments of this application, for the relevant data of the target object, since the relevant data can contain many different categories of data, such as according to different objects, the relevant data can include the data of the target object (i.e., the data of the first object), the transaction data between the target object and the associated object, the data of the transaction object, etc. For each category of data, one category of data can be used as one dimension of data, and the data of one category of data can be further divided, that is, the data of one category of data can also be divided into multiple dimensions of data. For example, when the target object is a merchant, the information belonging to the static category of merchant can be divided into merchant name, merchant registration information, etc., and at this time, the merchant name and merchant registration information can be divided into multiple dimensions of data.

[0055] For data of different dimensions, the attributes of the target object can be represented from different dimensions. In order to achieve more granular and accurate risk assessment of the data, in this embodiment of the application, for data of different dimensions, the model corresponding to each dimension of the data can be trained separately. When conducting risk assessment of the target object, the model corresponding to each dimension of the data can be used to conduct risk assessment of the target object.

[0056] Optionally, the specific format of the risk assessment result corresponding to each dimension of data can be pre-configured, and this embodiment of the application does not limit it. For example, different risk levels can be pre-configured, and the first risk prediction model can determine the probability of the data in that dimension corresponding to each risk level based on the input data and output it as the risk assessment result. The first risk prediction model for each dimension can be implemented using models such as decision trees (label propagation) or random forests.

[0057] Step S103: Based on the risk assessment results of each dimension corresponding to the target object, determine the comprehensive risk assessment result of the target object through the second risk prediction model, and carry out risk control processing on the target object based on the risk assessment results of each dimension and the comprehensive risk assessment result.

[0058] To ensure a more accurate risk assessment of the target object, after obtaining detailed risk assessment results for each dimension of the target object, a comprehensive risk assessment can be performed on the target object using a second risk prediction model based on these results. The form in which the comprehensive risk assessment result is presented is not limited in this embodiment. For example, different risk levels can be pre-configured. Based on the risk assessment results for each dimension of the target object, the probability of the target object corresponding to different risk levels can be determined, and the risk level with the highest probability is taken as the comprehensive risk assessment result. When both the risk assessment results for each dimension and the comprehensive risk assessment result use risk level representation, the granularity of the risk assessment results for each dimension and the comprehensive risk assessment result can be the same or different; this is not limited in this embodiment. For example, they can be divided into 5 different levels, or the risk assessment results for each dimension of the target object can be divided into 7 different levels, and the comprehensive risk assessment result can be divided into 5 different levels.

[0059] Accordingly, after knowing the risk assessment results of each dimension and the comprehensive risk assessment of the target object, risk management and control measures can be implemented for the target object based on these results. The specific implementation method for risk management and control of the target object can be pre-configured, and is not limited in the embodiments of this application.

[0060] For example, when the target is merchants, they can be categorized into five levels—A, B, C, D, and E—based on the risk assessment results from various dimensions and the overall risk assessment, forming a five-level risk classification system. Each risk classification system corresponds to a different risk management approach. For instance, A-level merchants might be subject to withdrawal suspension penalties, B-level merchants to withdrawal limits, C-level merchants to withdrawal interception, while E-level merchants are protected using a lenient policy. Correspondingly, risk management can be implemented for each merchant based on the risk management approach corresponding to their respective risk classification system.

[0061] In this embodiment of the application, when performing risk management on a target object, relevant data of the target object can be obtained, and the risk assessment result corresponding to that dimension of the target object can be determined based on the relevant data of the target object. That is, the risk assessment result of the target object can be measured based on multiple dimensions, and the comprehensive risk assessment result of the target object can be determined based on the risk assessment results of each dimension corresponding to the target object. At this time, the determined comprehensive risk assessment result integrates the measurement results of multiple dimensions, which can further improve the accuracy of the determined comprehensive risk assessment result. Accordingly, when performing risk management on the target object based on the risk assessment results of each dimension and the comprehensive risk assessment result, the risk can be effectively reduced.

[0062] In optional embodiments of this application, the associated objects include target associated objects and non-target associated objects. The target associated object refers to an object that has a transaction with the target object in the target business. The transaction data includes first transaction data between the target object and the target associated object, and second transaction data between the target object and the non-target associated object.

[0063] Based on the risk assessment results of each dimension corresponding to the target object, the comprehensive risk assessment result of the target object is determined, including:

[0064] Based on the risk assessment results of each dimension corresponding to the target object, the comprehensive risk assessment result corresponding to the target business is determined.

[0065] In this context, "target related objects" refers to entities that have transacted with the target object in the target business. For example, when the target object is a merchant, and the target business is the merchant's corporate payment business, then the target related objects refer to merchants or users who have transacted with that merchant in corporate payment business. Non-target related objects refer to merchants or users who have transacted with that merchant in non-corporate payment business, which can be one or more transactions. Correspondingly, when determining the comprehensive risk assessment result of the target object based on the risk assessment results of each dimension corresponding to the target object, this can refer to determining the comprehensive risk assessment result of the target object corresponding to the target business. For example, it can refer to the comprehensive risk assessment result of the target object corresponding to corporate payment business.

[0066] In optional embodiments of this application, the comprehensive risk assessment result of the target object is determined based on the risk assessment results of each dimension corresponding to the target object, including:

[0067] Based on the second risk prediction model, feature extraction is performed on the risk assessment results of each dimension corresponding to the target object to obtain the fused feature vector corresponding to each dimension.

[0068] Based on the fused feature vectors, the comprehensive risk assessment results of the target object are determined.

