Risk assessment methods, risk assessment devices, equipment and storage media

Abstract

This invention provides a risk assessment method, a risk assessment device, a computer device, and a storage medium, relating to the field of artificial intelligence. The method includes: acquiring business data and extracting information from the business data to obtain business information corresponding to each business type and multiple user nodes; establishing association information between user nodes based on the business information corresponding to each business type, and generating a multi-layer topology network based on the association information; performing network fusion on the multi-layer topology network to obtain a vector matrix corresponding to the multi-layer topology network; detecting and filtering user nodes based on the vector matrix corresponding to the multi-layer topology network to obtain target user nodes; and performing risk prediction on user node groups associated with the target user nodes to obtain risk assessment results corresponding to the user node groups. This allows for rapid and accurate detection of fraud targets, effectively curbing insurance fraud and preventing losses to company interests.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence, and in particular to risk assessment methods, risk assessment devices, computer equipment, and storage media. Background Technology

[0002] With the development of the insurance business, insurance fraud has become increasingly serious. At the same time, fraudulent activities involving users colluding to deliberately create multiple insurance claims to obtain payouts from multiple insurance companies (such as health insurance and car insurance) are on the rise, causing significant economic losses to companies. Because the mechanisms of insurance fraud are similar across various insurance products, fraudsters often use the same method to commit crimes with multiple insurance products simultaneously. For example, user A might collude to create an accident to fraudulently obtain health insurance benefits, and then use the same collusion mechanism to fraudulently obtain car insurance benefits. Therefore, existing methods are unable to accurately detect fraudsters, making it impossible to curb insurance fraud and resulting in losses for companies. Summary of the Invention

[0003] This application provides a risk assessment method, risk assessment device, computer equipment, and storage medium, which aim to quickly and accurately detect fraudulent targets, thereby effectively curbing insurance fraud and preventing damage to company interests.

[0004] To achieve the above objectives, this application provides a risk assessment method, the method comprising:

[0005] Acquire business data and extract information from the business data to obtain business information corresponding to each business type and multiple user nodes;

[0006] Establish the association information between each user node based on the business information corresponding to each business type, and generate a multi-layer topology network based on the association information;

[0007] The multi-layer topological relationship network is fused to obtain the vector matrix corresponding to the multi-layer topological relationship network;

[0008] Based on a preset node detection model, each user node is detected and filtered according to the vector matrix corresponding to the multi-layer topology network to obtain the target user node.

[0009] Risk prediction is performed on the user node group associated with the target user node to obtain the risk assessment result corresponding to the user node group.

[0010] To achieve the above objectives, this application also provides a risk assessment device, the risk assessment device comprising:

[0011] The business information extraction module is used to acquire business data and extract information from the business data to obtain business information corresponding to each business type and multiple user nodes.

[0012] The network construction module is used to establish the association information between each user node according to the business information corresponding to each business type, and to generate a multi-layer topology network according to the association information.

[0013] The network fusion module is used to perform network fusion on the multi-layer topology network to obtain the vector matrix corresponding to the multi-layer topology network.

[0014] The node detection module is used to detect and filter each user node based on a preset node detection model and the vector matrix corresponding to the multi-layer topology network to obtain the target user node.

[0015] The risk assessment module is used to predict the risk of the user node group associated with the target user node and obtain the risk assessment result corresponding to the user node group.

[0016] In addition, to achieve the above objectives, this application also provides a computer device, the computer device including a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program and, when executing the computer program, implement any of the risk assessment methods provided in the embodiments of this application.

[0017] In addition, to achieve the above objectives, this application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to implement any of the risk assessment methods provided in the embodiments of this application.

[0018] The risk assessment method, risk assessment device, equipment, and storage medium disclosed in this application acquire business information corresponding to each business type and multiple user nodes; establish a multi-layer topology network based on the business information; perform network fusion on the multi-layer topology network to obtain a vector matrix corresponding to the multi-layer topology network; and perform risk prediction on user node groups associated with user nodes based on each vector matrix to obtain risk assessment results corresponding to user node groups associated with user nodes. This allows for risk assessment of users of each business type and accurate detection of fraud targets, thereby effectively protecting company interests and reducing company economic losses. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram illustrating a scenario of a risk assessment method provided in an embodiment of this application;

[0021] Figure 2 This is a flowchart illustrating a risk assessment method provided in an embodiment of this application;

[0022] Figure 3 This is a schematic block diagram of a risk assessment device provided in an embodiment of this application;

[0023] Figure 4 This is a schematic block diagram of a computer device provided in one embodiment of this application. Detailed Implementation

[0024] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0025] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily need to be performed in the described order. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation. Furthermore, although functional modules are divided in the device diagram, in some cases, a different module division may be used.

