Team user identification method and apparatus, electronic device, medium, and program product

Abstract

The application provides a team user identification method and device, electronic equipment, medium and program product, which can be used in the field of data processing. The method comprises the following steps: obtaining first behavior data and first portrait data of a plurality of first users; converting the first behavior data of the first users into first risk image data corresponding to the first users; determining first user feature vectors of the first users according to the first risk image data corresponding to the first users and the first portrait data; clustering the first user feature vectors corresponding to each of the first users to obtain a cluster group, and determining the first users corresponding to the first user feature vectors in the cluster group as users belonging to the same risk team. In the application, the determination accuracy of the team users is improved.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to a team user identification method, device, electronic device, medium and program product. Background Art

[0002] Team arbitrage hinders the development of Internet marketing, preventing businesses from obtaining expected profit results and causing them to suffer many losses including financial and reputational losses. How to accurately identify members of arbitrage teams is an urgent problem that needs to be solved.

[0003] In the exemplary technology, the behavioral trajectories of users within a team converge, which means that the behaviors of the users belonging to the same team are highly similar. Therefore, the similarity between the behavioral data of the users can be used to determine the users belonging to the same team.

[0004] However, the above method only uses the similarity between behavioral data to determine team users, ignoring information other than behavioral data, which will result in low accuracy in determining team users. Summary of the Invention

[0005] The present application provides a team user identification method, device, electronic device, medium and program product, which solves the problem of low accuracy in determining team users.

[0006] In a first aspect, the present application provides a team user identification method, comprising:

[0007] Obtaining first behavior data and first portrait data of a plurality of first users;

[0008] converting the first behavior data of the first user into first risk image data corresponding to the first user;

[0009] determining a first user feature vector of the first user based on the first risk image data and the first portrait data corresponding to the first user;

[0010] The first user feature vectors corresponding to the first users are clustered to obtain cluster groups, and the first users corresponding to the first user feature vectors in the cluster groups are determined as users belonging to the same risk team.

[0011] In some embodiments, converting the first behavior data of the first user into first risk image data corresponding to the first user includes:

[0012] Obtaining each first behavior event and the corresponding occurrence time in the first behavior data of the first user;

[0013] Determining a first pixel point of the first behavior event in the initial two-dimensional image according to the first behavior event and the corresponding occurrence time;

[0014] Assigning pixel values ​​to first pixel points corresponding to each first behavior event in the initial two-dimensional image according to encoding parameters corresponding to each first behavior event to obtain an intermediate two-dimensional image;

[0015] First risk image data corresponding to the first user is determined according to the intermediate two-dimensional image.

[0016] In some embodiments, determining first risk image data corresponding to the first user based on the intermediate two-dimensional image includes:

[0017] determining a risk type corresponding to each of the first behavioral events;

[0018] According to the risk type of each first behavior event, a numerical value is assigned to the first pixel point of each first behavior event in the intermediate two-dimensional image to obtain a dual-channel image as the first risk image data corresponding to the first user.

[0019] In some embodiments, determining the risk type corresponding to each first behavioral event includes:

[0020] determining an event type of the first behavior event according to the event identifier of the first behavior event, and acquiring event data associated with the first behavior event;

[0021] Acquire event features based on the event data, and determine similarities between the event features and risk features associated with the event type;

[0022] A risk type of the first behavior event is determined according to the similarity.

[0023] In some embodiments, determining a first user feature vector of the first user based on the first risk image data and the first portrait data corresponding to the first user includes:

[0024] The first risk image data and the first portrait data corresponding to the first user are input into a user feature extraction model to obtain a first user feature vector corresponding to the first user output by the user feature extraction model.

[0025] In some embodiments, before inputting the first risk image data and the first portrait data corresponding to the first user into the user feature extraction model, the method further includes:

[0026] Acquire a plurality of training samples, where the training samples are constructed from the second risk image data and the second portrait data of the second user;

[0027] The first preset model and the second preset model are jointly trained according to each of the training samples to obtain the user feature extraction model, wherein the first preset model is used to process the second risk image data, and the second preset model is used to process the second portrait data.

[0028] In some embodiments, obtaining a plurality of training samples includes:

[0029] Obtaining second behavior data and second portrait data of multiple second users;

[0030] converting the second behavior data of the second user into second risk image data corresponding to the second user;

[0031] A training sample corresponding to the second user is constructed based on the second risk image data and the second portrait data corresponding to the second user.

[0032] In some embodiments, converting the second behavior data of the second user into second risk image data corresponding to the second user includes:

[0033] Obtaining each second behavior event and the corresponding occurrence time in the second behavior data of the second user;

[0034] Determining a second pixel point of the second behavior event in the original two-dimensional image according to the second behavior event and the corresponding occurrence time;

[0035] Assigning pixel values ​​to second pixel points corresponding to each second behavior event in the initial two-dimensional image according to encoding parameters corresponding to each second behavior event to obtain a target two-dimensional image;

[0036] Second risk image data corresponding to the second user is determined according to the target two-dimensional image.

[0037] In some embodiments, determining second risk image data corresponding to the second user based on the target two-dimensional image includes:

[0038] determining a risk type corresponding to each of the second behavioral events;

[0039] According to the risk type of each second behavior event, a numerical value is assigned to the second pixel point of each second behavior event in the target two-dimensional image to obtain a dual-channel image as the second risk image data corresponding to the second user.

[0040] In some embodiments, the jointly training the first preset model and the second preset model according to each of the training samples includes:

[0041] Inputting the training sample into a joint model, where the joint model is constructed based on the first preset model and the second preset model;

[0042] Obtaining a second user feature vector output by the joint model and a first team to which the second user corresponding to the training sample belongs;

[0043] The parameters of the joint model are updated according to the label in the training sample and the first team, where the label is used to indicate the second team to which the second user belongs.

[0044] In some embodiments, determining the first users corresponding to the respective first user feature vectors in the cluster group as users belonging to the same risk team includes:

[0045] Obtaining a target number of the first user feature vectors in the cluster group;

[0046] In response to the target number being greater than a preset number, the first users corresponding to the first user feature vectors in the cluster group are determined to be users belonging to the same risk team.

[0047] In some embodiments, determining a first user feature vector of the first user based on the first risk image data and the first portrait data corresponding to the first user includes:

[0048] determining, based on the first risk image data corresponding to the first user and the first portrait data, a first user feature vector of the first user and an identifier of a risk team to which the first user belongs;

[0049] After determining the first users corresponding to the first user feature vectors in the cluster group as users belonging to the same risk team, the method further includes:

[0050] determining a processing strategy corresponding to the cluster group according to a risk team identifier of the first user corresponding to the first user feature vector in the cluster group;

[0051] The processing strategy is performed on the account associated with the first user corresponding to each first user feature vector in the cluster group.

[0052] In some embodiments, obtaining the first behavior data of the plurality of first users includes:

[0053] Obtaining first behavior events of multiple first users in a set historical time period and the occurrence time of each first behavior event;

[0054] First behavior data corresponding to the first user is determined according to each of the first behavior events and the corresponding occurrence time.

