Data processing method and device, equipment and medium

Abstract

The embodiment of the invention provides a data processing method and device, equipment and a medium, which are used for improving the management efficiency of business accounts. The method comprises the following steps: if an asset transfer relationship exists between a business account i belonging to a first business type and a business account j belonging to a second business type in a business account set, establishing a connection relationship between the business account i and the business account j, and determining a first edge weight according to a feedback asset data volume of the business account i and an asset transfer volume between the business account i and the business account j; if the business account x and the business account y which belong to the same business type in the business account set have an environment interaction relationship, establishing a connection relationship between the business account x and the business account y, and setting a second edge weight; generating an account association graph according to the business account set, the connection relationship associated with the business account set, the first edge weight and the second edge weight; and according to the association degree magnitude of each service account in the account association graph, performing sub-graph division on the account association graph to obtain N service account sub-graphs.

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and in particular to a data processing method, device, equipment and medium. Background Art

[0002] With the development of Internet technology, cyber black industries have emerged. Cyber ​​black industries refer to illegal activities that use the Internet as a medium and network technology as the main means to bring major security risks to the security of computer information systems and the management order of cyberspace. In the current management of black industry accounts, a large number of business accounts on various platforms are usually repeatedly checked. After screening out accounts suspected of engaging in black industry activities from a large number of business accounts, it is necessary to implement relevant disposal measures for the accounts suspected of engaging in black industry activities one by one (for example, further tracing and screening, account blocking, etc.). It can be seen that the screening of black industry accounts requires a series of tedious operations, and after screening out accounts suspected of engaging in black industry activities, these accounts suspected of engaging in black industry activities need to be managed independently, which takes a lot of time and results in low management efficiency of black industry accounts. Summary of the invention

[0003] The embodiments of the present application provide a data processing method, apparatus, device and medium, which can improve the management efficiency of business accounts.

[0004] On the one hand, an embodiment of the present application provides a data processing method, including:

[0005] Get the business account collection;

[0006] If there is an asset transfer relationship between a business account i belonging to the first business type and a business account j belonging to the second business type in the business account set, a connection relationship between the business account i and the business account j is established, and the first edge weight between the business account i and the business account j is determined according to the amount of feedback asset data associated with the business account i and the amount of asset transfer between the business account i and the business account j;

[0007] If there is an environment interaction relationship between business account x and business account y of the same business type in the business account set, a connection relationship between business account x and business account y is established, and a second edge weight between business account x and business account y is set;

[0008] Generate an account association graph according to the business account set, the connection relationship between each business account in the business account set, the first edge weight, and the second edge weight;

[0009] According to the association metric values ​​of the business accounts in the business account set in the account association graph, the account association graph is divided into sub-graphs to obtain N business account sub-graphs; the N business account sub-graphs are used as the basis for grouping management corresponding to the business account set, and N is a positive integer.

[0010] An embodiment of the present application provides a data processing device, including:

[0011] A business account set acquisition module is used to acquire a business account set;

[0012] A first edge weight determination module is used to establish a connection relationship between the business account i and the business account j if there is an asset transfer relationship between the business account i belonging to the first business type and the business account j belonging to the second business type in the business account set, and determine the first edge weight between the business account i and the business account j according to the amount of feedback asset data associated with the business account i and the amount of asset transfer between the business account i and the business account j;

[0013] A second edge weight determination module, configured to establish a connection relationship between business account x and business account y if business account x and business account y of the same business type in the business account set have an environment interaction relationship, and set a second edge weight between business account x and business account y;

[0014] An account association graph generation module, used to generate an account association graph according to a business account set, a connection relationship between each business account in the business account set, a first edge weight, and a second edge weight;

[0015] The first sub-graph partitioning processing module is used to perform sub-graph partitioning processing on the account association graph according to the association metric values ​​of the business accounts in the business account set in the account association graph to obtain N business account sub-graphs; the N business account sub-graphs are used as the basis for grouping management corresponding to the business account set, and N is a positive integer.

[0016] The first edge weight determination module determines the first edge weight between the business account i and the business account j according to the amount of feedback asset data associated with the business account i and the amount of asset transfer between the business account i and the business account j, including:

[0017] Obtain the amount of feedback asset data associated with business account i, and the amount of asset transfer between business account i and business account j;

[0018] If the asset transfer amount is less than the amount of feedback asset data, the ratio between the asset transfer amount and the amount of feedback asset data is determined as the first edge weight between business account i and business account j;

[0019] If the asset transfer amount is greater than or equal to the amount of feedback asset data, and the asset transfer amount is less than or equal to the asset transfer threshold, the first constant parameter is determined as the first edge weight between business account i and business account j; the asset transfer threshold is the product of the first business coefficient and the amount of feedback asset data;

[0020] If the asset transfer amount is less than the asset transfer threshold, the first edge weight between the business account i and the business account j is determined according to the feedback asset data amount, the asset transfer amount, the first business coefficient and the second business coefficient.

[0021] If the asset transfer amount is less than the asset transfer threshold, the first edge weight determination module determines the first edge weight between the business account i and the business account j according to the amount of feedback asset data, the asset transfer amount, the first business coefficient, and the second business coefficient, including:

[0022] If the asset transfer amount is less than the asset transfer threshold, the difference between the first business coefficient and the second business coefficient is determined as the third business coefficient, and the ratio between the second constant parameter and the third business coefficient is determined as the edge weight coefficient;

[0023] The product of the ratio between the asset transfer amount and the feedback asset data amount and the edge weight coefficient is determined as the first candidate edge weight;

[0024] The ratio of the second business coefficient to the third business coefficient is determined as the second candidate edge weight, and the difference between the first candidate edge weight and the third candidate edge weight is determined as the first edge weight between the business account i and the business account j.

[0025] The data processing device further comprises:

[0026] A first account information acquisition module is used to acquire historical login device information and account registration information associated with a business account in a business account set;

[0027] The first interaction relationship determination module is used to determine that business account x and business account y have an environment interaction relationship if business account x and business account y belonging to the same business type in the business account set have the same historical login device information, or business account x and business account y have the same account registration information.

[0028] The data processing device further comprises:

[0029] A second account information acquisition module, configured to acquire a first feedback account set associated with the business account x and a second feedback account set associated with the business account y if the business account x and the business account y belong to the first business type;

[0030] A third account information acquisition module is used to acquire a first business account set having an asset transfer relationship with the business account x, and a second business account set having an asset transfer relationship with the business account y;

[0031] The second interaction relationship determination module is used to determine that the business account x has an environment interaction relationship with the business account y if there is an intersection between the first feedback account set and the second feedback account set, or if there is an intersection between the first business account set and the second business account set.

[0032] The data processing device further comprises:

[0033] An abnormal account determination module is used to obtain the sum of edge weights corresponding to each business account in the account association graph, and determine the business account whose sum of edge weights is less than a first threshold as an abnormal account;

[0034] A path removal module is used to take the path of the abnormal account in the account association map as the path to be pruned, remove the path to be pruned from the account association map, and obtain a pruned account association map;

[0035] The second subgraph partitioning processing module is used to perform game subgraph partitioning processing on the target association graph and subgraph partitioning processing on the account association graph according to the association metric values ​​of the business accounts in the business account set in the pruned account association graph, so as to obtain N business account subgraphs.

[0036] The first subgraph partitioning processing module performs subgraph partitioning processing on the account association graph according to the association metric values ​​of the business accounts in the business account set in the account association graph to obtain N business account subgraphs, including:

[0037] Obtain a neighbor account set of a business account a in the business account set, and obtain an association metric value between the business account a and each neighbor account in the neighbor account set;

[0038] Determine the neighbor account corresponding to the largest association metric value in the neighbor account set as a candidate neighbor account, and if the association metric value between the business account a and the candidate neighbor account is greater than a second threshold, merge the business account a and the candidate neighbor account into a first candidate group;

[0039] Obtain a set of neighbor candidate groups of the first candidate group in the account association graph, and obtain association metrics between the first candidate group and each neighbor candidate group in the set of neighbor candidate groups;

[0040] Determine the neighbor candidate group corresponding to the largest association metric value in the neighbor candidate group set as the pending neighbor candidate group, and if the association metric value between the first candidate group and the pending neighbor candidate group is greater than a second threshold, merge the first candidate group and the pending neighbor candidate group into a second candidate group;

[0041] According to the second candidate group included in the account association graph, the account association graph is divided into sub-graphs to obtain N business account sub-graphs; wherein the association metric values ​​between different business account sub-graphs are all less than or equal to the second threshold.

[0042] The neighbor account set includes the business account b; the first subgraph partitioning processing module obtains the association metric values ​​between the business account a and each neighbor account in the neighbor account set, including:

[0043] Get the graph weight corresponding to the account association graph, get the edge weight between business account a and business account b, get the sum of edge weights corresponding to business account a, and get the sum of edge weights corresponding to business account b;

[0044] The product of the sum of edge weights corresponding to business account a and the sum of edge weights corresponding to business account b is determined as the first candidate association metric value, and the ratio between the first candidate association metric value and the graph weight corresponding to the account association graph is determined as the second candidate association metric value;

[0045] Determine the difference between the edge weight between the business account a and the business account b and the second candidate association metric value as the third candidate association metric value;

[0046] The association metric value between business account a and business account b is determined according to the graph weight corresponding to the account association graph and the third candidate association metric value.

[0047] The first subgraph partitioning processing module performs subgraph partitioning processing on the account association graph according to the association metric values ​​of the business accounts in the business account set in the account association graph to obtain N business account subgraphs, including:

[0048] According to the association metric values ​​of the business accounts in the business account set in the account association graph, the account association graph is divided into subgraphs to obtain M initial account subgraphs; M is an integer greater than 1;

[0049] Combine the M initial account subgraphs to obtain multiple account subgraph pairs, and obtain a similarity evaluation value between two initial account subgraphs included in each account subgraph pair;

[0050] According to the similarity evaluation value associated with each account subgraph pair, the M initial account subgraphs are merged to obtain N business account subgraphs.

[0051] The first subgraph division processing module performs subgraph merging processing on the M initial account subgraphs according to the similarity evaluation value associated with each account subgraph pair, and obtains N business account subgraphs, including:

[0052] Acquire first multimedia feedback data corresponding to a business account in an initial account subgraph p, and acquire second multimedia feedback data corresponding to a business account in an initial account subgraph q; the initial account subgraph p and the initial account subgraph q belong to any one of a plurality of account subgraph pairs;

[0053] Determine a first subgraph similarity between an initial account subgraph p and an initial account subgraph q according to the first multimedia feedback data and the second multimedia feedback data;

[0054] Obtaining a second subgraph similarity between the initial account subgraph p and the initial account subgraph q according to the geographic locations associated with the business accounts in the initial account subgraph p and the business accounts in the initial account subgraph q;

[0055] According to the total amount of feedback assets associated with the business accounts in the initial account subgraph p and the business accounts in the initial account subgraph q, the third subgraph similarity between the initial account subgraph p and the initial account subgraph q is obtained;

[0056] Perform weighted summation on the first subgraph similarity, the second subgraph similarity, and the third subgraph similarity to obtain a similarity evaluation value between the initial account subgraph p and the initial account subgraph q;

[0057] If the similarity evaluation value between the initial account sub-graph p and the initial account sub-graph q is greater than the similarity threshold, the initial account sub-graph p and the initial account sub-graph q are merged into a business account sub-graph.

[0058] The first subgraph similarity includes a first cosine similarity, an intersection-over-union ratio, and a second cosine similarity; the first subgraph partitioning processing module determines the first subgraph similarity between the initial account subgraph p and the initial account subgraph q according to the first multimedia feedback data and the second multimedia feedback data, including:

[0059] Performing feature extraction processing on the first multimedia feedback data to obtain a first comprehensive multimedia representation feature corresponding to the first multimedia feedback data, and performing feature extraction processing on the second multimedia feedback data to obtain a second comprehensive multimedia representation feature corresponding to the second multimedia feedback data;

[0060] Obtaining a first cosine similarity between the first comprehensive multimedia representation feature and the second comprehensive multimedia representation feature;

[0061] Extracting key information from the first multimedia feedback data to obtain a first multimedia key feature corresponding to the first multimedia feedback data, and extracting key information from the second multimedia feedback data to obtain a second multimedia key feature corresponding to the second multimedia feedback data;

[0062] Obtaining an intersection-over-union ratio between the first multimedia key feature and the second multimedia key feature;

[0063] Performing dimensionality reduction processing on the first multimedia key feature to obtain a third comprehensive multimedia representation feature corresponding to the first multimedia key feature, and performing dimensionality reduction processing on the second multimedia key information to obtain a fourth comprehensive multimedia representation feature corresponding to the second multimedia key feature;

[0064] A second cosine similarity between the third comprehensive multimedia representation feature and the fourth comprehensive multimedia representation feature is obtained.

[0065] On one hand, an embodiment of the present application provides a computer device, including a memory and a processor, wherein the memory is connected to the processor, the memory is used to store a computer program, and the processor is used to call the computer program so that the computer device executes the method provided in the above aspect of the embodiment of the present application.

[0066] On one hand, an embodiment of the present application provides a computer-readable storage medium, in which a computer program is stored. The computer program is suitable for being loaded and executed by a processor, so that a computer device with a processor executes the method provided in the above aspect of the embodiment of the present application.

[0067] According to one aspect of the present application, a computer program product or computer program is provided, the computer program product or computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the method provided in the above aspect.