[0026] The term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items, as well as all possible combinations, and includes such combinations.

[0027] Because the mechanisms of insurance fraud are similar across various insurance products, perpetrators often use the same method to commit fraud across multiple insurance products simultaneously. For example, user A might stage an accident to fraudulently claim health insurance, and then use the same collusion mechanism to fraudulently claim car insurance. Therefore, existing methods are unable to accurately detect fraudsters, making it impossible to curb insurance fraud and resulting in losses for companies.

[0028] In existing technologies, there are two main strategies for detecting insurance fraud: (1) manually analyzing different types of insurance claims and making empirical judgments; (2) using machine learning-based clustering algorithms to detect fraud in each type of insurance. However, both strategies still have some drawbacks. Manually analyzing claims often requires a huge amount of manpower, and the number of cases that can be handled manually is limited, thus it cannot be widely applied to large-scale claims. Although existing machine learning-based clustering algorithms can quickly detect fraud in a single type of insurance, they ignore the behavior of fraud in multiple insurance schemes simultaneously. This makes it impossible for existing models to effectively model the various behavioral characteristics of fraud, thus reducing the accuracy of fraud detection.

[0029] To address the aforementioned issues, this application provides a risk assessment method applied to a server, which can quickly and accurately detect fraudulent activities, thereby effectively curbing insurance fraud and preventing damage to the company's interests.

[0030] The server can be a single server or a server cluster. However, for ease of understanding, the following embodiments will be described in detail using risk assessment methods applied to servers.

[0031] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0032] like Figure 1 As shown, the risk assessment method provided in this application embodiment can be applied to, for example, Figure 1The application environment shown includes a terminal device 110 and a server 120. The terminal device 110 can communicate with the server 120 via a network. Specifically, the server 120 obtains business information corresponding to each business type and multiple user nodes; establishes a multi-layer topology network based on the business information; performs network fusion on the multi-layer topology network based on a preset network fusion algorithm to obtain a vector matrix corresponding to the multi-layer topology network; performs risk prediction on the user node groups associated with each user node based on each vector matrix to obtain the risk assessment results corresponding to the user node groups associated with each user node, and sends the risk assessment results corresponding to the user node groups associated with each user node to the terminal device 110. The server 120 can be an independent server or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms. The terminal device 110 can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, etc., but is not limited to these. Terminals and servers can be connected directly or indirectly via wired or wireless communication, and this application does not impose any restrictions on this.

[0033] Please see Figure 2 , Figure 2 This is a schematic flowchart illustrating a risk assessment method provided in an embodiment of this application. This risk assessment method can be applied to a server, thereby accurately detecting fraudulent targets, effectively protecting company interests, reducing economic losses, and improving user experience.

[0034] like Figure 2 As shown, the risk assessment method includes steps S101 to S105.

[0035] S101. Obtain business data and extract information from the business data to obtain business information corresponding to each business type and multiple user nodes.

[0036] The business data may include data corresponding to insurance business. The business type may be an insurance type, such as health insurance and auto insurance. The business information may be an insurance work order, where each insurance type corresponds to an insurance work order. Specifically, the business information may include information such as the policyholder and the insured amount. The user node is an abstract node formed for the policyholder or beneficiary in the insurance work order corresponding to each insurance type.

[0037] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use 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 obtain optimal results.

[0038] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0039] In some embodiments, the business data is classified based on preset business categories to obtain business information corresponding to each business type; the business information is then traversed to extract multiple user nodes corresponding to each business information. This allows for accurate extraction of user nodes from the business information corresponding to each business type.

[0040] The preset business category is a pre-set insurance type, which can be set by the user.

[0041] Specifically, the business data is classified and processed, and insurance policies corresponding to preset business categories are filtered out to obtain business information corresponding to each business type; the user information is extracted from the business information and abstracted into user nodes, thereby obtaining multiple user nodes corresponding to each business information.

