A method, apparatus, electronic device, and storage medium for identifying a target merchant.

Abstract

This application relates to the field of computer technology and provides a method, device, electronic device, and storage medium for identifying target merchants. It constructs a merchant network graph by using merchants in a transaction dataset as nodes and their common transaction objects as edges. The merchant network graph is then divided into multiple communities based on a modularity algorithm. Each community is input into a distributed representation learning model to obtain a graph structure feature vector for each community. This vector is then concatenated with the transaction features of the merchants to obtain a concatenated result. This concatenated result is then input into an isolated forest classification model to identify unknown target merchant communities. This application's embodiments can identify intermediary merchants and / or individuals involved in abnormal transactions without requiring any black tags, and can also identify unknown abnormal transactions, uncovering uncovered high-risk merchants, thereby further improving the management level of the transaction system.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a method, apparatus, electronic device, and storage medium for identifying target merchants. Background Technology

[0002] With the continuous innovation and rapid development of instant messaging systems, payment services based on these systems are becoming increasingly sophisticated and mature. However, various irregular transactions are emerging, posing a significant threat to the security of payment services. Existing abnormal transaction identification solutions generally involve training sample financial transaction data using deep learning or machine learning models, and then predicting abnormal transactions based on the trained deep learning module or machine learning model and the financial transaction data to be predicted. This process primarily relies on the transaction characteristics in the financial transaction data for training and prediction. However, the selection of these transaction characteristics often requires prior experience, making it difficult to detect unknown abnormal transactions. Summary of the Invention

[0003] To address the problem that existing technologies, which predict abnormal transactions based on transaction characteristics, struggle to detect unknown abnormal transactions, this application provides a method, apparatus, electronic device, and storage medium for identifying target merchants:

[0004] According to a first aspect of this application, a method for identifying a target merchant is provided, comprising:

[0005] The transaction data set is used to determine the set of merchants to be identified and the set of transaction objects of each merchant in the set of merchants to be identified; the transaction data set includes the transaction data generated by the merchants to be identified within a preset time period.

[0006] Based on the set of merchants to be identified and the set of transaction objects of each merchant to be identified, a first target graph is constructed. The first node in the first target graph represents a merchant to be identified, and the edge in the first target graph represents that there are common transaction objects between two connected merchants to be identified. The weight of the edge represents the proportion of common transaction objects.

[0007] The first nodes in the first target graph are grouped according to the weights corresponding to the edges in the first target graph, and a second target graph is generated based on the grouping results; the second nodes in the second target graph represent the first node group, and the first node group includes multiple first nodes.

[0008] Based on the second target map, each second node in the second target map is represented by a vector to obtain the group features of the merchant group to be identified indicated by each second node;

[0009] The group features and transaction features of each merchant group to be identified are concatenated. Based on the concatenation results of each merchant group to be identified, the merchant groups to be identified are classified to obtain the classification results corresponding to each merchant group to be identified. The classification results indicate whether the merchant group to be identified is a normal transaction merchant.

[0010] According to a second aspect of this application, a target merchant identification device is provided, comprising:

[0011] The determination module is used to determine the set of merchants to be identified and the set of transaction objects of each merchant to be identified in the set of merchants to be identified from the transaction data set; the transaction data set includes the transaction data generated by the merchants to be identified within a preset time period;

[0012] The construction module is used to construct a first target graph based on the set of merchants to be identified and the set of transaction objects of each merchant to be identified. The first node in the first target graph represents a merchant to be identified, and the edge in the first target graph represents that there is a common transaction object between two connected merchants to be identified. The weight of the edge represents the proportion of the common transaction object.

[0013] The generation module is used to group the first nodes in the first target graph according to the weights corresponding to the edges in the first target graph, and generate a second target graph based on the grouping result; the second nodes in the second target graph represent the first node group, and the first node group includes multiple first nodes.

[0014] The vector representation module is used to perform vector representation on each second node in the second target map based on the second target map, so as to obtain the group features of the merchant group to be identified indicated by each second node;

[0015] The classification processing module is used to concatenate the group features and transaction features of each merchant group to be identified. Based on the concatenation result of each merchant group to be identified, the module performs classification processing to obtain the classification result corresponding to each merchant group to be identified. The classification result indicates whether the merchant group to be identified is a normal transaction merchant.

[0016] Optionally, a construction module is used to determine a common transaction object for any two merchants in the set of merchants to be identified, based on the transaction object set corresponding to each of the two merchants to be identified, so as to obtain a common transaction object set for any two merchants to be identified.

[0017] The weight of the edge between any two merchants to be identified is determined by the ratio of the number of common transaction objects in the common transaction object set to the total number of transaction objects; the total number of transaction objects represents the total number of transaction objects in the transaction object set corresponding to any two merchants to be identified.

[0018] Using the merchants to be identified in the set of merchants to be identified as the first node, construct the first target graph according to the weights of the edges between any two merchants to be identified.

[0019] Optionally, the generation module includes:

[0020] The first determination submodule is used to determine multiple target starting nodes corresponding to the current grouping process from the first target map; the target starting node is any one of the first nodes;

[0021] The construction submodule is used to construct the group corresponding to each target starting node, starting from each target starting node, to obtain the group map corresponding to multiple target starting nodes;

[0022] The first determination submodule is used to determine the modularity of the current grouping process based on the weights of the edges between any two target starting nodes in the grouped graph. The modularity represents the tightness of the connection between any two target starting nodes in the grouped graph.

[0023] The second determination submodule is used to determine the maximum modularity among the multiple modularities corresponding to multiple grouping processes, and to determine the grouping graph in the grouping process corresponding to the maximum modularity as the target grouping graph.

[0024] The aggregation submodule is used to aggregate the set of associated nodes corresponding to each target starting node in the target group map, to obtain the first node group corresponding to each target starting node, and generate the second target map.

[0025] Optionally, a submodule is constructed to determine, from the first node connected to each target starting node, the set of associated nodes corresponding to each target starting node in the current clustering process;

[0026] Clustering is performed on the target starting node and the set of associated nodes corresponding to the target starting node to determine the grouping corresponding to each target starting node in the current clustering process;

[0027] Clustering is performed on the group corresponding to each target starting node in the current clustering process to obtain the clustering group corresponding to the current clustering process;

[0028] The cluster groups obtained from multiple clustering processes are used as the grouping graphs corresponding to multiple target starting nodes.

[0029] Optionally, a submodule is constructed to determine, from the first nodes connected to each target starting node, the first set of associated nodes corresponding to each target starting node in the current clustering process, and to determine, from the first nodes connected to each first associated node, the second set of associated nodes corresponding to each first associated node in the current clustering process.

[0030] Wherein, the weight of the edge between each first associated node in the first associated node set and the target starting node is greater than the first preset weight threshold, the weight of the edge between each second associated node in the second associated node set and the corresponding first associated node is greater than the second preset weight threshold, and the second preset weight threshold is less than or equal to the first preset weight threshold.