[0068] In the embodiment of the present application, when two business accounts belonging to different business types in the business account set have an asset transfer relationship, a connection relationship between the two business accounts can be established, and the first edge weight between the two business accounts can be determined according to the amount of feedback asset data and the amount of asset transfer associated with the two business accounts; when two business accounts belonging to the same business type in the business account set have an environment interaction relationship, a connection relationship between the two business accounts can be established, and the second edge weight between the two business accounts can be set. It can be seen that the account association graph generated according to the first edge weight and the second edge weight can deeply mine the association relationship between each business account, so multiple business accounts can be managed at the same time, which helps to improve the management efficiency of business accounts. Relative to the account association graph, the business accounts contained in the business account subgraph are significantly smaller than the business accounts contained in the account association graph. By dividing a large and complex account association graph into multiple business account subgraphs, the management pressure of the business accounts in the business account set can be reduced. In addition, the business accounts contained in the business account subgraph are more closely connected, so the business account subgraph can provide a more effective basis for the group management of the business account set, thereby improving the management efficiency of the business account. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0070] Figure 1 This is an architecture diagram of a cloud application processing system provided in an embodiment of the present application;

[0071] Figure 2 This is a flow diagram of a data processing method provided in an embodiment of the present application. Figure 1 ;

[0072] Figure 3 It is a schematic diagram of dividing an account association map provided by an embodiment of the present application;

[0073] Figure 4 It is a schematic diagram of a graph pruning provided in an embodiment of the present application;

[0074] Figure 5 This is a flow diagram of a data processing method provided in an embodiment of the present application. Figure 2 ;

[0075] Figure 6 is a schematic diagram of sub-graph division provided in an embodiment of the present application;

[0076] Figure 7 is a schematic diagram of sub-graph merging provided in an embodiment of the present application;

[0077] Figure 8 is a schematic diagram of a training text processing model provided in an embodiment of the present application;

[0078] Fig. 9 It is a schematic diagram of generating a business account sub-graph provided in an embodiment of the present application;

[0079] Fig.10 is a structural schematic diagram of a data processing device provided in an embodiment of the present application;

[0080] Fig.11 It is a structural diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0081] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0082] See also Figure 1 , Figure 1 1 is a schematic diagram of a network architecture provided by an embodiment of the present application. The network architecture may include a server 10d and a terminal cluster. The terminal cluster may include one or more terminal devices. The number of terminal devices included in the terminal cluster is not limited. Figure 1 As shown, the terminal cluster may specifically include terminal device 10a, terminal device 10b, and terminal device 10c, etc.; all terminal devices in the terminal cluster (for example, including terminal device 10a, terminal device 10b, and terminal device 10c, etc.) may be connected to the server 10d through a network connection, so that each terminal device may exchange data with the server 10d through the network connection.

[0083] in, Figure 1The server 10d shown can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms. This application does not limit the type of server.

[0084] Figure 1 The terminal devices in the terminal cluster shown may include but are not limited to: smart phones, tablet computers, laptop computers, PDAs, mobile internet devices (MIDs), wearable devices (such as smart watches, smart bracelets, etc.), smart voice interaction devices, smart home appliances (such as smart TVs, etc.), vehicle-mounted devices, aircraft and other electronic devices. This application does not limit the type of terminal devices. It is understandable that, if Figure 1 Each terminal device in the terminal cluster shown in FIG. 1 can be installed with an application client (application platform). When the application client runs in each terminal device, it can be respectively connected to the above-mentioned Figure 1 The server 10d shown in the figure performs data interaction. The application client running in each terminal device may be an independent client or an embedded sub-client integrated in a certain client, which is not limited in the present application.

[0085] The application client may specifically include but is not limited to: a vehicle-mounted client, a smart home client, an entertainment client (e.g., a game client), a multimedia client (e.g., a video client), an interactive client, and an information client (e.g., a news client), etc., which provide social conversion promotion activities (also referred to as reflux activities). If the terminal device included in the terminal cluster is a vehicle-mounted device, then the vehicle-mounted device may be a smart terminal in a smart transportation scenario, and the application client running in the vehicle-mounted device may be referred to as a vehicle-mounted client.

[0086] In the embodiment of the present application, the server 10d and the terminal device (for example, the terminal device 10a, etc.) can be used independently to obtain the business account sub-map, and the server 10d and the terminal device (for example, the terminal device 10a, etc.) can also be used in collaboration to obtain the business account sub-map, wherein the business account sub-map can be used as a basis for group management corresponding to the business account set. Figure 1Taking the server 10d shown as an example, the server 10d can obtain a business account set, which may include business accounts of a first business type and business accounts of a second business type; when two business accounts of different business types in the business account set have an asset transfer relationship, the server 10d can establish a connection relationship between the two business accounts, and determine the first edge weight between the two business accounts according to the amount of feedback asset data and the amount of asset transfer associated with the two business accounts; when two business accounts of the same business type in the business account set have an environment interaction relationship, the server 10d can establish a connection relationship between the two business accounts and set the second edge weight between the two business accounts; further, the server 10d can construct an account association graph according to the business account set, the connection relationship between each business account in the business account set, the first edge weight and the second edge weight; and then perform sub-graph division processing on the account association graph according to the association metric values ​​of the business accounts in the business account set in the account association graph to obtain a business account sub-graph.

[0087] It is understandable that, relative to the account association graph, the business accounts included in the business account sub-graph are significantly smaller than those included in the account association graph. Therefore, using the business account sub-graph as the basis for account management can reduce the management pressure of business accounts. In addition, the business accounts included in the business account sub-graph are more closely connected. Therefore, the business account sub-graph can provide a more effective basis for the group management of business account sets, thereby improving the management efficiency of business accounts.

[0088] See also Figure 2 , Figure 2 This is a flow diagram of a data processing method provided in an embodiment of the present application. Figure 1 It can be understood that the data processing method is executed by a computer device, which may be a terminal device (e.g., Figure 1 The terminal device 10a of the corresponding embodiment, or it can be a server (for example, Figure 1 The data processing method may include the following steps S101 to S105:

[0089] Step S101: Obtain a business account set.

[0090] See also Figure 3 , Figure 3 Schematic diagram of dividing an account association map provided by an embodiment of the present application. Figure 3As shown, a business account set 20 can be obtained, and the business accounts of the business account set 20 may include business accounts belonging to the first business type, such as business account 20a, business account 20d, and business account 20f, etc.; and business accounts belonging to the second business type, such as business account 20b, business account 20c, and business account 20e, etc. The business account of the first business type may refer to an illegal account engaged in black market activities, and the business account of the second business type may refer to an illegal account having an asset transfer relationship with the business account of the first business type. Generally, the business account of the second business type is the asset transfer object of the business account of the first business type.

[0091] The embodiment of the present application does not specifically limit the number of business accounts of the first business type and the number of business accounts of the second business type in the business account set 20. For example, the business account set 20 may include multiple business accounts belonging to the first business type and one business account belonging to the second business type, or may include multiple business accounts belonging to the first business type and multiple business accounts belonging to the second business type, or may include one business account belonging to the first business type and multiple business accounts belonging to the second business type, etc.

[0092] In order to improve the timeliness of business account management, the business accounts associated with the feedback activities performed by the feedback accounts within the time range G on the application platform can be added to the business account set 20. Among them, the time range G can represent 24 hours, 48 ​​hours, 72 hours, etc. The specific value of the time range G can be determined according to actual conditions, and the embodiments of the present application do not limit this. The business accounts in the business account set can come from a single application platform or from multiple related application platforms, and the facts of the present application do not limit this. Among them, the above-mentioned application platform can be a terminal device (for example, Figure 1 The embodiment of the present application does not limit the type of application platform.

[0093] When it is detected within the time range G that a feedback account (e.g., feedback account a) has reported on the application platform that a certain business account (e.g., business account 20a) is engaged in illegal activities and assets have been transferred to business account 20a, it can be determined that business account 20a is a business account of the first business type, and business account 20a is added to business account set 20, and the attribute tag of business account 20a is set to the first business type. The feedback account refers to a business account that has suffered asset losses due to illegal activities and has provided feedback to the application platform.

[0094] Further, the asset transfer data corresponding to the business account 20a can be obtained, and the illegal account having an asset transfer relationship with the business account 20a can be determined according to the asset transfer data, and the illegal account is a business account belonging to the second business type. The asset transfer data corresponding to the business account 20a can include the asset transfer time associated with the business account 20a, the asset transfer object associated with the business account 20a, and the like.

[0095] Assuming that the transaction time of the asset transfer from feedback account a to business account 20a occurs at H hour on T day, then for business account 20a, the assets from feedback account a can be transferred after H hour on T day. In other words, the asset transfer object after H hour on T day is very likely to be a business account of the second business type, so the asset transfer object of business account 20a after H hour on T day (for example, business account 20b and business account 20c) can be determined as a business account of the second business type, and then business account 20b and business account 20c are added to business account set 20, and the attribute tags of business account 20b and business account 20c are set to the second business type.

[0096] Normally, when the feedback behavior of feedback account a is detected, the application platform will freeze the assets of business account 20a, so business account 20a will transfer the assets from feedback account a as quickly as possible. In other words, the closer the asset transfer time of business account 20a is to H time on T day, the higher the possibility that the corresponding asset transfer object is a business account of the second business type. Therefore, in order to improve the accuracy of account acquisition, the asset transfer object of business account 20a after H time on T day and close to H time on T day can be determined as a business account of the second business type. For example, the asset transfer object of business account 20a after H time on T day and before 0 time on T+1 day can be determined as a business account of the second business type, or the asset transfer object of business account 20a after H time on T day and before 24 hours on T+1 day can be determined as a business account of the second business type, etc., which can be set according to actual conditions, and the embodiment of the present application does not limit this.

[0097] Step S102: If there is an asset transfer relationship between a business account i belonging to the first business type and a business account j belonging to the second business type in the business account set, a connection relationship between the business account i and the business account j is established, and the first edge weight between the business account i and the business account j is determined according to the amount of feedback asset data associated with the business account i and the amount of asset transfer between the business account i and the business account j.

[0098] Business account i is any business account belonging to the first business type in the business account set, and business account j is any business account belonging to the second business type in the business account set. When there is an asset transfer relationship between business account i and business account j, a connection relationship between business account i and business account j is established, otherwise a connection relationship between business account i and business account j is not established; in other words, when business account i transfers assets to business account j, business account i and business account j are connected to obtain a connection relationship between business account i and business account j.

[0099] Take business account i as Figure 3 The business account number 20a shown in the figure, and the business account number j are Figure 3 Assume that there is an asset transfer relationship between business account 20a and business account 20c, such as Figure 3 As shown, at this time, a connection relationship between business account 20a and business account 20c can be established; assuming that there is no asset transfer relationship between business account 20a and business account 20e, such as Figure 3 As shown, at this time, the business account 20a and the business account 20e are not connected.

[0100] In the embodiment of the present application, the amount of feedback asset data associated with business account i refers to the total amount of asset losses reported by the feedback account corresponding to business account i on the application platform. There can be one or more feedback accounts corresponding to business account i, and the embodiment of the present application does not limit this. The amount of asset transfer between business account i and business account j refers to the total amount of assets transferred from business account i to business account j.

[0101] It can be known through analyzing historical data that when business account i makes complaints against multiple business accounts, it will generally receive feedback from 1 / x feedback accounts on the application platform for business account i. Therefore, it is generally believed that the asset transfer amount associated with business account i is reasonable between [feedback asset data amount, feedback asset data amount*x]. If the asset transfer amount associated with business account i is much smaller than the feedback asset data amount, or much larger than the feedback asset data amount*x, it is highly likely that it is not reasonable. Among them, x represents the first business coefficient, and x can be determined based on business knowledge. For example, the specific value of x can be 2, 3, 4..., and this embodiment of the present application does not limit this. Therefore, according to the above relationship, when there is an asset transfer relationship between business account i and business account j, the first edge weight between business account i and business account j is determined based on the feedback asset data amount associated with business account i and the asset transfer amount between business account i and business account j, which can accurately characterize the association relationship between business account i and business account j.

[0102] The process of determining the first edge weight between business account i and business account j may include: obtaining the amount of feedback asset data associated with business account i and the amount of asset transfer between business account i and business account j; if the amount of asset transfer is less than the amount of feedback asset data, determining the ratio between the amount of asset transfer and the amount of feedback asset data as the first edge weight between business account i and business account j; if the amount of asset transfer is greater than or equal to the amount of feedback asset data, and the amount of asset transfer is less than or equal to the asset transfer threshold, determining the first constant parameter as the first edge weight between business account i and business account j; the asset transfer threshold is the product of the first business coefficient and the amount of feedback asset data, the first business coefficient can be represented by x, and the asset transfer threshold can be represented by: feedback asset data amount*x. The first constant parameter may be a pre-set parameter, and its specific value may be 0.5, 0.8 or 1, etc., which is not limited in the embodiments of the present application.

[0103] If the asset transfer amount is less than the asset transfer threshold, the first edge weight between business account i and business account j is determined according to the amount of feedback asset data, the asset transfer amount, the first business coefficient, and the second business coefficient. The second business coefficient is a parameter pre-set according to business knowledge, and the second business coefficient can be determined according to the actual business scenario. For example, the specific value of the second business coefficient can be 5, 8, 10 or 12, etc., which is not limited in the embodiment of the present application.

