Object behavior mode determination method and apparatus, computer device, and storage medium

Abstract

The application relates to an object behavior mode determination method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: obtaining behavior data of each behavior generated by a target object, wherein the behavior data comprises behavior description information of at least one dimension; performing data coding on each behavior data to obtain multi-dimensional coding data matched with each behavior; the dimensions of the multi-dimensional coding data comprise the dimension union of the dimensions of the behavior description information comprised by each behavior data; performing data mapping processing on at least one dimension of a coding data sequence composed of each multi-dimensional coding data to obtain a behavior sequence feature under the at least one dimension; and determining a behavior mode of the target object based on the behavior sequence feature under the at least one dimension. The method can realize accurate analysis of the behavior mode.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method, apparatus, computer device, storage medium, and computer program product for determining object behavior patterns. Background Technology

[0002] With the development of internet technology, various applications have emerged, enabling users to shop online, make online payments, and communicate, greatly facilitating their lives. However, due to the virtual nature of the internet, among the many internet users, there are inevitably some risky users whose risky behaviors (such as telecommunications fraud) pose a security threat to other users. To improve the security of the interactive environment, it is necessary to analyze the behavior of users or accounts to identify risky individuals.

[0003] However, in the process of behavioral analysis, existing technologies mainly rely on historical experience to mine behavioral feature information. For different behaviors triggered in different scenarios, it is difficult to achieve effective integration of relevant data, resulting in low accuracy. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, apparatus, computer device, computer-readable storage medium, and computer program product for determining object behavior patterns that can improve the accuracy of behavior pattern analysis results, in response to the above-mentioned technical problems.

[0005] Firstly, this application provides a method for determining object behavior patterns. The method includes:

[0006] Acquire behavioral data for each behavior generated by the target object, wherein the behavioral data contains behavioral description information in at least one dimension.

[0007] Each of the aforementioned behavioral data is encoded to obtain multidimensional encoded data matching each behavior; the dimensions of the multidimensional encoded data include the union of the dimensions of the behavioral description information contained in each of the behavioral data.

[0008] The encoded data sequence composed of the multidimensional encoded data is subjected to data mapping processing in at least one dimension to obtain behavioral sequence features in the at least one dimension;

[0009] The behavioral pattern of the target object is determined based on the behavioral sequence features under the at least one dimension.

[0010] In one embodiment, acquiring behavioral data for each behavior generated by the target object includes:

[0011] Obtain the raw behavioral data of each action produced by the target object;

[0012] The behavior description information of the same dimension in each of the original behavior data is normalized to obtain the behavior data of each behavior.

[0013] In one embodiment, the method further includes:

[0014] The multidimensional encoded data is sorted according to the chronological order of the behaviors represented by the multidimensional encoded data to obtain an encoded data sequence.

[0015] In one embodiment, dividing the multidimensional encoded data with the same target encoded character into the same group to obtain a multidimensional encoded data group includes: sorting the multidimensional encoded data with the same target encoded character according to the occurrence time of the behavior represented by the multidimensional encoded data to obtain a multidimensional encoded data group.

[0016] Secondly, this application also provides an apparatus for determining object behavior patterns. The apparatus includes:

[0017] The data acquisition module is used to acquire behavioral data of each behavior generated by the target object, wherein the behavioral data contains behavioral description information of at least one dimension.

[0018] The data encoding module is used to encode each of the behavioral data separately to obtain multidimensional encoded data that matches each behavior; the dimension of the multidimensional encoded data includes the union of the dimensions of the behavioral description information contained in each behavioral data.

[0019] The data mapping module is used to perform data mapping processing on the encoded data sequence composed of the multidimensional encoded data in at least one dimension to obtain behavioral sequence features in the at least one dimension.

[0020] The behavior pattern determination module is used to determine the behavior pattern of the target object based on the behavior sequence features under the at least one dimension.

[0021] In one embodiment, the object behavior pattern determination device further includes an arrangement information determination module; the arrangement information determination module is used to determine the dimensional union of the dimensions to which each behavior description data belongs based on the behavior description information contained in each behavior data; and to arrange the dimensions in the dimensional union to determine the arrangement information of each dimension.

[0022] The data encoding module is further configured to encode the behavior description information contained in each behavior data to obtain the encoded characters of the behavior description information; and to arrange the encoded characters based on the arrangement information of the dimension to which the behavior description information belongs to obtain the multidimensional encoded data of the behavior data.

[0023] In one embodiment, the data encoding module is further configured to arrange the encoded characters based on the arrangement information of the dimension to which the behavior description information belongs, to obtain initial multidimensional encoded data containing partial encoding; and to fill the dimensions of the initial multidimensional encoded data with target data to obtain updated multidimensional encoded data.

[0024] In one embodiment, the data acquisition module is further configured to acquire the original behavior data of each behavior generated by the target object; and to perform format normalization processing on the behavior description information of the same dimension in each of the original behavior data to obtain the behavior data of each behavior.

[0025] In one embodiment, the multidimensional encoded data in the encoded data sequence is arranged in chronological order according to the occurrence of the behavior represented by the multidimensional encoded data; the data mapping module is further configured to perform data mapping processing on the encoded data sequence in at least one dimension according to the arrangement order of each multidimensional encoded data in the encoded data sequence, to obtain behavioral sequence features in at least one dimension.

[0026] In one embodiment, the object behavior pattern determination device further includes a sorting module; the sorting module is used to sort the multidimensional encoded data according to the chronological order of the occurrence of the behavior represented by the multidimensional encoded data to obtain an encoded data sequence.

[0027] In one embodiment, the data acquisition module is further configured to acquire a sequence of behavioral data of the target object; the behavioral data in the sequence of behavioral data is used to characterize the behavior generated by the target object, and each behavioral data is sorted according to the occurrence time of the behavior; the sorting position of the multidimensional encoded data in the sequence of encoded data is the same as the sorting position of the behavioral data matched by the multidimensional encoded data in the sequence of behavioral data.

[0028] In one embodiment, the data mapping processing dimension includes a time dimension; the data mapping module further includes a time dimension data mapping unit.

[0029] The time-dimensional data mapping unit is used to acquire time-dimensional operators; the time-dimensional operators represent data mapping relationships in the time dimension; determine the duration conditions matched by the time-dimensional operators; determine continuous multi-dimensional encoded data whose time data meets the duration conditions from the encoded data sequence; and perform data mapping processing on the continuous multi-dimensional encoded data based on the time-dimensional operators to obtain the behavioral sequence features of the target object in the time dimension.

[0030] In one embodiment, the time dimension data mapping unit is further configured to determine the time window matched by the time dimension operator; based on the dynamically sliding time window in the encoded data sequence, each multidimensional encoded data in the same time window is regarded as continuous multidimensional encoded data.

[0031] In one embodiment, the time-dimensional data mapping unit is further configured to extract behavioral temporal features from continuous multidimensional encoded data within the same time window based on the time-dimensional operator, thereby obtaining continuous behavioral sequence features of the target object.

[0032] In one embodiment, the data mapping processing dimension includes the object dimension; the data mapping module further includes an object dimension data mapping unit.

[0033] The object dimension data mapping unit is used to obtain object dimension operators; the object dimension operators represent the data mapping relationship of the object dimension; determine the object selection conditions matched by the object dimension operators; filter out multi-dimensional encoded data containing object information and satisfying the object selection conditions from the encoded data sequence; and perform data mapping processing on the filtered multi-dimensional encoded data based on the object dimension operators to obtain the behavioral sequence features of the target object in the object dimension.

[0034] In one embodiment, the object selection condition is the object that performs the interactive behavior; the object dimension data mapping unit is further used to obtain the target encoded characters representing the interactive object in the multidimensional encoded data, divide the multidimensional encoded data with the same target encoded characters into the same group to obtain a multidimensional encoded data group; for each multidimensional encoded data group, the interactive behavior feature is extracted to obtain the interactive behavior sequence feature of the target object.