[0069] Optionally, this embodiment also includes a second risk prediction model. After knowing the risk assessment results for each dimension of the target object, features can be extracted from the risk assessment results for each dimension based on the second risk prediction model to obtain a fused feature vector for each dimension. Then, based on the extracted fused feature vector, the comprehensive risk assessment result of the target object is determined. Since the feature vector used to determine the comprehensive risk assessment result of the target object is obtained by feature extraction based on the risk assessment results for each dimension of the target object, this feature vector includes multi-dimensional information, thus the determined comprehensive risk assessment result can be more accurate.

[0070] In optional embodiments of this application, the target object corresponds to the risk assessment results of each dimension or the comprehensive risk assessment results, including the risk assessment level. The method further includes at least one of the following:

[0071] Provide users with risk assessment results corresponding to the target objects across various dimensions;

[0072] Provide users with comprehensive risk assessment results.

[0073] Optionally, after obtaining the risk assessment results and comprehensive risk assessment results for the target object corresponding to each dimension, the risk assessment results and comprehensive risk assessment results for the target object corresponding to each dimension can be provided to the user. For example, the risk assessment results and comprehensive risk assessment results for the target object corresponding to each dimension can be displayed on the terminal device. At this time, the user can know the risk assessment results of the target object at any time based on the displayed information, and perform real-time risk management of the target object according to the risk assessment results to meet the user's actual needs.

[0074] Specifically, when the target object corresponds to risk assessment results across various dimensions or the comprehensive risk assessment result includes a risk assessment level, relevant attribute information of the target object belonging to each risk assessment level can be obtained and provided to the user to enable better risk management of the target object. Optionally, the relevant attribute information may include the number of target objects belonging to each risk assessment level, transaction information of the target object, and transaction information of related objects of the target object.

[0075] In optional embodiments of this application, the method further includes:

[0076] For any dimension, if there are multiple target objects, obtain the risk assessment result for each target object corresponding to that dimension;

[0077] Acquire and monitor transaction data of target objects belonging to the specified risk assessment results in real time;

[0078] When the transaction data belonging to the target object of the specified risk assessment result meets the optimization conditions, the first risk prediction model corresponding to the data in that dimension is optimized.

[0079] Optionally, when there are multiple target objects requiring risk monitoring, for any dimension, data corresponding to that dimension for each target object can be obtained, and a risk assessment result for each target object corresponding to that dimension can be obtained based on the first risk prediction model corresponding to that dimension. Further, transaction data of the target objects can be obtained and statistically analyzed to obtain transaction data of target objects belonging to a specified risk assessment result, and this data can be monitored in real time. If the transaction data of target objects belonging to a specified risk assessment result meets the optimization conditions, the first risk prediction model corresponding to the data of that dimension can be optimized. The specific types of the target object's transaction data and the specified risk assessment result can be pre-configured; for example, the target object's transaction data could be the target object's withdrawal amount, and the specified risk assessment result could be a malicious merchant, etc. This embodiment of the application does not limit this.

[0080] In one example, assuming the target is a merchant, the target's transaction data is the amount withdrawn by the target, and the risk assessment results include malicious merchants, unknown merchants, and trusted merchants, the risk assessment result is specified as malicious merchant. At this point, for any given dimension, data corresponding to that dimension for each merchant can be obtained. Based on the first risk prediction model corresponding to that dimension, the risk assessment result for each merchant corresponding to that dimension (i.e., whether the merchant is a malicious merchant, an unknown merchant, or a trusted merchant) can be obtained. Further, the withdrawal amount of merchants with a risk assessment result of malicious is obtained, and the proportion of the withdrawal amount of malicious merchants to the total withdrawal amount of all merchants is monitored in real time. When the proportion surges or continues to increase over a period of time, it is considered that the optimization conditions have been met, indicating a change in the malicious merchant's methods. In this case, the first risk prediction model corresponding to the data for that dimension needs to be optimized. When the specified risk assessment result is an unknown merchant, the withdrawal amount of merchants with a risk assessment result of unknown is obtained, and the proportion of the withdrawal amount of unknown merchants to the total withdrawal amount of all merchants is monitored in real time. When the proportion surges or continues to increase over a period of time, it is considered that the optimization conditions have been met, indicating that the malicious merchant may have shifted to an unknown merchant, meaning that the existing risk assessment system needs optimization. In this case, the first risk prediction model corresponding to the data for that dimension needs to be optimized.

[0081] Optionally, the method of optimizing the first risk prediction model is not limited in this application embodiment. For example, the first risk prediction model can be retrained by re-acquiring training data to achieve the purpose of optimization.

[0082] In an optional embodiment of this application, the first risk prediction model corresponding to each dimension is trained in the following manner:

[0083] Obtain the training dataset, which includes multiple first training samples. Each first training sample includes a sample object corresponding to a sub-sample of each dimension. A sub-sample of a dimension includes sample-related data of that dimension and risk label corresponding to that dimension.

[0084] For each dimension, the initial risk prediction model for that dimension is iteratively trained based on the subsamples of that dimension in each first training sample to obtain the first risk prediction model corresponding to that dimension.

[0085] Optionally, the first risk prediction model for each dimension can be obtained by iteratively training the initial risk prediction model for each dimension based on the acquired training dataset. The training dataset includes multiple first training samples, each containing a sample object corresponding to a sub-sample for each dimension. A sub-sample for a dimension includes sample-related data for that dimension and the corresponding risk label. Further, for each dimension, the first risk prediction model for that dimension can be obtained by iteratively training the initial risk prediction model for that dimension using the sub-samples from each of the first training samples.