[0042] For example, if we need to conduct risk assessments on business data related to health insurance, property insurance, and auto insurance, we would categorize the business data, filtering out the insurance policies corresponding to health insurance, property insurance, and auto insurance respectively, thus obtaining the business information for each type of insurance. We would then iterate through the insurance policies, extract user information, and abstract this information into user nodes, resulting in multiple user nodes corresponding to each type of business information.

[0043] S102. Establish the association information between each user node according to the business information corresponding to each business type, and generate a multi-layer topology network according to the association information.

[0044] The association information is used to indicate the relationships between user nodes and to generate a multi-layered topological network. This multi-layered topological network is constructed using user nodes as entity nodes and represents the collaborative relationships between them. The network models interpersonal relationships in different insurance incidents, reflecting the shared collaborative relationships among relevant personnel in various types of insurance claims. The number of layers in the topological network is related to the number of business types.

[0045] For example, if there are three types of business categories, then the number of layers in the topology network is three.

[0046] In some embodiments, target user nodes are determined from a plurality of user nodes based on service information corresponding to each service type, and association information between the target user nodes is established; association edges between the target user nodes are generated based on the association information; and a network is constructed for the plurality of target user nodes based on the association edges between the target user nodes to obtain a multi-layer topology network. This allows for the accurate construction of a multi-layer topology network, thereby accurately representing the collaborative relationships between user nodes.

[0047] The target user node is a user node with an association relationship, and the term used to represent the association relationship between user nodes is used to generate a multi-layer topology network.

[0048] Specifically, the business information corresponding to each business type is traversed to determine whether there is a relationship between any two user nodes. If there is a relationship, these two user nodes are taken as target user nodes, and the association information between each target user node is established. If there is no relationship, the relationship between two new user nodes is re-determined, and so on, to obtain the target user nodes.

[0049] For example, the existence of a relationship between each pair of user nodes can be determined by determining whether there is cooperation or a relationship between a suspect and a victim.

[0050] Specifically, after obtaining the association information between each target user node, an inverted index table or a Cartesian product table can be generated based on the association information; a first association edge between each user node can be generated based on the inverted index table; a second association edge between each user node can be generated based on the Cartesian product table; and a network can be constructed for the multiple target user nodes based on the first and second association edges between each user node to obtain a multi-layer topology network.

[0051] The first associated edge can be an equal associated edge, generated from an inverted index table, and the second associated edge can be a similar associated edge, generated from a Cartesian product table. Thus, a multi-layered topological network can be constructed by using equal and similar associated edges to build a network for multiple user nodes.

[0052] It should be noted that the associated edges generated by the business information corresponding to each type of business are used to construct the topological relationship network of each layer. By doing so, the topological relationship network of each layer is constructed to obtain a multi-layer topological relationship network.

[0053] S103. Perform network fusion on the multi-layer topology network to obtain the vector matrix corresponding to the multi-layer topology network.

[0054] The vector matrix is ​​a low-dimensional representation matrix. Specifically, a preset network fusion algorithm can be used to perform network fusion on the multi-layer topological relationship network to obtain the vector matrix corresponding to the multi-layer topological relationship network. The network fusion algorithm can fuse multiple low-dimensional representation matrices corresponding to the multi-layer network into a final low-dimensional representation matrix to enhance the model's ability to represent users (i.e., user nodes). This provides data guidance for detecting potential fraud targets, and enables members of the fraud targets to learn similar vector representations. This greatly improves the performance of downstream clustering, thereby achieving accurate detection of insurance fraud targets.

[0055] In some embodiments, vector transformation processing is performed on the topological relationship networks of each layer to obtain vector matrices corresponding to the topological relationship networks of each layer; the vector matrices corresponding to the topological relationship networks of each layer are input into a preset graph attention model to obtain the weight ratios corresponding to each vector matrix; the vector matrices corresponding to the topological relationship networks of each layer are weighted according to the weight ratios corresponding to each vector matrix to obtain the vector matrix corresponding to the multi-layer topological relationship network. This allows the vector matrices corresponding to the topological relationship networks of each layer to be fused, accurately obtaining the vector matrix corresponding to the multi-layer topological relationship network.

[0056] The preset graph attention model is used to assign different learning weights to the vector matrices corresponding to each layer of the topological relationship network.