[0104] Specifically, if the asset transfer amount is less than the asset transfer threshold, the difference between the first business coefficient and the second business coefficient is determined as the third business coefficient, and the ratio between the second constant parameter and the third business coefficient is determined as the edge weight coefficient. The second constant parameter can be a pre-set parameter, and its specific value can be 0.2, 0.5, 0.8 or 1, etc., which is not limited in the embodiment of the present application; the ratio between the asset transfer amount and the amount of feedback asset data and the product of the edge weight coefficient are determined as the first candidate edge weight; the ratio between the second business coefficient and the third business coefficient is determined as the second candidate edge weight, and then the difference between the first candidate edge weight and the first candidate edge weight can be determined as the first edge weight between business account i and business account j. Among them, the first edge weight between business account i and business account j can be shown as the following formula (1):

[0105]

[0106] Wherein, weight1 in formula (1) represents the first edge weight between business account i and business account j; amount represents the asset transfer amount between business account i and business account j; y represents the amount of feedback asset data associated with business account i; x represents the first business coefficient; xy represents the asset transfer threshold; k represents the second business coefficient; xk represents the third business coefficient; 1 represents the second constant parameter; represents the edge weight coefficient; represents the first candidate edge weight; Represents the second candidate edge weight.

[0107] According to formula (1), the first edge weight between any two business accounts associated with asset transfer in the business account set 20 can be obtained. For example, according to formula (1), it can be determined Figure 3 A first edge weight 21a between a business account 20a and a business account 20c is shown.

[0108] Step S103: If there is an environment interaction relationship between business account x and business account y belonging to the same business type in the business account set, a connection relationship between business account x and business account y is established, and a second edge weight between business account x and business account y is set.

[0109] Business account x and business account y are any two business accounts with the same business type in the business account set. The two business accounts having an environment interaction relationship can be understood as the two business accounts having the same login environment, having the same account registration information, etc. For example, in a possible implementation, a method for determining whether business account x and business account y have an environment interaction relationship may include: obtaining historical login device information and account registration information associated with the business accounts in the business account set; wherein the historical login device information may include but is not limited to: identification information of the electronic device used by the business account, Internet Protocol (IP) information used by the business account, and local area network information where the business account is located, etc.; the account registration information may include but is not limited to: basic attribute information, certificate information, communication information, etc. used when the business account is registered. If business account x and business account y belonging to the same business type in the business account set have the same historical login device information, or business account x and business account y have the same account registration information, it can be determined that business account x and business account y have an environment interaction relationship.

[0110] It is understandable that the above method of determining whether two business accounts have an environment interaction relationship is also applicable to two business accounts of different business types in the business account set. For example, business account i and business account j are two business accounts of different business types in the business account set. If business account i and business account j have the same historical login device information, or business account i and business account j have the same account registration information, it can be determined that business account i and business account j have an environment interaction relationship.

[0111] In a possible implementation, the method of determining whether business account x and business account y of the same business type have an environment interaction relationship may also include: if business account x and business account y belong to the first business type, obtaining a first feedback account set associated with business account x and a second feedback account set associated with business account y; obtaining a first business account set having an asset transfer relationship with business account x and a second business account set having an asset transfer relationship with business account y. The first feedback account set and the second feedback account set may be compared to see whether there is an intersection, or the first business account set and the second business account set may be compared to see whether there is an intersection, so as to determine whether business account x and business account y have an environment interaction relationship.

[0112] Assume that the first feedback account set includes feedback account a and feedback account b, and the second feedback account set includes feedback account b and feedback account c. Both the first feedback account set and the second feedback account set include feedback account b. Therefore, there is an intersection between the first feedback account set and the second feedback account set, and it can be determined that business account x and business account y have an environmental interaction relationship.

[0113] Assume that the first business account set includes business account a and business account b, and the second business account set includes business account b and business account c. Both the first business account set and the second business account set include business account b. Therefore, there is an intersection between the first business account set and the second business account set, and it can be determined that business account x and business account y have an environmental interaction relationship.

[0114] In a possible implementation, the method of determining whether a business account x and a business account y of the same business type have an environment interaction relationship may also include: if the business account x and the business account y belong to the second business type, obtaining a third business account set having an asset transfer relationship with the business account x and a fourth business account set having an asset transfer relationship with the business account y. The third business account set and the fourth business account set may be compared to see whether there is an intersection. When the third business account set and the fourth business account set have an intersection, it may be determined that the business account x and the business account y have an environment interaction relationship.

[0115] When business account x and business account y both belong to the first business type, business account x can be used as Figure 3 The business account number 20a shown is Figure 3 Assume that there is an environment interaction relationship between business account 20a and business account 20d, such as Figure 3 As shown, at this time, a connection relationship between the business account 20a and the business account 20d can be established; assuming that there is no environment interaction relationship between the business account 20a and the business account 20f, such as Figure 3 As shown, at this time, the business account 20a and the business account 20f are not connected.

[0116] Similarly, when both business account x and business account y belong to the second business type, business account x can be used as Figure 3 The business account number 20c shown is Figure 3 Assume that there is an environment interaction relationship between business account 20c and business account 20b, such as Figure 3 As shown, at this time, a connection relationship between the business account 20c and the business account 20b can be established; assuming that there is no environment interaction relationship between the business account 20c and the business account 20e, such as Figure 3 As shown, at this time, the business account 20c and the business account 20e are not connected.

[0117] It is understandable that, for two business accounts of different business types in the business account set 20, it is also possible to determine whether to establish a connection relationship between the two business accounts based on whether the two have an environment interaction relationship. Figure 3 As shown, business account 20a belongs to a business account of the first business type, and business account 20c and business account 20e belong to business accounts of the second business type. Assume that business account 20a and business account 20c have an environment interaction relationship, such as Figure 3 As shown, at this time, a connection relationship between business account 20a and business account 20c can be established; assuming that there is no environment interaction relationship between business account 20a and business account 20e, such as Figure 3 As shown, at this time, the business account 20a and the business account 20e are not connected.

[0118] When business account x and business account y have a connection relationship, a second edge weight can be set for business account x and business account y. The second edge weight can be a pre-set parameter, and its specific value can be 0.7, 0.8 or 1, etc., which is not limited in this embodiment of the present application. Figure 3As shown, assuming that there is an environmental interaction relationship between business account 20a and business account 20d, a connection relationship between business account 20a and business account 20d can be established, and a second edge weight 21c can be set for business account 20a and business account 20d; assuming that there is an environmental interaction relationship between business account 20c and business account 20b, a connection relationship between business account 20c and business account 20b can be established, and a second edge weight 21b can be set for business account 20c and business account 20b.

[0119] Optionally, assuming that there are two business accounts 20f and 20b belonging to different business types in the business account set 20, if there is no asset transfer relationship between business account 20f and business account 20b, but there is an environment interaction relationship, in this case, a connection relationship between business account 20f and business account 20b can also be established, and a second edge weight can be set for business account 20f and business account 20b, thereby enhancing the association relationship between the business accounts.

[0120] Step S104: Generate an account association graph according to the business account set, the connection relationship between each business account in the business account set, the first edge weight, and the second edge weight.

[0121] like Figure 3 As shown, the business accounts in the business account set can be used as nodes, the connection relationship between the business accounts in the business account set can be used as edges, and the first edge weight and the second edge weight can be used as edge weights to generate an account association graph 21. It can be understood that the account association graph generated according to the first edge weight and the second edge weight can deeply mine the association relationship between the business accounts in the business account set, so that multiple business accounts can be managed at the same time, which helps to improve the management efficiency of the business accounts.

[0122] Step S105: According to the association metric values ​​of the business accounts in the business account set in the account association graph, the account association graph is divided into sub-graphs to obtain N business account sub-graphs.

[0123] Among them, N business account subgraphs can be used as the basis for group management corresponding to the business account set, N is a positive integer, and the specific value of N can be 1, 2, 3, 4... The association metric value can be used as the basis for subgraph division. It can be understood that, relative to the account association graph, the business accounts included in the business account subgraph are significantly smaller than the business accounts included in the account association graph. Therefore, using the business account subgraph as the basis for account management can reduce the management pressure of business accounts. In addition, the business accounts included in the business account subgraph are more closely connected. Therefore, the business account subgraph can provide a more effective basis for the group management of the business account set, thereby improving the management efficiency of the business account.

[0124] like Figure 3 As shown, the account association graph 21 can be divided into sub-graphs according to the association metric values ​​of the business accounts in the business account set in the account association graph 21 to obtain the business account sub-graph 22a and the business account sub-graph 22b. The business account sub-graph 22a and the business account sub-graph 22b are used as the basis for grouping management corresponding to the business account set 20, and then according to the number of business accounts, total assets, etc. in the business account sub-graph 22a and the business account sub-graph 22b, relevant disposal measures are implemented for the business accounts in the business account sub-graph 22a and the business account sub-graph 22b, for example, further tracing and investigation, blocking accounts, and sorting clues and outputting them to authoritative agencies.

[0125] When the account association graph 21 is divided into sub-graphs according to the association metric values ​​of the business accounts in the business account set 20 in the account association graph 21, one or more algorithms such as the community discovery algorithm (Fast Unfolding algorithm, also known as Louvain algorithm), the KL algorithm, the overlapping path reconstruction clustering (COPAR) algorithm, the latent factor model (Latent FactorModel, LFM) algorithm and the information map (InfoMap) algorithm can be used to perform sub-graph division processing on the account association graph to obtain business account sub-graphs, and the specific implementation method can be described in detail below.

[0126] It is understandable that the account association map contains a large number of business accounts. Optionally, a part of the business accounts that are sparsely associated with other business accounts can be eliminated first, and the business accounts with close associations can be retained, thereby improving the efficiency of subgraph division. Specifically, the sum of the edge weights corresponding to each business account in the account association map can be obtained, and the business account whose sum of edge weights is less than the first threshold can be determined as an abnormal account; the path of the abnormal account in the account association map is used as the path to be pruned, and the path to be pruned is eliminated from the account association map to obtain the pruned account association map. Among them, the first threshold can be a pre-set parameter, and its specific value can be determined according to the actual application. For ease of understanding, the embodiment of the present application is illustrated by setting the first threshold to 0.5.

[0127] Please also see Figure 4 , Figure 4 Schematic diagram of a graph pruning provided in an embodiment of the present application. Figure 4As shown, the account association graph 60a includes business account 61a, business account 61b and business account 61c, edge ab between business account 61a and business account 61b, edge bc between business account 61b and business account 61c, and edge weight 62a corresponding to edge ab, and edge weight 62b corresponding to edge bc. Since business account 61a, business account 61b and business account 61c are all business accounts of the first business type, edge weight 62a and edge weight 62b both belong to the second edge weight.

[0128] like Figure 4 As shown, for business account 60b, its neighbor accounts are business account 60a and business account 60c. Assuming that the edge weight 62a between business account 60b and business account 60a is 0.1, and the edge weight 62b between business account 60b and business account 60c is 0.3, then the sum of the edge weights corresponding to business account 60b is 0.4. In this case, the sum of the edge weights corresponding to business account 60b is less than the first threshold, so business account 60b can be determined as an abnormal account. The path of business account 60b in the account association graph 60a is: business account 60a->edge ab->business account 60b->edge bc. Business account 60a->edge ab->business account 60b->edge bc can be used as the path to be pruned, and the path to be pruned: business account 60a->edge ab->business account 60b->edge bc can be removed from the account association graph 60a, and the following is obtained: Figure 4 The pruned account association graph 60b is shown. That is, relative to the business game graph, the pruned account association graph 60b does not include the business account 60a and the business account 60b, the edge ab and the edge bc, and the edge weight 62a and the edge weight 62b.

[0129] In the embodiment of the present application, when two business accounts belonging to different business types in the business account set have an asset transfer relationship, a connection relationship between the two business accounts can be established, and the first edge weight between the two business accounts can be determined according to the amount of feedback asset data and the amount of asset transfer associated with the two business accounts; when two business accounts belonging to the same business type in the business account set have an environment interaction relationship, a connection relationship between the two business accounts can be established, and the second edge weight between the two business accounts can be set. It can be seen that the account association graph generated according to the first edge weight and the second edge weight can deeply mine the association relationship between each business account, so multiple business accounts can be managed at the same time, which helps to improve the management efficiency of business accounts. Relative to the account association graph, the business accounts contained in the business account subgraph are significantly smaller than the business accounts contained in the account association graph. By dividing a large and complex account association graph into multiple business account subgraphs, the management pressure of the business accounts in the business account set can be reduced. In addition, the business accounts contained in the business account subgraph are more closely connected, so the business account subgraph can provide a more effective basis for the group management of the business account set, thereby improving the management efficiency of the business account.

[0130] See also Figure 5 , Figure 5 This is a flow diagram of a data processing method provided in an embodiment of the present application. Figure 2 It can be understood that the data processing method is executed by a computer device, which may be a terminal device (e.g., Figure 1 The terminal device 10a of the corresponding embodiment, or it can be a server (for example, Figure 1 The data processing method may include the following steps S201 to S211:

[0131] Step S201: Obtain a business account set.

[0132] Step S202: If there is an asset transfer relationship between a business account i belonging to the first business type and a business account j belonging to the second business type in the business account set, a connection relationship between the business account i and the business account j is established, and the first edge weight between the business account i and the business account j is determined according to the amount of feedback asset data associated with the business account i and the amount of asset transfer between the business account i and the business account j.

[0133] Step S203: If there is an environment interaction relationship between business account x and business account y belonging to the same business type in the business account set, a connection relationship between business account x and business account y is established, and a second edge weight between business account x and business account y is set.

[0134] Step S204: Generate an account association graph according to the business account set, the connection relationship between each business account in the business account set, the first edge weight, and the second edge weight.

[0135] The specific implementation process of step S201 and step S204 can be found in the above Figure 3 The description of steps S101 to S104 in the corresponding embodiment will not be repeated here.