[0031] Clustering is performed on each target starting node, its corresponding first set of associated nodes, and its corresponding second set of associated nodes to obtain the grouping corresponding to each target starting node;

[0032] For any two target starting nodes, if one of the two target starting nodes is the first or second associated node of the other target starting node, perform clustering on the groups corresponding to the two target starting nodes to obtain the clustered groups corresponding to the current clustering process.

[0033] The cluster groups obtained from multiple clustering processes are used as the grouping graphs corresponding to multiple target starting nodes.

[0034] Optionally, the first determining submodule is used to determine, for any two target starting nodes in the grouped graph, a first ratio of the weight corresponding to the edge between any two target starting nodes to the total weight; the total weight represents the sum of the weights corresponding to the edges in the grouped graph.

[0035] Determine the difference between the weight of the edge between any two target starting nodes and the first ratio;

[0036] The difference is calculated by multiplying the difference by the category parameters corresponding to any two target starting nodes to obtain the sum of the differences; the category parameters indicate whether any two target starting nodes belong to the same cluster group.

[0037] Determine the second ratio of the difference sum to the total weight to obtain the modularity corresponding to the grouped map.

[0038] Optionally, the vector representation module is used to determine the root node corresponding to the current root subgraph construction from the second target graph; the root node is any second node;

[0039] The set of first connected nodes for constructing the root subgraph is determined from the second nodes connected to the root node, and the set of second connected nodes for constructing the root subgraph is determined from the second nodes connected to each of the first connected nodes, thus obtaining the set of connected nodes corresponding to the root node; the set of connected nodes includes second nodes of a preset depth that are sequentially connected to the root node, and the preset depth represents the preset number of sequentially connected second nodes;

[0040] Based on the root node as the starting point and the set of connected nodes corresponding to the root node, construct the current root subgraph and construct multiple corresponding root subgraphs; the root node and the set of connected nodes corresponding to the root node in each root subgraph constitute the merchant group to be identified;

[0041] Multiple root subgraphs are constructed from multiple root subgraphs and then used as a root subgraph set.

[0042] Each root subgraph in the root subgraph set is represented by a vector, and the group characteristics of the merchant group to be identified corresponding to each root subgraph are determined, thus obtaining the graph structure vector set corresponding to each merchant group to be identified.

[0043] Optionally, the identification device for the target merchant may also include:

[0044] The third determination module is used to determine the transaction frequency information, transaction amount information, merchant stay duration information and transaction time information corresponding to the merchant group to be identified from the transaction data set;

[0045] The normalization module is used to normalize the transaction frequency information, transaction amount information, merchant stay duration information, and transaction time information corresponding to the merchant group to be identified, so as to obtain the transaction characteristics corresponding to the merchant group to be identified.

[0046] Optionally, the classification processing module is used to input the splicing result of each merchant group to be identified into the classification model to obtain the classification result output by the classification model.

[0047] According to a third aspect of this application, an electronic device is provided, comprising a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor to implement the target merchant identification method of the first aspect of this application.

[0048] According to a fourth aspect of this application, a computer storage medium is provided, wherein at least one instruction or at least one program is stored in the storage medium, and the at least one instruction or at least one program is loaded and executed by a processor to implement the method for identifying the target merchant of the first aspect of this application.

[0049] According to a fifth aspect of this application, a computer program product is provided, comprising at least one instruction or at least one program segment, wherein the at least one instruction or at least one program segment is loaded and executed by a processor to implement the target merchant identification method of the first aspect of this application.

[0050] The target merchant identification method, apparatus, electronic device, and storage medium provided in this application have the following technical effects:

[0051] The process involves determining the set of merchants to be identified and the set of transaction objects for each merchant within the set of merchants to be identified from a transaction data set. The transaction data set includes transaction data generated by the merchants to be identified within a preset time period. Based on the set of merchants to be identified and the set of transaction objects for each merchant, a first target graph is constructed. The first node in the first target graph represents a merchant to be identified, and the edges in the first target graph represent that two connected merchants have common transaction objects. The weight of the edge represents the proportion of common transaction objects. The first nodes in the first target graph are grouped according to the weights of the edges in the first target graph. The process involves generating a second target graph based on the grouping results. Second nodes in the second target graph represent first node groups, each containing multiple first nodes. Each second node in the second target graph is represented by a vector, yielding the group features of the merchant group to be identified. These group features are then concatenated with the transaction features of the merchant group to be identified. Based on the concatenation results, the merchant groups are classified to obtain the classification results for each group. The classification results indicate whether the merchant group to be identified is a normal transaction merchant. By combining the group features extracted from the distributed representation learning model of the graph with the transaction features of the merchant group to be identified (i.e., combining structured and unstructured features), and using an isolated forest model as input to determine target merchants, abnormal intermediary merchants and / or individuals can be identified without any black labels. Furthermore, unknown abnormal transactions can be identified, and uncovered high-risk merchants that have not yet been identified, further improving the management level of the transaction system. Attached Figure Description

[0052] To more clearly illustrate the technical solutions and advantages in the embodiments of this application 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 only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is a schematic diagram of an application environment provided in an embodiment of this application;

[0054] Figure 2 This is a flowchart illustrating a method for identifying a target merchant provided in an embodiment of this application;

[0055] Figure 3 This is a schematic diagram of a process for constructing a first target map provided in an embodiment of this application;

[0056] Figure 4This is a schematic diagram of a process for generating a second target map provided in an embodiment of this application;

[0057] Figure 5 This is a schematic diagram of a process for determining a grouped map of multiple target starting nodes provided in an embodiment of this application;

[0058] Figure 6 This is a flowchart illustrating a doc2vec algorithm provided in an embodiment of this application;

[0059] Figure 7 This is a flowchart illustrating a distributed representation learning algorithm for graphs provided in an embodiment of this application;

[0060] Figure 8 This is a flowchart illustrating an isolated forest model provided in an embodiment of this application;

[0061] Figure 9 This is a schematic diagram of the structure of a target merchant identification device provided in an embodiment of this application;

[0062] Figure 10 This is a schematic diagram of the hardware structure of an electronic device used to implement the target merchant identification method provided in this application embodiment. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings. Obviously, the described embodiments are merely one embodiment of this application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0064] The term "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of this application. In the description of the embodiments of this application, it should be understood that the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. Furthermore, the terms "first," "second," etc., are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data used can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. In addition, the terms "comprising," "having," and "being," and any variations thereof, are intended to cover non-exclusive inclusion.

[0065] It is understood that in the specific implementation of this application, transaction data and other related data are involved. When the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0066] Before providing a further detailed description of the embodiments of this application, the nouns and terms involved in the embodiments of this application will be explained, and the nouns and terms involved in the embodiments of this application shall be interpreted as follows.

[0067] The Graph2vec (Learning Distributed Representations of Graphs) model borrows ideas from the doc2vec paragraph vector method, treating the graph as a document and the rooted subgraphs surrounding each node as words. A vectorized representation of each rooted subgraph is obtained through the skip-gram neural network model.