[0035] In one embodiment, the object dimension data mapping unit is further configured to sort the multidimensional encoded data with the same target encoded character according to the occurrence time of the behavior represented by the multidimensional encoded data to obtain a multidimensional encoded data group.

[0036] In one embodiment, the data mapping processing dimension includes a cumulative behavior dimension; the data mapping module further includes a cumulative behavior dimension data mapping unit.

[0037] The behavior cumulative dimension data mapping unit is used to determine multidimensional encoded data representing the same type of behavior from the multidimensional encoded data in the encoded data sequence; for each type of behavior, the repetitive behavior feature is extracted from the multidimensional encoded data representing the behavior to obtain the repetitive behavior sequence feature of the target object.

[0038] In one embodiment, the data mapping processing dimension includes a business dimension; the data mapping module further includes a business dimension data mapping unit.

[0039] The business dimension data mapping unit is used to filter out multi-dimensional encoded data representing the same business from the multi-dimensional encoded data in the encoded data sequence; and to perform business behavior feature extraction processing on the multi-dimensional encoded data representing the same business to obtain the business behavior sequence features of the target object.

[0040] In one embodiment, the object behavior pattern determination device further includes a classification module, used to acquire an object classification model trained based on behavior sequence feature sample data, wherein the behavior sequence feature sample data carries object category labels; and based on the object classification model, classify the target object according to the behavior sequence features of the target object to obtain the classification result of the target object.

[0041] In one embodiment, the target object includes a target account; the object classification model is a binary classification model that classifies accounts into risky accounts or non-risky accounts; the object behavior pattern determination device further includes an associated account tagging module;

[0042] The associated account marking module is used to obtain associated accounts that belong to the same entity as the target account when the classification result of the target account is a risky account; and to mark the behavioral sequence features of the associated accounts as risky.

[0043] In one embodiment, the data acquisition module is further configured to acquire behavioral data of each behavior generated by the target object in each time period;

[0044] The data mapping module is further configured to determine the encoded data sequence consisting of multi-dimensional encoded data corresponding to behaviors within the same time period; perform data mapping processing on the encoded data sequence of each time period according to at least one dimension to obtain the behavior sequence features of each time period; and determine the behavior pattern of the target object based on the behavior sequence features of each time period.

[0045] In one embodiment, the target object includes a target account; the object behavior pattern determination device further includes a matching module, used to profile the user of the target account based on the behavior pattern of the target account, and obtain user profile data; when the user profile data fails to match the authentication object of the target account, the user is marked based on the user profile data to mark the behavior pattern of the target account as a user.

[0046] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0047] Acquire behavioral data for each behavior generated by the target object, wherein the behavioral data contains behavioral description information in at least one dimension.

[0048] Each of the aforementioned behavioral data is encoded to obtain multidimensional encoded data matching each behavior; the dimensions of the multidimensional encoded data include the union of the dimensions of the behavioral description information contained in each of the behavioral data.

[0049] The encoded data sequence composed of the multidimensional encoded data is subjected to data mapping processing in at least one dimension to obtain behavioral sequence features in the at least one dimension;

[0050] The behavioral pattern of the target object is determined based on the behavioral sequence features under the at least one dimension.

[0051] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:

[0052] Acquire behavioral data for each behavior generated by the target object, wherein the behavioral data contains behavioral description information in at least one dimension.

[0053] Each of the aforementioned behavioral data is encoded to obtain multidimensional encoded data matching each behavior; the dimensions of the multidimensional encoded data include the union of the dimensions of the behavioral description information contained in each of the behavioral data.

[0054] The encoded data sequence composed of the multidimensional encoded data is subjected to data mapping processing in at least one dimension to obtain behavioral sequence features in the at least one dimension;

[0055] The behavioral pattern of the target object is determined based on the behavioral sequence features under the at least one dimension.

[0056] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0057] Acquire behavioral data for each behavior generated by the target object, wherein the behavioral data contains behavioral description information in at least one dimension.

[0058] Each of the aforementioned behavioral data is encoded to obtain multidimensional encoded data matching each behavior; the dimensions of the multidimensional encoded data include the union of the dimensions of the behavioral description information contained in each of the behavioral data.

[0059] The encoded data sequence composed of the multidimensional encoded data is subjected to data mapping processing in at least one dimension to obtain behavioral sequence features in the at least one dimension;

[0060] The behavioral pattern of the target object is determined based on the behavioral sequence features under the at least one dimension.

[0061] The aforementioned object behavior pattern determination method, apparatus, computer equipment, storage medium, and computer program product acquire behavior data of each behavior generated by the target object, and use the union of the dimensions of the behavior description information contained in the behavior data as the dimension of the multidimensional encoded data. This achieves dimensional unification of the multidimensional encoded data, enabling the multidimensional encoded data obtained from data encoding based on behavior data to have a unified data expression. This facilitates data mapping processing according to one or more dimensions. By expressing the order of the multidimensional encoded data through the encoded data sequence, the mapping processing yields behavior sequence features. This not only covers all behaviors of the target object but also gives the features the characteristic of sequential expression. Consequently, the behavior pattern obtained based on the behavior sequence features can accurately express the behavioral characteristics of the target object. Attached Figure Description

[0062] Figure 1 This is a diagram illustrating the application environment of an object behavior pattern determination method in one embodiment.

[0063] Figure 2 This is a flowchart illustrating a method for determining object behavior patterns in one embodiment;

[0064] Figure 3 This is a schematic diagram illustrating parallel data encoding in one embodiment;

[0065] Figure 4 This is a schematic diagram illustrating data encoding in batches in one embodiment;

[0066] Figure 5 This is a schematic diagram of multidimensional encoded data in one embodiment;

[0067] Figure 6 This is a schematic diagram of multidimensional encoded data in another embodiment;

[0068] Figure 7 This is a schematic diagram of multidimensional encoded data filled with target data in one embodiment;

[0069] Figure 8This is a schematic diagram illustrating the relationships between the behaviors of objects in one embodiment.

[0070] Figure 9 This is a schematic diagram of a dynamically sliding time window in one embodiment;

[0071] Figure 10 This is a flowchart illustrating the process of extracting behavioral sequence features based on multiple operators in one embodiment;

[0072] Figure 11 This is a flowchart illustrating the object behavior pattern determination method in another embodiment;

[0073] Figure 12 This is a schematic diagram illustrating the application of the object behavior pattern determination method to object classification in one embodiment;

[0074] Figure 13 A structural block diagram of an object behavior pattern determination device in one embodiment;

[0075] Figure 14 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0076] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0077] The object behavior pattern determination method provided in this application embodiment can be applied to, for example, Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store data that server 104 needs to process, such as the behavioral data of the target object. The data storage system can be integrated onto server 104 or placed in the cloud or on another server. The target object can be the user of terminal 102; the action triggered by the target object through terminal 102 constitutes the target object's behavior. Server 104 acquires behavioral data for each behavior generated by the target object on terminal 102. The behavioral data contains behavioral description information in at least one dimension. Server 104 encodes each behavioral data separately to obtain multi-dimensional encoded data matching each behavior. The dimensions of the multi-dimensional encoded data include the union of the dimensions of the behavioral description information contained in each behavioral data. Server 104 performs data mapping processing on the encoded data sequence composed of each multi-dimensional encoded data in at least one dimension to obtain behavioral sequence features in at least one dimension. Based on the behavioral sequence features in at least one dimension, server 104 determines the behavioral pattern of the target object. Furthermore, the server can also determine whether the target object is a risky user based on the target object's behavioral pattern and then take corresponding processing measures.