[0055] In a second aspect, the present application provides a team user identification device, comprising:

[0056] An acquisition module, configured to acquire first behavior data and first portrait data of a plurality of first users;

[0057] a conversion module, configured to convert the first behavior data of the first user into first risk image data corresponding to the first user;

[0058] a determination module, configured to determine a first user feature vector of the first user based on the first risk image data and the first portrait data corresponding to the first user;

[0059] The clustering module is configured to cluster the first user feature vectors corresponding to the first users to obtain cluster groups, and determine the first users corresponding to the first user feature vectors in the cluster groups as users belonging to the same risk team.

[0060] In a third aspect, the present application provides an electronic device, comprising: a processor, and a memory and a communication interface communicatively connected to the processor;

[0061] The communication interface is used to communicate with other communication devices;

[0062] The memory is used to store computer-executable instructions;

[0063] The processor is used to execute the computer-executable instructions stored in the memory to implement the team user identification method provided in the first aspect.

[0064] In a fourth aspect, the present application provides a computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are executed by a processor, the team user identification method provided in the first aspect is implemented.

[0065] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the team user identification method provided in the first aspect.

[0066] The team user identification method, device, electronic device, medium and program product provided in this application convert the user's behavioral data into risk image data, and obtain the user's user feature vector based on the user's risk image data and the user's portrait data, and then cluster each user feature vector, so as to determine the users belonging to the same risk team based on the user feature vectors in the clustering group obtained by clustering. That is, by combining the user's portrait data with the user's behavioral data, the team to which the user belongs is accurately determined, thereby improving the accuracy of determining team users. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0068] Figure 1 A schematic diagram of the scenarios involved in the user identification method of this application team;

[0069] Figure 2 This is a schematic diagram of the steps of the team user identification method in this application embodiment. Figure 1 ;

[0070] Figure 3 This is a schematic diagram of the steps of the team user identification method in this application embodiment. Figure 2 ;

[0071] Figure 4 A schematic diagram of the process of constructing a dual-channel image according to an embodiment of the present application;

[0072] Figure 5 This is a schematic diagram of the steps of the team user identification method in this application embodiment. Figure 3 ;

[0073] Figure 6 A schematic diagram of the workflow of the user feature extraction model involved in the embodiment of the present application;

[0074] Figure 7 This is a schematic diagram of the steps of the team user identification method in this application embodiment. Figure 4 ;

[0075] Figure 8 This is a schematic diagram of the steps of the team user identification method in this application embodiment. Figure 5 ;

[0076] Figure 9 This is a schematic diagram of the steps of the team user identification method in this application embodiment. Figure 6 ;

[0077] Figure 10 A schematic diagram of a program module of a team user identification device provided in an embodiment of the present application;

[0078] Figure 11 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application.

[0079] The above drawings illustrate specific embodiments of the present application, which will be described in more detail below. These drawings and the textual description are not intended to limit the scope of the present application in any way, but rather to illustrate the concepts of the present application to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0080] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are 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 making creative work are within the scope of protection of this application. In addition, although the disclosure in this application is introduced according to one or several exemplary examples, it should be understood that each aspect of these disclosures can also constitute a complete implementation method separately.

[0081] It should be noted that the brief descriptions of terms in this application are only for the purpose of facilitating the understanding of the embodiments described below, and are not intended to limit the embodiments of this application. Unless otherwise specified, these terms should be understood according to their ordinary and usual meanings.

[0082] In addition, the terms "comprises" and "comprising" and any variations thereof are intended to cover but not exclude inclusion, for example, a product or device comprising a list of components is not necessarily limited to those components expressly listed but may include other components not expressly listed or inherent to such product or device.

[0083] The term "module" used in the embodiments of this application refers to any known or later developed hardware, software, firmware, artificial intelligence, fuzzy logic or combination of hardware and / or software code that can perform the functions associated with the component.

[0084] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0085] Team arbitrage hinders the development of Internet marketing, preventing businesses from obtaining expected profit results and causing them to suffer many losses including financial and reputational losses. How to accurately identify members of arbitrage teams is an urgent problem that needs to be solved.

[0086] In the exemplary technology, the behavioral trajectories of users within a team converge, which means that the behaviors of the users belonging to the same team are highly similar. Therefore, the similarity between the behavioral data of the users can be used to determine the users belonging to the same team.

[0087] The inventors of this application found that extracting behavioral features from behavioral data as one-dimensional data makes it impossible to fully mine behavioral information. In addition, only utilizing the similarity of behavioral data while ignoring information other than the behavioral sequence will lead to low accuracy in determining team users.

[0088] Therefore, the inventor of the present application came up with the idea of ​​converting the user's behavioral data into risk image data, and obtaining the user's user feature vector based on the user's risk image data and the user's portrait data, and then clustering each user feature vector, so as to determine the users belonging to the same risk team based on the each user feature vector in the cluster group obtained by clustering. That is, by combining the user's portrait data with the user's behavioral data, the team to which the user belongs can be accurately determined, thereby improving the accuracy of determining team users.

[0089] Reference Figure 1 , Figure 1 This is a scenario diagram of the team user identification method of the present application. The team user identification device 100 obtains the behavior data and portrait data of multiple users, and the behavior data and portrait data can be obtained from the database 200. The team user identification device 100 converts the behavior data into risk image data, and determines the user feature vector corresponding to the user based on the user's risk image data and portrait data. The team user identification device 100 clusters each user feature vector to obtain a cluster group. The users corresponding to each user feature vector in the cluster group are confirmed to be users of the same risk team. The team user identification device 100 outputs prompt information, which is the information of each user belonging to the same risk team. The user information is, for example, identification information such as the user's account number.

[0090] Combine Figure 1 It should be noted that the following embodiments can exist independently or in combination with each other, and the same or similar contents will not be repeated in different embodiments.

[0091] Reference Figure 2 , Figure 2 Schematic diagram of the process of the team user identification method provided in the embodiment of this application Figure 1 ,Team user identification methods include:

[0092] Step S201: Acquire first behavior data and first portrait data of multiple first users.

[0093] In this embodiment, the execution entity is a team user identification device. For ease of description, the term "device" will be used to refer to the team user identification device. The device can be a server or a terminal device equipped with team user identification capabilities. A team user refers to a user belonging to a risk team, which is a group of organized users engaged in organized arbitrage activities.

[0094] The device obtains data of multiple users, where the user is defined as a first user. The data of the first user includes first behavioral data and first profile data of the first user. In one example, the first behavioral data includes multiple behavioral events of the first user and the time at which the behavioral events occurred. A behavioral event may be, for example, the first user receiving a coupon, such as a discount coupon or instant discount coupon. A behavioral event may also be an event in which the first user purchases goods using the received coupon, or an event in which the first user purchases goods using other preferential methods.

[0095] In another example, the device obtains first behavior events for multiple users within a set historical time period and the occurrence time of each first behavior event. The set historical time period can be a recent time period, such as the last week, the last day, or the last month, and the end time point of the recent time period can be today's midnight. The device determines the first behavior data corresponding to the first user based on the first behavior events and the corresponding occurrence times.

[0096] Additionally, primary behavior data can include behavioral events, occurrences, event types, and whether or not there is arbitrage risk. Event types include logins, coupon redemption, and QR code transactions. Occurrences can be specified down to the smallest detail, such as 2024-10-01-08. Furthermore, if a more specific description of user behavior is required, event types can be further categorized. For example, a coupon redemption event can include an activity ID, expanding the coupon redemption event from "receive coupon" to "receive coupon-activity ID."