[0068] Isolation forest model: Used for data mining, suitable for anomaly detection in continuous numerical data. Anomalies are defined as "outliers more likely to be separated," which can be understood as sparsely distributed points far from denser clusters. Statistically, in the data space, sparsely distributed regions indicate a low probability of data occurring in those regions; therefore, data falling within these regions can be considered anomalies.

[0069] Please see Figure 1 , Figure 1 This is a schematic diagram of an application environment provided in an embodiment of this application. The application environment may include a client 10 and a server 20. The client 10 and the server 20 can be directly or indirectly connected via wired or wireless communication.

[0070] In some possible embodiments, each client 10 can send a subset of transaction data to the server 20, and the server 20 can receive the set of transaction data generated by the client 10. The server 20 can provide a target merchant identification service, constructing a merchant network graph with merchants in the transaction data set as nodes and common transaction objects of merchants as edges. Then, based on a modularity algorithm, the merchant network graph is divided into multiple communities. Each community is input into a distributed representation learning model of the graph to obtain a graph structure feature vector for each community. This feature vector is then concatenated with the transaction features of the merchants to obtain a concatenated result. This concatenated result is then input into an isolated forest classification model to identify unknown target merchant communities.

[0071] Client 10 can be a physical device such as a smartphone, computer (e.g., desktop computer, tablet, laptop), augmented reality (AR) / virtual reality (VR) device, digital assistant, smart voice interaction device (smart speaker), smart wearable device, smart home appliance, or terminal; or it can be software running on the physical device, such as a computer program. The operating system corresponding to the client can be Android, iOS (a mobile operating system developed by Apple), Linux, Microsoft Windows, etc.

[0072] Server 20 can be a standalone physical server, a service cluster or distributed system composed of multiple physical servers, 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 (CDNs), and big data and artificial intelligence platforms. The server may include network communication units, processors, and memory, etc. The server can provide backend services to corresponding clients.

[0073] In one exemplary implementation, both client 10 and server 20 can be node devices in the blockchain system, capable of sharing acquired and generated information with other node devices in the blockchain system, thus enabling information sharing among multiple node devices. Multiple node devices in the blockchain system can be configured with the same blockchain, which consists of multiple blocks, and adjacent blocks are related, ensuring that any data tampering in any block can be detected by the next block. This prevents data tampering in the blockchain and guarantees the security and reliability of the data in the blockchain.

[0074] The following describes a specific embodiment of a target merchant identification method according to this application. Figure 2 This is a flowchart illustrating a method for identifying a target merchant provided in an embodiment of this application. This specification provides the method operation steps shown in the embodiments or flowcharts, but based on conventional or non-inventive labor, more or fewer operation steps may be included. The order of steps listed in the embodiments is merely one of many execution orders and does not represent the only execution order. In actual execution, the methods can be executed in the order shown in the embodiments or drawings, or in parallel (e.g., in a parallel processor or multi-threaded processing environment).

[0075] Specific examples Figure 2 As shown, the target merchant identification method may include:

[0076] S201: Determine the set of merchants to be identified and the set of transaction objects of each merchant to be identified from the transaction data set.

[0077] In this embodiment of the application, the transaction data set may include transaction data generated by the merchant to be identified within a preset time period.

[0078] In some possible implementations, the transaction data set can be the transaction data generated by each merchant in the merchant-to-be-identified set within a preset time period using an instant messaging payment system. The instant messaging payment system can be WeChat Pay or other systems with payment services; this application embodiment does not specifically limit the specific implementation. Each client's WeChat Pay system can upload each transaction record to the server. The server can store transaction records uploaded by multiple clients within the past month, or within the past year. In practical applications, the server can extract merchants with transactions within the past month from the transaction records within the past year as merchants to be identified, and then use these merchants and / or individual transactors within the past month as a transaction object set.

[0079] S203: Construct the first target graph based on the set of merchants to be identified and the set of transaction objects of each merchant to be identified.

[0080] In this embodiment of the application, the first target graph may include multiple first nodes and multiple edges. Each edge may connect two first nodes, and each edge may have a corresponding weight. Specifically, the first nodes in the first target graph may represent merchants to be identified, the edges in the first target graph may represent that two connected merchants to be identified have a common transaction partner, and the weight of the edge may represent the proportion of the common transaction partner.

[0081] Figure 3 This is a schematic diagram illustrating a process for constructing a first target atlas, provided in an embodiment of this application. In practical applications, it can be based on... Figure 3 The method and steps shown are as follows to construct the first target map:

[0082] S301: For any two merchants to be identified in the set of merchants to be identified, determine the common transaction object based on the transaction object set corresponding to each of the two merchants to be identified, and obtain the common transaction object set corresponding to the two merchants to be identified.

[0083] In some possible implementations, after extracting the merchants to be identified from the transaction flow, pairwise matching can be performed on these merchants to obtain merchant relationship pairs. The transaction merchants and / or transaction individuals for each merchant in each merchant relationship pair, as well as the common transaction merchants and / or transaction individuals for the two merchants in each merchant relationship pair, can then be identified. In each determination process, any two merchants can be randomly selected from the set of merchants to be identified. Based on the transaction merchants and / or transaction individuals corresponding to each of these two merchants, the common transaction merchants and / or individuals for these two merchants can be determined. Through multiple determination processes, the common transaction merchants and / or individuals corresponding to any two merchants in the set of merchants to be identified can be obtained.

[0084] The following example illustrates a specific implementation method for determining the common transaction object set corresponding to any two merchants to be identified.

[0085] In this determination process, a first merchant A and a second merchant B can be identified from the set of merchants to be identified, where both are any one of the merchants to be identified in the set. Then, based on the trading merchants and / or individuals {a,b} of the first merchant A and the trading merchants and / or individuals {a,b,c} of the second merchant B, the common trading merchants and / or individuals {a,b} of the first and second merchants to be identified can be determined.

[0086] S303: Determine the weight of the edge between any two merchants to be identified based on the ratio of the number of common trading objects in the common trading object set to the total number of trading objects.

[0087] In this embodiment of the application, the total number of transaction objects represents the total number of transaction objects in the transaction object set corresponding to any two merchants to be identified.

[0088] In some possible implementations, the weights of the edges between two unidentified merchants in a merchant relationship pair can be determined based on the common trading merchants and / or trading individuals of the two unidentified merchants in the pair, and the sum of the trading merchants and / or trading individuals of each unidentified merchant in the pair. When the weight of the edge between two unidentified merchants is zero, that is, when the two unidentified merchants have no common trading merchants and / or trading individuals, then there is no edge between the two unidentified merchants. In application, the weight ω of the edge between any two unidentified merchants can be determined using equation (1).

[0089]

[0090] For example, based on the examples listed above, the weight ω of the edge between the first merchant A to be identified and the second merchant B to be identified. AB It can be:

[0091]

[0092] S305: Using the merchants to be identified in the set of merchants to be identified as the first node, construct the first target graph according to the weights of the edges between any two merchants to be identified.