[0078] The terminal 102 can be, but is not limited to, various desktop computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle systems. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0079] In one embodiment, such as Figure 2 As shown, a method for determining object behavior patterns is provided. Taking the application of this method to a computer device as an example, the computer device can be... Figure 1 The method for determining the behavior pattern of an object in a server context specifically includes the following steps:

[0080] Step 202: Obtain behavioral data for each behavior generated by the target object. The behavioral data contains behavioral description information in at least one dimension.

[0081] The target object is the entity that generates a series of behaviors and requires behavioral pattern analysis. Specifically, the target object can be a single user or an account used by that user. For users with multiple registered accounts, each account can be treated as a separate object for behavioral pattern analysis. The analysis of the target object's behavioral patterns can specifically focus on the user's behavior on a designated platform. This designated platform can be a single platform or multiple interconnected platforms. A platform can be an application, a webpage, or a sub-application running within a parent application's runtime environment. Taking an application as an example, the behaviors generated by the target object can be those generated within application A, or those generated across multiple interconnected applications. For instance, if a user needs to pay after shopping, the shopping application and the payment application can be interconnected. Similarly, if the first and second payment applications are both linked to the same bank card, then the first and second payment applications are also interconnected applications.

[0082] The actions performed by the target audience can specifically refer to the actions taken by the target audience to achieve a specific purpose through interaction. Taking a user of a payment application as an example, the actions performed by the target audience can specifically be behaviors triggered within the payment application. These include payment behaviors such as receiving payments, making payments, topping up, and withdrawing funds, as well as operations such as binding a bank card, canceling a bank card, logging in, and changing a password. Taking a user of a shopping application as an example, the actions performed by the target audience can specifically be behaviors triggered within the shopping application, such as submitting an order and requesting a refund. Taking a user of a social application as an example, the actions performed by the target audience can specifically be behaviors triggered within the social application, such as adding friends, transferring money to friends, sending red envelopes, and deleting friends.

[0083] Behavioral data describes the specific content of behaviors triggered by a target object across various dimensions. Each piece of behavioral data contains at least one dimension of behavioral description data, which is attribute information specific to one dimension of the behavioral data. It should be noted that the dimensions of the behavioral description data for different behaviors can be the same, different, or only partially the same. For example, for a transfer behavior initiated by a target object, the behavioral data for this transfer behavior includes behavioral description information across multiple dimensions such as the transfer behavior identifier, the transfer recipient, the transfer amount, and the source of funds (e.g., wallet balance, bank card, etc.). As another example, for a friend-adding behavior initiated by a target object, the behavioral data for this friend-adding behavior includes behavioral description information across multiple dimensions such as the friend-adding behavior identifier, the method of adding (e.g., adding via phone number search, adding via business card, etc.), and the recipient.

[0084] Specifically, a computer device can obtain behavioral data for each behavior generated by the target object based on the data collection results obtained under the condition that the target object is authorized. In a specific application, when the computer device performing behavioral pattern analysis on the target object is the same computer device that triggers the target object's behavior based on interaction, it can directly collect the behavioral data of the target object to obtain the behavioral data of each behavior generated by the target object and perform subsequent processing. When the first computer device performing behavioral pattern analysis on the target object and the second computer device that triggers the target object's behavior based on interaction are different computer devices, the first computer device can collect the behavioral data of the behavior generated by the target object and then transmit the collected behavioral data directly or indirectly to the second computer device, so that the second computer device can obtain the behavioral data of the behavior generated by the target object and perform subsequent processing. Indirect transmission can involve first storing the collected behavioral data and then retrieving the behavioral data from the stored data.

[0085] Step 204: Encode each behavior data separately to obtain multidimensional encoded data that matches each behavior; the dimensions of the multidimensional encoded data include the union of the dimensions of the behavior description information contained in each behavior data.

[0086] Data encoding is a data processing procedure that encodes and unifies the dimensions of behavioral data with different dimensions of behavioral description information. The behavioral data of each behavior generated by the target object can be processed synchronously or asynchronously. For example, ... Figure 3 As shown, computer devices can synchronously encode data using data processing threads equal to the amount of behavioral data, thereby improving data encoding efficiency. For example, such as... Figure 4 As shown, computer devices can also encode behavioral data in batches using at least two data processing threads, balancing data encoding efficiency with the number of data processing threads required. Alternatively, the computer device can encode each piece of behavioral data sequentially to reduce the number of data processing threads needed. The specific data encoding method can be selected based on the actual application scenario. For example, for scenarios with high time requirements, a multi-threaded approach can be used, while for computer devices with lower data processing load capacity, a fewer-threaded approach can be used.

[0087] Multidimensional encoded data is the result of encoding behavioral data. Each behavioral data point corresponds to one multidimensional data code, and the number of multidimensional encoded data points is the same as the number of behavioral data points. The dimension of multidimensional encoded data is the union of the dimensions of the behavioral description information contained in each behavioral data point, and the dimension arrangement of each multidimensional encoded data point is the same.

[0088] Specifically, the dimensions and their arrangement in multidimensional encoded data can be pre-configured. When encoding behavioral data, the computer can directly encode the behavioral data based on the pre-configured dimension configuration parameters, according to the dimensional composition and arrangement represented by the dimension configuration parameters. This eliminates the need for real-time confirmation of the dimensions and their arrangement, effectively improving data encoding efficiency. In practical implementation, since the types of behaviors a target object can produce are limited, the dimensions in the multidimensional encoded data can be directly obtained by taking the union of dimensions based on the behavioral data of each possible behavior of the target object, according to the dimensions to which the behavioral description information belongs. By pre-configuring the dimension configuration parameters of multidimensional encoded data, it is also applicable to scenarios requiring encoding data based on different batches of behavioral data. This ensures that the multidimensional encoded data obtained from encoding different batches of behavioral data has the same dimensions, improving the universality of the multidimensional encoded data obtained from encoding different batches of data.

[0089] In one specific application, the target object may exhibit 20 possible behaviors. The dimensions of the multidimensional encoded data can be obtained by calculating the union of the dimensions of the behavioral description information contained in the behavioral data of these 20 behaviors, for example, 100 dimensions. The target object may actually exhibit 10 behaviors, and the union of the dimensions of the behavioral description information contained in the behavioral data of these 10 behaviors might contain 80 dimensions. In another specific application, the computer device can also directly analyze the dimensions of the behavioral description information contained in each behavior data before data encoding to obtain the dimension parameters of the multidimensional encoded data. Compared to pre-configuring the dimension parameters of the multidimensional encoded data, directly determining the dimension parameters results in fewer dimensions in the multidimensional encoded data, reducing the amount of data processing required for subsequent behavioral pattern analysis of the target object based on the behavioral data and improving data processing efficiency.

[0090] Step 206: Perform data mapping processing on the encoded data sequence composed of multi-dimensional encoded data in at least one dimension to obtain behavioral sequence features in at least one dimension.

[0091] In this context, the encoded data sequence refers to the result of arranging the multidimensional encoded data in sequence. The multidimensional encoded data in the encoded data sequence can be sorted according to the timestamps they carry, either by timestamp order or by the generation order of the matched behavioral data. The encoded data sequence can characterize the sequential relationship of each multidimensional encoded data, ensuring that the features obtained after data mapping processing have the same sequential relationship.

[0092] Data mapping is a process of transforming data from one representation to another. Specifically, data mapping can transform all data to obtain another, or it can transform a portion of the data. The portion of data can be selected from the entire dataset based on a predefined data mapping relationship. Data mapping can be implemented using operators that represent data mapping relationships or specific data transformation functions.

[0093] The dimension for data mapping processing can be a single dimension included in the union of the aforementioned dimensions, i.e., the smallest granularity data dimension, or it can be an aggregated dimension of multiple dimensions included in the union of the aforementioned dimensions, i.e., a larger granularity dimension. For example, data mapping processing can be the dimension of the time when the behavior occurs, which is a dimension in the union of dimensions. Data mapping processing can also be an aggregated dimension of the interaction object, which includes object dimensions such as the transfer recipient object in the transfer behavior, the transfer initiator object in the receiving behavior, and the add object in the friend adding behavior.