[0097] The first profile data includes the first user's basic attributes, user behavior tags, and risk characteristics. Basic attributes include the account's registration time, registration source, and usual location; user behavior tags include the account's active time, cumulative payments, cumulative views, payment preferences, etc.; risk characteristics include whether the first user's account is blacklisted, recent disposal type, device risk characteristics, and geolocation tampering. Recent disposal type refers to whether the account has been recently warned or banned, etc. Device risk characteristics refer to whether the device where the first user's account is located is at risk.

[0098] Step S202: converting the first behavior data of the first user into first risk image data corresponding to the first user.

[0099] After obtaining the first behavior data, the device converts the first behavior data into risk image data corresponding to the first user, and the risk image data is defined as the first risk image data. Exemplarily, the first behavior data includes a first behavior event and an occurrence time. The first behavior event and the occurrence time are converted into pixels based on a mapping relationship, that is, each first behavior event has multiple pixels, and each pixel is constructed into a risk image. For example, the row number of the date where the occurrence time of the first behavior event is located, and the occurrence sequence number of the occurrence time on that date is the column number. For example, the first user generates 7 first behavior events on date A, and date A corresponds to row number 1. The 7 first behavior events have sequence numbers, and each sequence number is obtained by sorting the occurrence time. For example, the occurrence time of the first behavior event a is the earliest among the first behavior events in the 7, then the sequence number of the first behavior event a is 1, and therefore the column number of the first behavior event a is column number 1. In addition, the specific parameters in the first behavior event are represented as pixel points corresponding to the first behavior event. In this way, the first behavior data can be converted into an image, and the image is used as the first risk image data.

[0100] Step S203: Determine a first user feature vector of the first user based on the first risk image data and the first portrait data corresponding to the first user.

[0101] After obtaining the first risk image data, the device constructs a first user feature vector for the first user based on the first risk image data and the first portrait data. Exemplarily, the device constructs a first feature vector corresponding to the first risk image data, constructs a second feature vector based on the first portrait data, and then concatenates the first feature vector and the second feature vector to obtain a user feature vector, which is defined as the first user feature vector.

[0102] Step S204 : clustering the first user feature vectors corresponding to the first users to obtain cluster groups, and determining the first users corresponding to the first user feature vectors in the cluster groups as users belonging to the same risk team.

[0103] After obtaining the first user feature vectors of each first user, the device clusters the first user feature vectors. Exemplarily, the similarity between each pair of first user feature vectors is calculated, and two first user feature vectors with a similarity greater than a preset similarity are grouped together. In this manner, the first user features are clustered to form cluster groups, where the similarity between any two first user feature vectors in the cluster group is greater than the preset similarity.

[0104] In one example, after the cluster groups are obtained, the first users corresponding to the first user feature vectors in the cluster groups may be determined as users belonging to the same risk team.

[0105] In another example, a threshold value K is set in the device. The threshold value K is a preset number. The device obtains a target number of first user feature vectors in a cluster group. When the target number exceeds the preset number, the first users corresponding to each first user feature vector in the cluster group are determined to belong to the same risk group. Furthermore, if there are multiple cluster groups, cluster groups corresponding to a number less than or equal to the preset target number are eliminated.

[0106] The device may output prompt information, where the prompt information is used to prompt that the first users of each first user feature vector in the cluster group are users of the same risk team.

[0107] In this embodiment, the user's behavioral data is converted into risk image data, and the user's user feature vector is obtained based on the user's risk image data and the user's portrait data, and then each user feature vector is clustered. Based on the user feature vectors in the clustering group obtained by clustering, users belonging to the same risk team are determined. That is, by combining the user's portrait data with the user's behavioral data, the team to which the user belongs is accurately determined, thereby improving the accuracy of determining team users.

[0108] Reference Figure 3 , Figure 3 This is a schematic diagram of the user identification process for this application team Figure 2 ,based on Figure 2 In the illustrated embodiment, step S202 includes:

[0109] Step S301: Acquire each first behavior event and the corresponding occurrence time in the first behavior data of the first user.

[0110] Step S302: determining a first pixel point of the first behavior event in the initial two-dimensional image according to the first behavior event and the corresponding occurrence time.

[0111] In this embodiment, the device obtains each first behavior event and the occurrence time corresponding to each first behavior event from the first behavior data, and thereby determines the first pixel point of the first behavior event in the initial two-dimensional image based on the first behavior event and the occurrence time.

[0112] For example, the initial two-dimensional image is a blank two-dimensional image with multiple pixels. The first behavior event has a corresponding pixel in the initial two-dimensional image, and the pixel is defined as the first pixel. The first behavior data includes each first behavior event of the first user within a week and the time of occurrence. The first behavior events are sorted by the time of occurrence. The sorting table is shown in Table 1:

[0113] Table-1

[0114]

[0115] Referring to the table above, five first-behavior events occurred on the date 20240925, in descending order of occurrence time. These are A, B, C, D, and E, representing the event types of first-behavior events. The date 20240925 corresponds to the first row of the initial two-dimensional image. The first-behavior event A on the date 20240925 is the earliest occurrence on that date, meaning that the sequence number for this event is 1, which corresponds to the first column. Therefore, the first-behavior event A on the date 20240925 is located at the pixel point in the first row and first column of the initial two-dimensional image.

[0116] By the above method, the first pixel point corresponding to each first behavior event in the first behavior data in the initial two-dimensional image can be determined.

[0117] Step S303 : assigning pixel values ​​to first pixels corresponding to each first behavior event in the initial two-dimensional image according to encoding parameters corresponding to each first behavior event to obtain an intermediate two-dimensional image.

[0118] Each first behavior event corresponds to an encoding parameter, and the encoding parameter is determined based on the event type of the first behavior event.

[0119] For example, if the event type of the first behavior event is A, the encoding parameter is 1;

[0120] If the event type of the first behavior event is B, the encoding parameter is 2; if the event type of the first behavior event is C, the encoding parameter is 3;

[0121] The event type of the first behavior event is D, so the encoding parameter is 4.

[0122] For example, based on Table 1, the encoding parameters of each first behavior event in the first behavior data are shown in Table 2, specifically as shown in Table 2:

[0123] Table-2

[0124] 1 2 3 4 5 1 3 4 5 7 2 3 4 5 6 4 7 8 11 0 3 4 5 8 11 2 3 4 5 0 1 3 4 0 0

[0125] 0 indicates that there is no first action event at that position. For example, if the date is 20241001, and only three first action events, A, C, and D, occur, then the fourth position in Table 1, which is the date of 20241001, is blank, and the encoding parameter for the corresponding position in Table 2 is 0.

[0126] After obtaining the encoding parameters for each first behavior event, a pixel value is assigned to the first pixel based on the encoding parameters. For example, if the encoding parameter of the first pixel in the first row and first column is 1, the pixel value of the first pixel is 1. After completing the pixel value assignment for each first pixel in each initial two-dimensional image, an intermediate two-dimensional image is obtained.

[0127] Step S304: determining first risk image data corresponding to the first user based on the intermediate two-dimensional image.

[0128] After obtaining the intermediate two-dimensional image, first risk image data corresponding to the first user is determined based on the intermediate two-dimensional image.

[0129] In one example, the intermediate two-dimensional image is used as the first risk image data.