[0086] In an optional embodiment of this application, each first training sample further includes a comprehensive risk label corresponding to the sample object; the second risk prediction model is trained in the following manner:

[0087] Using the first risk prediction model for each dimension, obtain the risk assessment results for each first training sample corresponding to each dimension;

[0088] Based on the risk assessment results and comprehensive risk labels corresponding to each dimension of the first training sample, each second training sample is obtained, wherein each second training sample includes a sample object corresponding to the risk assessment results and comprehensive risk labels of each dimension;

[0089] The initial second risk prediction model is iteratively trained based on each second training sample to obtain the second risk prediction model.

[0090] Each first training sample also includes a comprehensive risk label corresponding to the sample object, which represents the true risk assessment result of the sample object. Optionally, after iteratively training the initial risk prediction model for each dimension based on the sub-samples of each dimension in each first training sample to obtain the first risk prediction model corresponding to each dimension, for each dimension, the sub-samples corresponding to that dimension can be input into the first risk prediction model trained for that dimension, and each resulting first training sample corresponds to the risk assessment result for that dimension. Further, a sample object corresponding to the risk assessment result of each dimension and the corresponding comprehensive risk label can be combined into a second training sample, and then the initial second risk prediction model can be iteratively trained based on the obtained second training samples to obtain the second risk prediction model.

[0091] In an optional embodiment of this application, for each dimension, an initial risk prediction model for that dimension is trained based on sub-samples of that dimension in each first training sample to obtain a first risk prediction model corresponding to that dimension, including:

[0092] Each subsample of this dimension is divided into multiple training subsets. At least two datasets are constructed based on the multiple training subsets. Each dataset includes at least two training subsets. One training subset from the at least two training subsets is used as the prediction set. The training subsets other than the prediction set are used as the training set. The prediction sets are different for different datasets.

[0093] The initial risk prediction model for that dimension is trained based on the training set in each dataset to obtain the model parameters for each dataset.

[0094] The model parameters of the first risk prediction model are determined based on the model parameters corresponding to each training set, and the first risk prediction model corresponding to this dimension is obtained.

[0095] Using the first risk prediction model for each dimension, obtain the risk assessment results for each first training sample corresponding to each dimension, including:

[0096] For each dimension, each prediction set of that dimension is passed through a first risk prediction model based on the model parameters corresponding to that prediction set to obtain the risk assessment result for each first training sample corresponding to each dimension.

[0097] Optionally, for each dimension's subsamples, the subsamples of that dimension can be randomly divided into multiple training subsets, and at least two datasets can be constructed based on these training subsets. Then, each dataset can be used as training data to train the initial risk prediction model for that dimension. Optionally, each dataset can include at least two training subsets. One training subset from the at least two training subsets is used as the prediction set, and the remaining training subsets are used as the training set. Then, the initial risk prediction model for that dimension is trained based on the training set in each dataset, obtaining the model parameters corresponding to each dataset. Optionally, the number of datasets equals the number of training subsets. Different datasets have different prediction sets; that is, each test subset is used as a prediction set once. Further, after obtaining the model parameters corresponding to each dataset, the model parameters of the first risk prediction model corresponding to that dimension can be determined based on the model parameters corresponding to each dataset, thus obtaining the first risk prediction model corresponding to that dimension.

[0098] Accordingly, when obtaining the risk assessment results of each first training sample corresponding to each dimension through the first risk prediction model of each dimension, for each dimension, each prediction set can be input into the first risk prediction model based on the model parameters corresponding to that prediction set to obtain the risk assessment results of each first training sample corresponding to each dimension.

[0099] In optional embodiments of this application, the method further includes:

[0100] Obtain the test dataset corresponding to the training dataset, and test the first risk prediction model for each dimension based on the test dataset;

[0101] Each test sample in the test dataset is input into the first risk prediction model based on the model parameters corresponding to each training subset, to obtain at least two first risk assessment results for each test sample;

[0102] For each test sample, the at least two first risk assessment results corresponding to the test sample are fused to obtain the second risk assessment result corresponding to the test sample;

[0103] Based on the second risk assessment results corresponding to each test sample, the test dataset corresponding to the second risk prediction model is obtained.

[0104] In practical applications, after iteratively training the initial second risk prediction model based on the obtained second training samples to obtain the second risk prediction model, a corresponding test dataset can be obtained to test the second risk prediction model. Specifically, when obtaining the test dataset corresponding to the second risk prediction model, the test dataset corresponding to the training dataset can be obtained, which includes test samples corresponding to each dimension. Further, for each dimension's test sample, the test sample can be input into the corresponding first risk prediction model based on the model parameters corresponding to each training subset for that dimension, obtaining at least two first risk assessment results for that test sample. Correspondingly, for each dimension's test sample, the at least two first risk assessment results corresponding to that test sample are fused to obtain the second risk assessment result corresponding to that test sample. For example, the at least two first risk assessment results corresponding to that test sample can be averaged, and the resulting average is used as the second risk assessment result corresponding to that test sample. Further, based on the second risk assessment results corresponding to the test samples for each dimension, a test dataset for testing the second risk prediction model can be obtained. The specific implementation method for the second risk assessment results based on the test samples corresponding to each dimension can be configured according to actual needs. For example, the second risk assessment results corresponding to the test samples corresponding to each dimension can be concatenated to obtain the test dataset corresponding to the second risk prediction model.

[0105] Optionally, to better understand the process of training the first and second risk prediction models for each dimension, a detailed explanation is provided. In this example, it is assumed that there are Q dimensions of data corresponding to the first risk prediction modules, namely model_1, model_2, ..., model_Q. The following explanation uses the training of the first risk prediction module (i.e., model_1) corresponding to one dimension of data and the training of the second risk prediction module as examples.

[0106] First, a training dataset can be obtained, which includes multiple first training samples. Each first training sample includes a sample object corresponding to a subsample of each dimension. A subsample of a dimension includes sample-related data for that dimension and the risk label corresponding to that dimension.