[0057] Specifically, the topological relationship network of each layer is transformed into a vector matrix. Then, the vector matrix corresponding to each layer of the topological relationship network is input into a preset graph attention model to assign attention weights to each vector matrix, thereby obtaining the weight ratio of each vector matrix. Finally, the vector matrices corresponding to each layer of the topological relationship network are weighted and summed using the weight ratio of each vector matrix to obtain the final vector matrix corresponding to the multi-layer topological relationship network.

[0058] In obtaining the low-dimensional representation matrix of each layer Then, the low-dimensional representation matrix of each layer is... The input is given to the graph attention model Q defined in this scheme. att In this process, the weight ratios of the low-dimensional representation matrices corresponding to different layers are obtained through training. The overall calculation method is as follows:

[0059]

[0060] Among them, l i This represents the i-th layer of the network. denoted by , where represents the weight ratio corresponding to each vector matrix, and p represents the number of layers in the network. Let N represent the low-dimensional network representation matrix corresponding to the i-th layer of the network. The number of rows in this matrix is ​​equal to the number of nodes N. Each row vector in the matrix represents the low-dimensional representation row vector of a node.

[0061] Specifically, the vector matrix corresponding to the multi-layer topological network, i.e., the final single-layer network representation matrix, can be obtained by weighted summation of the low-dimensional representation matrices of all layers according to their respective weight ratios, that is:

[0062]

[0063] Where H is the vector matrix corresponding to the multi-layer topological network, and p is the number of layers in the multi-layer network.

[0064] In some embodiments, an activation function is used to perform a linear transformation on each vector matrix to obtain the attention coefficients corresponding to each vector matrix; regularization is then performed on the attention coefficients corresponding to each vector matrix to obtain the weight ratios corresponding to each vector matrix. This allows for the accurate determination of the weight ratios corresponding to each vector matrix.

[0065] The activation function may include functions such as the hyperbolic tangent activation function, the sigmoid function, and the linear rectified function. The attention coefficient is the similarity coefficient between the vector matrices, used to determine the weight ratio of each vector matrix.

[0066] For example, taking the hyperbolic tangent activation function as the activation function, the hyperbolic tangent activation function can be used to perform a linear transformation on each vector matrix to obtain the attention coefficients corresponding to each vector matrix; then, the normalized exponential function is used to regularize the attention coefficients corresponding to each vector matrix to obtain the weight ratios corresponding to each vector matrix.

[0067] Specifically, the attention coefficient can be calculated using the following formula:

[0068]

[0069] in, denoted as the attention coefficient, N as the number of nodes in the network, tanh as the hyperbolic tangent activation function, and Wsem as the learnable linear transformation parameter matrix. This represents the row vector of the nth node in the i-th layer of the network, whose value is also equal to... n rows of data in the file. For learnable attention vectors, This is the bias vector.

[0070] Specifically, the weight ratios corresponding to each vector matrix can be obtained by the following formula:

[0071]

[0072] in, This represents the weight ratio corresponding to each vector matrix, and softmax is the normalization exponential function. This represents the attention coefficient.

[0073] By using the above formula, the weight ratios corresponding to each vector matrix can be accurately obtained.

[0074] S104. Based on the preset node detection model, the user nodes are detected and filtered according to the vector matrix corresponding to the multi-layer topology network to obtain the target user nodes.

[0075] The node detection model is used to detect the user type of each user node and determine whether the user corresponding to the user node is a fraudulent user. The user corresponding to the target user node is a fraudulent user.

[0076] In some embodiments, user type identification is performed on each user node based on the vector matrix corresponding to the multi-layer topology network to obtain the user type of each user node; if the user type of the user node is a preset user type, then the user node is taken as the target user node.

[0077] The user types for user nodes include fraudulent users and non-fraudulent users. The preset user type is fraudulent users.

[0078] Specifically, the vector matrix corresponding to the multi-layer topology network is input into the node detection model to identify the user type of each user node and obtain the user type of each user node; if the user type of the user node is a preset user type, then the user node is used as the target user node; if the user type of the user node is not a preset user type, then the user node is not used as the target user node.

[0079] For example, information about fraudsters that is already known can be used to identify potential targets for fraud.

[0080] S105. Perform risk prediction on the user node group associated with the target user node to obtain the risk assessment result corresponding to the user node group.

[0081] The user node group is a combination of multiple user nodes that are associated with the target user node.

[0082] In some embodiments, the target user nodes are clustered based on a preset clustering algorithm to obtain user node groups associated with the target user nodes; and a risk assessment result corresponding to the user node group is generated based on the number of target user nodes in the user node group.