[0093] In some possible implementations, the merchants to be identified in the set of merchants to be identified can be used as first nodes, and the two first nodes corresponding to two merchants to be identified that have common transactions and / or individuals can be connected. In practical applications, the connection between the two first nodes corresponding to two merchants to be identified that have common transactions and / or individuals can be of any length, or it can be a length proportional to the weight of the edge between the two merchants to be identified. This application does not impose specific limitations on this.

[0094] S205: Group the first nodes in the first target graph according to the weights corresponding to the edges in the first target graph, and generate the second target graph based on the results of the grouping process.

[0095] In this embodiment of the application, the second target graph may include multiple second nodes and multiple edges, and each edge may connect two second nodes. The second nodes in the second target graph may represent a first node group, and the first node group may include multiple first nodes.

[0096] Modularity is an important metric for evaluating the quality of grouping processing; a higher modularity in the resulting grouped graph indicates better processing performance. The Fast Unfolding algorithm is based on modularity for grouping. The goal of grouping processing is to make the connections between each cluster in the resulting grouped graph more compact, while the connections between multiple clusters are more sparse. The Fast Unfolding algorithm consists of two phases. The first phase, called Modularity Optimization, assigns each first node to the cluster containing its connected target starting node, increasing the modularity of the grouped graph. The second phase, called Community Aggregation, aggregates the clusters from the first phase into a single point, essentially reconstructing the second target graph based on the grouped graph generated in the previous phase. This process is repeated until the network structure no longer changes.

[0097] Figure 4 This is a schematic diagram of a process for generating a second target map provided in an embodiment of this application. In practical applications, it can be based on... Figure 4 The method steps shown generate the second target map. The specific steps are as follows:

[0098] S401: Determine the starting nodes of multiple targets corresponding to the current grouping process from the first target map.

[0099] In this embodiment of the application, the target starting node can be any one of the first nodes in the first target map, and multiple target starting nodes can be any multiple first nodes in the first target map.

[0100] S403: Construct a group corresponding to each target starting node, starting from each target starting node, to obtain a group map corresponding to multiple target starting nodes.

[0101] In one possible implementation, the set of associated nodes corresponding to each target starting node in the current clustering process can be determined from the first node connected to each target starting node. Then, clustering can be performed on the target starting node and its associated node set to determine the group corresponding to each target starting node in the current clustering process. Furthermore, clustering can be performed on the group corresponding to each target starting node in the current clustering process to obtain the clustered groups corresponding to the current clustering process. The clustered groups obtained from multiple clustering processes can then be used as a grouping map corresponding to multiple target starting nodes.

[0102] Figure 5This is a schematic flowchart illustrating a grouping graph for determining multiple target starting nodes, provided in an embodiment of this application. The first target graph includes 16 first nodes, i.e., dots numbered 0 to 15. In the current clustering process, any 6 first nodes can be arbitrarily selected from these 16 first nodes as target starting nodes. For example, the first nodes numbered 0, 2, 7, 8, 10, and 13 can be selected as target starting nodes. Then, the first node numbered 5 can be used as the set of associated nodes for target starting node 0, the first nodes numbered 1 and 4 as the set of associated nodes for target starting node 2, the first nodes numbered 3 and 6 as the set of associated nodes for target starting node 7, the first nodes numbered 9 and 15 as the set of associated nodes for target starting node 8, the first nodes numbered 12 and 14 as the set of associated nodes for target starting node 10, and the first node numbered 11 as the set of associated nodes for target starting node 13. Next, the color or shape of the first node with numbers 0 and 5 can be unified; the color or shape of the first node with numbers 1, 2, and 4 can be unified; the shape of the first node with numbers 3, 6, and 7 can be unified; the shape of the first node with numbers 8, 9, and 15 can be unified; the shape of the first node with numbers 10, 12, and 14 can be unified; and the shape of the first node with numbers 11 and 13 can be unified. Then, the shape of the first node in the cluster containing numbers 0 and 5 and the cluster containing numbers 1, 2, and 4 can be unified; and the shape of the first node in the cluster containing numbers 8, 9, and 15 and the cluster containing numbers 10, 12, and 14 can be unified. Then, the cluster groups containing numbers 0, 1, 2, 4, 5, the cluster groups containing numbers 3, 6, 7, the cluster groups containing numbers 8, 9, 10, 12, 14, 15, and the cluster groups containing numbers 11, 13 can be used as the corresponding target starting nodes 0, 2, 7, 8, 10, 13 in the cluster map.

[0103] The above example illustrates the grouped graphs corresponding to multiple target starting nodes obtained after two clustering processes. In practical applications, since the multiple target starting nodes are arbitrarily determined, in the current clustering process, any three first nodes can be randomly selected from these 16 first nodes as target starting nodes. For example, the first nodes with the numbers 0, 6, and 9 can be selected as target starting nodes, and the above steps can be repeated to obtain the grouped graphs corresponding to target starting nodes 0, 6, and 9. Furthermore, since the set of associated nodes corresponding to each target starting node is arbitrarily determined, in the secondary clustering process of the example listed above, the first node with numbers 3, 4, and 5 can be used as the set of associated nodes for target starting node 0; the first node with number 1 can be used as the set of associated nodes for target starting node 2; the first node with number 6 can be used as the set of associated nodes for target starting node 7; the first nodes with numbers 9 and 15 can be used as the set of associated nodes for target starting node 8; the first nodes with numbers 12 and 14 can be used as the set of associated nodes for target starting node 10; and the first node with number 11 can be used as the set of associated nodes for target starting node 13. The above steps are repeated to obtain the grouped graphs corresponding to target starting nodes 0, 2, 7, 8, 10, and 13. Thus, by repeatedly determining arbitrarily different multiple target starting nodes and / or the set of associated nodes corresponding to each target starting node, grouped graphs corresponding to different multiple target starting nodes can be obtained in the first stage.

[0104] In some possible implementations, a first set of associated nodes corresponding to each target starting node in the current clustering process can be determined from the first nodes connected to each target starting node, and a second set of associated nodes corresponding to each first associated node in the current clustering process can be determined from the first nodes connected to each first associated node. Then, clustering can be performed on each target starting node, its corresponding first set of associated nodes, and its corresponding second set of associated nodes to obtain a group corresponding to each target starting node. For any two groups corresponding to target starting nodes, if one of the two target starting nodes is a first or second associated node of the other target starting node, clustering can be performed on the groups corresponding to the two target starting nodes to obtain the clustered group corresponding to the current clustering process. The clustered groups obtained from multiple clustering processes can then be used as a grouping map corresponding to multiple target starting nodes.

[0105] Wherein, the weight of the edge between each first associated node in the first associated node set and the target starting node is greater than the first preset weight threshold, the weight of the edge between each second associated node in the second associated node set and the corresponding first associated node is greater than the second preset weight threshold, and the second preset weight threshold is less than or equal to the first preset weight threshold.

[0106] S405: For any two target starting nodes in the grouped graph, determine the modularity of the current grouping process based on the weights of the edges between the two target starting nodes; the modularity characterizes the tightness of the connection between any two target starting nodes in the grouped graph.