[0130] In another example, the first behavior data includes a label indicating whether the first behavior event is an arbitrage risk, and the risk type is determined by the label of the first behavior event. For example, if the label indicates that there is an arbitrage risk, the risk type is an arbitrage risk; if the label indicates that there is no arbitrage risk, the risk type is a non-arbitrage type. The device assigns a numerical value to the first pixel point based on the risk type to obtain a dual-channel image as the first risk image data. For example, if the risk type is an arbitrage risk, the assigned value is 1; if the risk type is a non-arbitrage risk, the assigned value is 0. Figure 4 , based on each behavioral event in the behavioral sequence, a dual-channel image is constructed, Figure 4 The box in the middle represents the behavior events and the time of occurrence in the behavior sequence, and the pixel values ​​of the pixels in the two-dimensional image are assigned. Figure 4 The middle circle represents the risk feature vector (risk type) of the behavioral event and assigns the risk image channel to obtain a dual-channel image.

[0131] The risk type can also be determined based on similarity. Specifically, the first behavior event corresponds to an event type. The device can determine the event type of the first behavior event based on the event identifier of the first behavior event, obtain data associated with the first behavior event, obtain event features based on the event type, and determine the similarity between the event features and the risk features of the event type, thereby determining the risk type of the first behavior event based on the similarity. For example, the event type is a QR code transaction, and the risk features are that the frequency of the QR code transaction exceeds the preset frequency, the number of QR code refreshes in the QR code transaction exceeds the preset number, and the transaction amount in the QR code transaction exceeds the target amount; the event features extracted based on the event data include the number of QR code refreshes, the transaction amount, and the transaction frequency. When the transaction frequency exceeds the preset frequency, the transaction amount is greater than the target amount, and the number of refreshes is greater than the preset number, the similarity between the event feature and the risk feature is 100%, then the risk type is that there is arbitrage; if only the transaction frequency exceeds the preset frequency, then the similarity between the event feature and the risk feature is 33.3%, and the risk type is that there is no arbitrage.

[0132] In this embodiment, considering that the collection of behavioral data is consistent for all users, the types of behavioral events included in the collection range will be extracted, resulting in redundant or interfering information mixed into the behavioral sequence. In this regard, in the risk control scenario, normal users and risk team users may also perform the same operations at a certain stage. Therefore, with the help of the multi-channel characteristics of the image, a risk image channel is added to the above-mentioned two-dimensional image, and a numerical value is given to the two-dimensional image pixel based on the risk type obtained from the accumulated risk data. This value reflects the risk characteristics of the behavioral event corresponding to the pixel, such as whether there is arbitrage, the degree of arbitrage or the type of risk. The behavior before and after the behavioral event with risk characteristics can better reflect the risk information. In this way, the multi-channel characteristics of the image can be brought into play, and more risk feature information can be further incorporated into the behavioral sequence, thereby improving the accuracy of feature extraction.

[0133] Reference Figure 5 , Figure 5 This is a schematic diagram of the user identification process for this application team Figure 3 ,based on Figure 2 or Figure 3 In the illustrated embodiment, step S203 includes:

[0134] Step S501: input the first risk image data and the first portrait data corresponding to the first user into a user feature extraction model to obtain a first user feature vector corresponding to the first user output by the user feature extraction model.

[0135] In this embodiment, a user feature extraction model is provided in the device. The device inputs the first risk image data and the first portrait data into the user feature extraction model, and the user feature extraction model outputs a first user feature vector corresponding to the first user.

[0136] Furthermore, the device determines the first user feature vector of the first user and the risk team identification to which the first user belongs based on the first risk image data and the first portrait data of the first user. Specifically, the device inputs the first risk image data and the first portrait data into the user feature extraction model to obtain the first user feature vector and the risk team identification output by the user feature extraction model. The user feature extraction model is composed of a nonlinear model and a linear model. The linear model is used to process the first portrait data to obtain a vector representation, and the nonlinear model is used to process the first risk image data to obtain a vector representation. These two parts of the vector representation are spliced ​​and merged to complete the fusion of the behavioral data represented by the portrait data and the risk image data. Subsequently, the user feature vector is obtained by successively passing through the nonlinear transformation of several fully connected layers, and finally the risk team identification to which the first user belongs is given under the action of the Softmax function of the output layer. For details, refer to Figure 6 The portrait data is processed using a linear model, while the risk image data is processed using a deep neural network using a nonlinear model. The data from both models is processed through a fully connected layer and input into the Softmax output layer. The Softmax output layer generates user feature vectors and predicted values ​​for each team type. The team type with the largest predicted value is used as the risk team identifier. The nonlinear model includes a convolutional neural network (CNN), which extracts deep behavioral features of user behavior sequences from dual-channel images, thereby improving the generalization ability of the user feature extraction model.

[0137] After the first users corresponding to the first user feature vectors in the cluster group are determined to be users belonging to the same risk team, the device determines the processing strategy corresponding to the cluster group based on the risk team identifier of the first user corresponding to the first user feature vector in the cluster group, thereby processing the processing strategy based on the account associated with the first user corresponding to each first user feature vector in the cluster group. Exemplarily, the cluster group is a, and when the risk team identifier of cluster group a represents that the risk team is risk team A, and the processing strategy of risk team A is to ban the account, then the account of each first user in cluster group a is banned; the cluster group is b, and when the risk team identifier of cluster group b represents that the risk team is risk team B, and the processing strategy of risk team B is to warn the account, then an alarm message is sent to the account of each first user in cluster group b.

[0138] In this embodiment, the device quickly obtains the first user feature vector of the first user based on the user feature extraction model.

[0139] Reference Figure 7 , Figure 7 This is a schematic diagram of the user identification process for this application team Figure 4 ,based on Figure 5 In the embodiment shown, before step S501, the following steps are further included:

[0140] Step S701: Acquire multiple training samples, where the training samples are constructed from the second risk image data and second portrait data of the second user.

[0141] In this embodiment, the device can also train a user feature extraction model. Specifically, the device obtains multiple training samples, and the training samples are constructed by the second risk image data and the second portrait data of the second user. The second portrait data includes the basic attributes of the second user, user behavior labels, and risk characteristics. Basic attributes include the registration time, registration source, and usual place of the account; user behavior labels include the active time of the account, the cumulative payment of the account, the cumulative number of views, payment preferences, etc.; risk characteristics include whether the second user's account is a blacklisted account, recent disposal type, device risk characteristics, geographic location tampering, etc. The recent disposal type refers to whether the account has been warned or banned recently, etc. The device risk characteristics refer to whether the device where the second user's account is located is at risk.

[0142] The second risk image data is obtained from the second user's second behavior data. In one example, the second behavior data includes multiple behavior events of the second user and the time of occurrence of the behavior events. A behavior event may include the second user redeeming a coupon, such as a discount coupon or instant discount coupon. A behavior event may also include the second user purchasing goods using the received coupon, or using other preferential methods to purchase goods.

[0143] In another example, the device obtains second behavior events for multiple users within a set historical time period and the occurrence time of each second behavior event. The set historical time period can be a recent time period, such as the last week, the last day, or the last month. The end time point of the recent time period can be today's midnight. The device determines the second behavior data corresponding to the second user based on the second behavior events and the corresponding occurrence times.