[0107] Furthermore, such as Figure 2As shown, for each dimension, the subsamples of that dimension in each first training sample can be divided into 5 training subsets (subset 1 to subset 5 respectively). Based on these 5 training subsets, 5 datasets are formed, and 5 training sessions are performed on each dataset (e.g., j=1 to j=5 represent the 1st to 5th training sessions respectively). For the first training iteration (j=1), subset 5 can be used as the prediction set, and subsets 1 to 4 as the training set. The initial risk prediction model is then trained using the training set to obtain Model 1 (the model parameters corresponding to the dataset mentioned earlier). Subset 5 (the prediction set) is then input into Model 1 to obtain the prediction result for the first training iteration. For the second training iteration (j=2), subset 4 can be used as the prediction set, and subsets 1 to 3 and subset 5 as the training set. The initial risk prediction model is then trained using the training set to obtain Model 2. Subset 4 (the prediction set) is then input into Model 2 to obtain the prediction result for the second training iteration. For the third training iteration (j=3), subset 3 can be used as the prediction set, and subsets 1, 2, 4 to 5 as the training set. The initial risk prediction model is then trained using the training set to obtain Model 3. Subset 3 (the prediction set) is then input into Model 2 to obtain the prediction result for the second training iteration. Inputting the data into Model 3 yields the prediction results for the third training iteration. For the fourth training iteration (j=4), subset 2 can be used as the prediction set, and subsets 1, 3 to 5 can be used as the training set. The initial risk prediction model is then trained based on the training set to obtain Model 4. Subset 2 (the prediction set) is then input into Model 4 to obtain the prediction results for the fourth training iteration. For the fifth training iteration (j=5), subset 1 can be used as the prediction set, and subsets 2 to 5 can be used as the training set. The initial risk prediction model is then trained based on the training set to obtain Model 5. Subset 1 (the prediction set) is then input into Model 5 to obtain the prediction results for the fifth training iteration. At this point, five prediction results for this dimension can be obtained (i.e., each first training sample in the previous text corresponds to the risk assessment results for each dimension, P_1 in the figure). Simultaneously, the first risk prediction module for this dimension can be obtained based on the obtained Models 1 to 5.

[0108] Furthermore, the test dataset corresponding to that dimension in the training dataset (i.e., the test data in the figure) can be obtained. The test dataset includes each test sample, and each test sample includes test data corresponding to that dimension for the sample object. Then, the test data is input into the trained models 1 to 5 respectively to obtain 5 initial test results corresponding to that dimension (i.e., the first risk assessment result). Then, the average of the 5 initial test results corresponding to that dimension is processed to obtain the test result corresponding to that dimension (i.e., the second risk assessment result, T_1 in the figure). Based on the same training method, models 2...Q can be obtained, as well as the prediction results (i.e., P_2...P_Q) and test results (i.e., T_2...T_Q) corresponding to each of models 2...Q.

[0109] Correspondingly, such as Figure 3 As shown, the prediction results (i.e., P_1, P_2, ..., P_Q) corresponding to each of Model_1, Model_2, ..., Model_Q can be concatenated to obtain the second training samples used to train the second risk prediction model. The test results (i.e., T_1, T_2, ..., T_Q) corresponding to each of Model_1, Model_2, ..., Model_Q can be concatenated to obtain the test dataset used to test the second risk prediction model. Then, the initial second risk prediction model is iteratively trained based on the obtained second training samples to obtain the second risk prediction model. Finally, the second risk prediction model is tested based on the obtained test dataset.

[0110] The aforementioned sample dataset can be obtained from a high-performance server with Python (a computer programming language) and data dependency components installed. The sample objects in the training dataset include positive sample objects and negative sample objects. Positive sample objects can refer to merchants of a certain type of malicious activity that have been identified in the past, while negative sample objects can refer to trusted merchants that have been identified in the past. Furthermore, when training the first risk prediction model and the second risk prediction model based on the training dataset, a k-fold cross-validation method (k=5 in this embodiment) is used to effectively prevent model overfitting.

[0111] Optionally, to better understand the method provided in the embodiments of this application, the method will be described below in conjunction with a specific application scenario of risk assessment for enterprise payment business by merchants. In this example, the target object is the merchant, the non-target associated objects of the target object include the merchant's transaction users, and the target associated objects of the target object include the merchant's fund-contributing users.

[0112] Correspondingly, relevant merchant data can be obtained, specifically such as Figure 4As shown, the relevant data for merchants can include merchant business payment cash flow (i.e., merchant business collection cash flow), merchant enterprise payment cash flow, information data of merchants, fund providers and transaction users, and text data, including merchant complaints, other data from official accounts, etc. Furthermore, relevant merchant data can be aggregated and integrated, including merchant commercial payment cash flow, merchant enterprise payment cash flow, information data of merchants, fund providers, and transaction users, as well as text data, including specific data from merchant complaints and other data from public accounts, which can be categorized into different modes. For example, merchant status information, merchant registration information, merchant rating information, and merchant name can be categorized into the merchant static information mode; normal payment transaction behavior, abnormal payment transaction behavior, and transaction complaint information related to merchants can be categorized into the merchant commercial payment side behavior mode; normal withdrawal behavior, abnormal withdrawal behavior, and text information at the time of withdrawal can be categorized into the merchant commercial payment side behavior mode; tag information and team information related to merchant transaction users can be categorized into the user-side lead mode; and information such as the merchant's associated lists and public account scenarios associated with the merchant can be categorized into other lead modes.

