Card raising user identification method and device, equipment, storage medium and program product

Abstract

The invention relates to the technical field of artificial intelligence, and provides a card-raising user identification method, device and equipment, a storage medium and a program product, and the method comprises the steps: building a heterogeneous graph according to the feature data of a user and the relation data between the user and an agent, a base station and a terminal; in the heterogeneous graph, the user, the agent, the base station and the terminal are all used as nodes; a meta-path in the heterogeneous graph is determined, the meta-path is a connection mode of two nodes in the heterogeneous graph, and different meta-paths correspond to different semantic information; determining an embedded representation of the node according to the meta-path and an attention mechanism; and inputting the embedded representation as a feature vector into a card-raising user identification model to obtain an identification result output by the user identification model. According to the method, the heterogeneous graph with the users, the agents, the base stations and the terminals as nodes is constructed, on the basis of the isolated features of the users, the structural features of the graph, namely the relation features between the users, are fused, the card-raising users are classified and recognized, and the recognition precision is improved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to methods, devices, equipment, storage media, and program products for identifying cardholders. Background Technology

[0002] In pursuit of profit, some social channel agents use other people's information to fraudulently open SIM cards and simulate real user activity to evade operator audits. This behavior not only siphons large sums of money from operators and wastes marketing resources, but also allows criminals to use these fake SIM cards for internet registration, marketing, and fraudulent transactions, severely disrupting market order.

[0003] Traditional methods for identifying credit card users mainly rely on features extracted from isolated samples to build identification models. This approach often overlooks the potential features of relationships between users, resulting in incomplete information extraction. Summary of the Invention

[0004] This invention provides a method, apparatus, device, storage medium, and program product for identifying cardholders, which solves the shortcomings of existing identification methods that establish cardholder identification models based on isolated features of samples, neglecting the relationship features between users and resulting in insufficient information extraction.

[0005] This invention provides a method for identifying card-using users, comprising: establishing a heterogeneous graph based on user feature data and relationship data between users and agents, base stations, and terminals; in the heterogeneous graph, users, agents, base stations, and terminals are all treated as nodes; determining meta-paths in the heterogeneous graph, where a meta-path is the connection method between two nodes in the heterogeneous graph, and different meta-paths correspond to different semantic information; determining the embedding representation of nodes based on meta-paths and an attention mechanism; and inputting the embedding representation as a feature vector into a card-using user identification model to obtain the identification result output by the user identification model.

[0006] According to the present invention, a method for identifying cardholders is provided to determine meta-paths in a heterogeneous graph, including: determining basic meta-paths based on the relationships between nodes in the heterogeneous graph; and expanding the basic meta-paths based on edge weight information to obtain derived meta-paths as meta-paths in the heterogeneous graph.

[0007] According to the present invention, a method for identifying cardholders based on credit cards determines the embedding representation of nodes based on meta-paths and attention mechanisms, including: using a self-control mechanism to determine the weight relationship between nodes in a derived meta-path to obtain node-level attention; calculating the attention of the derived meta-path to obtain the importance of the derived meta-path; and determining the embedding representation of nodes based on the node-level attention and the importance of the derived meta-path. The embedding representation includes the degree of mutual influence between users under the same meta-path and the importance of the meta-path.

[0008] According to the present invention, a method for identifying cardholders is provided, which determines the embedding representation of a node based on the importance of node-level attention and the derivative path, including: normalizing the importance of the derivative path to obtain a normalized first weight of the derivative path; and weighting the node-level attention according to the first weight to obtain the embedding representation of the node.

[0009] According to the present invention, a method for identifying cardholders is provided, which expands the basic meta-path based on edge weight information to obtain a derived meta-path as a meta-path in a heterogeneous graph. The method includes: dividing the edge weights of the heterogeneous graph into multiple intervals; performing path derivation based on the multiple intervals; and combining the derived meta-path with the basic meta-path to obtain the derived meta-path as a meta-path in the heterogeneous graph.