[0144] Secondary behavioral data can also include behavioral events, occurrences, event types, and whether or not there's an arbitrage risk indicator. Event types include logins, coupon redemption, and QR code transactions. Occurrences can be detailed down to the smallest detail, such as 2024-10-01-08. Furthermore, if a more specific description of user behavior is required, event types can be further categorized. For example, a coupon redemption event can include an activity ID, expanding the coupon redemption event from "receive coupon" to "receive coupon-activity ID."

[0145] After obtaining the second behavior data, the device converts the second behavior data into risk image data corresponding to the second user, and the risk image data is defined as the second risk image data. Exemplarily, the second behavior data includes a second behavior event and an occurrence time. The second behavior event and the occurrence time are converted into pixels based on a mapping relationship, that is, each second behavior event has multiple pixels, and each pixel is constructed into a risk image. For example, the row number of the date where the occurrence time of the second behavior event is located, and the occurrence sequence number of the occurrence time on that date is the column number. For example, the second user generates 7 second behavior events on date A, and date A corresponds to row number 1. The 7 second behavior events have sequence numbers, and each sequence number is obtained by sorting the occurrence time. For example, the occurrence time of second behavior event a is the earliest among the 7 second behavior events, then the sequence number of second behavior event a is 1, and therefore the column number of second behavior event a is column number 1. In addition, the specific parameters in the second behavior event are represented as pixel points corresponding to the second behavior event. In this way, the second behavior data can be converted into an image, and the image is used as the second risk image data.

[0146] Step S702: Jointly train the first preset model and the second preset model according to each training sample to obtain a user feature extraction model, wherein the first preset model is used to process the second risk image data, and the second preset model is used to process the second portrait data.

[0147] After obtaining each training sample, the first preset model and the second preset model can be jointly trained to obtain a user feature extraction model. The first preset model is used to process the second risk image data, and the second preset model is used to process the second portrait data.

[0148] In addition, the second preset model is a linear model, which is implemented using a logistic regression model, but can actually be other models. The output of the second preset model, that is, the output of the logistic regression model, is expressed by the following formula:

[0149] Y=XW+b

[0150] Where X is the feature matrix input to the second preset model, each row represents the profile data of a user's account to be identified, W is the weight matrix, and b is the bias vector. Matrix W and vector b are the parameters to be learned by the second preset model.

[0151] The first preset model is a nonlinear model, which uses a convolutional neural network (CNN). Other image feature extraction models, such as improved CNN models and Transformer models, can also be used. The CNN network structure consists of an input layer, a convolutional layer, a pooling layer, and an output layer. Each neuron in the input layer corresponds to each pixel in the input image, passing the behavior sequence graph data to the convolutional layer. The convolutional layer performs a convolution operation to extract features from the input image, and then enters the pooling layer to reduce the feature dimension, remove redundancy, and retain important information. After passing through multiple convolutional and pooling layers, the data enters the output layer, which combines the input features and outputs a fixed-size vector.

[0152] During the data processing stage, necessary feature engineering is performed on the portrait data and dual-channel images obtained above based on business prior knowledge and feature engineering technology. The former focuses on text feature processing, while the latter focuses on image feature processing, including but not limited to data cleaning, normalization, feature selection, feature combination and feature dimensionality reduction.

[0153] Furthermore, the device inputs the training sample into the joint model, and the joint model is constructed based on the first preset model and the second preset model. The device obtains the second user feature vector output by the joint model and the first team to which the second user corresponding to the training sample belongs, and updates the parameters of the joint model based on the label in the training sample and the first team. The label is used to indicate the second team to which the second user belongs. The label in the training sample is manually labeled. The label is, for example, the type of team to which the second user belongs, such as real-person crowdsourcing and group control operation. The label and the first team update the parameters of the joint model, which refers to the difference between the labeled risk team (the value of the labeled risk team is 1) and the predicted value of the predicted risk team type, and updates the loss function in the joint model.

[0154] It should be noted that the label of the training sample can indicate whether the user belongs to a team. If the user belongs to a team, the label contains a serial number, which represents the type of team. For example, if the label of the training sample is serial number 1, it represents that the second user corresponding to the training sample is a member of team A. If the label of the training sample is serial number 3, it represents that the second user corresponding to the training sample is a member of team C. If the label of the training sample is a preset value, such as 0, it represents that the second user corresponding to the training sample is not a member of the team. Similarly, the predicted value obtained based on the training sample is a serial number and a probability value. The serial number represents the team predicted by the model to which the user belongs, and the probability value is the probability that the user belongs to the team corresponding to the serial number. The device uses the training sample and the first team to update the parameters of the joint model, which means that the device selects the maximum probability value and the label of the training sample to update the loss function in the joint model. The team with the serial number corresponding to the maximum probability value is the first team.

[0155] During joint model training, the second and first preset models of the joint model process the portrait data and risk image data respectively to obtain their own vector representations. These two vector representations are then concatenated and merged, and then passed through several fully connected layers for nonlinear transformation to further enhance the generalization ability of the model. The expression of the fully connected layer is as follows:

[0156] a (l+1) =f(W (l) a (l) +b (l) )

[0157] In the formula, l represents the number of hidden layers; f represents the activation function, which is generally ReLU; a (l) 、W (l) and b (l) The output, weight matrix, and bias vector of the lth hidden layer are the parameters that the joint model needs to learn. The fully connected layer is followed by the output layer. The output layer uses the Softmax function to obtain the probability that the second user's account belongs to each team type based on the output vector of the fully connected layer. The probability calculation formula is as follows:

[0158]

[0159] Where z i is the input value of the i-th node, and C is the number of output nodes, or in other words, the number of team types. The larger the probability value, the greater the probability that the second user belongs to the corresponding team type. The joint model ultimately selects the team type with the highest probability value as the team type for the second user's account to be identified. To minimize the loss function, the joint model uses a joint training method, with the parameters of the second and first preset models iteratively optimized simultaneously. The loss function uses the cross-entropy loss commonly used in multi-classification tasks. After sample training is complete, the optimal parameters are selected and imported into the joint model, ultimately resulting in a user feature extraction model.

[0160] In this embodiment, the device trains the joint model by acquiring various training samples, thereby obtaining a user feature extraction model, and then quickly acquires a user feature vector corresponding to the user based on the user feature extraction model.

[0161] Reference Figure 8 , Figure 8 This is a schematic diagram of the user identification process for this application team Figure 5 ,based on Figure 7 In the illustrated embodiment, step S701 includes:

[0162] Step S801: Obtain second behavior data and second portrait data of multiple second users.

[0163] Step S802: Convert the second behavior data of the second user into second risk image data corresponding to the second user.

[0164] Step S803: construct a training sample corresponding to the second user based on the second risk image data and the second portrait data corresponding to the second user.

[0165] In this embodiment, the training sample is composed of the second risk image data and the second image data. The second portrait data includes the basic attributes of the second user, user behavior labels and risk characteristics. Basic attributes include the registration time, registration source and usual place of the account; user behavior labels include the active time of the account, the cumulative payment of the account, the cumulative number of views, payment preferences, etc.; risk characteristics include whether the second user's account is a blacklisted account, recent disposal type, device risk characteristics, geographic location tampering, etc. The recent disposal type refers to whether the account has been warned or banned recently, etc. The device risk characteristics refer to whether the device where the second user's account is located is at risk.