[0113] Furthermore, a primary risk prediction model can be constructed based on the acquired merchant data, corresponding to data from various dimensions. For example, such as... Figure 5 As shown, merchant status information can be processed (i.e., data processing in the diagram) to obtain processed data (i.e., feature engineering in the diagram), and a subject-based investigation status model can be built based on the processed data; merchant registration information (such as registered bank cards, ID cards, mobile phone numbers, etc.) can be processed to obtain a correlation graph (i.e., feature engineering in the diagram), and a registration graph model can be built based on the obtained correlation graph; commercial payment side fund flow information can be processed (i.e., data processing in the diagram) to obtain processed data (i.e., feature engineering in the diagram), and a malicious abnormal transaction behavior identification model (or community detection model) can be built based on the processed data; and merchant enterprise payment fund flow information can be processed (i.e., data processing in the diagram) to obtain processed data (i.e., feature engineering in the diagram), and a malicious abnormal payout behavior identification model (or a game-related normal payout behavior identification model) can be built based on the processed data.

[0114] Correspondingly, a second risk prediction model can be obtained by stacking models based on the subject's investigation status model, registration graph model, malicious abnormal transaction behavior identification model, and malicious abnormal transaction behavior identification model. This second risk prediction model, along with the subject-based investigation status model, registration graph model, malicious abnormal transaction behavior identification model, and malicious abnormal transaction behavior identification model, are then deployed to a server. The subject-based investigation status model, registration graph model, malicious abnormal transaction behavior identification model, and malicious abnormal transaction behavior identification model are the first risk prediction models corresponding to the data of each dimension described above. For ease of description, the first risk prediction models corresponding to the data of each dimension will be collectively referred to as the base models, and the second risk prediction model will be referred to as the fusion model.

[0115] Furthermore, such as Figure 6 As shown, when the base models and fusion models are obtained, risk assessment and prediction of merchants' enterprise payment business can be performed based on the pre-configured implementation strategy engine and the base models and fusion models. In this example, the risk assessment results for the target object corresponding to this dimension include various sub-risk levels, and the comprehensive risk assessment results include five levels: A, B, C, D, and E. Specifically, they may include:

[0116] (1) It can obtain data of merchants in various dimensions. For each dimension of data, based on the base model corresponding to the data of that dimension, it can determine the subdivided risk level of the merchant corresponding to that dimension.

[0117] (2) Write the merchant’s subdivided risk level corresponding to each dimension into the implementation strategy engine. The fusion model can obtain the merchant’s subdivided risk level corresponding to each dimension from the implementation strategy engine, and determine the merchant’s comprehensive risk assessment result based on the obtained subdivided risk level corresponding to each dimension, that is, which risk level A, B, C, D and E the merchant belongs to.

[0118] (3) Write the merchant’s subdivided risk level for each dimension and the comprehensive risk assessment results (i.e., acquire data) into the online visualization platform corresponding to the implementation strategy engine and display it to the user, and monitor the merchant’s subdivided risk level for each dimension and the comprehensive risk assessment results in real time.

[0119] (4) When a merchant has a business payment transaction, the implementation strategy engine can be triggered to determine which risk level (A, B, C, D, or E) the merchant belongs to, and risk management can be carried out on the merchant according to the risk management strategy corresponding to the specific risk level. For example, if the merchant belongs to any level (A, B, C, or D) (i.e., it is a malicious merchant), the merchant's business payment transaction will be blocked; if the merchant belongs to the trusted level (i.e., it is not a malicious merchant), the merchant's business payment transaction will be allowed.

[0120] Optionally, in this embodiment, the base models can be implemented using methods including, but not limited to, xgboost (eXtremeGradient Boosting), GBDT (Gradient Boosting Decision Tree), and random forest. Furthermore, to better manage merchant risk, relevant statistics for different risk levels and across the five risk levels, such as the number of merchants and the amount paid by merchants, can be entered into an online visualization platform to monitor the variation of merchants within different risk levels.

[0121] Based on the methods provided in the embodiments of this application, merchants can gain a better understanding and control of the various sub-risks and overall risks in enterprise payment business. Furthermore, the risk can be automatically managed according to the merchant's risk level, effectively reducing the risk. In addition, the various sub-risk levels and the five-level classification can be monitored on a visual page, which is conducive to timely detection of risk points.

[0122] This application provides a risk assessment device for an object, such as... Figure 7 As shown, the risk assessment device 60 for this object may include: a data acquisition module 601, a risk assessment result determination module 602, and a risk control processing module 603, wherein,

[0123] The data acquisition module 601 is used to acquire relevant data of the target object in multiple dimensions. The relevant data in multiple dimensions includes the first object data of the target object and the transaction data between the target object and related objects. Among them, related objects include objects that have transactions with the target object.

[0124] The risk assessment result determination module 602 is used to determine the risk assessment result of the target object corresponding to that dimension based on the first risk prediction model corresponding to the data of each dimension in the relevant data.

[0125] The risk management and control module 603 is used to determine the comprehensive risk assessment result of the target object based on the risk assessment results of each dimension corresponding to the target object through the second risk prediction model, and to carry out risk management and control processing on the target object based on the risk assessment results of each dimension and the comprehensive risk assessment result.

[0126] Optionally, the associated objects include target associated objects and non-target associated objects. Target associated objects refer to objects that have transactions with the target object in the target business. The transaction data includes first transaction data between the target object and the target associated objects, and second transaction data between the target object and non-target associated objects.

[0127] When determining the comprehensive risk assessment result of a target object based on the risk assessment results of each dimension corresponding to the target object, the risk management and processing module is specifically used for:

[0128] Based on the risk assessment results of each dimension corresponding to the target object, the comprehensive risk assessment result corresponding to the target business is determined.

[0129] Optionally, the relevant data may also include second object data of the associated object.