[0094] Behavioral sequence features are the results obtained after feature extraction from an encoded data sequence. Specifically, behavioral sequence features can be feature vectors, containing either a portion of the encoded data sequence or the results of data processing on a portion of the encoded data sequence. Behavioral sequence features in different dimensions are used to represent different combinations of behaviors, which in turn characterize the behavioral patterns of the target object.

[0095] Specifically, based on the commonalities in the dimensions of the behavioral patterns of the target object to be analyzed, the computer device determines the dimensions for data mapping processing. This dimension can be a single dimension, multiple dimensions, or multiple categories of dimensions. The computer device acquires a sequence of encoded data composed of various multi-dimensional encoded data, and performs data mapping processing on the encoded data sequence according to the determined dimensions to obtain the behavioral sequence features of the target object under the corresponding dimensions. In practical applications, the behavioral sequence features of the target object under different dimensions can be applied to case analysis scenarios targeting the target object to achieve a visual representation of the behavioral sequence features.

[0096] Step 208: Determine the behavior pattern of the target object based on the behavior sequence features in at least one dimension.

[0097] Among them, behavioral sequence features represent a combination of multiple behaviors. Computer devices can determine a behavioral pattern based on a single behavioral sequence feature, or they can determine a behavioral pattern based on multiple behavioral sequence features that satisfy similarity conditions. Specifically, similarity conditions can be calculated to be greater than a set threshold, or the proportion of identical components reaching a set ratio, etc.

[0098] In a specific application, computer devices can perform data mapping processing on encoded data sequences along dimensions such as time and object to obtain behavioral sequence features in these dimensions. This allows for the determination of fine-grained behavioral patterns of the target object in these dimensions, and further analysis of the similarity of behavioral sequence features across these dimensions reveals the overall behavioral pattern of the target object. Because behavioral sequence features inherently possess correlations between different behaviors, they offer stronger feature representation and interpretability. Furthermore, by using data encoding, data mapping does not rely on historical data or supervised learning, enabling real-time computation and analysis of behavioral data and improving the timeliness of behavioral pattern analysis of target objects.

[0099] The aforementioned method for determining object behavior patterns acquires behavioral data for each behavior generated by the target object, and uses the union of the dimensions of the behavioral description information contained in the behavioral data as the dimension of the multidimensional encoded data. This achieves dimensional unification of the multidimensional encoded data, enabling the multidimensional encoded data obtained from data encoding based on behavioral data to have a unified data expression. This facilitates data mapping processing by one or more dimensions. By expressing the order of the multidimensional encoded data through the encoded data sequence, the mapping processing yields behavioral sequence features that not only cover all behaviors of the target object but also give the features the characteristic of sequential expression. Consequently, the behavioral patterns obtained based on the behavioral sequence features can accurately express the behavioral characteristics of the target object.

[0100] In one embodiment, the object behavior pattern determination method further includes: determining the dimensional union of the dimensions to which each behavior description data belongs based on the behavior description information contained in each behavior data; arranging the dimensions in the dimensional union to determine the arrangement information of each dimension.

[0101] The dimension union is the result of taking the union of the sets of dimensions to which the behavioral description information belongs for each behavioral data point. For example, the dimensions to which the behavioral description information belongs for the first behavioral data point constitute dimension set A, the dimensions to which the behavioral description information belongs for the second behavioral data point constitute dimension set B, and so on. The computer device combines all the dimensions from each dimension set to form the dimension union. The dimension union contains the dimensions to which each behavioral description information belongs for each behavioral data point. For each dimension in the dimension union, its arrangement position in the multidimensional encoded data needs to be defined so that behavioral description information of the same dimension is in the same position in the multidimensional encoded data. The arrangement information of the dimensions is used to indicate the sorting position of the encoded character corresponding to the behavioral description information of that dimension in the multidimensional encoded data.

[0102] Furthermore, each behavioral data is encoded separately to obtain multidimensional encoded data matching each behavior, including: for each behavioral data, encoding the behavioral description information contained in the behavioral data to obtain encoded characters of the behavioral description information; arranging the encoded characters based on the arrangement information of the dimension to which the behavioral description information belongs to obtain multidimensional encoded data of the behavioral data.

[0103] Among these, behavioral description information refers to attribute information of a specific dimension within the behavioral data of the target object. Encoding behavioral description information involves converting it from one form or format to another using a pre-defined encoding method. Encoding multiple behavioral descriptions contained within the same behavioral data can be performed synchronously or asynchronously. For example, a computer device can encode each behavioral description within the same behavioral data synchronously, in batches, or sequentially. The specific configuration can be determined based on the actual scenario requirements.

[0104] Encoded characters are the result of encoding behavioral description information. Different behavioral description information yields different encoded characters after encoding. Encoded characters can be decoded to retrieve the behavioral description information; decoding is the reverse process of encoding. For example... Figure 5 As shown, encoded characters can be represented by characters, such as... Figure 6 As shown, encoded characters can be represented by numbers.

[0105] Specifically, the computer device can determine the set of dimensions corresponding to each behavior data based on the behavior description information contained in each behavior data, according to the dimension to which the behavior description data belongs. Then, it determines the union of the dimensions of each behavior data's respective set of dimensions. Based on a dimension arrangement rule, the computer device arranges the dimensions in the union of dimensions, and then determines the arrangement information of each dimension based on the positional relationship between the arranged dimensions. The dimension arrangement rule can be pre-configured, for example, arranging the dimensions of multiple behavior description information of the same category, or it can be a random arrangement. For each behavior data, the computer device encodes the behavior description information contained in the behavior data to obtain the coded characters matching each behavior description information. Then, it arranges the coded characters according to the arrangement information of the dimensions to which the behavior description information belongs, obtaining the multi-dimensional coded data matching each behavior data.

[0106] In this embodiment, the computer device determines the arrangement information of the dimensions so that the encoded characters corresponding to the behavioral description information of the dimension are arranged according to the arrangement information of the dimension to which the behavioral description information belongs. This realizes that the encoded characters matched by the behavioral description information of the same dimension in different behavioral data are located in the same position in different multidimensional encoded data, which facilitates data mapping processing by dimension and improves data processing efficiency and the accuracy of the results obtained after data mapping processing.

[0107] In one embodiment, based on the arrangement information of the dimension to which the behavior description information belongs, the encoded characters are arranged to obtain multidimensional encoded data of the behavior data, including: arranging the encoded characters based on the arrangement information of the dimension to which the behavior description information belongs to obtain initial multidimensional encoded data containing partial encoding; and filling the dimensions with missing encoded characters in the initial multidimensional encoded data with target data to obtain updated multidimensional encoded data.

[0108] The initial multidimensional encoded data consists of data with only some dimensions of encoded characters filled in; the filled dimensions are those to which the behavioral description information belongs. For dimensions lacking encoded characters, the computer device can fill them with target data to achieve a unified data format across each dimension of the multidimensional encoded data, facilitating subsequent data mapping processing. The target data can be data belonging to the same category as the encoded characters but different from any single encoded character, ensuring uniformity in the category of the encoded characters across the multidimensional encoded data. Specifically, it can be a target numerical value or a target symbol, for example... Figure 7 As shown, the target data to be filled can be the number "0", in other embodiments. The target data to be filled can also be the symbol " / ", etc.

[0109] In other embodiments, for the arrangement information based on the dimension to which the behavioral description information belongs, the coded characters are arranged, and the dimension with missing coded characters is determined. The computer device obtains multidimensional coded data by emptying the dimension with missing coded characters. The emptying process retains the information filling position for that dimension but does not fill it with data. By emptying the dimension with missing coded characters, data simplification can be achieved, improving data processing efficiency.