[0010] According to the present invention, a method for identifying cardholders is provided, which establishes a heterogeneous graph based on user feature data and relationship data between users and agents, base stations, and terminals. The method includes: determining the node types of the heterogeneous graph based on the relationships between users, wherein the node types include users, agents, base stations, and terminals; acquiring user feature data, wherein the feature data includes at least one of basic features, business features, and behavioral features; acquiring relationship data between users and agents, base stations, and terminals within a preset period; and forming a heterogeneous graph with relationship weights based on the feature data and relationship data. In the heterogeneous graph, a line connecting two points indicates a relationship, and the value of the line represents the weight.

[0011] This invention also provides a card-paying user identification device, comprising: a data acquisition and heterogeneous graph generation module, used to establish a heterogeneous graph based on user feature data and relationship data between users and agents, base stations, and terminals; in the heterogeneous graph, users, agents, base stations, and terminals are all treated as nodes; a meta-path generation module, used to determine meta-paths in the heterogeneous graph, where a meta-path is the connection method between two nodes in the heterogeneous graph, and different meta-paths correspond to different semantic information; and a node embedding representation and identification module, used to determine the embedding representation of nodes based on meta-paths and an attention mechanism; and inputting the embedding representation as a feature vector into a card-paying user identification model to obtain the identification result output by the user identification model.

[0012] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement any of the above-described methods for identifying cardholders.

[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the cardholder identification methods described above.

[0014] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements any of the above-described methods for identifying cardholders.

[0015] This invention provides a method, apparatus, device, storage medium, and program product for identifying card-farming users. The method includes: establishing a heterogeneous graph based on user feature data and relationship data between users and agents, base stations, and terminals; in the heterogeneous graph, users, agents, base stations, and terminals are all treated as nodes; determining meta-paths in the heterogeneous graph, where a meta-path is the connection method between two nodes, and different meta-paths correspond to different semantic information; determining the embedding representation of nodes based on meta-paths and an attention mechanism; and inputting the embedding representation as a feature vector into a card-farming user identification model to obtain the identification result output by the user identification model. Through the above method, this invention proposes constructing a heterogeneous graph with users, agents, base stations, and terminals as nodes, integrating the structural features of the graph, i.e., the relationship features between users, into the isolated features of users to classify and identify card-farming users, thereby improving identification accuracy. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the cardholder identification method provided in this embodiment of the invention.

[0018] Figure 2 This is a schematic diagram of the heterogeneous graph containing four types of nodes: users, agents, base stations, and terminals, provided in an embodiment of the present invention.

[0019] Figure 3 This is a schematic diagram of the structure of the card-holding user identification device provided in an embodiment of the present invention.

[0020] Figure 4 This is a schematic diagram of the physical structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0022] In the description of the embodiments of the present invention, it should be noted that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the embodiments of the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the embodiments of the present invention. In addition, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0023] In the description of the embodiments of the present invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "connected" and "linked" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in the embodiments of the present invention based on the specific circumstances.

[0024] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0025] This invention provides a method for identifying card-farming users, which takes into account the relationship characteristics between users, extracts sufficient information, and can incorporate multiple relationship types, thus fully mining relationship information. By constructing a heterogeneous graph with users, agents, base stations, and terminals as nodes, and incorporating the structural features of the graph, the method classifies and identifies card-farming users, thereby improving the identification accuracy.

[0026] Please see Figure 1 , Figure 1 This is a flowchart illustrating the method for identifying card-farming users provided in an embodiment of the present invention. In this embodiment, the method for identifying card-farming users may include steps S110 to S140, each step of which is as follows: S110: Based on user characteristic data and relationship data between users and agents, base stations, and terminals, a heterogeneous graph is established; in the heterogeneous graph, users, agents, base stations, and terminals are all treated as nodes.

[0027] S120: Determine the meta-path in the heterogeneous graph. The meta-path is the way two nodes are connected in the heterogeneous graph. Different meta-paths correspond to different semantic information.

[0028] S130: Determine the embedding representation of the node based on the metapath and attention mechanism.

[0029] S140: Input the embedded representation as a feature vector into the card user identification model to obtain the identification result output by the user identification model.

[0030] Through the above methods, this embodiment of the invention proposes to construct a heterogeneous graph with users, agents, base stations, and terminals as nodes. Based on the isolated features of users, the structural features of the graph, namely the relationship features between users, are incorporated to classify and identify card-maintaining users, thereby improving the identification accuracy.