[0166] The second risk image data is obtained from the second user's second behavior data. In one example, the second behavior data includes multiple behavior events of the second user and the time of occurrence of the behavior events. A behavior event may include the second user redeeming a coupon, such as a discount coupon or instant discount coupon. A behavior event may also include the second user purchasing goods using the received coupon, or using other preferential methods to purchase goods.

[0167] In another example, the device obtains second behavior events for multiple users within a set historical time period and the occurrence time of each second behavior event. The set historical time period can be a recent time period, such as the last week, the last day, or the last month. The end time point of the recent time period can be today's midnight. The device determines the second behavior data corresponding to the second user based on the second behavior events and the corresponding occurrence times.

[0168] Secondary behavioral data can also include behavioral events, occurrences, event types, and whether or not there's an arbitrage risk indicator. Event types include logins, coupon redemption, and QR code transactions. Occurrences can be detailed down to the smallest detail, such as 2024-10-01-08. Furthermore, if a more specific description of user behavior is required, event types can be further categorized. For example, a coupon redemption event can include an activity ID, expanding the coupon redemption event from "receive coupon" to "receive coupon-activity ID."

[0169] After obtaining the second behavior data, the device converts the second behavior data into risk image data corresponding to the second user, and the risk image data is defined as the second risk image data. Exemplarily, the second behavior data includes a second behavior event and an occurrence time. The second behavior event and the occurrence time are converted into pixels based on a mapping relationship, that is, each second behavior event has multiple pixels, and each pixel is constructed into a risk image. For example, the row number of the date where the occurrence time of the second behavior event is located, and the occurrence sequence number of the occurrence time on that date is the column number. For example, the second user generates 7 second behavior events on date A, and date A corresponds to row number 1. The 7 second behavior events have sequence numbers, and each sequence number is obtained by sorting the occurrence time. For example, the occurrence time of second behavior event a is the earliest among the 7 second behavior events, then the sequence number of second behavior event a is 1, and therefore the column number of second behavior event a is column number 1. In addition, the specific parameters in the second behavior event are represented as pixel points corresponding to the second behavior event. In this way, the second behavior data can be converted into an image, and the image is used as the second risk image data.

[0170] In this embodiment, the device constructs training samples based on the second behavior data and the second portrait data of the second user, and trains the model based on each training sample to obtain a user feature extraction model.

[0171] Reference Figure 9 , Figure 9 This is a schematic diagram of the user identification process for this application team Figure 6 ,based on Figure 8 In the illustrated embodiment, step S802 includes:

[0172] Step S901: Acquire each second behavior event and the corresponding occurrence time in the second behavior data of the second user.

[0173] Step S902 : determining a second pixel point of the second behavior event in the original two-dimensional image according to the second behavior event and the corresponding occurrence time.

[0174] In this embodiment, the device obtains each second behavior event and the occurrence time corresponding to each second behavior event from the second behavior data, and thereby determines the second pixel point of the second behavior event in the original two-dimensional image based on the second behavior event and the occurrence time.

[0175] For example, the original two-dimensional image is a blank two-dimensional image having multiple pixels. The second behavior event has a corresponding pixel in the original two-dimensional image, and the pixel is defined as the second pixel. The second behavior data includes each second behavior event of the second user within a week and the time of occurrence. The second behavior events are sorted by occurrence time, and the sorting table is shown in Table 1 above.

[0176] Referring to Table 1, five second-behavior events occurred in descending order on the date 20240925. These events are designated A, B, C, D, and E, representing the event types of the second-behavior events. The date 20240925 corresponds to the first row of the original 2D image. The second-behavior event A on the date 20240925 is the earliest occurrence on that date, meaning its sequence number is 1, which corresponds to the first column. Therefore, the second-behavior event A on the date 20240925 is located at the pixel in the first row and first column of the original 2D image.

[0177] By the above method, the second pixel point corresponding to each second behavior event in the second behavior data in the original two-dimensional image can be determined.

[0178] Step S903 : assigning pixel values ​​to second pixel points corresponding to each second behavior event in the initial two-dimensional image according to encoding parameters corresponding to each second behavior event, to obtain a target two-dimensional image.

[0179] Each second behavior event has a corresponding encoding parameter, which is determined based on the event type of the second behavior event. For example, if the event type of the second behavior event is A, the encoding parameter is 1; if the event type of the second behavior event is B, the encoding parameter is 2; if the event type of the second behavior event is C, the encoding parameter is 3; and if the event type of the second behavior event is D, the encoding parameter is 4. For example, based on Table 1, the encoding parameters of each second behavior event in the second behavior data are shown in Table 2.

[0180] After obtaining the encoding parameters for each second behavior event, a pixel value is assigned to the second pixel based on the encoding parameters. For example, if the encoding parameter of the second pixel in the first row and first column is 1, the pixel value of the second pixel is 1. After completing the pixel value assignment for each second pixel in each original two-dimensional image, the target two-dimensional image is obtained.

[0181] Step S904: determining second risk image data corresponding to the second user according to the target two-dimensional image.

[0182] After obtaining the target two-dimensional image, second risk image data corresponding to the second user is determined based on the target two-dimensional image.

[0183] In one example, the target two-dimensional image is used as the second risk image data.

[0184] In another example, the second behavior data includes a label indicating whether the second behavior event is an arbitrage risk, and the risk type is determined by the labeling of the second behavior event. For example, if the label indicates that there is an arbitrage risk, the risk type is an arbitrage risk; if the label indicates that there is no arbitrage risk, the risk type is a non-arbitrage type. The device assigns a numerical value to the second pixel point based on the risk type to obtain a dual-channel image as the second risk image data. For example, if the risk type is an arbitrage risk, the assigned value is 1; if the risk type is a non-arbitrage risk, the assigned value is 0. Figure 4 , based on each behavioral event in the behavioral sequence, a dual-channel image is constructed, Figure 4 The box in the middle represents the behavior events and the time of occurrence in the behavior sequence, and the pixel values ​​of the pixels in the two-dimensional image are assigned. Figure 4 The middle circle represents the risk type of the behavioral event, and the value of another channel in the two-dimensional image is assigned to obtain a dual-channel image.

[0185] In this embodiment, considering that the collection of behavioral data is consistent for all users, the types of behavioral events included in the collection range will be extracted, resulting in redundant or interfering information mixed into the behavioral sequence. In this regard, in the risk control scenario, normal users and risk team users may also perform the same operations at a certain stage. Therefore, with the help of the multi-channel characteristics of the image, a risk image channel is added to the above-mentioned two-dimensional image, and a numerical value is given to the two-dimensional image pixel based on the risk type obtained from the accumulated risk data. This value reflects the risk characteristics of the behavioral event corresponding to the pixel, such as whether there is arbitrage, the degree of arbitrage or the type of risk. The behavior before and after the behavioral event with risk characteristics can better reflect the risk information. In this way, the multi-channel characteristics of the image can be brought into play, and more risk feature information can be further incorporated into the behavioral sequence, thereby improving the accuracy of feature extraction and constructing accurate training samples.