[0107] In some possible implementations, for any two target starting nodes in the grouped graph, a first ratio of the weight of the edge between the two target starting nodes to the total weight can be determined, where the total weight can represent the sum of the weights of the edges in the grouped graph. Then, the difference between the weight of the edge between the two target starting nodes and the first ratio can be determined, and the product of the difference and the category parameters corresponding to the two target starting nodes can be determined to obtain the difference sum, where the category parameters can represent whether the two target starting nodes belong to the same cluster group. Then, a second ratio of the difference sum to the total weight can be determined to obtain the modularity corresponding to the grouped graph. In application, equation (2) can be used to determine the modularity Q corresponding to the grouped graph.

[0108] The weight between the starting node i and the target starting node j, k i k can represent the sum of the weights of the edges connected to the target starting node i. j It can represent the sum of the weights of the edges connected to the target starting node j. If σ = 1, it means that the target starting node i and the target starting node are in the same cluster group. If σ = 0, it means that the target starting node i and the target starting node are in two different cluster groups.

[0109] S407: Based on the multiple modularities corresponding to multiple grouping processes, determine the maximum modularity among the multiple modularities, and determine the grouping graph corresponding to the maximum modularity in the grouping process as the target grouping graph.

[0110] S409: Aggregate the set of associated nodes corresponding to each target starting node in the target group map to obtain the first node group corresponding to each target starting node, and generate the second target map.

[0111] Based on the example above of determining the grouped graphs corresponding to multiple target starting nodes in the first stage, in the second stage, clusters containing numbers 0, 1, 2, 4, and 5 can be aggregated into one point; clusters containing numbers 3, 6, and 7 can be aggregated into one point; clusters containing numbers 8, 9, 10, 12, 14, and 15 can be aggregated into one point; and clusters containing numbers 11 and 13 can be aggregated into one point. Then, the two aggregated points containing numbers 0, 1, 2, 4, and 5 and containing numbers 3, 6, and 7, and the two aggregated points containing numbers 8, 9, 10, 12, 14, and 15 and containing numbers 11 and 13 are aggregated to generate the second target graph.

[0112] S207: Based on the second target map, each second node in the second target map is represented by a vector to obtain the group features of the merchant group to be identified indicated by each second node.

[0113] In this embodiment, the root node corresponding to the current root subgraph construction can be determined from the second target graph, where the root node is any second node. Then, the first set of connecting nodes for the current root subgraph construction can be determined from the second nodes connected to the root node, and the second set of connecting nodes for the current root subgraph construction can be determined from the second nodes connected to each first connecting node, resulting in the connection node set corresponding to the root node. The connection node set includes second nodes sequentially connected to the root child node at a preset depth, where the preset depth represents the preset number of sequentially connected second nodes. Subsequently, multiple root subgraphs corresponding to the current root subgraph construction can be constructed based on the root node as the starting point and the connection node set corresponding to the root node. The root node and the connection node set corresponding to the root node in each root subgraph constitute the merchant group to be identified. The multiple root subgraphs corresponding to multiple root subgraph constructions are then used as a root subgraph set, and each root subgraph in the root subgraph set is represented by a vector to determine the group features of the merchant group to be identified corresponding to each root subgraph, resulting in the graph structure vector set corresponding to each merchant group to be identified.

[0114] Currently, distributed representation learning of graphs mainly involves learning the latent vectors of nodes or subgraphs within the graph. However, many real-world tasks, such as graph classification or clustering, require encoding the entire graph into a fixed-length vector. Furthermore, graph kernel-based methods suffer from poor versatility due to the use of custom features. Borrowing from the idea of ​​doc2vec, graph2vec can encode graphs of arbitrary size, representing the topological results of each graph as a vector. Figure 6 This is a flowchart illustrating a doc2vec algorithm provided in an embodiment of this application. Figure 7This is a flowchart illustrating a distributed representation learning algorithm for graphs provided in an embodiment of this application. The second graph generated in step S205 can be input into graph2vec for training, and the output is a set of rooted subgraphs corresponding to the second graph, such as rooted subgraph 1, rooted subgraph 2...rooted subgraph c.

[0115] S209: The group features of each merchant group to be identified are concatenated with the transaction features of the merchant group to be identified. Based on the concatenation result of each merchant group to be identified, the merchant group to be identified is classified to obtain the classification result corresponding to each merchant group to be identified. The classification result indicates whether the merchant group to be identified is a normal transaction merchant.

[0116] In this embodiment, while obtaining the vector representation of each merchant group to be identified, transaction characteristics of each merchant group can be introduced. Optionally, transaction frequency information, transaction amount information, merchant stay duration information, and transaction time information corresponding to the merchant group to be identified can be determined from the transaction data set. The transaction frequency information, transaction amount information, merchant stay duration information, and transaction time information corresponding to the merchant group to be identified can be normalized to obtain the transaction characteristics corresponding to the merchant group to be identified. Specifically, the transaction frequency information can be the average transaction frequency of multiple merchants in the merchant group to be identified; the transaction amount information can be the average transaction amount of multiple merchants in the merchant group to be identified; the merchant stay duration information can be the average stay duration of multiple merchants in the merchant group to be identified; and the transaction time information can be a transaction time period determined based on the transaction times of multiple merchants in the merchant group to be identified.

[0117] In this embodiment, after normalizing the variables such as transaction frequency, transaction amount, merchant stay duration, and transaction time for each merchant group to be identified, these variables can be concatenated with the group features extracted by the graph2vec model to obtain the concatenation result for each merchant group to be identified. Then, the concatenation result for each merchant group to be identified can be input into a classification model to obtain the classification result output by the model. The classification result can characterize whether the merchant group to be identified consists of normal transaction merchants. For example, if the classification result shows that the average transaction frequency of merchants in the merchant group to be identified is too high, the average transaction data is too large, the merchant stay duration is short, and the transaction time generally occurs at night, the merchant group to be identified can be determined as the target merchant group.

[0118] In some possible implementations, the vector representation of the merchant group to be identified can be a 16-dimensional vector, and the transaction features of the merchant group to be identified can include 5 transaction features, ultimately resulting in a 21-dimensional vector.

[0119] In some possible implementations, the classification model can be an isolation forest model. After obtaining the stitched results, the stitched results of each merchant group to be identified can be input into the isolation forest model to obtain the classification results output by the isolation forest model.

[0120] Figure 8 This is a flowchart illustrating an isolated forest model provided in an embodiment of this application. The specific computational process of the isolated forest model is as follows: Given a dataset of n data points, a batch of samples can be drawn from these n data points, typically using sampling without replacement. Assume the number of samples is ψ. Randomly select a feature from the concatenated results as the starting node, and randomly select a value from the range of that feature. Perform a binary split on the ψ samples, assigning samples with values ​​less than the assigned value to the left branch and samples with values ​​greater than the assigned value to the right branch. Repeat the binary split operation on the left and right branches until the dataset contains only one data point or the tree's height limit is reached.