[0031] Specifically, the cardholder identification method in this embodiment is implemented based on an attention-based graph neural network framework. First, data acquisition and heterogeneous graph generation are required.

[0032] In some embodiments, the step of establishing a heterogeneous graph based on user characteristic data and relationship data between users and agents, base stations, and terminals may specifically include: Based on the relationships between users, the node types of the heterogeneous graph are determined, including users, agents, base stations, and terminals; user feature data is obtained, including at least one of basic features, business features, and behavioral features; relationship data between users and agents, base stations, and terminals within a preset period is obtained; and a heterogeneous graph with relationship weights is formed based on the feature data and relationship data; in the heterogeneous graph, a line between two points indicates a relationship, and the value of the line indicates the weight.

[0033] The node types of the heterogeneous graph can be determined based on the possible relationships between users. In this embodiment, users, agents, base stations, and terminals are used as nodes.

[0034] Once the node type is determined, the following data will be collected: 1) Node feature data, such as user node feature data including but not limited to age, place of origin, package, consumption, etc., and agent node feature data including but not limited to star rating, years of franchise, monthly business volume, scale, etc. After the node feature data is collected, a node feature vector is formed.

[0035] 2) Relationship data between nodes, such as whether a user has opened an account with a certain agent, the number of days a user has accessed a certain base station within a certain period, and the number of days a user has used a certain terminal within a certain period, and form a weighted heterogeneous graph based on these data, with the weights being the relationship values ​​such as whether an account has been opened and the number of days.

[0036] Based on any of the above embodiments, the step of determining the meta-path in the heterogeneous graph may specifically include: The basic meta-paths are determined based on the relationships between nodes in the heterogeneous graph; the derived meta-paths are then expanded based on edge weight information on the basic meta-paths, which serve as meta-paths in the heterogeneous graph.

[0037] Based on any of the above embodiments, the step of determining the embedding representation of a node according to the meta-path and attention mechanism may specifically include: The weight relationships between nodes in the derived metapath are determined using a self-control mechanism to obtain node-level attention; the importance of the derived metapath is obtained by calculating the attention of the derived metapath; and the embedding representation of the node is determined based on the node-level attention and the importance of the derived metapath; wherein the embedding representation includes the degree of mutual influence between users under the same metapath and the importance of the metapath.

[0038] Based on any of the above embodiments, the step of determining the embedding representation of a node according to the importance of node-level attention and derived meta-paths may specifically include: The importance of the derived metapaths is normalized to obtain the normalized first weight of the derived metapaths; the node-level attention is weighted according to the first weight to obtain the embedding representation of the node.

[0039] Based on any of the above embodiments, the step of expanding the basic meta-path according to edge weight information to obtain the derived meta-path as a meta-path in the heterogeneous graph may specifically include: The edge weights of the heterogeneous graph are divided into multiple intervals; path derivation is performed based on the multiple intervals, and derived meta-paths are obtained by combining the basic meta-paths, which serve as meta-paths in the heterogeneous graph.

[0040] This embodiment constructs a heterogeneous graph with users, agents, terminals, and base stations as nodes, and uses a graph neural network based on a two-layer attention mechanism to fully explore the relationships between users, forming an embedded representation of user nodes for use by downstream recognition tasks.

[0041] During the generation of metapaths, the basic type of metapath can be determined with the user as the target node. For example, users may have opened accounts with the same agent, communicate under the same base station, use the same terminal, or combinations of the above relationships. In order to introduce relationship weights into the model while taking into account the computational complexity, the weights are discretized into several levels. Combined with the basic type of metapath, more metapaths are formed to mine deeper semantic information.

[0042] In the process of node embedding representation and recognition, the attention value between nodes based on meta-paths is calculated to obtain the node embedding representation based on meta-paths; the attention values ​​of meta-paths are then calculated and weighted to obtain the final embedding representation of the node. Node embeddings are then fused with the node's own feature vectors for classifier learning.

[0043] Heterogeneous graphs can contain more comprehensive information and richer semantics. In the scenario of identifying credit card fraud users, there are relationships between users, agents, base stations, and terminals. These relationships contain rich semantic information, such as users opening accounts with agents, communicating under the same base station, and using the same terminal. These are common anomalies in credit card fraud scenarios.