[0136] Step S205: Obtain a neighbor account set of the business account a in the business account set, and obtain an association metric value between the business account a and each neighbor account in the neighbor account set.

[0137] In the embodiment of the present application, business account a is any business account in the business account set. The neighbor account set of business account a includes one or more neighbor accounts of business account a. The neighbor account of business account a refers to the business account that has a first-order neighbor relationship with business account a in the account association graph. For the convenience of description, business account a is taken as Figure 3 The business account 20a shown is used as an example. Figure 3 As shown, in the account association graph 21, business account 20a has a first-order neighbor relationship with business account 20b, business account 20c and business account 20d. Therefore, business account 20b, business account 20c and business account 20d can be added to the neighbor account set of business account 20a.

[0138] For ease of description, the embodiment of the present application takes the Fast Unfolding algorithm as an example to describe the process of subgraph division in detail. In the embodiment of the present application, when the account association graph is divided into subgraphs, the association metric value can be understood as the modularity change value in the Fast Unfolding algorithm. In order to facilitate the understanding of the definition of the modularity change value, the concept of modularity is first introduced. Modularity, also known as modularity metric value, is a method of measuring the strength of the graph structure. If business accounts with denser connections or closer associations in the account association graph are divided into one group, the modularity value will increase, and the group with the largest modularity will eventually be the optimal group.

[0139] Generally, the goal of subgraph division is to make the connections within each business account subgraph closer, while the connections between each business account subgraph are sparser; in other words, the goal of subgraph division is to make the associations between the business accounts contained in each business account subgraph closer, while the relationships between business accounts in different business account subgraphs are more distant. In the embodiment of the present application, the greater the modularity, the closer the business account connections within the obtained business account subgraph, and the stronger the internal business account associations. The modularity can be shown as the following formula (2):

[0140]

[0141] Where Q in formula (1) represents modularity; m represents the graph weight corresponding to the account association graph. A ab represents the weight between business account a and business account b; k a represents the sum of edge weights associated with business account b, k a =∑ b A ab ;k b represents the sum of edge weights associated with business account a, k b =∑ a A ab ;c a Indicates the group to which business account a is assigned; c b Indicates the group to which the business account b is assigned; δ(c a ,c b ) is used to determine whether business account a and business account b are classified into the same group. If they are classified into the same group, then δ(c a ,c b ) takes the value of 1, otherwise (c a ,c b ) is 0. Business account a is any business account in the account association graph, and the neighbor account set of business account a includes business account b. That is, in the account association graph, business account a and business account b are neighbor accounts of each other.

[0142] Furthermore, formula (2) can be simplified, and the modularity can be expressed as the following formula (3):

[0143]

[0144] Among them, ∑ in formula (3) in represents the sum of edge weights within group c; ∑ tot Indicates the sum of edge weights associated with business accounts.

[0145] It can be seen from the above formulas (2) and (3) that the size of the modularity Q mainly depends on the grouping of business accounts in the account association graph. In other words, the grouping of the account association graph can be used to quantitatively measure the quality of the graph grouping. The closer its value is to 1, the stronger the strength of the grouping structure obtained by the account association graph, that is, the better the grouping quality. Therefore, the optimal grouping can be obtained by maximizing the modularity Q. It can be understood that when business account a is assigned to group c where business account b is located, the change in modularity Q is only related to the group of business account a and business account b, and has nothing to do with other groups.

[0146] Specifically, the specific method of obtaining the association measurement value (taking the modularity change value as an example) between business account a and business account b may include: obtaining the graph weight corresponding to the account association graph, obtaining the edge weight between business account a and business account b, obtaining the sum of edge weights corresponding to business account a, and obtaining the sum of edge weights corresponding to business account b; determining the product of the sum of edge weights corresponding to business account a and the sum of edge weights corresponding to business account b as the first candidate association measurement value, and determining the ratio between the first candidate association measurement value and the graph weight corresponding to the account association graph as the second candidate association measurement value; determining the difference between the edge weight between business account a and business account b and the second candidate association measurement value as the third candidate association measurement value; and then determining the association measurement value between business account a and business account b based on the graph weight corresponding to the account association graph and the third candidate association measurement value.

[0147] The association metric between business account a and business account b can be expressed as the following formula (4):

[0148]

[0149] Wherein, △Q in formula (4) represents the association metric between business account a and business account b; m represents the graph weight corresponding to the account association graph; k a,in represents the edge weight between business account a and business account b, ∑ tot represents the sum of edge weights corresponding to business account a; k a represents the sum of edge weights corresponding to business account b; ∑ tot k a represents the first candidate association metric value; represents a second candidate association metric value; Represents the third candidate association metric value.

[0150] It is understandable that the association metric values ​​between other business accounts in the account association map and the method for obtaining the association metric values ​​between each first candidate group can refer to the method for obtaining the association metric value between business account a and business account b, and will not be repeated here.

[0151] Step S206: determine the neighbor account corresponding to the largest association metric value in the neighbor account set as a candidate neighbor account. If the association metric value between the business account a and the candidate neighbor account is greater than the second threshold, merge the business account a and the candidate neighbor account into a first candidate group.

[0152] Step S207: Obtain a set of neighbor candidate groups of the first candidate group in the account association graph, and obtain association metrics between the first candidate group and each neighbor candidate group in the set of neighbor candidate groups.

[0153] Step S208: Determine the neighbor candidate group corresponding to the largest association metric value in the neighbor candidate group set as the pending neighbor candidate group. If the association metric value between the first candidate group and the pending neighbor candidate group is greater than the second threshold, merge the first candidate group and the pending neighbor candidate group into the second candidate group.

[0154] Please also see Figure 6 , Figure 6 Schematic diagram of a sub-graph division provided in an embodiment of the present application. Figure 6 As shown, the account association graph 40a may include business account a, and the neighbor account set of business account a may include business account b, business account c and business account e. The edge weight between business account a and business account b is 0.5, the edge weight between business account a and business account c is 0.4, and the edge weight between business account a and business account e is 0.1. According to formula (4), the calculation results of the association metric between business account a and each neighbor account in the neighbor account set are as follows:

[0155]

[0156]

[0157]

[0158] Among them, the association metric value between business account a and business account b is the largest, so business account b can be determined as a candidate neighbor account. The second threshold can be a pre-set parameter, and its specific value can be determined according to actual conditions. For ease of understanding, the following example is taken as 0. Since the association metric value △Q between business account a and business account b is ab =△Q ba=0.044, which is greater than the second threshold, so the business account a and the business account b can be combined into the first candidate group 41a. Similarly, the business account e and the business account f in the account association graph 40a can be combined into the first candidate group 41b, and the business account c and the business account d can be combined into the first candidate group 41c in the same manner.

[0159] like Figure 6 As shown, the first candidate group 41a, the first candidate group 41b and the first candidate group 41c are neighbor candidate groups, and the neighbor candidate group set corresponding to the first candidate group 41a includes the first candidate group 41b and the first candidate group 41c. According to formula (4), the association metric value △Q between the first candidate group 41a and the first candidate group 41b is ab,ef =△Q ef,ba , and the association metric value △Q between the first candidate group 41a and the first candidate group 41c ab,cd =△Q cd,ba The calculation results are as follows:

[0160]

[0161]

[0162] The correlation metric value ΔQ between the first candidate group 41a and the first candidate group 41c is ab,ef =△Q ef,ba The maximum value is, therefore, the first candidate group 40c can be determined as the pending neighbor candidate group. However, the association metric value ΔQ between the first candidate group 41a and the first candidate group 41c is ab,ef =△Q ef,ba =-0.015, which is less than the second threshold, so the first candidate group 41a and the first candidate group 41c cannot be merged into the second candidate group. And after calculation, the association metric value between the first candidate group 41b and the first candidate group 41c is also less than the second threshold, so, Figure 6 The number of second candidate groups included in the corresponding account association graph 40a is 0. It is understandable that if the association metric between the first candidate group and the pending neighbor candidate group in the account association graph is greater than the second threshold, the first candidate group and the pending neighbor candidate group can be merged into the second candidate group.

[0163] Optionally, when the account association graph is divided into sub-graphs, part or all of the first candidate groups in the account association graph may be determined as business account sub-graphs. Figure 6As shown, the first candidate group 41a, the first candidate group 41b and the first candidate group 41c can be determined as the business account sub-graph; or part of the first candidate groups can be selected from the first candidate group 41a, the first candidate group 41b and the first candidate group 41c as the business account sub-graph, for example, the first candidate group 41a and the first candidate group 41b are determined as the business account sub-graph.

[0164] Step S209: According to the second candidate group included in the account association graph, the account association graph is divided into sub-graphs to obtain M initial account sub-graphs.

[0165] When the account association graph contains M second candidate groups, the M second candidate groups can be determined as the initial account subgraph, and then the multiple initial account subgraphs are merged to obtain the business account subgraph. Wherein, M is a positive integer greater than 1, and the specific value of M can be 2, 3, 4... Alternatively, the M second candidate groups can be determined as N business account subgraphs, and the values ​​of M and N are the same. In this case, the association metric values ​​between different business account subgraphs in the N business account subgraphs are less than or equal to the second threshold, so that the business account connections within the same business account subgraph are closer, and the business account connections within different business account subgraphs are sparser. When the account association graph contains a second candidate group, the second candidate group can be determined as the business account subgraph.

[0166] Optionally, the association metric value between each second candidate group in the account association graph can also be calculated. If the association metric value is greater than the second threshold, the merging operation is continued until the subgraph converges. The condition for the subgraph convergence can be set as the association metric values ​​between each group obtained in the account association graph are less than or equal to the second threshold, or the number of merges reaches a specified number, then the merging is stopped. When the number of groups finally obtained by the division is M, the M groups finally obtained by the division can be determined as the initial account subgraph, and then the multiple initial account subgraphs are subjected to subgraph merging processing to obtain a business account subgraph. Alternatively, the M groups finally obtained by the division can also be determined as N business account subgraphs, and in this case, the values ​​of M and N are the same. When the number of groups finally obtained by the division is 1, the group finally obtained by the division can be determined as the business account subgraph.

[0167] Step S210: Combine the M initial account subgraphs to obtain a plurality of account subgraph pairs, and obtain a similarity evaluation value between two initial account subgraphs included in each account subgraph pair.

[0168] The accuracy of subgraph division depends on the accuracy of the account association graph. That is to say, if the connection relationship and edge weights between the business accounts in the account association graph are not accurate enough, when the account association graph is divided into subgraphs, the number of subgraphs may increase, while the number of business accounts contained in the subgraphs is small, which is not conducive to the management of business accounts. Therefore, the embodiment of the present application deeply mines the correlation between the initial account subgraphs and merges the subgraphs of the initial account subgraphs with strong correlations, which can reduce the number of subgraphs and make a business account subgraph contain as many business accounts with close relationships as possible, thereby reducing the management pressure of business accounts.

[0169] See also Figure 7 , Figure 7 is a schematic diagram of a subgraph merging provided in an embodiment of the present application. In the embodiment of the present application, the value of M is 3, and the merging process of the initial account subgraph is described. Figure 7 As shown, after the account association graph 30 is processed by sub-graph division, the initial account sub-graph 30a, the initial account sub-graph 30b and the initial account sub-graph 30c are obtained. The initial account sub-graph 30a, the initial account sub-graph 30b and the initial account sub-graph 30c can be combined to obtain multiple account sub-graph pairs, wherein one account sub-graph pair includes two different initial account sub-graphs. Figure 7 As shown, the initial account subgraph 30a, the initial account subgraph 30b and the initial account subgraph 30c can be combined in pairs to obtain three account subgraph pairs. For example, the initial account subgraph 30a and the initial account subgraph 30b can be combined to obtain the account subgraph pair 31a, the initial account subgraph 30a and the initial account subgraph 30c can be combined to obtain the account subgraph pair 31b, and the initial account subgraph 30b and the initial account subgraph 30c can be combined to obtain the account subgraph pair 31c.

[0170] Optionally, the initial account sub-graph 30a, the initial account sub-graph 30b and the initial account sub-graph 30c may be randomly combined to obtain a plurality of account sub-graph pairs. For example, the initial account sub-graph 30a and the initial account sub-graph 30b may be combined to obtain the account sub-graph pair 31a, and the initial account sub-graph 30a and the initial account sub-graph 30c may be combined to obtain the account sub-graph pair 31b; or, the initial account sub-graph 30a and the initial account sub-graph 30c may be combined to obtain the account sub-graph pair 31b, and the initial account sub-graph 30b and the initial account sub-graph 30c may be combined to obtain the account sub-graph pair 31c, etc. The embodiment of the present application does not limit the combination method of the initial account sub-graphs.

[0171] like Figure 7As shown, after obtaining the account subgraph pair 31a, the account subgraph pair 31b and the account subgraph pair 31c, the similarity evaluation value between the two initial account subgraphs contained in each account subgraph pair can be obtained. Among them, the similarity evaluation value can be understood as the subgraph similarity between the two initial account subgraphs contained in the account subgraph pair. The similarity evaluation value can be visualized in the form of a numerical value. For example, the similarity evaluation value can be a numerical value between 0 and 1, or it can also be a numerical value between 0 and 100. The larger the numerical value, the larger the similarity evaluation value, the greater the subgraph similarity between the two initial account subgraphs contained in the account subgraph pair, the stronger the correlation between the two initial account subgraphs, and the greater the probability of performing a subgraph merge.

[0172] Assuming that the initial account subgraph p and the initial account subgraph q belong to any one of multiple account subgraph pairs, for the convenience of description, the embodiment of the present application takes the initial account subgraph p and the initial account subgraph q as examples to describe the process of obtaining the similarity evaluation value between two initial account subgraphs in an account subgraph pair.