[0130] Optionally, when determining the comprehensive risk assessment result of the target object based on the risk assessment results of each dimension corresponding to the target object, the risk management processing module is specifically used for:

[0131] Based on the second risk prediction model, feature extraction is performed on the risk assessment results of each dimension corresponding to the target object to obtain the fused feature vector corresponding to each dimension.

[0132] Based on the fused feature vectors, the comprehensive risk assessment results of the target object are determined.

[0133] Optionally, the data providing module of the device is used for:

[0134] Provide users with risk assessment results corresponding to the target objects across various dimensions;

[0135] Provide users with comprehensive risk assessment results.

[0136] Optionally, the device also includes an optimization module for:

[0137] For any dimension, if there are multiple target objects, obtain the risk assessment result for each target object corresponding to that dimension;

[0138] Acquire and monitor transaction data of target objects belonging to the specified risk assessment results in real time;

[0139] When the transaction data belonging to the target object of the specified risk assessment result meets the optimization conditions, the first risk prediction model corresponding to the data in that dimension is optimized.

[0140] Optionally, the device also includes a model training module for training a first risk prediction model for each dimension in the following manner:

[0141] Obtain the training dataset, which includes multiple first training samples. Each first training sample includes a sample object corresponding to a sub-sample of each dimension. A sub-sample of a dimension includes sample-related data of that dimension and risk label corresponding to that dimension.

[0142] For each dimension, the initial risk prediction model for that dimension is iteratively trained based on the subsamples of that dimension in each first training sample to obtain the first risk prediction model corresponding to that dimension.

[0143] Optionally, each first training sample also includes a comprehensive risk label corresponding to the sample object; the second risk prediction model is trained in the following way:

[0144] Using the first risk prediction model for each dimension, obtain the risk assessment results for each first training sample corresponding to each dimension;

[0145] Based on the risk assessment results and comprehensive risk labels corresponding to each dimension of the first training sample, each second training sample is obtained, wherein each second training sample includes a sample object corresponding to the risk assessment results and comprehensive risk labels of each dimension;

[0146] The initial second risk prediction model is iteratively trained based on each second training sample to obtain the second risk prediction model.

[0147] Optionally, for each dimension, when the model training module trains the initial risk prediction model for that dimension based on the subsamples of that dimension in each of the first training samples to obtain the first risk prediction model corresponding to that dimension, it is specifically used for:

[0148] Each subsample of this dimension is divided into multiple training subsets. At least two datasets are constructed based on the multiple training subsets. Each dataset includes at least two training subsets. One training subset from the at least two training subsets is used as the prediction set. The training subsets other than the prediction set are used as the training set. The prediction sets are different for different datasets.

[0149] The initial risk prediction model for that dimension is trained based on the training set in each dataset to obtain the model parameters for each dataset.

[0150] The model parameters of the first risk prediction model are determined based on the model parameters corresponding to each training set, and the first risk prediction model corresponding to this dimension is obtained.

[0151] Using the first risk prediction model for each dimension, obtain the risk assessment results for each first training sample corresponding to each dimension, including:

[0152] For each dimension, each prediction set of that dimension is passed through a first risk prediction model based on the model parameters corresponding to that prediction set to obtain the risk assessment result for each first training sample corresponding to each dimension.

[0153] Optionally, the model training module is also used for:

[0154] Obtain the test dataset corresponding to the training dataset, and test the first risk prediction model for each dimension based on the test dataset;

[0155] Each test sample in the test dataset is input into the first risk prediction model based on the model parameters corresponding to each training subset, to obtain at least two first risk assessment results for each test sample;

[0156] For each test sample, the at least two first risk assessment results corresponding to the test sample are fused to obtain the second risk assessment result corresponding to the test sample;

[0157] Based on the second risk assessment results corresponding to each test sample, the test dataset corresponding to the second risk prediction model is obtained.

[0158] The object risk assessment device in this application embodiment can execute an object risk assessment method provided in this application embodiment, and its implementation principle is similar, so it will not be described again here.

[0159] The risk assessment device for an object can be a computer program (including program code) running on a computer device, for example, the risk assessment device for an object is an application software; the device can be used to perform the corresponding steps in the method provided in the embodiments of this application.

[0160] In some embodiments, the risk assessment device for an object provided in this invention can be implemented using a combination of hardware and software. As an example, the risk assessment device for an object provided in this invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the risk assessment method for an object provided in this invention. For example, the processor in the form of a hardware decoding processor can be one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0161] In other embodiments, the risk assessment device for the object provided in this invention can be implemented in software. Figure 7 A risk assessment device 60 for an object is shown. It can be software in the form of programs and plug-ins, and includes a series of modules, including a data acquisition module 601, a risk assessment result determination module 602, and a risk control processing module. The data acquisition module 601, the risk assessment result determination module 602, and the risk control processing module are used to implement the risk assessment method for an object provided in the embodiments of the present invention.

[0162] This application provides an electronic device, such as... Figure 8 As shown, Figure 8 The illustrated electronic device 2000 includes a processor 2001 and a memory 2003. The processor 2001 and the memory 2003 are connected, for example, via a bus 2002. Optionally, the electronic device 2000 may also include a transceiver 2004. It should be noted that in practical applications, the transceiver 2004 is not limited to one type, and the structure of this electronic device 2000 does not constitute a limitation on the embodiments of this application.

[0163] In this embodiment, the processor 2001 is used to implement... Figure 7 The functions of each module are shown.

[0164] Processor 2001 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 2001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0165] Bus 2002 may include a pathway for transmitting information between the aforementioned components. Bus 2002 may be a PCI bus or an EISA bus, etc. Bus 2002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 8 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0166] The memory 2003 may be ROM or other type of static storage device capable of storing static information and computer programs, RAM or other type of dynamic storage device capable of storing information and computer programs, or EEPROM, CD-ROM or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing or in the form of a desired computer program and capable of being accessed by a computer, but not limited thereto.