[0044] In this embodiment, users, agents, base stations, and terminals are selected as graph nodes, and feature data of user nodes and relationship data between users and agents, base stations, and terminals are collected to construct a heterogeneous graph.

[0045] First, user node feature data (i.e., the user feature data mentioned above) is collected. User node feature data can be extracted from several aspects, including basic features, business features, and behavioral features. The feature selection should reflect the differences in users' personal information as much as possible. Specific examples are as follows:

[0046] After the above feature data is collected, the user's initial feature vector is formed.

[0047] Furthermore, data on the relationships between users, agents, base stations, and terminals within a certain period is collected, as follows:

[0048] After the above data collection is completed, a heterogeneous graph with relational weights is generated. Please refer to [link / reference]. Figure 2 , Figure 2 This is a schematic diagram of the heterogeneous graph containing four types of nodes: users, agents, base stations, and terminals, provided in an embodiment of the present invention.

[0049] in On behalf of users 、 Representative Agent 、 Representing base stations 、 This represents the terminal. A line connecting two points indicates a relationship, and the line has a numerical weight.

[0050] The relationships between User Agents (UAs) represent account opening relationships; if such a relationship exists, its weight is 1. For example... and between, and between, and Both are 1.

[0051] The relationship between UBs represents the communication relationship between users under this base station, with the weight being the number of communication days, for example... and The number of days of communication between them is 3. and The number of days of communication between them is 5. and The number of days for communication between them is 10.

[0052] The relationship between UTs represents the usage of a user on this terminal, with the weight being the number of days used. For example... and The number of days of use between them is 1. and The number of days of use between them is 2. and The usage period is 7 days. and The usage period is 4 days.

[0053] A meta-path is a connection between two nodes in a heterogeneous graph. For example, when two users connect through the same terminal (UTU), it constitutes a meta-path. Different meta-paths correspond to different semantic information, representing a certain relationship between users.

[0054] In this embodiment, to integrate the edge weight information of heterogeneous graphs into the embedded representation of nodes, two concepts are proposed: basic meta-paths and derived meta-paths. Basic meta-paths are meta-paths that do not contain edge weight information and only represent the relationships between nodes. Derived meta-paths are derived from basic meta-paths by discretizing the edge weight values, thus expanding to include more meta-paths.

[0055] Examples of basic metapath types are as follows:

[0056] First, divide the edge weights in the heterogeneous graph into multiple intervals. For example, if the edge weight of UB is the number of communication days, it can be divided into the following three intervals. ,in , Based on an empirically set threshold, edges are classified into three categories—low, medium, and high—according to their weight range. Unbounded edges can then be derived as follows: , , Similarly, UT edges can be derived as , , Since the weight of UA is 1, no derivation is needed.

[0057] The above methods can be used to derive more meta-paths, incorporating edge weight information from heterogeneous graphs into the meta-paths to improve node representation. The derived meta-paths can be represented as follows: In this embodiment, the node embedding representation can be accomplished using a two-layer attention mechanism. The first layer is node-level attention, where each node's neighbors based on derivative meta-paths influence whether it is a card-maintenance user. The importance of these derivative meta-path neighbors to each node in the heterogeneous graph can be learned, and the representations of these meaningful neighbors are aggregated to form the node embedding. The second layer is derivative meta-path-level attention. A node embedding under a given derivative meta-path can only reflect a node from one perspective. Each derivative meta-path contains different semantic information. To obtain a more comprehensive node embedding, the semantic information of multiple derivative meta-paths needs to be fused.

[0058] 1. Node-level attention Self-attention mechanisms can be used to learn the weight relationships between nodes, given a derived meta-path. Its node pairs Node-level attention express right The degree of influence can be expressed as follows: ; Here, ‖ represents the concatenation operation. It is a derivative path Node-level attention vectors. (Note) It is asymmetric, nodes For nodes Importance and Nodes For nodes Their importance differs.

[0059] To facilitate model training, it can be done via... The function is normalized to obtain the node pairs. Based on Derivative Path Weighting coefficients , means as follows: ; in, Represents a node The set of all neighbors of the derived meta-path φ, including itself. This is the activation function.

[0060] Based on Derivative Path nodes The embedding of a can be represented by the feature vectors of all its neighbors: .