[0110] In this embodiment, the computer device first obtains initial multidimensional encoded data containing partial encoding, determines the dimensions of missing encoded characters, and then processes the dimensions of missing encoded characters accordingly. This ensures the uniformity of the data format across each dimension of the multidimensional encoded data, facilitating subsequent data mapping processing and improving the accuracy of the data mapping results. Specifically, the processing performed on the dimensions of missing encoded characters can involve filling in target data to achieve complete filling of encoded characters in each dimension of the multidimensional encoded data.

[0111] In one embodiment, obtaining behavioral data for each behavior generated by the target object includes: obtaining the original behavioral data for each behavior generated by the target object; and performing format normalization processing on the behavioral description information of the same dimension in each original behavioral data to obtain behavioral data for each behavior.

[0112] Among them, raw behavioral data can be the raw data collected. Since different behaviors may be obtained based on different collection methods or collection paths, the raw behavioral data collected for different behaviors, and for the same type of data in the same dimension, may have different data expression forms. For example, the time information collected through the first collection path is January 1, 2022, 00:10:30, while the time information collected through the second collection path is 2022 / 01 / 01 00:10:50. They both represent time information, but they have different data formats. For another example, the financial information collected through the first collection path is 500.00 yuan, while the financial information collected through the second collection path is 400 yuan. They both represent monetary information, but they have different precision decimal places.

[0113] Specifically, behavioral description information of the same dimension represents information of the same category. Computer devices determine the target dimension that needs to be standardized in terms of data format and determine the target data format of the target dimension. The target format can be any of the data formats that appear in the dimension, such as the format corresponding to the data with the highest sort position, or the format that appears most frequently. By performing format normalization processing on the behavioral description information of the same dimension, the computer devices can ensure that the behavioral description information of the same dimension should have the same data format, avoid the inability to identify similar data due to different data formats, and improve data accuracy.

[0114] In one embodiment, the multidimensional encoded data in the encoded data sequence are arranged in chronological order of the behaviors represented by the multidimensional encoded data. The encoded data sequence, composed of the multidimensional encoded data, is subjected to at least one dimension of data mapping processing to obtain behavioral sequence features in at least one dimension. This includes: performing at least one dimension of data mapping processing on the encoded data sequence according to the order of the multidimensional encoded data in the encoded data sequence to obtain behavioral sequence features in at least one dimension.

[0115] Each multidimensional encoded data in the encoded data sequence is obtained by encoding behavioral data. Behavioral data is used to characterize the behavior generated by the target object. Since the behavior of the target object is generated at different points in time, by obtaining the encoded data sequence arranged in the order of the occurrence time of the behavior matched by the multidimensional encoded data, the encoded data sequence can characterize the temporal relationship of the target object's behavior.

[0116] Furthermore, the encoded data sequence can be achieved by first sorting the behavioral data according to the time of occurrence of the behavior, and then encoding the behavioral data according to the sorting order to obtain multidimensional encoded data with the same sorting order, thus obtaining the encoded data sequence. This is achieved by sorting first and then encoding, which advances the sorting process, reduces the data processing steps after encoding, and improves the response speed of subsequent processing. Alternatively, the encoded data sequence can be based on encoding the behavioral data to obtain multidimensional encoded data, and then directly sorting the multidimensional encoded data according to the time of occurrence of the behavior. This method of encoding first and then sorting can improve the data processing speed from acquiring behavioral data to data encoding, and reduce the order conditions that need to be followed in the data encoding process.

[0117] Specifically, by acquiring a sequence of encoded data arranged chronologically according to the occurrence time of the behavior, computer equipment can directly determine the temporal relationship between the corresponding behavior and other behaviors based on the order of the multidimensional encoded data in the encoded data sequence. By performing data mapping processing on the encoded data sequence in at least one dimension, the behavioral sequence features obtained in at least one dimension conform to the temporal order of the behavior, thereby demonstrating whether the behaviors of the target object are related in the corresponding dimension, which facilitates accurate judgment of the behavior pattern of the target object.

[0118] In a specific application, the process of obtaining the encoded data sequence includes: sorting the multidimensional encoded data according to the chronological order of the behavior represented by the multidimensional encoded data to obtain the encoded data sequence.

[0119] Specifically, the computer device acquires multidimensional encoded data obtained by encoding each behavioral data. At least one of the multidimensional encoded data and behavioral data carries the occurrence time of the behavior it represents. The computer device can sort the multidimensional encoded data based on the chronological order of occurrence time to obtain an encoded data sequence that can represent the order in which the behaviors of the target object are generated. By directly encoding the multidimensional encoded data, the data encoding process does not need to follow the sequential relationship of each behavioral data. This makes it convenient for the computer device to simultaneously encode behavioral data and store multidimensional encoded data in multiple ways, and it can be widely applied to computer devices with different data processing and computing capabilities.

[0120] In one embodiment, acquiring behavior data for each behavior generated by the target object includes: acquiring a behavior data sequence of the target object; the behavior data in the behavior data sequence is used to characterize the behavior generated by the target object, and each behavior data is sorted according to the occurrence time of the behavior; the sorting position of the multidimensional encoded data in the encoded data sequence is the same as the sorting position of the behavior data matched by the multidimensional encoded data in the behavior data sequence.

[0121] The behavioral data sequence includes multiple behavioral data sets used to characterize the behaviors generated by the target object, such as... Figure 8 As shown, each behavioral data point is sorted according to the occurrence time of the behavior it represents. Specifically, by acquiring the behavioral data sequence, the computer device can directly obtain the behavioral data sorted in chronological order of occurrence. Furthermore, based directly on the order of the behavioral data in the behavioral data sequence, it can determine the order of the multidimensional encoded data obtained by encoding the behavioral data, thus obtaining the encoded data sequence.

[0122] In a specific application, a computer device can encode each behavior data in the behavior data sequence according to the sorting order of the behavior data to obtain multidimensional encoded data of each behavior data. Based on the generation order of the multidimensional encoded data, an encoded data sequence that can represent the order in which the behavior occurs can be obtained.

[0123] In another specific application, the computer device can also determine an empty sequence that can store the same amount of multidimensional encoded data based on the number of behavioral data in the behavioral data sequence. Then, based on the correspondence between the empty sequence and each sorting position in the behavioral data sequence, the multidimensional encoded data of the behavioral data is filled into the same sorting position in the empty sequence according to the sorting position of the behavioral data in the behavioral data sequence.

[0124] In this embodiment, the computer device can directly obtain the sequence of behavioral data representing the occurrence time of the behavior by directly acquiring the sequence of behavioral data that can characterize the behavior, and determine the order of the multidimensional encoded data based on the order of the behavioral data in the sequence. This can directly obtain the encoded data sequence representing the occurrence time of the behavior, thereby improving the data processing efficiency of obtaining the encoded data sequence.

[0125] Data mapping of coded data sequences can be performed from different dimensions, specifically time, object, cumulative behavior, business, and other dimensions, or a combination of these dimensions, to analyze the behavioral patterns of the target object. To facilitate a clear understanding of the data mapping process across different dimensions, the following will primarily focus on data mapping from four dimensions: time, object, cumulative behavior, and business, to illustrate the behavioral sequence characteristics obtained in each dimension.

[0126] In one embodiment, taking the time dimension as an example of data mapping processing, the data sequence is processed according to the order of the multidimensional encoded data in the encoded data sequence, performing data mapping processing on the encoded data sequence in at least one dimension to obtain behavioral sequence features in at least one dimension. This includes: obtaining a time dimension operator; the time dimension operator representing the data mapping relationship in the time dimension; determining the duration condition matched by the time dimension operator; identifying continuous multidimensional encoded data in the encoded data sequence whose time data meets the duration condition; and performing data mapping processing on the continuous multidimensional encoded data based on the time dimension operator to obtain the behavioral sequence features of the target object in the time dimension.