[0121] Because there is relatively little abnormal data, and the feature values ​​differ significantly from normal data, abnormal data is closer to the root and normal data is farther from the root when constructing the tree. The classification result of a single tree is often unreliable. Therefore, the iForest algorithm can be used to sample multiple times, constructing multiple binary trees. Finally, the results of all trees are integrated, and the average depth is taken as the final output depth. This allows for the calculation of abnormal branches in the data.

[0122] The isolated forest model uses an abnormal value function s(x,n) to measure whether the data is abnormal. In the application process, equation (3) can be used to measure whether the data is abnormal.

[0123]

[0124] Here, E(h(x)) represents the expected path length of data x in multiple trees, H(*) represents the harmonic number, and θ represents Euler's constant, approximately 0.5772156649. As s approaches 1, the more abnormal the data becomes.

[0125] In this embodiment, after obtaining the classification results corresponding to each merchant group to be identified, the merchant groups can be pushed to a manual review process for merchant characterization, and high-risk merchants can be identified from the merchant groups. Alternatively, the tag information of each merchant in the merchant group to be identified can be obtained, and high-risk merchants can be identified from the merchant group based on the tag information of each merchant. The tag information can represent that the merchant to be identified belongs to the target merchant group during the historical identification process.

[0126] The target merchant identification method provided in this application combines the group features of the merchant group to be identified with the transaction features of the merchant group to be identified by the distributed representation learning model of the graph, and uses the isolated forest model as input to determine the target merchants. Without the need for any black labels, it can identify intermediary merchants and / or individuals with high abnormal transactions, and can also identify unknown abnormal transactions and discover uncovered high-risk merchants, which can further improve the management level of the transaction system.

[0127] This application also provides a target merchant identification device. Figure 9 This is a schematic diagram of the structure of a target merchant identification device provided in an embodiment of this application, as shown below. Figure 9 As shown, the identification device for the target merchant may include:

[0128] The determination module 901 is used to determine the set of merchants to be identified and the set of transaction objects of each merchant to be identified in the set of merchants to be identified from the transaction data set; the transaction data set includes the transaction data generated by the merchants to be identified within a preset time period;

[0129] The construction module 903 is used to construct a first target graph based on the set of merchants to be identified and the set of transaction objects of each merchant to be identified. The first node in the first target graph represents a merchant to be identified, and the edge in the first target graph represents that there is a common transaction object between two connected merchants to be identified. The weight of the edge represents the proportion of the common transaction object.

[0130] The generation module 905 is used to group the first nodes in the first target graph according to the weights corresponding to the edges in the first target graph, and generate a second target graph based on the grouping result; the second nodes in the second target graph represent the first node group, and the first node group includes multiple first nodes.

[0131] The vector representation module 907 is used to perform vector representation on each second node in the second target map based on the second target map, so as to obtain the group features of the merchant group to be identified indicated by each second node;

[0132] The classification processing module 909 is used to concatenate the group features of each merchant group to be identified with the transaction features of the merchant group to be identified, and to classify the merchant group to be identified based on the concatenation result of each merchant group to be identified, so as to obtain the classification result corresponding to each merchant group to be identified; the classification result indicates whether the merchant group to be identified is a normal transaction merchant.

[0133] In some possible implementations, the construction module 903 is used to determine a common transaction object for any two merchants to be identified in the set of merchants to be identified, based on the transaction object set corresponding to each of the two merchants to be identified, so as to obtain a common transaction object set corresponding to the two merchants to be identified.

[0134] The weight of the edge between any two merchants to be identified is determined by the ratio of the number of common transaction objects in the common transaction object set to the total number of transaction objects; the total number of transaction objects represents the total number of transaction objects in the transaction object set corresponding to any two merchants to be identified.

[0135] Using the merchants to be identified in the set of merchants to be identified as the first node, construct the first target graph according to the weights of the edges between any two merchants to be identified.

[0136] In some possible implementations, generation module 905 includes:

[0137] The first determination submodule is used to determine multiple target starting nodes corresponding to the current grouping process from the first target map; the target starting node is any one of the first nodes;

[0138] The construction submodule is used to construct the group corresponding to each target starting node, starting from each target starting node, to obtain the group map corresponding to multiple target starting nodes;

[0139] The second determination submodule is used to determine the modularity of the current grouping process based on the weights of the edges between any two target starting nodes in the grouped graph. The modularity represents the tightness of the connection between any two target starting nodes in the grouped graph.

[0140] The third determination submodule is used to determine the maximum modularity among the multiple modularities corresponding to multiple grouping processes, and to determine the grouping graph in the grouping process corresponding to the maximum modularity as the target grouping graph.

[0141] The aggregation submodule is used to aggregate the set of associated nodes corresponding to each target starting node in the target group map, to obtain the first node group corresponding to each target starting node, and generate the second target map.

[0142] In some possible implementations, a submodule is constructed to determine, from the first node connected to each target starting node, the set of associated nodes corresponding to each target starting node in the current clustering process;

[0143] Clustering is performed on the target starting node and the set of associated nodes corresponding to the target starting node to determine the grouping corresponding to each target starting node in the current clustering process;

[0144] Clustering is performed on the group corresponding to each target starting node in the current clustering process to obtain the clustering group corresponding to the current clustering process;

[0145] The cluster groups obtained from multiple clustering processes are used as the grouping graphs corresponding to multiple target starting nodes.

[0146] In some possible implementations, a submodule is constructed to determine, from the first nodes connected to each target starting node, a first set of associated nodes corresponding to each target starting node in the current clustering process, and to determine, from the first nodes connected to each first associated node, a second set of associated nodes corresponding to each first associated node in the current clustering process.

[0147] Wherein, the weight of the edge between each first associated node in the first associated node set and the target starting node is greater than the first preset weight threshold, the weight of the edge between each second associated node in the second associated node set and the corresponding first associated node is greater than the second preset weight threshold, and the second preset weight threshold is less than or equal to the first preset weight threshold.

[0148] Clustering is performed on each target starting node, its corresponding first set of associated nodes, and its corresponding second set of associated nodes to obtain the grouping corresponding to each target starting node;

[0149] For any two target starting nodes, if one of the two target starting nodes is the first or second associated node of the other target starting node, perform clustering on the groups corresponding to the two target starting nodes to obtain the clustered groups corresponding to the current clustering process.

[0150] The cluster groups obtained from multiple clustering processes are used as the grouping graphs corresponding to multiple target starting nodes.

[0151] In some possible implementations, the second determining submodule is used to determine, for any two target starting nodes in the grouped graph, a first ratio of the weight corresponding to the edge between any two target starting nodes to the total weight; the total weight represents the sum of the weights corresponding to the edges in the grouped graph.

[0152] Determine the difference between the weight of the edge between any two target starting nodes and the first ratio;

[0153] The difference is calculated by multiplying the difference by the category parameters corresponding to any two target starting nodes; the category parameters indicate whether any two target starting nodes belong to the same group.

[0154] Determine the second ratio of the difference sum to the total weight to obtain the modularity corresponding to the grouped map.