[0167] The memory 2003 stores computer programs for executing the application scheme of this application, and its execution is controlled by the processor 2001. The processor 2001 executes the computer programs for the application scheme stored in the memory 2003 to implement… Figure 7 The operation of the risk assessment device for the object provided in the illustrated embodiment.

[0168] This application provides an electronic device including a processor and a memory: the memory is configured to store a computer program, which, when executed by the processor, causes the processor to perform any of the methods described in the above embodiments.

[0169] This application provides a computer-readable storage medium for storing a computer program that, when run on a computer, enables the computer to perform any of the methods described in the above embodiments.

[0170] According to one aspect of this application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various alternative implementations described above.

[0171] The specific terms and implementation principles involved in the computer-readable storage medium in this application can be found in the risk assessment method for an object in the embodiments of this application, and will not be repeated here.

[0172] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0173] The above are only some embodiments of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for risk assessment of an object, characterized in that, include: Obtain relevant data for the target object from multiple dimensions, including first object data of the target object and transaction data between the target object and associated objects, wherein the associated objects include objects that have transactions with the target object; For each dimension of the relevant data, based on the first risk prediction model corresponding to the data of that dimension, the risk assessment result of the target object corresponding to that dimension is determined; Based on the risk assessment results of each dimension corresponding to the target object, the comprehensive risk assessment result of the target object is determined by the second risk prediction model, and the target object is subject to risk management and control based on the risk assessment results of each dimension and the comprehensive risk assessment result. The second risk prediction model was trained in the following way: Obtain a training dataset, wherein the training dataset includes multiple first training samples; each first training sample includes a sample object corresponding to a sub-sample of each dimension and a comprehensive risk label corresponding to the sample object, and the sub-sample of each dimension includes sample-related data of that dimension; Based on the sub-samples corresponding to each dimension in each of the first training samples, the risk assessment results corresponding to each dimension of each first training sample are obtained through the first risk prediction model of each corresponding dimension; Each of the first training samples corresponds to the risk assessment results and comprehensive risk labels of each dimension, and is used as a second training sample; The initial second risk prediction model is iteratively trained based on each of the second training samples to obtain the second risk prediction model.

2. The method according to claim 1, characterized in that, The associated objects include target associated objects and non-target associated objects. The target associated objects refer to objects that have transactions with the target object in the target business. The transaction data includes first transaction data between the target object and the target associated objects, and second transaction data between the target object and the non-target associated objects. The step of determining the comprehensive risk assessment result of the target object based on the risk assessment results of each dimension corresponding to the target object through a second risk prediction model includes: Based on the risk assessment results of each dimension corresponding to the target object, the comprehensive risk assessment result of the target object corresponding to the target business is determined.

3. The method according to claim 1, characterized in that, The related data also includes the second object data of the associated object.

4. The method according to any one of claims 1 to 3, characterized in that, The step of determining the comprehensive risk assessment result of the target object based on the risk assessment results of each dimension corresponding to the target object includes: Based on the second risk prediction model, feature extraction is performed on the risk assessment results of each dimension corresponding to the target object to obtain the fused feature vector corresponding to each dimension. Based on the fused feature vector, the comprehensive risk assessment result of the target object is determined.

5. The method according to claim 1, characterized in that, The method further includes at least one of the following: The risk assessment results corresponding to each dimension of the target object are provided to the user; The comprehensive risk assessment results will be provided to the user.

6. The method according to claim 1, characterized in that, The method further includes: For any dimension, if there are multiple target objects, obtain the risk assessment result for each target object corresponding to that dimension; Acquire and monitor transaction data of target objects belonging to the specified risk assessment results in real time; When the transaction data belonging to the target object of the specified risk assessment result meets the optimization conditions, the first risk prediction model corresponding to the data in that dimension is optimized.

7. The method according to any one of claims 1 to 6, characterized in that, Each dimension's subsample also includes the corresponding risk label for that dimension; The first risk prediction model corresponding to each of the aforementioned dimensions is trained in the following way: For each dimension, the initial risk prediction model for that dimension is iteratively trained based on the sub-samples of that dimension in each of the first training samples to obtain the first risk prediction model corresponding to that dimension.

8. The method according to claim 7, characterized in that, For each dimension, the initial risk prediction model for that dimension is trained based on sub-samples of that dimension in each of the first training samples to obtain the first risk prediction model corresponding to that dimension, including: Each subsample of this dimension is divided into multiple training subsets. At least two datasets are constructed based on the multiple training subsets. Each dataset includes at least two training subsets. One training subset in the at least two training subsets is used as the prediction set, and the training subsets other than the prediction set are used as the training set. The prediction sets of different datasets are different. The initial risk prediction model for each dimension is trained based on the training set in each of the datasets to obtain the model parameters corresponding to each dataset. The model parameters of the first risk prediction model are determined based on the model parameters corresponding to each training set, and the first risk prediction model corresponding to that dimension is obtained. The step of obtaining the risk assessment result for each of the first training samples corresponding to each of the said dimensions through the first risk prediction model of each dimension includes: For each dimension, each prediction set of that dimension is passed through a first risk prediction model based on the model parameters corresponding to that prediction set to obtain the risk assessment result of each first training sample corresponding to each dimension.

9. The method according to claim 8, characterized in that, Also includes: Obtain the test dataset corresponding to the training dataset, and test the first risk prediction model corresponding to each dimension based on the test dataset; Each test sample in the test dataset is input into a first risk prediction model based on the model parameters corresponding to each training subset, to obtain at least two first risk assessment results for each test sample; For each test sample, at least two first risk assessment results corresponding to the test sample are fused to obtain a second risk assessment result corresponding to the test sample. Based on the second risk assessment results corresponding to each of the test samples, the test dataset corresponding to the second risk prediction model is obtained.