[0061] To reduce the variance of the embeddings and make the training process more stable, a multi-head attention mechanism can be introduced. This involves calculating and aggregating attention multiple times to obtain richer information. The expression for merging secondary attention calculations is as follows: .

[0062] 2. Derivative Meta-Path Attention The importance of a derived meta-path can be obtained by calculating its attention. To achieve this, node-level embeddings can be transformed into meta-path-level attention vectors through a nonlinear transformation. q Similarity, the importance of the derived meta-path p It is expressed as follows: ; in V For the derivative metapath p A set of nodes. W It is a weight matrix. b It is the deviation vector. q These are attention vectors, all of which can be learned and updated. It is the hyperbolic tangent function, a type of activation function.

[0063] After obtaining the importance of each derived metapath, through The function normalizes these values ​​to obtain the normalized weights of the derived path p. .

[0064] ; Where N is the number of derivative paths.

[0065] Normalized weights for each derivative path After the calculation is complete, then apply node-level attention. The final embedding of the node is obtained by weighting the result.

[0066] .

[0067] The final embedding of the node can be used for downstream classification tasks. Guided by the labeled data, the loss function is minimized through the backpropagation algorithm to optimize the relevant parameters of the model, which will not be elaborated here.

[0068] In summary, this invention proposes a method for identifying card-farming users based on an attention-based graph neural network. It constructs a heterogeneous graph with users, agents, base stations, and terminals as nodes. Building upon the isolated features of users, it incorporates structural features of the graph for classification and identification of card-farming users. This method not only uses isolated user features for identification but also considers the relationship features between users and agents, base stations, and terminals, improving identification accuracy. Furthermore, it constructs a heterogeneous graph with multiple relationship types as nodes (users, agents, base stations, and terminals) and employs a two-layer attention mechanism algorithm to mine the semantic information of the graph, forming the final embedding of user nodes for downstream identification. Finally, it proposes the concept of derived meta-paths, utilizing edge weight information in the graph to further expand the meta-paths and enrich the semantic information of the graph.

[0069] On the one hand, this invention, based on the isolated features of user nodes, uses a two-layer attention mechanism at the node level and the meta-path level to calculate the degree of influence between users and the importance of meta-paths, and integrates this information with the isolated user features to form the final embedding of user nodes, resulting in more comprehensive information mining. On the other hand, it fully considers the influencing factors of the credit card maintenance scenario, constructs a heterogeneous graph of users, agents, terminals, and base stations, which is more comprehensive, and the use of the graph attention mechanism algorithm is more advanced.

[0070] Specifically, node-level attention aims to learn the importance of a node and its meta-path-based neighbors, while path-level attention learns the importance of different meta-paths. By leveraging the importance learned from node-level and path-level attention, the importance of both nodes and meta-paths can be fully considered. The proposed model then generates node embeddings by hierarchically aggregating features from meta-path-based neighbors.

[0071] The present invention also provides a card-raising user identification device. The card-raising user identification device provided by the present invention will be described below. The card-raising user identification device described below can be referred to in correspondence with the card-raising user identification method described above.

[0072] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of the card-maintaining user identification device provided in an embodiment of the present invention. In this embodiment, the card-maintaining user identification device includes a data acquisition and heterogeneous graph generation module 310, a meta-path generation module 320, and a node embedding representation and identification module 330.

[0073] The data acquisition and heterogeneous graph generation module 310 is used to build a heterogeneous graph based on the user's characteristic data and the relationship data between the user and the agent, base station and terminal; in the heterogeneous graph, the user, agent, base station and terminal are all nodes.

[0074] The meta-path generation module 320 is used to determine the meta-path in the heterogeneous graph. The meta-path is the connection method between two nodes in the heterogeneous graph, and different meta-paths correspond to different semantic information.

[0075] The node embedding representation and recognition module 330 is used to determine the embedding representation of a node based on the meta-path and attention mechanism; the embedding representation is input as a feature vector into the card user recognition model to obtain the recognition result output by the user recognition model.

[0076] In the data acquisition and heterogeneous graph generation module 310, a heterogeneous graph is constructed, including user U, agent A, base station B and terminal T; in the meta-path generation module 320, multiple meta-paths can be generated; in the node embedding representation and recognition module, node embedding needs to be determined through node attention and path attention, so that the recognition result of card-maintaining users can be obtained after neural network training.