[0083] The clustering algorithm may include K-means clustering algorithm, mean shift clustering algorithm, etc.

[0084] Specifically, based on a preset clustering algorithm, the target user nodes are clustered to obtain user node groups associated with the target user nodes; if a preset number of target user nodes exist in the user node group, the user node group is determined to be at risk; if a preset number of target user nodes do not exist in the user node group, the user node group is determined to be at no risk.

[0085] The preset quantity can be any quantity, and no specific limit is set here.

[0086] For example, taking a preset quantity of 2 as an example, if there are 3 target user nodes in the user node group, it is determined that the user node group has a risk; if there is only 1 target user node in the user node group, it is determined that the user node group has no risk.

[0087] It should be noted that the parameters of the node detection model, graph attention model, and algorithm in this application need to be fine-tuned to achieve a balance between accuracy and preventing overfitting. Since the clustering coefficient in K-means is directly related to the performance of the results, it can be adjusted to observe the accuracy of the model output, thereby selecting the optimal model parameters to achieve the best clustering performance while ensuring efficiency, thus enabling efficient detection of fraudulent targets.

[0088] Please see Figure 3 , Figure 3 This is a schematic block diagram of a risk assessment device provided in an embodiment of this application. The risk assessment device can be configured in a server to perform the aforementioned risk assessment method.

[0089] like Figure 3 As shown, the risk assessment device 200 includes: a business information extraction module 201, a network construction module 202, a network fusion module 203, a node detection module 204, and a risk assessment module 205.

[0090] The business information extraction module 201 is used to acquire business data and extract information from the business data to obtain business information corresponding to each business type and multiple user nodes.

[0091] The network construction module 202 is used to establish the association information between each user node according to the business information corresponding to each business type, and to generate a multi-layer topology network according to the association information.

[0092] Network fusion module 203 is used to perform network fusion on the multi-layer topology network to obtain a vector matrix corresponding to the multi-layer topology network;

[0093] The node detection module 204 is used to detect and filter each user node based on a preset node detection model and the vector matrix corresponding to the multi-layer topology network to obtain the target user node.

[0094] The risk assessment module 205 is used to perform risk prediction on the user node group associated with the target user node and obtain the risk assessment result corresponding to the user node group.

[0095] The business information extraction module 201 is also used to classify the business data based on a preset business category to obtain business information corresponding to each business type; and to traverse the business information to extract multiple user nodes corresponding to each business information.

[0096] The network construction module 202 is further configured to determine target user nodes from multiple user nodes based on the service information corresponding to each service type, and establish association information between each target user node; generate association edges between each target user node based on the association information between each target user node; and construct a network for the multiple target user nodes based on the association edges between each target user node to obtain a multi-layer topology network.

[0097] The network fusion module 203 is further configured to perform vector transformation processing on the topological relationship network of each layer to obtain the vector matrix corresponding to the topological relationship network of each layer; input the vector matrix corresponding to the topological relationship network of each layer into a preset graph attention model to obtain the weight ratio corresponding to each vector matrix; and perform weighted processing on the vector matrix corresponding to the topological relationship network of each layer according to the weight ratio corresponding to each vector matrix to obtain the vector moment corresponding to the multi-layer topological relationship network.

[0098] The network fusion module 203 is further configured to perform linear transformation processing on each of the vector matrices through an activation function to obtain the attention coefficients corresponding to each of the vector matrices; and to perform regularization processing on the attention coefficients corresponding to each of the vector matrices to obtain the weight ratios corresponding to each of the vector matrices.

[0099] The node detection module 204 is also used to identify the user type of each user node based on the vector matrix corresponding to the multi-layer topology network, and obtain the user type of each user node.

[0100] If the user type of the user node is a preset user type, then the user node will be used as the target user node.

[0101] The risk assessment module 205 is also used to perform clustering processing on the target user nodes based on a preset clustering algorithm to obtain user node groups associated with the target user nodes; and to generate risk assessment results corresponding to the user node groups based on the number of target user nodes in the user node groups.

[0102] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the above-described apparatus and its modules and units can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0103] The methods and apparatus of this application can be used in a wide variety of general-purpose or special-purpose computing system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer terminal devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.

[0104] For example, the above-described method and apparatus can be implemented as a computer program, which can be used in, for example... Figure 4 It runs on the computer device shown.