10. A risk assessment device for an object, characterized in that, include: The data acquisition module is used to acquire relevant data of the target object in multiple dimensions. The relevant data in multiple dimensions includes the first object data of the target object and the transaction data between the target object and related objects. The related objects include objects that have transactions with the target object. The risk assessment result determination module is used to determine the risk assessment result of the target object corresponding to each dimension of the relevant data, based on the first risk prediction model corresponding to the data of that dimension; The risk management and control module is used to determine the comprehensive risk assessment result of the target object based on the risk assessment results of each dimension corresponding to the target object through a second risk prediction model, and to perform risk management and control on the target object based on the risk assessment results of each dimension and the comprehensive risk assessment result; The model training module is used to acquire a training dataset, wherein the training dataset includes multiple first training samples; each first training sample includes a sample object corresponding to a sub-sample of each dimension and a comprehensive risk label corresponding to the sample object, and the sub-sample of each dimension includes sample-related data of that dimension; Based on the sub-samples corresponding to each dimension in each of the first training samples, the risk assessment results corresponding to each dimension of each first training sample are obtained through the first risk prediction model of each corresponding dimension; Each of the first training samples corresponds to the risk assessment results and comprehensive risk labels of each dimension, and is used as a second training sample; The initial second risk prediction model is iteratively trained based on each of the second training samples to obtain the second risk prediction model.

11. The apparatus according to claim 10, characterized in that, The associated objects include target associated objects and non-target associated objects. The target associated objects refer to objects that have transactions with the target object in the target business. The transaction data includes first transaction data between the target object and the target associated objects, and second transaction data between the target object and the non-target associated objects. When the risk management processing module determines the comprehensive risk assessment result of the target object based on the risk assessment results of each dimension corresponding to the target object through the second risk prediction model, it is specifically used for: Based on the risk assessment results of each dimension corresponding to the target object, the comprehensive risk assessment result of the target object corresponding to the target business is determined.

12. The apparatus according to claim 10, characterized in that, The related data also includes the second object data of the associated object.

13. The apparatus according to any one of claims 10-12, characterized in that, When the risk management processing module determines the comprehensive risk assessment result of the target object based on the risk assessment results of each dimension corresponding to the target object through the second risk prediction model, it is specifically used for: Based on the second risk prediction model, feature extraction is performed on the risk assessment results of each dimension corresponding to the target object to obtain the fused feature vector corresponding to each dimension. Based on the fused feature vector, the comprehensive risk assessment result of the target object is determined.

14. The apparatus according to claim 10, characterized in that, The device further includes a data providing module for: The risk assessment results corresponding to each dimension of the target object are provided to the user; The comprehensive risk assessment results will be provided to the user.

15. The apparatus according to claim 10, characterized in that, The device further includes an optimization module for: For any dimension, if there are multiple target objects, obtain the risk assessment result for each target object corresponding to that dimension; Acquire and monitor transaction data of target objects belonging to the specified risk assessment results in real time; When the transaction data belonging to the target object of the specified risk assessment result meets the optimization conditions, the first risk prediction model corresponding to the data in that dimension is optimized.

16. The apparatus according to any one of claims 10-15, characterized in that, Each dimension's subsample also includes the corresponding risk label for that dimension; The first risk prediction model corresponding to each of the aforementioned dimensions is trained by the model training module in the following manner: For each dimension, the initial risk prediction model for that dimension is iteratively trained based on the sub-samples of that dimension in each of the first training samples to obtain the first risk prediction model corresponding to that dimension.

17. The apparatus according to claim 16, characterized in that, For each dimension, when the model training module trains the initial risk prediction model for that dimension based on sub-samples of that dimension in each of the first training samples to obtain the first risk prediction model corresponding to that dimension, it is specifically used for: Each subsample of this dimension is divided into multiple training subsets. At least two datasets are constructed based on the multiple training subsets. Each dataset includes at least two training subsets. One training subset in the at least two training subsets is used as the prediction set, and the training subsets other than the prediction set are used as the training set. The prediction sets of different datasets are different. The initial risk prediction model for each dimension is trained based on the training set in each of the datasets to obtain the model parameters corresponding to each dataset. The model parameters of the first risk prediction model are determined based on the model parameters corresponding to each training set, and the first risk prediction model corresponding to that dimension is obtained. When the model training module obtains the risk assessment result for each of the first training samples corresponding to each of the dimensions using the first risk prediction model for each of the dimensions, it is specifically used for: For each dimension, each prediction set of that dimension is passed through a first risk prediction model based on the model parameters corresponding to that prediction set to obtain the risk assessment result of each first training sample corresponding to each dimension.

18. The apparatus according to claim 17, characterized in that, The model training module is also used for: Obtain the test dataset corresponding to the training dataset, and test the first risk prediction model corresponding to each dimension based on the test dataset; Each test sample in the test dataset is input into a first risk prediction model based on the model parameters corresponding to each training subset, to obtain at least two first risk assessment results for each test sample; For each test sample, at least two first risk assessment results corresponding to the test sample are fused to obtain a second risk assessment result corresponding to the test sample. Based on the second risk assessment results corresponding to each of the test samples, the test dataset corresponding to the second risk prediction model is obtained.

19. An electronic device, characterized in that, Including the processor and memory: The memory is configured to store a computer program that, when executed by the processor, causes the processor to perform the method according to any one of claims 1-9.

20. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program that, when run on a computer, enables the computer to perform the method described in any one of claims 1-9.

Images