[0155] In some possible implementations, the vector representation module 907 is used to determine the root node corresponding to the current root subgraph construction from the second target graph; the root node is any second node;

[0156] The set of first connected nodes for constructing the root subgraph is determined from the second nodes connected to the root node, and the set of second connected nodes for constructing the root subgraph is determined from the second nodes connected to each of the first connected nodes, thus obtaining the set of connected nodes corresponding to the root node; the set of connected nodes includes second nodes of a preset depth that are sequentially connected to the root node, and the preset depth represents the preset number of sequentially connected second nodes;

[0157] Based on the root node as the starting point and the set of connected nodes corresponding to the root node, construct the current root subgraph and construct multiple corresponding root subgraphs; the root node and the set of connected nodes corresponding to the root node in each root subgraph constitute the merchant group to be identified;

[0158] Multiple root subgraphs are constructed from multiple root subgraphs and then used as a root subgraph set.

[0159] Each root subgraph in the root subgraph set is represented by a vector, and the group characteristics of the merchant group to be identified corresponding to each root subgraph are determined, thus obtaining the graph structure vector set corresponding to each merchant group to be identified.

[0160] In some possible implementations, the target merchant identification device further includes:

[0161] The third determination module is used to determine the transaction frequency information, transaction amount information, merchant stay duration information and transaction time information corresponding to the merchant group to be identified from the transaction data set;

[0162] The normalization module is used to normalize the transaction frequency information, transaction amount information, merchant stay duration information, and transaction time information corresponding to the merchant group to be identified, so as to obtain the transaction characteristics corresponding to the merchant group to be identified.

[0163] In some possible implementations, the classification processing module 909 is used to input the splicing result of each merchant group to be identified into the classification model to obtain the classification result output by the classification model.

[0164] The apparatus and method embodiments in this application are based on the same application concept.

[0165] This application provides an electronic device including a processor and a memory. The memory stores at least one instruction or at least one program, which is loaded and executed by the processor to implement the target merchant identification method provided in the above method embodiments.

[0166] Figure 10This is a schematic diagram of the hardware structure of an electronic device for implementing the target merchant identification method provided in this application embodiment. The electronic device can participate in or include the target merchant identification device provided in this application embodiment. Figure 10 As shown, the electronic device may include one or more processors 1001 (shown as 1001a and 1001b in the figure) 1001 (processor 1001 may include, but is not limited to, a microprocessor 1001 MCU or a programmable logic device FPGA, etc.), a memory 1003 for storing data, and a transmission device 1005 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, and / or a power supply. Those skilled in the art will understand that... Figure 10 The structure shown is for illustrative purposes only and does not limit the structure of the electronic device described above. For example, the electronic device may also include components that are more... Figure 10 The more or fewer components shown, or having the same Figure 10 The different configurations shown.

[0167] It should be noted that the aforementioned one or more processors 1001 and / or other data processing circuits are generally referred to as "data processing circuits" in this application. The data processing circuit can be embodied, in whole or in part, in software, hardware, firmware, or any other combination. Furthermore, the data processing circuit can be a single, independent processing module, or it can be integrated, in whole or in part, into any other element within an electronic device (or mobile device). As involved in the embodiments of this application, the data processing circuit acts as a processor 1001 for control (e.g., selection of a variable resistor termination path connected to an interface).

[0168] The memory 1003 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the target merchant identification method in this embodiment. The processor 1001 implements the above-mentioned target merchant identification method by running the software programs and modules stored in the memory 1003 and executing various functional applications and data processing. The memory 1003 may include high-speed random access memory, and may also include non-volatile random access memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some possible embodiments, the memory 1003 may further include remotely configured memory 1003 relative to the processing, which can be connected to electronic devices via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0169] The transmission device 1005 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the electronic device. In one example, the transmission device 1005 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 1005 may be a radio frequency (RF) module used for wireless communication with the Internet.

[0170] The display can be, for example, a touchscreen liquid crystal display (LED), which allows users to interact with the user interface of an electronic device (or mobile device).

[0171] This application provides a computer-readable storage medium that can be disposed in an electronic device to store at least one instruction or at least one program related to implementing a target merchant identification method in the method embodiment. The at least one instruction or at least one program is loaded and executed by the processor to implement the target merchant identification method provided in the above method embodiment.

[0172] Optionally, in this embodiment, the storage medium may be located at at least one of the multiple network servers in a computer network. Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0173] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, while this specification describes specific embodiments, other embodiments are also within the scope of the appended claims. In some cases, the actions or steps described in the claims can be performed in the order shown in different embodiments and still achieve the desired results. Additionally, the processes depicted in the drawings do not necessarily require a specific order or sequence of connections to achieve the desired results; in some implementations, parallel processing of multiple tasks is possible or may be advantageous.

[0174] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, embodiments of apparatus and electronic devices are described simply because they are based on similar method embodiments; relevant parts can be referred to the descriptions of the method embodiments.

[0175] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for identifying target merchants, characterized in that, include: The set of merchants to be identified and the set of transaction objects of each merchant to be identified in the set of transaction data are determined from the transaction data set; the transaction data set includes the transaction data generated by the merchants to be identified within a preset time period. Based on the set of merchants to be identified and the set of transaction objects of each of the merchants to be identified, a first target map is constructed; The first node in the first target graph represents the merchant to be identified, the edge in the first target graph represents that there is a common transaction object between two connected merchants to be identified, and the weight corresponding to the edge represents the proportion of the common transaction object. The first node in the first target graph is grouped according to the weights corresponding to the edges in the first target graph, and a second target graph is generated based on the results of the grouping process. The second node in the second target map represents the first node group, and the first node group includes multiple first nodes; Based on the second target map, each second node in the second target map is represented by a vector to obtain the group features of the merchant group to be identified indicated by each second node; The group features of each merchant group to be identified are concatenated with the transaction features of the merchant group to be identified. Based on the concatenation result of each merchant group to be identified, the merchant groups to be identified are classified to obtain the classification result corresponding to each merchant group to be identified. The classification result indicates whether the merchant group to be identified is a normal transaction merchant.

2. The method according to claim 1, characterized in that, The step of constructing a first target graph based on the set of merchants to be identified and the set of transaction objects of each of the merchants to be identified includes: For any two merchants in the set of merchants to be identified, a common transaction object is determined based on the transaction object set corresponding to each of the two merchants to be identified, thus obtaining the common transaction object set corresponding to the two merchants to be identified. The weight of the edge between any two merchants to be identified is determined based on the ratio of the number of common trading objects in the common trading object set to the total number of trading objects; the total number of trading objects represents the total number of trading objects in the trading object set corresponding to any two merchants to be identified. Using the merchants to be identified in the set of merchants to be identified as the first node, the first target graph is constructed according to the weights of the edges between any two merchants to be identified.