[0127] In this context, the time dimension refers specifically to the time when the target object's behavior occurs. The time dimension operator is a mapping relationship that uses time data as a link to transform data in one form into another. The data mapping relationship represented by the time dimension operator can be set based on actual feature extraction conditions. For example, the data mapping relationship represented by the time dimension operator could be that multidimensional encoded data within a set time range are treated as a group of data, and features are extracted from information in a specified dimension of this group of data. Here, the specified dimension can be the dimension involved in the behavioral pattern expected to be identified in the target scene. Continuous multidimensional encoded data refers to at least two multidimensional encoded data that are consecutively ordered in the encoded data sequence.

[0128] The duration condition is used to filter multidimensional encoded data sequences. For encoded data sequences arranged in chronological order of behavior generation, computer equipment can filter multidimensional encoded data based on the duration condition to obtain multiple sets of continuous multidimensional encoded data. In each set of continuous multidimensional encoded data, the time interval between the first and last multidimensional encoded data is no greater than the duration threshold in the duration condition, which can be 5 minutes, 10 minutes, etc.

[0129] Specifically, the computer device acquires time-dimensional operators representing data mapping relationships in the time dimension, determines the duration conditions matched by the time-dimensional operators, and obtains a duration threshold. For encoded data sequences ordered chronologically, based on the time data in the multidimensional encoded data, continuous multidimensional encoded data with a time span less than or equal to the duration threshold is obtained. Based on the data mapping relationships represented by the time-dimensional operators, the computer device performs data mapping processing on each set of continuous multidimensional encoded data to obtain behavioral sequence features in the time dimension. Each set of continuous multidimensional encoded data can correspond to one behavioral sequence feature.

[0130] In this embodiment, the computer device uses a time-dimensional operator to map continuous multidimensional encoded data that meets the duration condition to obtain behavioral sequence features in the time dimension. This can accurately identify the continuous behaviors of the target object within a time range, making it easier to accurately analyze the behavioral patterns of the target object based on the correlation between these behaviors, thereby improving the accuracy of the behavioral pattern analysis results.

[0131] In one embodiment, determining the duration condition matched by the time dimension operator includes: determining the time window matched by the time dimension operator. Determining continuous multidimensional encoded data that meets the duration condition from the encoded data sequence includes: based on a dynamically sliding time window in the encoded data sequence, identifying multidimensional encoded data within the same time window as continuous multidimensional encoded data.

[0132] The window width of the time window represents the duration condition matched by the time dimension operator, and it is used to characterize the time span. For example... Figure 9 As shown, the time window can dynamically slide in the encoded data sequence according to a set step size. In this embodiment, the sliding step size can be a preset number of multidimensional encoded data, specifically one multidimensional encoded data or multiple multidimensional encoded data.

[0133] In this embodiment, the computer device uses a dynamically sliding time window in the encoded data sequence to treat the multidimensional encoded data in the same time window as continuous multidimensional encoded data. This enables the traversal of the encoded data sequence and the efficient and accurate filtering of continuous multidimensional encoded data that meet the duration condition, thereby improving data processing efficiency and the accuracy of the filtering results.

[0134] In one embodiment, based on the time dimension operator, data mapping processing is performed on continuous multidimensional encoded data to obtain the behavioral sequence features of the target object in the time dimension, including: based on the time dimension operator, behavioral temporal feature extraction is performed on continuous multidimensional encoded data in the same time window to obtain the continuous behavioral sequence features of the target object.

[0135] Specifically, continuous behavior sequence features are feature data that characterize the continuous behaviors of a target object. Continuous behavior sequence features can be obtained by temporal feature extraction based on continuous multidimensional encoded data. Temporal feature extraction refers to the data processing procedure of extracting features from continuous multidimensional encoded data according to the temporal order and the mapping relationship represented by the time dimension operator.

[0136] In a specific application, the time dimension operator can be used to statistically analyze continuous user behavior patterns over a short period, such as rapid inflows and outflows of funds. Specifically, within a certain time window, statistical calculations of specific dimensional indicators are performed on the multidimensional encoded data matching sequential behaviors in the encoded data sequence, such as frequency statistics and extreme value statistics. The width of the time window is adjustable (e.g., adjustable to 1 minute or 5 minutes), and the time window can dynamically slide across the encoded data sequence to determine the continuous multidimensional encoded data corresponding to different start and end points.

[0137] In this embodiment, the computer device uses a time dimension operator to extract behavioral temporal features from continuous multidimensional encoded data within the same time window. This allows for accurate identification of the continuous behaviors of the target object within the time window, facilitating further analysis of the target object's behavioral patterns based on the relationships between these behaviors and improving the accuracy of the behavioral pattern analysis results.

[0138] In one embodiment, taking the object dimension as an example of data mapping processing, the data sequence is processed by at least one dimension of the encoded data sequence according to the order of the multidimensional encoded data in the encoded data sequence to obtain behavioral sequence features in at least one dimension. This includes: obtaining object dimension operators; object dimension operators representing the data mapping relationship of the object dimension; determining the object selection conditions matched by the object dimension operators; filtering out multidimensional encoded data containing object information and satisfying the object selection conditions from the encoded data sequence; and performing data mapping processing on the filtered multidimensional encoded data based on the object dimension operators to obtain the behavioral sequence features of the target object in the object dimension.

[0139] In this context, the object dimension refers to the objects contained in the multidimensional encoded data. Specifically, it can be the target object, the target object's interactive objects, such as interactive users or interactive accounts. The object dimension operator is a mapping relationship that uses objects as a link to transform data in one form into data in another. The data mapping relationship represented by the object dimension operator can be set based on actual feature extraction conditions. For example, the data mapping relationship represented by the object dimension operator could be based on the various behaviors generated by the interaction between the target object and the same interactive object, with the multidimensional encoded data matching each behavior treated as a group of data, and feature extraction performed on the information of a specified dimension in this group of data. Object selection conditions are used to filter the multidimensional encoded data sequence according to object information. For example, object selection conditions could be objects with interactive behaviors, meaning the multidimensional encoded data simultaneously contains the target object and the target object's interactive objects.

[0140] Specifically, the computer device acquires object dimension operators that represent data mapping relationships of object dimensions, obtains object selection conditions that match the object dimension operators, determines target multidimensional encoded data containing object information in the encoded data sequence, and then selects multidimensional encoded data whose object information satisfies the object selection conditions from the target multidimensional encoded data. Based on the data mapping relationships represented by the object dimension operators, the computer device performs data mapping processing on the selected multidimensional encoded data to obtain behavioral sequence features in the object dimension.

[0141] In this embodiment, the computer device uses an object dimension operator to perform data mapping on continuous multidimensional encoded data that meets the duration condition, thereby obtaining behavioral sequence features in the object dimension. This can accurately identify the continuous behaviors of the target object within a duration range, facilitating accurate analysis of the target object's behavioral patterns based on the correlation between these behaviors, and improving the accuracy of the behavioral pattern analysis results.

[0142] In one embodiment, the object selection condition is used as the object to which the interactive behavior occurs. The object behavior pattern determination method further includes: obtaining the target encoded characters representing the interactive object from the multidimensional encoded data, and dividing the multidimensional encoded data with the same target encoded characters into the same group to obtain a multidimensional encoded data group.

[0143] Furthermore, based on the object dimension operator, the selected multidimensional encoded data is processed by data mapping to obtain the behavioral sequence features of the target object in the object dimension, including: for each multidimensional encoded data group, the interactive behavior features of the multidimensional encoded data group are extracted to obtain the interactive behavior sequence features of the target object.

[0144] It's important to note that not all behaviors generated by the target object are interactions with other objects. For non-interactive behaviors, the matched multidimensional encoded data does not contain specific object data. The encoded characters in the object-related dimensions indicate the absence of an interactive object. Therefore, multidimensional encoded data lacking specific object data does not meet the object selection criteria. Multidimensional encoded data containing target encoded characters representing interactive objects indicates the existence of an interactive behavior with that object. The same target encoded characters indicate the same interactive object.