[0173] In an embodiment of the present application, the first multimedia feedback data corresponding to the business account in the initial account subgraph p can be obtained, and the second multimedia feedback data corresponding to the business account in the initial account subgraph q can be obtained, and then the first subgraph similarity between the initial account subgraph p and the initial account subgraph q can be determined based on the first multimedia feedback data and the second multimedia feedback data. Among them, the first multimedia feedback data refers to the feedback data of the feedback account in the application platform for the business account included in the initial account subgraph p; the first multimedia feedback data may include one or more of text feedback data, video feedback data, audio feedback data and image feedback data; the definition of the second multimedia feedback data can refer to the definition of the first multimedia feedback data, and will not be repeated here. The first subgraph similarity can be used to measure the similarity of the initial account subgraph p and the initial account subgraph q for multimedia feedback data.

[0174] Furthermore, feature extraction processing may be performed on the first multimedia feedback data to obtain a first comprehensive multimedia representation feature corresponding to the first multimedia feedback data, and feature extraction processing may be performed on the second multimedia feedback data to obtain a second comprehensive multimedia representation feature corresponding to the second multimedia feedback data. The first comprehensive multimedia representation feature may be a comprehensive feature used to characterize the first multimedia feedback data; and the second comprehensive multimedia representation feature may be a comprehensive feature used to characterize the second multimedia feedback data.

[0175] In an embodiment of the present application, if the first multimedia feedback data includes video feedback data or audio feedback data, the first multimedia feedback data can be converted into text feedback data and image feedback data. For example, the video feedback data in the first multimedia feedback data can be framed to obtain image feedback data corresponding to the video feedback data, and the audio and video processing tool (such as FFmpeg, etc.) can be used to strip the audio feedback data from the video feedback data, thereby obtaining image feedback data corresponding to the video feedback data. Further, a text recognition tool can be used to perform text recognition on the image feedback data in the first multimedia feedback data to obtain text feedback data, and the text recognition tool can include but is not limited to optical character recognition (OCR), natural language processing (NLP) model and other recognition tools; speech recognition technology can be used to perform speech recognition on the audio feedback data in the first multimedia feedback data to obtain text feedback data, and the speech recognition model used by the speech recognition technology can be a model constructed with a mixed Gaussian model-hidden Markov model (GMM-HMM), a recurrent neural network (RNN), a feedforward sequential memory neural network (FSMN) and other models as the basic framework. Similarly, if the second multimedia feedback data includes video feedback data or audio feedback data, the second multimedia feedback data may be converted into text feedback data and image feedback data in the same manner.

[0176] For the convenience of description, the text feedback data contained in the first multimedia feedback data may be referred to as first text feedback data, and the image feedback data contained in the first multimedia feedback data may be referred to as first image feedback data; the text feedback data contained in the second multimedia feedback data may be referred to as second text feedback data, and the image feedback data contained in the second multimedia feedback data may be referred to as second image feedback data.

[0177] Specifically, a text processing model may be used to extract features from the first text feedback data to obtain a first text representation feature, and a text processing model may be used to extract features from the second text feedback data to obtain a second text representation feature. The text processing model may include, but is not limited to: a sequential information maximization for compositional embedding spaces model (SIMCES), a Transformer (a network model using an attention mechanism), a GPT (Generative Pre-trained Transformer), a BERT (Bidirectional Encoder Representation from Transformers, a natural language processing model), an ALBERT (an improved natural language processing model based on BERT), etc., or may be a variation of any of the above text processing models.

[0178] Before using the text processing model to output the first text representation feature corresponding to the first text feedback data, the text processing model needs to be trained. For ease of understanding, the text processing model in the training phase can be referred to as the initial prediction model. Figure 8 The training process of the initial prediction model is described.

[0179] See also Figure 8 , Figure 8 Schematic diagram of a training text processing model provided in an embodiment of the present application. Figure 8 As shown, a sample text data set 50a can be obtained, and the sample text data set 50a can include sample text data 1, sample text data 2 and sample text data 3. The sample text data 1 can be input into the initial prediction model 51, and the text encoder 51a in the initial prediction model 51 performs text encoding processing on the sample text data 1 to obtain the sample text encoding feature 1, and the random deactivation (dropout) layer 51a in the initial prediction model 51 outputs the sample text representation feature 1 corresponding to the sample text encoding feature 1; and then the sample text data 1 is input into the initial prediction model 51 again, and the text encoder 51a in the initial prediction model 51 performs text encoding processing on the sample text data 1 to obtain the sample text encoding feature 1, and the dropout layer 51a in the initial prediction model 51 outputs the sample text representation feature 1' corresponding to the sample text encoding feature 1.

[0180] Among them, the text encoder 51a can be a Transformer structure encoder, BERT structure, GPT, BART structure, RNN model, LSTM (Long Short-Term Memory), etc., or it can be a deformation of any one of the above models, or a combination of any two or more of the above models, etc. The embodiment of the present application does not limit the network structure of the text encoder 51a.

[0181] The function of the dropout layer 51a is to randomly disable some neurons of the text encoder 51a so that the initial prediction model 51 will not be overly dependent on a specific neuron. Since the neurons that are randomly inactivated by the dropout layer 51a are not the same each time, two different sample text representation features can be obtained by inputting the same sample text data twice into the initial prediction model 51. Therefore, the sample text representation features can be expanded by the dropout layer 51a. For example, by inputting the sample text data 1 twice into the initial prediction model 51, two different sample text representation features can be obtained, namely, sample text representation feature 1 and sample text representation feature 1'. Similarly, sample text representation feature 2 and sample text representation feature 2' corresponding to sample text data 2, and sample text representation feature 3 and sample text representation feature 3' corresponding to sample text data 3 can be obtained; add sample text representation feature 1, sample text representation feature 1', sample text representation feature 2, sample text representation feature 2', sample text representation feature 3 and sample text representation feature 3' to the sample feature set 50b.

[0182] After obtaining the sample feature set 50b, the positive samples and negative samples of the initial prediction model 51 can be constructed according to the sample text representation features in the sample feature set 50b. Since the sample text representation feature 1 and the sample text representation feature 1' are sample text features obtained by extracting features from the same sample text data, the similarity between the sample text representation feature 1 and the sample text representation feature 1' is relatively high, so the sample text representation feature 1 and the sample text representation feature 1' can be determined as a positive sample pair and added to the positive sample set 50c. The sample text representation feature 2 and the sample text representation feature 2', as well as the sample text representation feature 3 and the sample text representation feature 3' are sample text features obtained by extracting features from the sample text data 2 and the sample text data 3, respectively, and the similarity with the sample text representation feature 1' is relatively low, so the sample text representation feature 1', the sample text representation feature 2, the sample text representation feature 2', the sample text representation feature 3 and the sample text representation feature 3' can be determined as negative samples and added to the negative sample set 50d. Similarly, the sample text representation feature 2 and the sample text representation feature 2' in the sample feature set 50b can also be used as positive samples, and the sample text representation feature 2', the sample text representation feature 1, the sample text representation feature 1', the sample text representation feature 3 and the sample text representation feature 3' can be used as negative samples, etc., and the embodiments of the present application do not limit this. Therefore, when training the initial prediction model 51, there is no need to label the sample features, and the training of the initial prediction model 51 can be completed by self-supervision.

[0183] Furthermore, the model loss corresponding to the initial prediction model 51 can be constructed by using the sample text representation features in the positive sample set 50c and the sample text representation features in the negative sample set 50d. The model loss corresponding to the initial prediction model 51 can be shown as the following formula (5):

[0184]

[0185] Among them, in formula (5) represents the model loss corresponding to the initial prediction model 51; h′ represents the sample text representation feature 1′; h represents the sample text representation feature 1; T represents the number of features of the sample text representation feature in the negative sample set 50d; h t Represents the feature of any sample text in the negative sample set 50d; cos() represents the cosine similarity between two features; τ represents the adjustment parameter, which is used to control the degree of attention paid by the initial prediction model 51 to difficult samples, and its specific value can be determined according to actual conditions; log() represents logarithmic operation.

[0186] like Figure 8As shown, after obtaining the model loss corresponding to the initial prediction model 51, the network parameters of the initial prediction model 51 can be iteratively trained according to the model loss until the updated network parameters meet the training end condition, and the training is stopped, and the initial prediction model 51 at the end of the training is determined as the information prediction model. Among them, the training end condition of the initial prediction model 51 may include but is not limited to: the number of training times of the initial prediction model 51 reaches the preset maximum number of iterations, or the update amplitude of the network parameters of the initial prediction model 51 is less than the preset error value, etc., and the embodiment of the present application does not limit this.

[0187] For example, in the training phase of the initial prediction model 51, an optimization method such as AdamW (an optimization method) or gradient descent with momentum can be used as an optimizer for training. In the training phase of the initial prediction model 51, the initial prediction model 51 can be set to train 10 epochs (iterative rounds), one epoch represents a complete sample text data set 50a, epoch=10 represents a complete training of the sample text data set 50a 10 times, and the initial prediction model 50 at the end of the 10th epoch training is determined as the text processing model.

[0188] After obtaining the text processing model, the first text feedback data can be input into the text processing model, and the first text representation feature corresponding to the first text feedback data can be output through the text processing model. It can be understood that when the first text feedback data includes sub-text feedback data corresponding to multiple feedback accounts, each sub-text feedback data can be input into the text processing model to obtain the sub-text representation feature corresponding to each sub-text feedback data, and then the average value of each sub-text representation feature is calculated, and the average value is determined as the first text representation feature corresponding to the first text feedback data; similarly, the second text representation feature corresponding to the second text feedback data can be obtained in the same way.

[0189] Furthermore, an image processing model may be used to extract features from the first image feedback data to obtain first image representation features, and an image processing model may be used to extract features from the second image feedback data to obtain second image representation features. The network structure of the image processing model may include but is not limited to ResNet (a residual network model), DenseNet (a densely connected network model), Transformer network model, Conformer (a network model composed of a convolutional neural network and a Transformer), Swin Transformer (a network model based on a Transformer structure), and a deformation of any of the above models, or a combination of any two or more of the above models, etc. The embodiment of the present application does not limit the network structure of the image processing model.

[0190] Optionally, the first text representation feature and the first image representation feature may be feature fused to obtain a first multimodal feature corresponding to the first multimedia feedback data. The feature fusion method may include but is not limited to: ① performing feature splicing processing on the first text representation feature and the first image representation feature, and determining the spliced ​​feature as the first multimodal feature; ② calculating the average value between the first text representation feature and the first image representation feature, and determining the average value as the first multimodal feature. The first multimodal feature corresponding to the first multimedia feedback data may be obtained by using one or more of the above feature fusion methods. Similarly, the second multimodal feature corresponding to the second multimedia feedback data may be obtained in the same manner.

[0191] Furthermore, the first text representation feature may be determined as the first comprehensive multimedia representation feature, and the second text representation feature may be determined as the second comprehensive multimedia representation feature; or the first image representation feature may be determined as the first comprehensive multimedia representation feature, and the second image representation feature may be determined as the second comprehensive multimedia representation feature; or the first multimodal feature may be determined as the first comprehensive multimedia representation feature, and the second multimodal feature may be determined as the second comprehensive multimedia representation feature. The embodiments of the present application do not limit the method for obtaining the first comprehensive multimedia representation feature and the second comprehensive multimedia representation feature.

[0192] After obtaining the first comprehensive multimedia representation feature and the second comprehensive multimedia representation feature, the first subgraph similarity between the initial account subgraph p and the initial account subgraph q can be represented by the distance between the first comprehensive multimedia representation feature and the second comprehensive multimedia representation feature. The larger the distance between the two, the smaller the corresponding first subgraph similarity; conversely, the smaller the distance between the two, the greater the corresponding first subgraph similarity. The method used to calculate the distance between the two first comprehensive multimedia representation features and the second comprehensive multimedia representation features may include but is not limited to: one or more of the methods such as Euclidian distance, Manhattan distance, Minkowski distance, and cosine similarity.

[0193] In a possible implementation, the first subgraph similarity may include a first cosine similarity, an intersection-over-union ratio, and a second cosine similarity. The first cosine similarity may be as shown in the following formula (6):

[0194] sim model1 =cos(V p ,V q ) (6)

[0195] Among them, sim in formula (6) model1 represents the first cosine similarity; V p Represents the first comprehensive multimedia presentation feature; V q represents the second comprehensive multimedia representation feature; cos() represents the cosine similarity between two features.

[0196] In a possible implementation, the method for obtaining the intersection-and-union ratio in the first subgraph similarity may include: extracting key information from the first multimedia feedback data to obtain a first multimedia key feature corresponding to the first multimedia feedback data, extracting key information from the second multimedia feedback data to obtain a second multimedia key feature corresponding to the second multimedia feedback data; and obtaining the intersection-and-union ratio between the first multimedia key feature and the second multimedia key feature. The intersection-and-union ratio may be as shown in the following formula (7):

[0197]

[0198] Among them, sim in formula (7) model2 represents the intersection-over-union ratio, also known as the Jaccard coefficient; K p Represents the first multimedia key feature; K q Represents the second multimedia key feature; |K p ∩K q | represents the intersection of the first multimedia key feature and the second multimedia key feature; |K p ∪K q | represents the union of the first multimedia key feature and the second multimedia key feature.