3. The method according to claim 1, characterized in that, The step of grouping the first nodes in the first target graph according to the weights corresponding to the edges in the first target graph, and generating a second target graph based on the result of the grouping process, includes: Multiple target starting nodes corresponding to the current grouping process are determined from the first target map; the target starting node is any one of the first nodes; Each target starting node is used as the starting point to construct a group corresponding to each target starting node, thereby obtaining a group map corresponding to multiple target starting nodes; For any two target starting nodes in the grouped graph, the modularity corresponding to the current grouping process is determined based on the weights of the edges between the two target starting nodes; the modularity characterizes the connection tightness between the two target starting nodes in the grouped graph. Based on the multiple modularity corresponding to multiple grouping processes, the maximum modularity among the multiple modularity is determined, and the grouping graph corresponding to the maximum modularity is determined as the target grouping graph. The associated node set corresponding to each target starting node in the target group map is aggregated to obtain the first node group corresponding to each target starting node, and the second target map is generated.

4. The method according to claim 3, characterized in that, The step of constructing a group corresponding to each target starting node, using each target starting node as the starting point, to obtain a grouped map corresponding to multiple target starting nodes, includes: Determine the set of associated nodes corresponding to each of the target starting nodes in the current clustering process from the first nodes connected to each of the target starting nodes; Clustering is performed on the target starting node and the set of associated nodes corresponding to the target starting node to determine the grouping corresponding to each target starting node in the current clustering process. Clustering is performed on the groups corresponding to each target starting node in the current clustering process to obtain the clustering groups corresponding to the current clustering process; The cluster groups obtained from multiple clustering processes are used as the grouping graphs corresponding to the multiple target starting nodes.

5. The method according to claim 3, characterized in that, The step of constructing a group corresponding to each target starting node, starting from each target starting node, to obtain a grouped map corresponding to multiple target starting nodes includes: A first set of associated nodes corresponding to each target starting node in the current clustering process is determined from the first nodes connected to each target starting node, and a second set of associated nodes corresponding to each first associated node in the current clustering process is determined from the first nodes connected to each first associated node. Wherein, the weight of the edge between each first associated node in the first associated node set and the target starting node is greater than the first preset weight threshold, the weight of the edge between each second associated node in the second associated node set and the corresponding first node is greater than the second preset weight threshold, and the second preset weight threshold is less than or equal to the first preset weight threshold. Clustering is performed on each target starting node, the corresponding first set of associated nodes, and the corresponding second set of associated nodes to obtain a group corresponding to each target starting node; For any two target starting nodes, if one of the target starting nodes is the first or second associated node of the other target starting node, clustering is performed on the groups corresponding to the two target starting nodes to obtain the clustered groups corresponding to the current clustering process. The cluster groups obtained from multiple clustering processes are used as the grouping graphs corresponding to the multiple target starting nodes.

6. The method according to claim 3, characterized in that, The step of determining the modularity of the grouped graph based on the weights of the edges between any two target starting nodes in the grouped graph includes: For any two target starting nodes in the grouped graph, determine a first ratio of the weight of the edge between the two target starting nodes to the total weight; the total weight represents the sum of the weights of the edges in the grouped graph. Determine the difference between the weight of the edge between any two target starting nodes and the first ratio; The product of the difference and the category parameters corresponding to any two target starting nodes is determined to obtain the sum of the differences; the category parameters characterize whether any two target starting nodes belong to the same cluster group. Determine the second ratio of the difference sum to the total weight to obtain the modularity corresponding to the grouped map.

7. The method according to claim 4 or 5, characterized in that, The step of representing each second node in the second target map with a vector based on the second target map to obtain the group features of the merchant group to be identified corresponding to each second node includes: The root node for constructing the current root subgraph is determined from the second target graph; the root node is any second node. The first set of connected nodes for the current root subgraph is determined from the second nodes connected to the root node, and the second set of connected nodes for the current root subgraph is determined from the second nodes connected to each of the first connected nodes, to obtain the set of connected nodes corresponding to the root node; the set of connected nodes includes second nodes of a preset depth that are sequentially connected to the root node, and the preset depth represents the preset number of sequentially connected second nodes; Based on the root node as the starting point and the set of connected nodes corresponding to the root node, the current root subgraph is constructed to construct multiple root subgraphs; the root node and the set of connected nodes corresponding to the root node in each root subgraph constitute the merchant group to be identified; Multiple root subgraphs are constructed from multiple root subgraphs and then used as a root subgraph set. Each root subgraph in the root subgraph set is represented by a vector, and the group features of the merchant group to be identified corresponding to each root subgraph are determined, thereby obtaining the graph structure vector set corresponding to each merchant group to be identified.

8. The method according to claim 1, characterized in that, The method further includes: From the transaction data set, determine the transaction frequency information, transaction amount information, merchant stay duration information, and transaction time information corresponding to the merchant group to be identified; The transaction frequency information, transaction amount information, merchant stay duration information, and transaction time information corresponding to the merchant group to be identified are normalized to obtain the transaction characteristics corresponding to the merchant group to be identified.

9. The method according to claim 1, characterized in that, The process of classifying the merchant groups to be identified based on the splicing results of each merchant group to be identified, to obtain the classification result corresponding to each merchant group to be identified, includes: The concatenation result of each of the merchant groups to be identified is input into the classification model to obtain the classification result output by the classification model.

10. A target merchant identification device, characterized in that, include: The determination module is used to determine the set of merchants to be identified and the set of transaction objects of each merchant to be identified in the set of merchants to be identified from the transaction data set; the transaction data set includes the transaction data generated by the merchants to be identified within a preset time period; The construction module is used to construct a first target graph based on the set of merchants to be identified and the set of transaction objects of each of the merchants to be identified; The first node in the first target graph represents the merchant to be identified, the edge in the first target graph represents that there is a common transaction object between two connected merchants to be identified, and the weight corresponding to the edge represents the proportion of the common transaction object. The generation module is used to group the first nodes in the first target graph according to the weights corresponding to the edges in the first target graph, and generate a second target graph based on the result of the grouping process. The second node in the second target map represents the first node group, and the first node group includes multiple first nodes; The vector representation module is used to perform vector representation on each second node in the second target map based on the second target map, so as to obtain the group features of the merchant group to be identified indicated by each second node; The classification processing module is used to concatenate the group features of each merchant group to be identified with the transaction features of the merchant group to be identified, and to classify the merchant group to be identified based on the concatenation result of each merchant group to be identified, so as to obtain the classification result corresponding to each merchant group to be identified; the classification result indicates whether the merchant group to be identified is a normal transaction merchant.

11. An electronic device, characterized in that, The electronic device includes a processor and a memory, the memory storing at least one instruction or at least one program, the at least one instruction or the at least one program being loaded and executed by the processor to implement the target merchant identification method as described in any one of claims 1-9.

12. A computer storage medium, characterized in that, The storage medium stores at least one instruction or at least one program, which is loaded and executed by a processor to implement the target merchant identification method as described in any one of claims 1-9.

13. A computer program product, characterized in that, The computer program product includes at least one instruction or at least one program segment, which is loaded and executed by a processor to implement the target merchant identification method as described in any one of claims 1-9.

Images