[0077] Based on any of the above embodiments, the meta-path generation module 320 can specifically be used for: The basic meta-paths are determined based on the relationships between nodes in the heterogeneous graph; the derived meta-paths are then expanded based on edge weight information on the basic meta-paths, which serve as meta-paths in the heterogeneous graph.

[0078] Based on any of the above embodiments, the node embedding representation and recognition module 330 can specifically be used for: The weight relationships between nodes in the derived metapath are determined using a self-control mechanism to obtain node-level attention; the importance of the derived metapath is obtained by calculating the attention of the derived metapath; and the embedding representation of the node is determined based on the node-level attention and the importance of the derived metapath; wherein the embedding representation includes the degree of mutual influence between users under the same metapath and the importance of the metapath.

[0079] Based on any of the above embodiments, the node embedding representation and recognition module 330 can specifically be used for: The importance of the derived metapaths is normalized to obtain the normalized first weight of the derived metapaths; the node-level attention is weighted according to the first weight to obtain the embedding representation of the node.

[0080] Based on any of the above embodiments, the meta-path generation module 320 can specifically be used for: The edge weights of the heterogeneous graph are divided into multiple intervals; path derivation is performed based on the multiple intervals, and derived meta-paths are obtained by combining the basic meta-paths, which serve as meta-paths in the heterogeneous graph.

[0081] Based on any of the above embodiments, the data acquisition and heterogeneous graph generation module 310 can be specifically used for: Based on the relationships between users, the node types of the heterogeneous graph are determined, including users, agents, base stations, and terminals; user feature data is obtained, including at least one of basic features, business features, and behavioral features; relationship data between users and agents, base stations, and terminals within a preset period is obtained; and a heterogeneous graph with relationship weights is formed based on the feature data and relationship data; in the heterogeneous graph, a line between two points indicates a relationship, and the value of the line indicates the weight.

[0082] In summary, this invention provides a device for identifying card-farming users, comprising: a data acquisition and heterogeneous graph generation module, used to establish a heterogeneous graph based on user feature data and relationship data between users and agents, base stations, and terminals; in the heterogeneous graph, users, agents, base stations, and terminals are all treated as nodes; a meta-path generation module, used to determine meta-paths in the heterogeneous graph, where a meta-path is the connection method between two nodes in the heterogeneous graph, and different meta-paths correspond to different semantic information; and a node embedding representation and recognition module, used to determine the embedding representation of nodes based on meta-paths and an attention mechanism; the embedding representation is input as a feature vector into a card-farming user identification model to obtain the recognition result output by the user identification model. Through the above method, this invention proposes to construct a heterogeneous graph with users, agents, base stations, and terminals as nodes, integrating the structural features of the graph, i.e., the relationship features between users, on top of the isolated features of users, to classify and identify card-farming users, thereby improving recognition accuracy.

[0083] On the other hand, embodiments of the present invention also provide an electronic device, please refer to... Figure 4 , Figure 4 This is a schematic diagram of the physical structure of the electronic device provided in the embodiments of the present invention, such as... Figure 4 As shown, the electronic device may include: a memory 420, a processor 410, and a computer program stored in the memory 420 and executable on the processor 410. When the processor 410 executes the program, it implements the card-based user identification methods provided by the methods described above.

[0084] Optionally, the electronic device may further include a communication bus 430 and a communication interface 440, wherein the processor 410, the communication interface 440, and the memory 420 communicate with each other via the communication bus 430. The processor 410 can call a computer program in the memory 420 to execute a card-based user identification method, which may include: Based on user feature data and relationship data between users and agents, base stations, and terminals, a heterogeneous graph is constructed. In the heterogeneous graph, users, agents, base stations, and terminals are all treated as nodes. Meta-paths in the heterogeneous graph are determined. A meta-path is the way two nodes are connected in the heterogeneous graph, and different meta-paths correspond to different semantic information. The embedding representation of the nodes is determined based on the meta-paths and the attention mechanism. The embedding representation is used as a feature vector and input into the card-paying user identification model to obtain the identification result output by the user identification model.

[0085] Furthermore, the logical instructions in the aforementioned memory 420 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0086] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the card user identification method provided by the above methods. The steps and principles of the method have been described in detail in the above methods and will not be repeated here.