[0186] Based on the content described in the above embodiment, the present application also provides a team user identification device, referring to Figure 10 , Figure 10 This is a schematic diagram of program modules of a team user identification device provided in an embodiment of the present application. In some embodiments, the team user identification device 1000 includes:

[0187] An acquisition module 1010 is configured to acquire first behavior data and first portrait data of a plurality of first users;

[0188] A conversion module 1020, configured to convert the first behavior data of the first user into first risk image data corresponding to the first user;

[0189] A determination module 1030 is configured to determine a first user feature vector of the first user based on the first risk image data and the first portrait data corresponding to the first user;

[0190] The clustering module 1040 is configured to cluster the first user feature vectors corresponding to the first users to obtain cluster groups, and determine the first users corresponding to the first user feature vectors in the cluster groups as users belonging to the same risk team.

[0191] In some embodiments, the team user identification device 1000 is specifically used to:

[0192] Obtaining each first behavior event and the corresponding occurrence time in the first behavior data of the first user;

[0193] Determine, according to the first behavior event and the corresponding occurrence time, a first pixel point of the first behavior event in the initial two-dimensional image;

[0194] Assigning pixel values ​​to first pixels corresponding to each first behavior event in the initial two-dimensional image according to encoding parameters corresponding to each first behavior event to obtain an intermediate two-dimensional image;

[0195] First risk image data corresponding to the first user is determined according to the intermediate two-dimensional image.

[0196] In some embodiments, the team user identification device 1000 is specifically used to:

[0197] Determine the risk type corresponding to each first behavioral event;

[0198] According to the risk type of each first behavior event, a numerical value is assigned to the first pixel point of each first behavior event in the intermediate two-dimensional image to obtain a dual-channel image as the first risk image data corresponding to the first user.

[0199] In some embodiments, the team user identification device 1000 is specifically used to:

[0200] Determine the event type of the first behavior event according to the event identifier of the first behavior event, and obtain event data associated with the first behavior event;

[0201] Obtaining event features based on event data and determining the similarity between the event features and risk features associated with the event type;

[0202] The risk type of the first behavioral event is determined based on the similarity.

[0203] In some embodiments, the team user identification device 1000 is specifically used to:

[0204] The first risk image data and the first portrait data corresponding to the first user are input into the user feature extraction model to obtain a first user feature vector corresponding to the first user output by the user feature extraction model.

[0205] In some embodiments, the team user identification device 1000 is specifically used to:

[0206] Acquire a plurality of training samples, where the training samples are constructed from the second risk image data and the second portrait data of the second user;

[0207] The first preset model and the second preset model are jointly trained according to each training sample to obtain a user feature extraction model, wherein the first preset model is used to process the second risk image data, and the second preset model is used to process the second portrait data.

[0208] In some embodiments, the team user identification device 1000 is specifically used to:

[0209] Obtaining second behavior data and second portrait data of multiple second users;

[0210] converting the second behavior data of the second user into second risk image data corresponding to the second user;

[0211] A training sample corresponding to the second user is constructed based on the second risk image data and the second portrait data corresponding to the second user.

[0212] In some embodiments, the team user identification device 1000 is specifically used to:

[0213] Obtain each second behavior event and the corresponding occurrence time in the second behavior data of the second user;

[0214] Determine, according to the second behavior event and the corresponding occurrence time, a second pixel point of the second behavior event in the original two-dimensional image;

[0215] Assigning pixel values ​​to second pixel points corresponding to each second behavior event in the initial two-dimensional image according to encoding parameters corresponding to each second behavior event to obtain a target two-dimensional image;

[0216] Second risk image data corresponding to the second user is determined according to the target two-dimensional image.

[0217] In some embodiments, the team user identification device 1000 is specifically used to:

[0218] Determine the risk type corresponding to each second behavioral event;

[0219] According to the risk type of each second behavior event, a numerical value is assigned to the second pixel point of each second behavior event in the target two-dimensional image to obtain a dual-channel image as the second risk image data corresponding to the second user.

[0220] In some embodiments, the team user identification device 1000 is specifically used to:

[0221] Inputting the training sample into the joint model, where the joint model is constructed based on the first preset model and the second preset model;

[0222] Obtain the second user feature vector output by the joint model and the first team to which the second user corresponding to the training sample belongs;

[0223] Parameters of the joint model are updated according to the label in the training sample and the first team, where the label is used to indicate the second team to which the second user belongs.

[0224] In some embodiments, the team user identification device 1000 is specifically used to:

[0225] Obtaining a target number of first user feature vectors in a cluster group;

[0226] In response to the target number being greater than the preset number, the first users corresponding to the respective first user feature vectors in the cluster group are determined to be users belonging to the same risk team.

[0227] In some embodiments, the team user identification device 1000 is specifically used to:

[0228] Determining a first user feature vector of the first user and an identifier of a risk team to which the first user belongs based on the first risk image data and the first portrait data corresponding to the first user;

[0229] After determining the first users corresponding to the first user feature vectors in the cluster group as users belonging to the same risk team, the method further includes:

[0230] Determining a processing strategy corresponding to the cluster group according to the risk team identifier of the first user corresponding to the first user feature vector in the cluster group;

[0231] The processing strategy is performed on the account associated with the first user corresponding to each first user feature vector in the cluster group.

[0232] In some embodiments, the team user identification device 1000 is specifically used to:

[0233] Obtaining first behavior events of multiple first users in a set historical time period and the occurrence time of each first behavior event;

[0234] First behavior data corresponding to the first user is determined according to each first behavior event and the corresponding occurrence time.

[0235] It should be noted that the various steps in the team user identification method executed by the team user identification device are specifically referred to the above embodiment and will not be described in detail here.

[0236] Furthermore, based on the contents described in the above embodiments, an electronic device is also provided in an embodiment of the present application, which includes at least one processor, and a communication interface and a memory connected to the processor; wherein the communication interface is used to communicate with other communication devices, and the memory stores computer-executable instructions; the above at least one processor executes the computer-executable instructions stored in the memory to implement the various steps in the team user identification method described in the above embodiments.

[0237] In order to better understand the embodiments of the present application, refer to Figure 11 , Figure 11 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application.

[0238] like Figure 11 As shown, the electronic device 1100 of this embodiment includes: a processor 1101, a memory 1102, and a communication interface 1104; wherein:

[0239] Memory 1102, for storing computer-executable instructions;

[0240] The communication interface 1104 is used to communicate with other communication devices;

[0241] The processor 1101 is configured to execute computer-executable instructions stored in the memory to implement the various steps of the query optimization method described in the above embodiment.

[0242] Optionally, the memory 1102 may be independent or integrated with the processor 1101 .

[0243] When the memory 1102 is independently provided, the device further includes a bus 1103 for connecting the memory 1102 , the communication interface 1104 and the processor 1101 .

[0244] An embodiment of the present application provides a computer-readable storage medium, which stores computer-executable instructions. When a processor executes the computer-executable instructions, the various steps in the team user identification method described in the above embodiment are implemented.

[0245] An embodiment of the present application provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the various steps in the team user identification method described in the above embodiment.

[0246] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or module, which can be electrical, mechanical or other forms.

[0247] Modules described as separate components may or may not be physically separate, and components shown as modules may or may not be physical units, that is, they may be located in one place or distributed across multiple network elements. Some or all of these modules may be selected to achieve the purpose of this embodiment based on actual needs.

[0248] In addition, the functional modules in the various embodiments of the present application may be integrated into a single processing unit, or each module may exist physically separately, or two or more modules may be integrated into a single unit. The above-mentioned modules may be implemented in the form of hardware or hardware plus software functional units.