[0199] In an embodiment of the present application, key information extraction can be performed on the first multimedia feedback data through a key information extraction model to obtain a first multimedia key feature corresponding to the first multimedia feedback data; and key information extraction can be performed on the second multimedia feedback data through a key information extraction model to obtain a second multimedia key feature corresponding to the second multimedia feedback data. The key information extraction model may include, but is not limited to, one or more of a term frequency-inverse document frequency (TFIDF) model, a BERT model, a latent Dirichlet allocation (LDA) model, and the like.

[0200] Among them, the TFIDF model can be used to evaluate the importance of a word for a file set or one of the files in a corpus. The importance of a word increases in direct proportion to the number of times it appears in the file, but at the same time it decreases in inverse proportion to the frequency of its appearance in the corpus. Among them, TF represents the frequency of an entry in a document. The main idea of ​​TFIDF is that if a word or phrase appears in an article with a high frequency TF and rarely appears in other articles, it is considered that the word or phrase has a good ability to distinguish categories. Therefore, when the first multimedia feedback data includes text feedback data, the TFIDF model can be used to extract the first keyword in the text feedback data, and the first keyword can be determined as the key information of the first multimedia feedback data. Further, the first multimedia key feature corresponding to the first keyword can be obtained, for example, the first keyword can be represented by a vector using one-hot (unique hot code) to obtain the first multimedia key feature; or the first keyword can be represented by a vector using word2vec (a word vector model) to obtain the first multimedia key feature. Similarly, the second multimedia key feature corresponding to the second multimedia feedback data can be obtained in the same way.

[0201] Furthermore, the first multimedia key feature can be reduced in dimension by one or more of the algorithms such as singular value decomposition (SVD) algorithm, principal component analysis (PCA) algorithm, non-negative matrix decomposition (NMF) algorithm, etc., to obtain a third comprehensive multimedia representation feature corresponding to the first multimedia key feature, and the second multimedia key information can be reduced in dimension to obtain a fourth comprehensive multimedia representation feature corresponding to the second multimedia key feature; and then the second cosine similarity in the first sub-graph similarity can be determined by the third comprehensive multimedia representation feature and the fourth comprehensive multimedia representation feature. The second cosine similarity can be shown as the following formula (8):

[0202] sim model3 =cos(M p ,M q ) (8)

[0203] Among them, sim in formula (8) model3 represents the second cosine similarity; M p Indicates the third comprehensive multimedia presentation feature; M q Represents the fourth comprehensive multimedia presentation feature.

[0204] Optionally, after obtaining the first sub-graph similarity between the initial account sub-graph p and the initial account sub-graph q, it can be determined as a similarity evaluation value between the initial account sub-graph p and the initial account sub-graph q.

[0205] Optionally, the geographic locations associated with the business accounts in the initial account subgraph p and the initial account subgraph q may also be obtained, and the second subgraph similarity between the initial account subgraph p and the initial account subgraph q may be obtained based on the geographic locations associated with the business accounts in the initial account subgraph p and the business accounts in the initial account subgraph q. The second subgraph similarity may be as shown in the following formula (9):

[0206]

[0207] Among them, sim in formula (9) feature1 Indicates the similarity of the second subgraph; L p represents the geographic location associated with the business account in the initial account subgraph p; q represents the geographic location associated with the business account in the initial account subgraph q; |L p ∩L q | represents the intersection of the geographic locations associated with the business accounts in the initial account subgraph p and the geographic locations associated with the business accounts in the initial account subgraph q; |L p ∪L q | represents the union of the geographic locations associated with the business accounts in the initial account subgraph p and the geographic locations associated with the business accounts in the initial account subgraph q.

[0208] Optionally, the total amount of feedback assets associated with the business accounts in the initial account subgraph p and the initial account subgraph q can also be obtained, and the third subgraph similarity between the initial account subgraph p and the initial account subgraph q can be obtained according to the total amount of feedback assets associated with the business accounts in the initial account subgraph p and the business accounts in the initial account subgraph q. The third subgraph similarity can be shown as the following formula (10):

[0209]

[0210] Among them, sim in formula (10) feature2 Indicates the similarity of the third subgraph; M p represents the total amount of feedback assets associated with the business accounts in the initial account subgraph p, wherein the total amount of feedback assets associated with the initial account subgraph p refers to the sum of the amount of feedback asset data associated with multiple business accounts contained in the initial account subgraph p; M q Represents the total amount of feedback assets associated with the business accounts in the initial account subgraph q, where the total amount of feedback assets associated with the initial account subgraph q refers to the sum of the feedback asset data volumes associated with multiple business accounts contained in the initial account subgraph q; |M p ∩M q | represents the intersection of the total amount of feedback assets associated with the business accounts in the initial account subgraph p and the total amount of feedback assets associated with the business accounts in the initial account subgraph q; |Mp ∪M q | represents the union of the total amount of feedback assets associated with the business accounts in the initial account subgraph p and the total amount of feedback assets associated with the business accounts in the initial account subgraph q.

[0211] After obtaining the first subgraph similarity, the second subgraph similarity, and the third subgraph similarity, a weighted sum of the first subgraph similarity, the second subgraph similarity, and the third subgraph similarity can be performed to obtain a similarity evaluation value between the initial account subgraph p and the initial account subgraph q. The similarity evaluation value between the initial account subgraph p and the initial account subgraph q can be shown as the following formula (11):

[0212] sim (p,q) =w1sim model1 +w2sim model2 +w3sim model3 +w4sim feature1 +w5sim feature2 (11)

[0213] Among them, sim in formula (11) (p,q) Represents the similarity evaluation value between the initial account subgraph p and the initial account subgraph q; w1 represents the first similarity weight coefficient; w2 represents the second similarity weight coefficient; w3 represents the third similarity weight coefficient; w4 represents the fourth similarity weight coefficient; w5 represents the fifth similarity weight coefficient; wherein, the first similarity weight coefficient, the second similarity weight coefficient, the third similarity weight coefficient, the fourth similarity weight coefficient and the fifth similarity weight coefficient are pre-set parameters, and the specific values ​​can be set according to actual conditions.

[0214] Step S211: According to the similarity evaluation value associated with each account sub-graph pair, the M initial account sub-graphs are subjected to sub-graph merging processing to obtain N business account sub-graphs.

[0215] like Figure 7 As shown, the similarity evaluation value 1 associated with the account subgraph pair 31a, the similarity evaluation value 2 associated with the account subgraph pair 31b, and the similarity evaluation value 3 associated with the account subgraph pair 31c can be obtained in the same manner as above. Then, the similarity evaluation value 1, the similarity evaluation value 2, and the similarity evaluation value 3 can be compared with the similarity threshold. If there is a similarity evaluation value greater than the similarity threshold, the two initial account subgraphs in the account subgraph pair corresponding to the similarity evaluation value are subjected to subgraph merging processing. Figure 7As shown, if the similarity evaluation value 2 is greater than the similarity threshold, the initial account subgraph 30a and the initial account subgraph 30c in the account subgraph pair 31b can be merged to obtain a candidate account subgraph 32a; and then the initial account subgraph 30b and the candidate account subgraph 32a can be determined as business account subgraphs. The similarity threshold is a limit or restriction value set to measure whether the two initial account subgraphs in an account subgraph pair are merged, and its specific value can be set according to actual applications.

[0216] In a possible implementation, the initial account subgraph 30a and the initial account subgraph 30c may be merged based on the connection relationship and edge weights of the various business accounts included in the initial account subgraph 30a and the various business accounts included in the initial account subgraph 30c in the account association graph 30 to obtain a candidate account subgraph 32a.

[0217] In a possible implementation, the initial account subgraph 30a and the initial account subgraph 30c may also be subjected to subgraph merging processing according to the account similarity between each business account included in the initial account subgraph 30a and each business account included in the initial account subgraph 30c. The calculation method of the account similarity may refer to the calculation method of the similarity evaluation value involved above. For example, the initial account subgraph 30a includes the business account c, and the initial account subgraph 30c includes the business account d. By calculating the account similarity cd between the business account c and the business account d, when the account similarity cd is greater than the merging threshold, a connection relationship between the business account c and the business account d may be established, and the account similarity cd may be determined as the edge weight between the business account c and the business account d. When the account similarity cd is less than or equal to the merging threshold, the connection relationship between the business account c and the business account d is not established. The definition of the merging threshold may refer to the definition of the similarity threshold.

[0218] Optionally, when there are multiple candidate account subgraphs, the similarity evaluation values ​​between the candidate account subgraphs can also be calculated. If the similarity evaluation value between two candidate account subgraphs is greater than the similarity threshold, the two candidate account subgraphs are merged until the merging stop condition is reached, then the merging is stopped and the final subgraph is determined as the business account subgraph. The merging stop condition can be set to the similarity evaluation values ​​between each subgroup being less than or equal to the similarity threshold, or the number of merging times reaches a specified number.

[0219] See also Fig. 9 , Fig. 9 This is a schematic diagram of generating a business account sub-graph provided by an embodiment of the present application. Fig. 9As shown, the generation process of the business account subgraph may include two modules: subgraph partitioning and subgraph merging: when two business accounts belonging to different business types in the business account set have an asset transfer relationship, the edge weights (i.e., the first edge weight) of the two can be set according to the amount of feedback asset data and the amount of asset transfer associated with the two business accounts; when two business accounts belonging to the same business type in the business account set have an environmental interaction relationship, the edge weights (i.e., the second edge weight) can be set for the two business accounts. Among them, the method for obtaining the first edge weight and the second edge weight can refer to the description above, and will not be repeated here. Then, according to the business account set, the connection relationship between each business account in the business account set, the first edge weight and the second edge weight, an account association graph can be constructed, and the account association graph can be subgraphed using a community partitioning algorithm (e.g., FastUnfolding algorithm, etc.) to obtain multiple initial account subgraphs.

[0220] Furthermore, the similarity evaluation values ​​between each initial account subgraph can be calculated based on the text processing model, the key information extraction model (for example, the TFIDF model and the SVD algorithm) and the account aggregation features (for example, the geographic information of the initial account subgraph, the total amount of asset feedback of the initial account subgraph, etc.), and the multiple initial account subgraphs are merged according to the similarity evaluation values ​​between the initial account subgraphs to obtain the business account subgraph. Among them, the subgraph merging method can refer to the description above and will not be repeated here. After obtaining the business account subgraph, relevant disposal measures can be implemented for the business accounts in the business account subgraph. For example, the account status information (for example, asset transfer status, login status, etc.) of the business accounts in each business account subgraph can be detected, and the account status information can be pushed to the corresponding account administrator, so as to closely monitor the black industry trends in the application platform, which helps to take effective management measures in time and prevent problems before they occur. It is also possible to implement corresponding management measures on the business accounts in each business account sub-graph, such as further tracing and investigation, account blocking, etc. It is also possible to organize each business account sub-graph into clues and output them to authoritative agencies, thereby effectively curbing illegal activities within the application platform.

[0221] In the embodiment of the present application, when two business accounts belonging to different business types in the business account set have an asset transfer relationship, a connection relationship between the two business accounts can be established, and the first edge weight between the two business accounts can be determined according to the amount of feedback asset data and the amount of asset transfer associated with the two business accounts; when two business accounts belonging to the same business type in the business account set have an environment interaction relationship, a connection relationship between the two business accounts can be established, and the second edge weight between the two business accounts can be set. It can be seen that the account association graph generated according to the first edge weight and the second edge weight can deeply mine the association relationship between each business account, so multiple business accounts can be managed at the same time, which helps to improve the management efficiency of business accounts. Relative to the account association graph, the business accounts contained in the business account subgraph are significantly smaller than the business accounts contained in the account association graph. By dividing a large and complex account association graph into multiple business account subgraphs, the management pressure of the business accounts in the business account set can be reduced. In addition, the business accounts contained in the business account subgraph are more closely connected, so the business account subgraph can provide a more effective basis for the group management of the business account set, thereby improving the management efficiency of the business account.

[0222] It is understandable that in the specific implementation of this application, relevant information of users in the application platform may be involved (for example, asset transfer information, feedback asset information, device login information, account registration information, feedback account information, etc.). When the above embodiments of this application are applied to specific products or technologies, the user's permission or consent is required, and the collection, use and processing of relevant data need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0223] See also Fig.10 , Fig.10 is a schematic diagram of the structure of a data processing device provided in an embodiment of the present application; it can be understood that the data processing device can be applied in Figure 1 Any terminal device in the terminal cluster in the corresponding embodiment (for example, terminal device 10a), or a server (for example Figure 1 The server 10d) shown in the figure is not limited here. Fig.10 As shown, the data processing device 1 may include a business account set acquisition module 11, a first edge weight determination module 12, a second edge weight determination module 13, an account association graph generation module 14 and a first subgraph partitioning processing module 15, wherein:

[0224] A business account set acquisition module 11 is used to acquire a business account set;

[0225] A first edge weight determination module 12 is configured to establish a connection relationship between the business account i and the business account j if there is an asset transfer relationship between the business account i belonging to the first business type and the business account j belonging to the second business type in the business account set, and determine the first edge weight between the business account i and the business account j according to the amount of feedback asset data associated with the business account i and the amount of asset transfer between the business account i and the business account j;

[0226] A second edge weight determination module 13 is configured to establish a connection relationship between business account x and business account y if business account x and business account y of the same business type in the business account set have an environment interaction relationship, and set a second edge weight between business account x and business account y;

[0227] An account association graph generation module 14, configured to generate an account association graph according to a business account set, a connection relationship between each business account in the business account set, a first edge weight, and a second edge weight;

[0228] The first sub-graph partitioning processing module 15 is used to perform sub-graph partitioning processing on the account association graph according to the association metric values ​​of the business accounts in the business account set in the account association graph to obtain N business account sub-graphs; the N business account sub-graphs are used as a basis for grouping management corresponding to the business account set, where N is a positive integer.