[0087] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the card user identification method provided by the above methods. The steps and principles of the method have been described in detail in the above methods and will not be repeated here.

[0088] Non-transitory computer-readable storage media can be any available medium or data storage device that can be accessed by a processor, including but not limited to magnetic storage (e.g., floppy disks, hard disks, magnetic tapes, magneto-optical disks (MOs), etc.), optical storage (e.g., CDs, DVDs, BDs, HVDs, etc.), and semiconductor storage (e.g., ROMs, EPROMs, EEPROMs, non-volatile memory (NAND flash), solid-state drives (SSDs)).

[0089] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0090] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying cardholders, characterized in that, include: A heterogeneous graph is established based on user characteristic data and the relationship data between the user and agents, base stations, and terminals. In the heterogeneous graph, the user, the agent, the base station, and the terminal all serve as nodes; Determine the meta-paths in the heterogeneous graph, where a meta-path is the connection method between two nodes in the heterogeneous graph, and different meta-paths correspond to different semantic information; The embedding representation of the node is determined based on the meta-path and attention mechanism; The embedded representation is input as the feature vector into the cardholder identification model to obtain the identification result output by the user identification model.

2. The method for identifying cardholders according to claim 1, characterized in that, Determining the meta-path in the heterogeneous graph includes: The basic metapath is determined based on the relationships between nodes in the heterogeneous graph; The basic meta-path is extended based on edge weight information to obtain derived meta-paths, which serve as meta-paths in the heterogeneous graph.

3. The method for identifying cardholders according to claim 2, characterized in that, The step of determining the node embedding representation based on the meta-path and attention mechanism includes: The weight relationships between nodes in the derived metapath are determined using a self-control mechanism to obtain node-level attention. The importance of the derivatized path is obtained by calculating the attention of the derivatized path; The embedding representation of the node is determined based on the importance of the node-level attention and the derived meta-path; The embedded representation includes the degree of mutual influence between users under the same metapath and the importance of the metapath.

4. The method for identifying cardholders according to claim 3, characterized in that, Determining the embedding representation of a node based on the importance of the node-level attention and the derived meta-path includes: The importance of the derived path is normalized to obtain the normalized first weight of the derived path; The node-level attention is weighted according to the first weight to obtain the embedded representation of the node.

5. The method for identifying cardholders according to claim 2, characterized in that, The step of expanding the basic meta-path based on edge weight information to obtain derived meta-paths as meta-paths in the heterogeneous graph includes: The edge weights of the heterogeneous graph are divided into multiple intervals; Path derivation is performed based on the multiple intervals, and the derived meta-paths are obtained by combining the basic meta-paths, which serve as meta-paths in the heterogeneous graph.

6. The method for identifying cardholders according to any one of claims 1 to 5, characterized in that, The step of establishing a heterogeneous graph based on user characteristic data and the relationship data between the user and agents, base stations, and terminals includes: Based on the relationships between users, the node types of the heterogeneous graph are determined, wherein the node types include users, agents, base stations, and terminals; Acquire user feature data; wherein the feature data includes at least one of basic features, business features, and behavioral features; Acquire data on the relationships between the user and the agent, base station, and terminal within a preset period; Based on the feature data and the relation data, a heterogeneous graph with relation weights is formed; In the heterogeneous graph, a line connecting two points indicates a relationship, and the value of the line represents the weight.

7. A user identification device for card-based services, characterized in that, include: The data acquisition and heterogeneous graph generation module is used to build a heterogeneous graph based on the user's feature data and the relationship data between the user and the agent, base station, and terminal. In the heterogeneous graph, the user, the agent, the base station, and the terminal all serve as nodes; The meta-path generation module is used to determine the meta-paths in the heterogeneous graph. The meta-path is the connection method between two nodes in the heterogeneous graph, and different meta-paths correspond to different semantic information. The node embedding representation and recognition module is used to determine the node embedding representation based on the meta-path and attention mechanism; and input the embedding representation as the feature vector into the card-holding user recognition model to obtain the recognition result output by the user recognition model.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the cardholder identification method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the cardholder identification method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the cardholder identification method as described in any one of claims 1 to 6.

Images