[0249] The above-mentioned integrated module implemented in the form of a software functional module can be stored in a computer-readable storage medium. The above-mentioned software functional module is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) or a processor to perform some steps of the methods of various embodiments of the present application.

[0250] It should be understood that the processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in the application may be directly implemented by a hardware processor or implemented by a combination of hardware and software modules in the processor.

[0251] The memory may include high-speed memory, and may also include non-volatile storage, such as at least one disk memory, and may also be a USB flash drive, a mobile hard disk, a read-only memory, a magnetic disk or an optical disk.

[0252] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. Buses can be classified into address buses, data buses, and control buses. For ease of illustration, the buses in the drawings of this application are not limited to just one bus or just one type of bus.

[0253] The storage medium may be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random access memory, electrically erasable programmable read-only memory, erasable programmable read-only memory, programmable read-only memory, read-only memory, magnetic storage, flash memory, magnetic disk, or optical disk. The storage medium may be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0254] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A team user identification method, characterized in that: include: Obtaining first behavior data and first portrait data of a plurality of first users; converting the first behavior data of the first user into first risk image data corresponding to the first user; determining a first user feature vector of the first user based on the first risk image data and the first portrait data corresponding to the first user; The first user feature vectors corresponding to the first users are clustered to obtain cluster groups, and the first users corresponding to the first user feature vectors in the cluster groups are determined as users belonging to the same risk team.

2. The method according to claim 1, characterized in that The converting the first behavior data of the first user into first risk image data corresponding to the first user includes: Obtaining each first behavior event and the corresponding occurrence time in the first behavior data of the first user; Determining a first pixel point of the first behavior event in the initial two-dimensional image according to the first behavior event and the corresponding occurrence time; Assigning pixel values ​​to first pixel points corresponding to each first behavior event in the initial two-dimensional image according to encoding parameters corresponding to each first behavior event to obtain an intermediate two-dimensional image; First risk image data corresponding to the first user is determined according to the intermediate two-dimensional image.

3. The method according to claim 2, characterized in that The determining, based on the intermediate two-dimensional image, first risk image data corresponding to the first user includes: determining a risk type corresponding to each of the first behavioral events; According to the risk type of each first behavior event, a numerical value is assigned to the first pixel point of each first behavior event in the intermediate two-dimensional image to obtain a dual-channel image as the first risk image data corresponding to the first user.

4. The method according to claim 3, characterized in that Determining the risk type corresponding to each first behavior event includes: determining an event type of the first behavior event according to the event identifier of the first behavior event, and acquiring event data associated with the first behavior event; Acquire event features based on the event data, and determine similarities between the event features and risk features associated with the event type; A risk type of the first behavior event is determined according to the similarity.

5. The method according to claim 1, wherein The determining, based on the first risk image data and the first portrait data corresponding to the first user, a first user feature vector of the first user includes: The first risk image data and the first portrait data corresponding to the first user are input into a user feature extraction model to obtain a first user feature vector corresponding to the first user output by the user feature extraction model.

6. The method according to claim 5, characterized in that Before inputting the first risk image data and the first portrait data corresponding to the first user into the user feature extraction model, the method further includes: Acquire a plurality of training samples, where the training samples are constructed from the second risk image data and the second portrait data of the second user; The first preset model and the second preset model are jointly trained according to each of the training samples to obtain the user feature extraction model, wherein the first preset model is used to process the second risk image data, and the second preset model is used to process the second portrait data.

7. The method according to claim 6, characterized in that The obtaining of multiple training samples includes: Obtaining second behavior data and second portrait data of multiple second users; converting the second behavior data of the second user into second risk image data corresponding to the second user; A training sample corresponding to the second user is constructed based on the second risk image data and the second portrait data corresponding to the second user.

8. The method according to claim 7, characterized in that The converting the second behavior data of the second user into second risk image data corresponding to the second user includes: Obtaining each second behavior event and the corresponding occurrence time in the second behavior data of the second user; Determining a second pixel point of the second behavior event in the original two-dimensional image according to the second behavior event and the corresponding occurrence time; Assigning pixel values ​​to second pixel points corresponding to each second behavior event in the initial two-dimensional image according to encoding parameters corresponding to each second behavior event to obtain a target two-dimensional image; Second risk image data corresponding to the second user is determined according to the target two-dimensional image.

9. The method according to claim 8, characterized in that The determining, based on the target two-dimensional image, second risk image data corresponding to the second user includes: determining a risk type corresponding to each of the second behavioral events; According to the risk type of each second behavior event, a numerical value is assigned to the second pixel point of each second behavior event in the target two-dimensional image to obtain a dual-channel image as the second risk image data corresponding to the second user.

10. The method according to claim 6, characterized in that The jointly training the first preset model and the second preset model according to each of the training samples includes: Inputting the training sample into a joint model, where the joint model is constructed based on the first preset model and the second preset model; Obtaining a second user feature vector output by the joint model and a first team to which the second user corresponding to the training sample belongs; The parameters of the joint model are updated according to the label in the training sample and the first team, where the label is used to indicate the second team to which the second user belongs.

11. The method according to claim 1, characterized in that The determining the first users corresponding to the first user feature vectors in the cluster group as users belonging to the same risk team includes: Obtaining a target number of the first user feature vectors in the cluster group; In response to the target number being greater than a preset number, the first users corresponding to the first user feature vectors in the cluster group are determined to be users belonging to the same risk team.

12. The method according to claim 1, characterized in that The determining, based on the first risk image data and the first portrait data corresponding to the first user, a first user feature vector of the first user includes: determining, based on the first risk image data corresponding to the first user and the first portrait data, a first user feature vector of the first user and an identifier of a risk team to which the first user belongs; After determining the first users corresponding to the first user feature vectors in the cluster group as users belonging to the same risk team, the method further includes: determining a processing strategy corresponding to the cluster group according to a risk team identifier of the first user corresponding to the first user feature vector in the cluster group; The processing strategy is performed on the account associated with the first user corresponding to each first user feature vector in the cluster group.

13. The method according to any one of claims 1 to 12, characterized in that The obtaining of first behavior data of a plurality of first users includes: Obtaining first behavior events of multiple first users in a set historical time period and the occurrence time of each first behavior event; First behavior data corresponding to the first user is determined according to each of the first behavior events and the corresponding occurrence time.

14. A team user identification device, characterized in that: include: An acquisition module, configured to acquire first behavior data and first portrait data of a plurality of first users; a conversion module, configured to convert the first behavior data of the first user into first risk image data corresponding to the first user; a determination module, configured to determine a first user feature vector of the first user based on the first risk image data and the first portrait data corresponding to the first user; The clustering module is configured to cluster the first user feature vectors corresponding to the first users to obtain cluster groups, and determine the first users corresponding to the first user feature vectors in the cluster groups as users belonging to the same risk team.

15. An electronic device, characterized in that: include: a processor, and a memory and a communication interface communicatively connected to the processor; The communication interface is used to communicate with other communication devices; The memory is used to store computer-executable instructions; The processor is used to execute the computer-executable instructions stored in the memory to implement the team user identification method as described in any one of claims 1-13.

16. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by the processor, the team user identification method according to any one of claims 1 to 13 is implemented.

17. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the team user identification method according to any one of claims 1 to 13 is implemented.

Images