[0229] In a possible implementation, the first edge weight determination module 12 determines the first edge weight between the business account i and the business account j according to the amount of feedback asset data associated with the business account i and the amount of asset transfer between the business account i and the business account j, including:

[0230] Obtain the amount of feedback asset data associated with business account i, and the amount of asset transfer between business account i and business account j;

[0231] If the asset transfer amount is less than the amount of feedback asset data, the ratio between the asset transfer amount and the amount of feedback asset data is determined as the first edge weight between business account i and business account j;

[0232] If the asset transfer amount is greater than or equal to the amount of feedback asset data, and the asset transfer amount is less than or equal to the asset transfer threshold, the first constant parameter is determined as the first edge weight between business account i and business account j; the asset transfer threshold is the product of the first business coefficient and the amount of feedback asset data;

[0233] If the asset transfer amount is less than the asset transfer threshold, the first edge weight between the business account i and the business account j is determined according to the feedback asset data amount, the asset transfer amount, the first business coefficient and the second business coefficient.

[0234] In a possible implementation manner, if the asset transfer amount is less than the asset transfer threshold, the first edge weight determination module 12 determines the first edge weight between the business account i and the business account j according to the feedback asset data amount, the asset transfer amount, the first business coefficient, and the second business coefficient, including:

[0235] If the asset transfer amount is less than the asset transfer threshold, the difference between the first business coefficient and the second business coefficient is determined as the third business coefficient, and the ratio between the second constant parameter and the third business coefficient is determined as the edge weight coefficient;

[0236] The product of the ratio between the asset transfer amount and the feedback asset data amount and the edge weight coefficient is determined as the first candidate edge weight;

[0237] The ratio of the second business coefficient to the third business coefficient is determined as the second candidate edge weight, and the difference between the first candidate edge weight and the third candidate edge weight is determined as the first edge weight between the business account i and the business account j.

[0238] In a possible implementation, the data processing device 1 may further include: a first account information acquisition module 16 and a first interaction relationship determination module 17, wherein:

[0239] A first account information acquisition module 16, used to acquire historical login device information and account registration information associated with a business account in a business account set;

[0240] The first interaction relationship determination module 17 is used to determine that business account x and business account y have an environment interaction relationship if business account x and business account y of the same business type in the business account set have the same historical login device information, or business account x and business account y have the same account registration information.

[0241] In a possible implementation, the data processing device 1 may further include: a second account information acquisition module 18, a third account information acquisition module 19, and a second interactive relationship determination module 20, wherein:

[0242] The second account information acquisition module 18 is used to acquire a first feedback account set associated with the business account x and a second feedback account set associated with the business account y if the business account x and the business account y belong to the first business type;

[0243] The third account information acquisition module 19 is used to acquire a first business account set having an asset transfer relationship with the business account x, and a second business account set having an asset transfer relationship with the business account y;

[0244] The second interaction relationship determination module 20 is configured to determine that the business account x has an environment interaction relationship with the business account y if there is an intersection between the first feedback account set and the second feedback account set, or if there is an intersection between the first business account set and the second business account set.

[0245] In a possible implementation, the data processing device 1 may further include: an abnormal account determination module 21, a path elimination module 22, and a second subgraph division processing module 23, wherein:

[0246] The abnormal account determination module 21 is used to obtain the sum of edge weights corresponding to each business account in the account association graph, and determine the business account whose sum of edge weights is less than a first threshold as an abnormal account;

[0247] A path removal module 22 is used to take the path of the abnormal account in the account association map as the path to be pruned, remove the path to be pruned from the account association map, and obtain a pruned account association map;

[0248] The second subgraph partitioning processing module 23 is used to perform game subgraph partitioning processing on the target association graph and subgraph partitioning processing on the account association graph according to the association metric values ​​of the business accounts in the business account set in the pruned account association graph, so as to obtain N business account subgraphs.

[0249] In a possible implementation, the first subgraph partitioning processing module 15 performs subgraph partitioning processing on the account association graph according to the association metric values ​​of the business accounts in the business account set in the account association graph to obtain N business account subgraphs, including:

[0250] Obtain a neighbor account set of a business account a in the business account set, and obtain an association metric value between the business account a and each neighbor account in the neighbor account set;

[0251] Determine the neighbor account corresponding to the largest association metric value in the neighbor account set as a candidate neighbor account, and if the association metric value between the business account a and the candidate neighbor account is greater than a second threshold, merge the business account a and the candidate neighbor account into a first candidate group;

[0252] Obtain a set of neighbor candidate groups of the first candidate group in the account association graph, and obtain association metrics between the first candidate group and each neighbor candidate group in the set of neighbor candidate groups;

[0253] Determine the neighbor candidate group corresponding to the largest association metric value in the neighbor candidate group set as the pending neighbor candidate group, and if the association metric value between the first candidate group and the pending neighbor candidate group is greater than a second threshold, merge the first candidate group and the pending neighbor candidate group into a second candidate group;

[0254] According to the second candidate group included in the account association graph, the account association graph is divided into sub-graphs to obtain N business account sub-graphs; wherein the association metric values ​​between different business account sub-graphs are all less than or equal to the second threshold.

[0255] In a possible implementation, the neighbor account set includes the business account b; the first subgraph partitioning processing module 15 obtains the association metric value between the business account a and each neighbor account in the neighbor account set, including:

[0256] Get the graph weight corresponding to the account association graph, get the edge weight between business account a and business account b, get the sum of edge weights corresponding to business account a, and get the sum of edge weights corresponding to business account b;

[0257] The product of the sum of edge weights corresponding to business account a and the sum of edge weights corresponding to business account b is determined as the first candidate association metric value, and the ratio between the first candidate association metric value and the graph weight corresponding to the account association graph is determined as the second candidate association metric value;

[0258] Determine the difference between the edge weight between the business account a and the business account b and the second candidate association metric value as the third candidate association metric value;

[0259] The association metric value between business account a and business account b is determined according to the graph weight corresponding to the account association graph and the third candidate association metric value.

[0260] In a possible implementation, the first subgraph partitioning processing module 15 performs subgraph partitioning processing on the account association graph according to the association metric values ​​of the business accounts in the business account set in the account association graph to obtain N business account subgraphs, including:

[0261] According to the association metric values ​​of the business accounts in the business account set in the account association graph, the account association graph is divided into subgraphs to obtain M initial account subgraphs; M is an integer greater than 1;

[0262] Combine the M initial account subgraphs to obtain multiple account subgraph pairs, and obtain a similarity evaluation value between two initial account subgraphs included in each account subgraph pair;

[0263] According to the similarity evaluation value associated with each account subgraph pair, the M initial account subgraphs are merged to obtain N business account subgraphs.

[0264] In a possible implementation, the first subgraph division processing module 15 performs subgraph merging processing on the M initial account subgraphs according to the similarity evaluation value associated with each account subgraph pair to obtain N business account subgraphs, including:

[0265] Acquire first multimedia feedback data corresponding to a business account in an initial account subgraph p, and acquire second multimedia feedback data corresponding to a business account in an initial account subgraph q; the initial account subgraph p and the initial account subgraph q belong to any one of a plurality of account subgraph pairs;

[0266] Determine a first subgraph similarity between an initial account subgraph p and an initial account subgraph q according to the first multimedia feedback data and the second multimedia feedback data;

[0267] Obtaining a second subgraph similarity between the initial account subgraph p and the initial account subgraph q according to the geographic locations associated with the business accounts in the initial account subgraph p and the business accounts in the initial account subgraph q;

[0268] According to the total amount of feedback assets associated with the business accounts in the initial account subgraph p and the business accounts in the initial account subgraph q, the third subgraph similarity between the initial account subgraph p and the initial account subgraph q is obtained;

[0269] Perform weighted summation on the first subgraph similarity, the second subgraph similarity, and the third subgraph similarity to obtain a similarity evaluation value between the initial account subgraph p and the initial account subgraph q;

[0270] If the similarity evaluation value between the initial account sub-graph p and the initial account sub-graph q is greater than the similarity threshold, the initial account sub-graph p and the initial account sub-graph q are merged into a business account sub-graph.

[0271] In a possible implementation, the first subgraph similarity includes a first cosine similarity, an intersection-over-union ratio, and a second cosine similarity;

[0272] The first subgraph division processing module 15 determines the first subgraph similarity between the initial account subgraph p and the initial account subgraph q according to the first multimedia feedback data and the second multimedia feedback data, including:

[0273] Performing feature extraction processing on the first multimedia feedback data to obtain a first comprehensive multimedia representation feature corresponding to the first multimedia feedback data, and performing feature extraction processing on the second multimedia feedback data to obtain a second comprehensive multimedia representation feature corresponding to the second multimedia feedback data;

[0274] Obtaining a first cosine similarity between the first comprehensive multimedia representation feature and the second comprehensive multimedia representation feature;

[0275] Extracting key information from the first multimedia feedback data to obtain a first multimedia key feature corresponding to the first multimedia feedback data, and extracting key information from the second multimedia feedback data to obtain a second multimedia key feature corresponding to the second multimedia feedback data;

[0276] Obtaining an intersection-over-union ratio between the first multimedia key feature and the second multimedia key feature;

[0277] Performing dimensionality reduction processing on the first multimedia key feature to obtain a third comprehensive multimedia representation feature corresponding to the first multimedia key feature, and performing dimensionality reduction processing on the second multimedia key information to obtain a fourth comprehensive multimedia representation feature corresponding to the second multimedia key feature;

[0278] A second cosine similarity between the third comprehensive multimedia representation feature and the fourth comprehensive multimedia representation feature is obtained.

[0279] According to an embodiment of the present application, the steps involved in the data processing method shown above can be Fig.10 The data processing device 1 shown in the figure is executed by each module. For example, Figure 2 Step S101 shown can be performed by Fig.10 The business account set acquisition module 11 is executed. Figure 2 Step S102 shown can be performed by Fig.10 The first edge weight determination module 12 is shown to execute, Figure 2 Step S103 shown can be performed by Fig.10 The second edge weight determination module 13 shown is executed, Figure 2 Step S104 shown can be performed by Fig.10 The account association graph generation module 14 is shown to execute, Figure 2 Step S105 shown can be performed by Fig.10 The first sub-graph partitioning processing module 15 shown is used to perform the above steps.

[0280] According to one embodiment of the present application, Fig.10 The various modules in the data processing device 1 shown can be separately or all combined into one or several units to constitute, or one (some) of the units can be further divided into at least two functionally smaller sub-units, which can achieve the same operation without affecting the realization of the technical effects of the embodiments of the present application. The above modules are divided based on logical functions. In practical applications, the functions of one module can also be implemented by at least two units, or the functions of at least two modules can be implemented by one unit. In other embodiments of the present application, the data processing device 1 may also include other units. In practical applications, these functions can also be implemented with the assistance of other units, and can be implemented by the collaboration of at least two units.

[0281] In the embodiment of the present application, when two business accounts belonging to different business types in the business account set have an asset transfer relationship, a connection relationship between the two business accounts can be established, and the first edge weight between the two business accounts can be determined according to the amount of feedback asset data and the amount of asset transfer associated with the two business accounts; when two business accounts belonging to the same business type in the business account set have an environment interaction relationship, a connection relationship between the two business accounts can be established, and the second edge weight between the two business accounts can be set. It can be seen that the account association graph generated according to the first edge weight and the second edge weight can deeply mine the association relationship between each business account, so multiple business accounts can be managed at the same time, which helps to improve the management efficiency of business accounts. Relative to the account association graph, the business accounts contained in the business account subgraph are significantly smaller than the business accounts contained in the account association graph. By dividing a large and complex account association graph into multiple business account subgraphs, the management pressure of the business accounts in the business account set can be reduced. In addition, the business accounts contained in the business account subgraph are more closely connected, so the business account subgraph can provide a more effective basis for the group management of the business account set, thereby improving the management efficiency of the business account.

[0282] See also Fig.11 , Fig.11 Schematic diagram of the structure of a computer device provided in an embodiment of the present application. Fig.11 As shown, the computer device 1000 can be Figure 1 Any terminal device in the terminal cluster in the corresponding embodiment (for example, terminal device 10a), or a server (for example Figure 1 The server 10d shown in the figure will not be limited here. For ease of understanding, the embodiment of the present application takes the computer device as a terminal device as an example. The computer device 1000 may include: a processor 1001, a network interface 1004 and a memory 1005. In addition, the computer device 1000 may also include: a user interface 1003, and at least one communication bus 1002. Among them, the communication bus 1002 is used to realize the connection and communication between these components. Among them, the user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory 1005 may be a high-speed RAM memory, or it may be a non-volatile memory (non-volatile memory), such as at least one disk storage. The memory 1005 may optionally also be at least one storage device located away from the aforementioned processor 1001. As Fig.11As shown, the memory 1005 as a computer-readable storage medium may include an operating system, a network communication module, a user interface module, and a device control application program.

[0283] The network interface 1004 in the computer device 1000 can also provide a network communication function. Fig.11 In the computer device 1000 shown, the network interface 1004 can provide a network communication function; the user interface 1003 is mainly used to provide an input interface for the user; and the processor 1001 can be used to call the device control application stored in the memory 1005 to achieve:

[0284] Get the business account collection;

[0285] If there is an asset transfer relationship between a business account i belonging to the first business type and a business account j belonging to the second business type in the business account set, a connection relationship between the business account i and the business account j is established, and the first edge weight between the business account i and the business account j is determined according to the amount of feedback asset data associated with the business account i and the amount of asset transfer between the business account i and the business account j;

[0286] If there is an environment interaction relationship between business account x and business account y of the same business type in the business account set, a connection relationship between business account x and business account y is established, and a second edge weight between business account x and business account y is set;

[0287] Generate an account association graph according to the business account set, the connection relationship between each business account in the business account set, the first edge weight, and the second edge weight;

[0288] According to the association metric values ​​of the business accounts in the business account set in the account association graph, the account association graph is divided into sub-graphs to obtain N business account sub-graphs; the N business account sub-graphs are used as the basis for grouping management corresponding to the business account set, and N is a positive integer.