[0145] Interaction behavior sequence features are characteristic data representing the continuous behaviors that occur between a target object and the same interactive object. Interaction behavior sequence features can be obtained by extracting interaction behavior features from a multi-dimensional encoded data set containing coded characters corresponding to the same interactive object. Interaction behavior feature extraction refers to the data processing procedure of extracting features from a multi-dimensional encoded data set containing coded characters corresponding to the same interactive object according to the mapping relationship represented by the time sequence and object dimension operators.

[0146] Specifically, the computer device can filter multidimensional encoded data based on object selection criteria to obtain target multidimensional encoded data containing target encoded characters representing interactive objects within the encoded data sequence. Based on the specific interactive objects represented by the target encoded characters, multidimensional encoded data with the same target encoded characters are grouped together to obtain multidimensional encoded data groups. Then, for each multidimensional encoded data group, interactive behavior features are extracted to obtain the interactive behavior sequence features of the target object. Furthermore, to ensure the temporal order of each group of multidimensional encoded data, the computer device can also sort the multidimensional encoded data with the same target encoded characters according to the occurrence time of the behavior represented by the multidimensional encoded data to obtain multidimensional encoded data groups.

[0147] In a specific application, object-dimensional operators can be used to statistically analyze the characteristic data of continuous behaviors between target objects and interaction objects (such as users interacting with each other), such as the fund transfers between users A and B within a certain time period. Specifically, computer devices can use object-dimensional operators to perform statistical calculations on specific dimensional indicators, such as frequency statistics and extreme value statistics, on multidimensional encoded data representing the interaction behaviors of interaction objects in an encoded data sequence within a certain time period, thereby obtaining the characteristics of the interaction behavior sequence.

[0148] In this embodiment, the computer device extracts interactive behavior features from a multi-dimensional encoded data group containing coded characters corresponding to the same interactive object. This can accurately identify the continuous behaviors between the target object and the same interactive object, facilitating further analysis of the target object's behavior patterns based on the correlation between these behaviors, thereby improving the accuracy of the behavior pattern analysis results.

[0149] In one embodiment, taking the data mapping processing dimension as the behavior accumulation dimension as an example, the data mapping processing of the encoded data sequence is performed on the encoded data sequence according to the arrangement order of each multidimensional encoded data in the encoded data sequence to obtain the behavior sequence features in at least one dimension. This includes: determining the multidimensional encoded data representing the same type of behavior from the multidimensional encoded data in the encoded data sequence; and extracting repeated behavior features from the multidimensional encoded data representing the behavior for each type of behavior to obtain the repeated behavior sequence features of the target object.

[0150] Among them, the behavior cumulative dimension refers to the behavior represented by multidimensional encoded data, which uses behavior identifiers to represent the behaviors it represents. Repeated behavior sequence features are feature data representing the recurring behaviors of a target object. Interactive behavior sequence features can be obtained by extracting repeated behavior features from multidimensional encoded data sets containing the same or similar behavior identifiers. Repeated behavior feature extraction can be a data processing procedure that extracts features from multidimensional encoded data sets containing the same or similar behavior identifiers according to the time sequence and the mapping relationship represented by the behavior cumulative dimension operator.

[0151] Feature extraction of repetitive behaviors can be achieved through the behavior cumulative dimension operator. This operator is a mapping relationship that transforms data from one representation to another, using behavior identifiers as the link. The data mapping relationship represented by the behavior cumulative dimension operator can be set based on the actual feature extraction conditions. For example, the data mapping relationship represented by the behavior cumulative dimension operator could be that multi-dimensional encoded data with the same or similar behavior identifiers are treated as a group of data, and features are extracted from information in a specified dimension of that group of data.

[0152] Specifically, the computer device obtains the behavior cumulative dimension operator, which represents the data mapping relationship of the behavior cumulative dimension, to determine the target multidimensional encoded data of the same behavior identifier or the same type of behavior identifier in the encoded data sequence. Then, based on the data mapping relationship represented by the behavior cumulative dimension operator, the target multidimensional encoded data of the same behavior identifier or the same type of behavior identifier is processed by data mapping to obtain the behavior sequence features under the behavior cumulative dimension.

[0153] In a specific application, the behavior cumulative dimension operator can be used to statistically analyze the characteristic data of frequently occurring behaviors of a target object, such as user A repeatedly making deposits or withdrawals within a certain time period. Specifically, computer equipment can use the behavior cumulative dimension operator to perform statistical calculations on specific dimension indicators, such as frequency statistics and extreme value statistics, on the multidimensional encoded data representing the repetitive behaviors of the target object in the encoded data sequence within a certain time period, thereby obtaining the repetitive behavior sequence characteristics of the target object.

[0154] In this embodiment, the computer device uses a behavior accumulation dimension operator to perform data mapping on multidimensional encoded data with the same behavior identifier or similar behavior identifiers to obtain behavior sequence features on the behavior accumulation dimension. This can accurately identify the same behaviors repeatedly generated by the target object, making it easier to accurately analyze the behavior patterns of the target object based on the correlation between these behaviors, thereby improving the accuracy of the behavior pattern analysis results.

[0155] In one embodiment, taking the data mapping processing dimension including the business dimension as an example; according to the arrangement order of each multidimensional encoded data in the encoded data sequence, the encoded data sequence is subjected to data mapping processing of at least one dimension to obtain behavioral sequence features under at least one dimension, including: filtering out multidimensional encoded data representing the same business from the multidimensional encoded data in the encoded data sequence; performing business behavior feature extraction processing on the multidimensional encoded data representing the same business to obtain the business behavior sequence features of the target object.

[0156] In this context, "business dimension" refers to the business represented by multidimensional encoded data, which uses business-related identifiers to represent its respective business. Business behavior sequence features are characteristic data representing various behaviors of a target object related to the same business. Interaction behavior sequence features can be obtained by extracting business behavior features from multidimensional encoded data containing the same or related business identifiers. Business behavior feature extraction can be a data processing procedure that extracts features from multidimensional encoded data containing the same or related business identifiers according to the time sequence and the mapping relationship represented by the business dimension operator.

[0157] Business behavior feature extraction can be achieved through business dimension operators. These operators are mapping relationships that transform data from one representation to another, using business identifiers as the link. The data mapping relationship represented by the business dimension operator can be set based on the actual feature extraction conditions. For example, the data mapping relationship represented by the business dimension operator could be that multi-dimensional encoded data with the same or related business identifiers are treated as a group of data, and features are extracted from information in a specified dimension of that group of data.

[0158] Specifically, the computer device obtains the business dimension operator that represents the data mapping relationship of the business dimension, determines the target multidimensional encoded data with the same business identifier or related business identifier in the encoded data sequence, and then performs data mapping processing on the target multidimensional encoded data with the same business identifier or related business identifier based on the data mapping relationship represented by the business dimension operator to obtain the behavioral sequence features under the business dimension.

[0159] In a specific application, business-dimensional operators can be used to statistically analyze the characteristic data of a target object's behavior within the same business scenario, such as a user making multiple payment and refund operations within a certain time period. Specifically, computer devices can use business-dimensional operators to perform statistical calculations on specific dimensional indicators, such as frequency statistics and extreme value statistics, on multi-dimensional encoded data representing the target object's behavior in the same business scenario within a certain time period, thereby obtaining the business behavior sequence characteristics of the target object.

[0160] In this embodiment, the computer device uses business dimension operators to map multidimensional encoded data with the same or similar behavioral identifiers to obtain behavioral sequence features in the business dimension. This can accurately identify the same behaviors repeatedly generated by the target object, making it easier to accurately analyze the behavioral patterns of the target object based on the correlation between these behaviors, thereby improving the accuracy of the behavioral pattern analysis results.