[0105] Please see Figure 4 , Figure 4 This is a schematic diagram of a computer device provided in an embodiment of this application. The computer device may be a server.

[0106] like Figure 4 As shown, the computer device includes a processor, memory, and network interface connected via a system bus, wherein the memory may include volatile storage media, non-volatile storage media, and internal memory.

[0107] Non-volatile storage media can store operating systems and computer programs. These computer programs include program instructions that, when executed, cause the processor to perform any risk assessment method.

[0108] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0109] Internal memory provides an environment for the execution of computer programs on non-volatile storage media, which, when executed by a processor, enable the processor to perform any risk assessment method.

[0110] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that the structure of this computer device is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0111] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0112] In some embodiments, the processor is used to run a computer program stored in a memory to perform the following steps: acquiring business data and extracting information from the business data to obtain business information corresponding to each business type and multiple user nodes; establishing association information between each user node based on the business information corresponding to each business type, and generating a multi-layer topology network based on the association information; performing network fusion on the multi-layer topology network to obtain a vector matrix corresponding to the multi-layer topology network; detecting and filtering each user node based on a preset node detection model and the vector matrix corresponding to the multi-layer topology network to obtain target user nodes; and performing risk prediction on the user node group associated with the target user node to obtain the risk assessment result corresponding to the user node group.

[0113] In some implementations, the processor is further configured to classify the business data based on a preset business category to obtain business information corresponding to each business type; and to traverse the business information to extract multiple user nodes corresponding to each business information.

[0114] In some implementations, the processor is further configured to determine target user nodes from multiple user nodes based on service information corresponding to each service type, and establish association information between each target user node; generate association edges between each target user node based on the association information between each target user node; and construct a network for the multiple target user nodes based on the association edges between each target user node to obtain a multi-layer topology network.

[0115] In some embodiments, the processor is further configured to perform vector transformation processing on the topological relationship network of each layer to obtain the vector matrix corresponding to the topological relationship network of each layer; input the vector matrix corresponding to the topological relationship network of each layer into a preset graph attention model to obtain the weight ratio corresponding to each vector matrix; and perform weighted processing on the vector matrix corresponding to the topological relationship network of each layer according to the weight ratio corresponding to each vector matrix to obtain the vector matrix corresponding to the multi-layer topological relationship network.

[0116] In some embodiments, the processor is further configured to perform linear transformation processing on each of the vector matrices through an activation function to obtain attention coefficients corresponding to each of the vector matrices; and to perform regularization processing on the attention coefficients corresponding to each of the vector matrices to obtain weight ratios corresponding to each of the vector matrices.

[0117] In some embodiments, the processor is further configured to identify the user type of each user node based on the vector matrix corresponding to the multi-layer topology network, and obtain the user type of each user node; if the user type of the user node is a preset user type, then the user node is used as the target user node.

[0118] In some implementations, the processor is further configured to perform clustering processing on the target user nodes based on a preset clustering algorithm to obtain user node groups associated with the target user nodes; and generate risk assessment results corresponding to the user node groups based on the number of target user nodes in the user node groups.

[0119] This application also provides a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed, implement any of the risk assessment methods provided in this application.

[0120] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiments, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device.

[0121] Furthermore, the computer-readable storage medium may primarily include a program storage area and a data storage area, wherein the program storage area may store the operating system, at least one application required for a function, etc.; and the data storage area may store data created based on the use of blockchain nodes, etc.

[0122] This invention refers to a novel application model of computer technologies such as storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms in the blockchain language model. A blockchain, essentially a decentralized database, is 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 an underlying blockchain platform, a platform product service layer, and an application service layer.

[0123] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A risk assessment method, characterized in that, The method includes: Acquire business data and extract information from the business data to obtain business information corresponding to each business type and multiple user nodes; Establish the association information between each user node based on the business information corresponding to each business type, and generate a multi-layer topology network based on the association information; The multi-layer topological relationship network is fused to obtain the vector matrix corresponding to the multi-layer topological relationship network; Based on a preset node detection model, each user node is detected and filtered according to the vector matrix corresponding to the multi-layer topology network to obtain the target user node. Risk prediction is performed on the user node group associated with the target user node to obtain the risk assessment result corresponding to the user node group; The step of performing network fusion on the multi-layer topology network to obtain the vector matrix corresponding to the multi-layer topology network includes: The topological relationship network of each layer is processed by vector transformation to obtain the vector matrix corresponding to the topological relationship network of each layer. Each row vector of the vector matrix is ​​used to represent the low-dimensional representation row vector of a user node. The vector matrices corresponding to the topological relationship network of each layer are input into a preset graph attention model, and attention weights are assigned to each vector matrix to obtain the weight ratios corresponding to each vector matrix. The vector matrices corresponding to each layer of the topological relationship network are weighted and summed according to the weight ratio of each vector matrix to obtain the vector matrix corresponding to the multi-layer topological relationship network. The vector matrix corresponding to the multi-layer topological relationship network has a stronger representation ability for the user node than the vector matrix corresponding to each layer of the topological relationship network.