[0289] It should be understood that the computer device 1000 described in the embodiment of the present application can execute the above Figure 2 and Figure 5 The description of the data processing method in any corresponding embodiment can also be performed as described above. Fig.10 The description of the data processing device 1 in the corresponding embodiment will not be repeated here. In addition, the description of the beneficial effects of the same method will not be repeated here either.

[0290] In addition, it should be pointed out that: the embodiment of the present application also provides a computer-readable storage medium, and the computer-readable storage medium stores a computer program executed by the data processing device 1 mentioned above, and the computer program includes program instructions. When the processor executes the program instructions, it can execute the above-mentioned Figure 2 and Figure 5 The description of the data processing method in any corresponding embodiment will not be repeated here. In addition, the description of the beneficial effects of the same method will not be repeated. Among them, the storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc. For technical details not disclosed in the computer-readable storage medium embodiment involved in this application, please refer to the description of the method embodiment of this application. As an example, program instructions can be deployed on a computer device for execution, or on multiple computer devices located at one location, or on multiple computer devices distributed at multiple locations and interconnected by a communication network. Multiple computer devices distributed at multiple locations and interconnected by a communication network can form a blockchain system.

[0291] In addition, it should be noted that: the embodiment of the present application also provides a computer program product or a computer program, which may include computer instructions, and the computer instructions may be stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor may execute the computer instructions, so that the computer device executes the above Figure 2 and Figure 5 The description of the data processing method in any corresponding embodiment will not be repeated here. In addition, the description of the beneficial effects of the same method will not be repeated. For technical details not disclosed in the computer program product or computer program embodiment involved in this application, please refer to the description of the method embodiment of this application.

[0292] It should be noted that, for the above-mentioned various method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that this application is not limited by the order of the actions described, because according to this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by this application.

[0293] The steps in the method of the embodiment of the present application can be adjusted in order, combined and deleted according to actual needs.

[0294] The modules in the device of the embodiment of the present application can be merged, divided and deleted according to actual needs.

[0295] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the computer program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0296] The above disclosure is only the preferred embodiment of the present application, which certainly cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. A data processing method, characterized in that: include: Get the business account collection; If there is an asset transfer relationship between a business account i belonging to the first business type and a business account j belonging to the second business type in the business account set, a connection relationship between the business account i and the business account j is established, and a first edge weight between the business account i and the business account j is determined according to the amount of feedback asset data associated with the business account i and the amount of asset transfer between the business account i and the business account j; If there is an environment interaction relationship between business account x and business account y belonging to the same business type in the business account set, a connection relationship between the business account x and the business account y is established, and a second edge weight between the business account x and the business account y is set; generating an account association graph according to the business account set, the connection relationship between each business account in the business account set, the first edge weight, and the second edge weight; According to association metric values ​​of business accounts in the business account set in the account association graph, the account association graph is divided into subgraphs to obtain N business account subgraphs; The N business account subgraphs are used as a basis for grouping management corresponding to the business account set, where N is a positive integer.

2. The method according to claim 1, characterized in that The determining the first edge weight between the business account i and the business account j according to the amount of feedback asset data associated with the business account i and the amount of asset transfer between the business account i and the business account j includes: Obtaining the amount of feedback asset data associated with the business account i, and the amount of asset transfer between the business account i and the business account j; If the asset transfer amount is less than the feedback asset data amount, the ratio between the asset transfer amount and the feedback asset data amount is determined as the first edge weight between the business account i and the business account j; If the asset transfer amount is greater than or equal to the feedback asset data amount, and the asset transfer amount is less than or equal to the asset transfer threshold, the first constant parameter is determined as the first edge weight between the business account i and the business account j; the asset transfer threshold is the product of the first business coefficient and the feedback asset data amount; If the asset transfer amount is less than the asset transfer threshold, the first edge weight between the business account i and the business account j is determined according to the feedback asset data amount, the asset transfer amount, the first business coefficient, and the second business coefficient.

3. The method according to claim 2, characterized in that If the asset transfer amount is less than the asset transfer threshold, determining a first edge weight between the business account i and the business account j according to the feedback asset data amount, the asset transfer amount, the first business coefficient, and the second business coefficient, including: If the asset transfer amount is less than the asset transfer threshold, the difference between the first business coefficient and the second business coefficient is determined as a third business coefficient, and the ratio between the second constant parameter and the third business coefficient is determined as an edge weight coefficient; Determine the product of the ratio between the asset transfer amount and the feedback asset data amount and the edge weight coefficient as the first candidate edge weight; The ratio of the second business coefficient to the third business coefficient is determined as the second candidate edge weight, and the difference between the first candidate edge weight and the first candidate edge weight is determined as the first edge weight between the business account i and the business account j.

4. The method according to claim 1, characterized in that: The method further comprises: Obtain historical login device information and account registration information associated with the business accounts in the business account set; If business account x and business account y belonging to the same business type in the business account set have the same historical login device information, or the business account x and the business account y have the same account registration information, it is determined that the business account x and the business account y have an environment interaction relationship.

5. The method according to claim 1, characterized in that The method further comprises: If the business account x and the business account y belong to the first business type, then obtaining a first feedback account set associated with the business account x and a second feedback account set associated with the business account y; Acquire a first business account set having an asset transfer relationship with the business account x, and a second business account set having an asset transfer relationship with the business account y; If the first feedback account set and the second feedback account set have an intersection, or the first business account set and the second business account set have an intersection, it is determined that the business account x and the business account y have an environment interaction relationship.

6. The method according to claim 1, characterized in that The method further comprises: Obtaining the sum of edge weights corresponding to each business account in the account association graph, and determining a business account whose sum of edge weights is less than a first threshold as an abnormal account; The path of the abnormal account in the account association graph is used as a path to be pruned, and the path to be pruned is removed from the account association graph to obtain a pruned account association graph; Then, according to the association metric values ​​of the business accounts in the business account set in the account association graph, the account association graph is divided into subgraphs to obtain N business account subgraphs, including: According to the association metric values ​​of the business accounts in the business account set in the pruned account association graph, the target association graph is divided into game subgraphs, and the account association graph is divided into subgraphs to obtain N business account subgraphs.

7. The method according to claim 1, characterized in that The process of performing sub-graph division processing on the account association graph according to the association metric values ​​of the business accounts in the business account set in the account association graph to obtain N business account sub-graphs includes: Obtaining a neighbor account set of a business account a in the business account set, and obtaining an association metric value between the business account a and each neighbor account in the neighbor account set; Determine the neighbor account corresponding to the largest association metric value in the neighbor account set as a candidate neighbor account, and if the association metric value between the business account a and the candidate neighbor account is greater than a second threshold, merge the business account a and the candidate neighbor account into a first candidate group; Obtain a set of neighbor candidate groups of a first candidate group in the account association graph, and obtain association metrics between the first candidate group and each neighbor candidate group in the set of neighbor candidate groups; Determine the neighbor candidate group corresponding to the largest association metric value in the neighbor candidate group set as the pending neighbor candidate group, and if the association metric value between the first candidate group and the pending neighbor candidate group is greater than the second threshold, merge the first candidate group and the pending neighbor candidate group into a second candidate group; According to the second candidate group included in the account association graph, the account association graph is divided into sub-graphs to obtain N business account sub-graphs; wherein the association metric values ​​between different business account sub-graphs are all less than or equal to the second threshold.

8. The method according to claim 7, characterized in that The neighbor account set includes business account b; The obtaining of association metrics between the service account a and each neighbor account in the neighbor account set includes: Obtaining the graph weight corresponding to the account association graph, obtaining the edge weight between the business account a and the business account b, obtaining the sum of the edge weights corresponding to the business account a, and obtaining the sum of the edge weights corresponding to the business account b; The product of the sum of the edge weights corresponding to the business account a and the sum of the edge weights corresponding to the business account b is determined as a first candidate association metric value, and the ratio between the first candidate association metric value and the graph weight corresponding to the account association graph is determined as a second candidate association metric value; Determine the difference between the edge weight between the business account a and the business account b and the second candidate association metric value as a third candidate association metric value; The association metric value between the business account a and the business account b is determined according to the graph weight corresponding to the account association graph and the third candidate association metric value.

9. The method according to claim 1, characterized in that: The process of performing sub-graph division processing on the account association graph according to the association metric values ​​of the business accounts in the business account set in the account association graph to obtain N business account sub-graphs includes: According to the association metric values ​​of the business accounts in the business account set in the account association graph, the account association graph is divided into subgraphs to obtain M initial account subgraphs, where M is an integer greater than 1; Combining the M initial account subgraphs to obtain a plurality of account subgraph pairs, and obtaining a similarity evaluation value between two initial account subgraphs included in each account subgraph pair; According to the similarity evaluation value associated with each account sub-graph pair, the M initial account sub-graphs are subjected to sub-graph merging processing to obtain N business account sub-graphs.

10. The method according to claim 9, characterized in that According to the similarity evaluation value associated with each account subgraph, the M initial account subgraphs are subjected to subgraph merging processing to obtain N business account subgraphs, including: Acquire first multimedia feedback data corresponding to the business account in the initial account subgraph p, and acquire second multimedia feedback data corresponding to the business account in the initial account subgraph q; the initial account subgraph p and the initial account subgraph q belong to any one of the multiple account subgraph pairs; Determining a first subgraph similarity between the initial account subgraph p and the initial account subgraph q according to the first multimedia feedback data and the second multimedia feedback data; Obtaining a second subgraph similarity between the initial account subgraph p and the initial account subgraph q according to the geographic locations associated with the business accounts in the initial account subgraph p and the business accounts in the initial account subgraph q; Obtaining a third subgraph similarity between the initial account subgraph p and the initial account subgraph q according to the total amount of feedback assets associated with the business accounts in the initial account subgraph p and the business accounts in the initial account subgraph q; Performing a weighted summation on the first subgraph similarity, the second subgraph similarity, and the third subgraph similarity to obtain a similarity evaluation value between the initial account subgraph p and the initial account subgraph q; If the similarity evaluation value between the initial account subgraph p and the initial account subgraph q is greater than the similarity threshold, the initial account subgraph p and the initial account subgraph q are merged into a business account subgraph.

11. The method according to claim 10, characterized in that The first subgraph similarity includes a first cosine similarity, an intersection-over-union ratio and a second cosine similarity; The determining, according to the first multimedia feedback data and the second multimedia feedback data, a first subgraph similarity between the initial account subgraph p and the initial account subgraph q includes: performing feature extraction processing on the first multimedia feedback data to obtain a first comprehensive multimedia representation feature corresponding to the first multimedia feedback data, and performing feature extraction processing on the second multimedia feedback data to obtain a second comprehensive multimedia representation feature corresponding to the second multimedia feedback data; Obtaining a first cosine similarity between the first comprehensive multimedia representation feature and the second comprehensive multimedia representation feature; Extracting key information from the first multimedia feedback data to obtain a first multimedia key feature corresponding to the first multimedia feedback data, and extracting key information from the second multimedia feedback data to obtain a second multimedia key feature corresponding to the second multimedia feedback data; Obtaining an intersection-over-union ratio between the first multimedia key feature and the second multimedia key feature; Performing dimensionality reduction processing on the first multimedia key feature to obtain a third comprehensive multimedia representation feature corresponding to the first multimedia key feature, and performing dimensionality reduction processing on the second multimedia key information to obtain a fourth comprehensive multimedia representation feature corresponding to the second multimedia key feature; A second cosine similarity between the third comprehensive multimedia representation feature and the fourth comprehensive multimedia representation feature is obtained.

12. A data processing device, characterized in that: include: A business account set acquisition module is used to acquire a business account set; A first edge weight determination module is configured to establish a connection relationship between the business account i and the business account j if there is an asset transfer relationship between the business account i belonging to the first business type and the business account j belonging to the second business type in the business account set, and determine a first edge weight between the business account i and the business account j according to the amount of feedback asset data associated with the business account i and the amount of asset transfer between the business account i and the business account j; A second edge weight determination module is configured to establish a connection relationship between the business account x and the business account y, and set a second edge weight between the business account x and the business account y, if there is an environment interaction relationship between the business account x and the business account y in the business account set, and set a second edge weight between the business account x and the business account y; an account association graph generation module, configured to generate an account association graph according to the business account set, the connection relationship between each business account in the business account set, the first edge weight, and the second edge weight; A first subgraph partitioning processing module, configured to perform subgraph partitioning processing on the account association graph according to association metric values ​​of the business accounts in the business account set in the account association graph, to obtain N business account subgraphs; The N business account subgraphs are used as a basis for grouping management corresponding to the business account set, where N is a positive integer.

13. A computer device, characterized in that: including memory and processor; The memory is connected to the processor, the memory is used to store a computer program, and the processor is used to call the computer program so that the computer device executes the method according to any one of claims 1 to 11.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is suitable for being loaded and executed by a processor, so that a computer device having the processor executes the method according to any one of claims 1 to 11.

15. A computer program product, characterized in that The method comprises a computer program / instruction, which implements the method according to any one of claims 1 to 11 when executed by a processor.

Images