2. The method according to claim 1, characterized in that, The process of extracting information from the business data to obtain business information corresponding to each business type and multiple user nodes includes: The business data is classified and processed based on preset business categories to obtain business information corresponding to each business type. The business information is traversed to extract multiple user nodes corresponding to each business information.

3. The method according to claim 1, characterized in that, The step of establishing association information between user nodes based on business information corresponding to each business type, and generating a multi-layer topology network based on the association information, includes: Based on the business information corresponding to each business type, target user nodes are determined from multiple user nodes, and association information between each target user node is established. Generate association edges between the target user nodes based on the association information between them; Based on the associated edges between the target user nodes, a network is constructed for the multiple target user nodes to obtain a multi-layer topological network.

4. The method according to claim 1, characterized in that, The step of inputting the vector matrices corresponding to the topological relationship network of each layer into a preset graph attention model to obtain the weight ratios corresponding to each vector matrix includes: By performing a linear transformation on each of the vector matrices using an activation function, attention coefficients corresponding to each of the vector matrices are obtained. The attention coefficients corresponding to each vector matrix are regularized to obtain the weight ratios corresponding to each vector matrix.

5. The method according to claim 1, characterized in that, The node detection model based on a preset method detects and filters each user node according to the vector matrix corresponding to the multi-layer topological relationship network to obtain the target user node, including: Based on the vector matrix corresponding to the multi-layer topology network, user type identification is performed on each user node to obtain the user type of each user node; If the user type of the user node is a preset user type, then the user node will be used as the target user node.

6. The method according to claim 1, characterized in that, The step of performing risk prediction on the user node group associated with the target user node to obtain the risk assessment result corresponding to the user node group includes: Based on a preset clustering algorithm, the target user nodes are clustered to obtain the user node groups associated with the target user nodes; Based on the number of target user nodes in the user node group, generate the risk assessment result corresponding to the user node group.

7. A risk assessment device, characterized in that, include: The business information extraction module is used to acquire business data and extract information from the business data to obtain business information corresponding to each business type and multiple user nodes. The network construction module is used to establish the association information between each user node according to the business information corresponding to each business type, and to generate a multi-layer topology network according to the association information. The network fusion module is used to perform network fusion on the multi-layer topology network to obtain the vector matrix corresponding to the multi-layer topology network. The node detection module is used to detect and filter each user node based on a preset node detection model and the vector matrix corresponding to the multi-layer topology network to obtain the target user node. The risk assessment module is used to predict the risk of the user node group associated with the target user node and obtain the risk assessment result corresponding to the user node group. The step of performing network fusion on the multi-layer topology network to obtain the vector matrix corresponding to the multi-layer topology network includes: The topological relationship network of each layer is processed by vector transformation to obtain the vector matrix corresponding to the topological relationship network of each layer. Each row vector of the vector matrix is ​​used to represent the low-dimensional representation row vector of a user node. The vector matrices corresponding to the topological relationship network of each layer are input into a preset graph attention model, and attention weights are assigned to each vector matrix to obtain the weight ratios corresponding to each vector matrix. The vector matrices corresponding to each layer of the topological relationship network are weighted and summed according to the weight ratio of each vector matrix to obtain the vector matrix corresponding to the multi-layer topological relationship network. The vector matrix corresponding to the multi-layer topological relationship network has a stronger representation ability for the user node than the vector matrix corresponding to each layer of the topological relationship network.

8. A computer device, characterized in that, The computer device includes a memory and a processor; The memory is used to store computer programs; The processor is configured to execute the computer program and, in executing the computer program, implement: The risk assessment method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, causes the processor to implement the risk assessment method as described in any one of claims 1-6.

Images