[0161] It is understandable that in some application scenarios, such as Figure 10 As shown, the computer device can simultaneously perform the four-dimensional data mapping processing listed in the above embodiments on the encoded data sequence of the target object, or it can perform data mapping processing on at least one of the four dimensions listed in the above embodiments. In other application scenarios, more dimensions of behavioral sequence features can be extracted based on different dimensional operators according to actual needs.

[0162] In one embodiment, the object behavior pattern determination method further includes: acquiring an object classification model trained based on behavior sequence feature sample data, wherein the behavior sequence feature sample data carries object category labels; and classifying the target object according to the behavior sequence features of the target object based on the object classification model to obtain the classification result of the target object.

[0163] Specifically, the object classification model can be obtained by training an initial classification model based on behavioral sequence feature sample data. The initial classification model can be any of the following: logistic regression, decision tree, support vector machine, Naive Bayes, etc. Logistic regression uses a combination of the sigmold function (the activation function of a neural network) and a linear regression equation to achieve classification. The classification principle of a classification tree is that, given an observation, the predicted value of the dependent variable is the class that most frequently occurs in the training set within its terminal node. Support vector machines are a type of generalized linear classifier that performs binary classification of data using supervised learning. Naive Bayes, based on Bayes' theorem, uses probability and statistics to classify sample datasets.

[0164] Specifically, the computer device uses the object classification model trained with behavioral sequence feature sample data as input data for the object classification model, and classifies the behavioral sequence features of the target object by the object classification model to obtain the classification result of the target object.

[0165] In a specific application, taking a fraud risk control system as an example, the core of the system is the fraud risk identification subsystem. Its main function is to detect and identify the fraud and being defrauded risks of both the payer and payee in a payment transaction. First, the behavioral data of the target object is preprocessed by sorting, encoding, and normalizing the data according to chronological order, resulting in an encoded data sequence containing multiple multi-dimensional encoded data. Then, based on this encoded data sequence, corresponding behavioral sequence features are automatically extracted using time-related, object-related, cumulative-related, business-related, and custom behavioral sequence feature extraction operators. Based on these extracted features, a supervised model is designed by constructing positive and negative samples to identify the fraud and being defrauded risks of the target object. Extracting behavioral sequence features using behavioral sequence feature extraction operators effectively avoids the problem of poor generality in manual feature extraction. Furthermore, each behavioral sequence feature has a corresponding relevance meaning, making it more interpretable for the risk control model. It also improves the lag in risk object identification by object classification models, enabling more real-time identification of fraud risks.

[0166] Furthermore, the target object includes the target account; the object classification model is a binary classification model that classifies accounts into risky accounts or non-risky accounts. The object behavior pattern determination method also includes: when the target account is classified as a risky account, obtaining related accounts that belong to the same entity as the target account; and marking the behavioral sequence characteristics of the related accounts as risky.

[0167] The binary classification model can output only two possible results, such as 0 and 1, where 0 represents a non-risk account and 1 represents a risky account. The target account owner refers to the user to whom the target account belongs. When a user has multiple registered accounts, and the target account is only one of them, the associated account can be any of the user's registered accounts other than the target account.

[0168] In practical applications, when one of a user's accounts is a risky account, it indicates that the user is a risky user. Therefore, other accounts under that user's name may also be operated by the risky user, which also poses a risk. By marking the behavioral sequence characteristics of related accounts for risk, we can expand the labeling of classification results based on data association, thereby increasing the attention paid to related accounts and reducing security risks.

[0169] In one embodiment, acquiring behavioral data for each behavior generated by the target object includes: acquiring behavioral data for each behavior generated by the target object within each time period. Further, performing at least one dimension of data mapping processing on the encoded data sequence composed of the multidimensional encoded data to obtain behavioral sequence features in at least one dimension includes: determining the encoded data sequence composed of multidimensional encoded data corresponding to behaviors within the same time period; performing data mapping processing on the encoded data sequence of each time period according to at least one dimension to obtain behavioral sequence features for each time period; and determining the behavioral pattern of the target object based on the behavioral sequence features of each time period.

[0170] The time period can be a day, a week, a month, etc. Different time periods can be set for different application scenarios. The time period can also be based on pre-configured multi-dimensional data mapping operations, using the maximum time range within each operator's time range as the time period to ensure the accuracy of the behavioral sequence features obtained in each dimension during multi-dimensional data mapping.

[0171] When it is necessary to perform behavioral pattern analysis on a target object within a target time period, and the target time period contains multiple time cycles, the computer device performs data mapping processing on the encoded data sequence of each time cycle to obtain the behavioral sequence features of each time cycle. By fusing the behavioral sequence features of each time cycle according to their order in the target time period, the behavioral pattern of the target object in the target time period can be obtained. This effectively realizes behavioral pattern analysis of the target object for different durations and improves the applicability in different scenarios.

[0172] In one embodiment, the target object includes a target account; the object behavior pattern determination method further includes: based on the behavior pattern of the target account, profiling the user of the target account to obtain user profile data; when the user profile data fails to match the authentication object of the target account, marking the user based on the behavior pattern of the target account based on the user profile data.

[0173] In this context, profiling is the process of characterizing and labeling the features of a specific object. For accounts logged in on computer devices, the actual user and the registrant may not be the same person. For example, user A registers an account on an application using their mobile phone number, which identifies user A. However, user A may not use the application frequently, while their child often uses the same computer device to log in and operate the account. By profiling the behavioral patterns of this account, it can be discovered that the user is a child, while the registration information shows that the registrant is an adult. Therefore, the object profiling data can be used to label the account's behavioral patterns, indicating a difference between the actual user and the registrant. This increases the level of attention paid to the account during analysis and reduces security risks.

[0174] In a specific embodiment, such as Figure 11 As shown, the method for determining object behavior patterns includes the following steps:

[0175] Step 1102: Obtain an initial behavior data sequence consisting of the original behavior data of each behavior generated by the target object. Each original behavior data is sorted according to the occurrence time of the behavior it represents. The original behavior data contains behavior description information of at least one dimension.

[0176] Step 1104: Normalize the format of the behavior description information of the same dimension in each original behavior data to obtain a behavior data sequence composed of updated behavior data.

[0177] Step 1106: Based on the behavioral description information contained in each behavioral data, determine the dimension union of each behavioral description data dimension, arrange the dimensions in the dimension union, and determine the arrangement information of each dimension.

[0178] Step 1108: For each behavior data, the behavior description information contained in the behavior data is encoded to obtain the encoded characters of the behavior description information.

[0179] Step 1110: Arrange the coded characters based on the arrangement information of the dimension to which the behavior description information belongs, and obtain the initial multidimensional coded data containing partial codes.

[0180] Step 1112: Fill the missing encoded characters in the dimensions of the initial multidimensional encoded data with target data to obtain updated multidimensional encoded data.

[0181] Step 1114: Obtain the time dimension operator representing the data mapping relationship of the time dimension, and determine the time window matched by the time dimension operator.

[0182] Step 1116: Based on the dynamically sliding time window in the encoded data sequence, each multidimensional encoded data in the same time window is regarded as continuous multidimensional encoded data.

[0183] Step 1118: Based on the time dimension operator, extract behavioral temporal features from continuous multidimensional encoded data within the same time window to obtain continuous behavioral sequence features of the target object.

[0184] Step 1120: Based on the target encoded characters representing interactive objects in the multidimensional encoded data, sort the multidimensional encoded data with the same target encoded characters according to the occurrence time of the behavior represented by the multidimensional encoded data to obtain a multidimensional encoded data group.

[0185] Step 1122: Based on the object dimension operator representing the data mapping relationship of the object dimension, perform interactive behavior feature extraction on each multidimensional encoded data group to obtain the interactive behavior sequence features of the target object.

[0186] Step 1124: Determine the multidimensional encoded data representing similar behaviors from the multidimensional encoded data in the encoded data sequence.

[0187] Step 1126: For each type of behavior, extract repetitive behavior features from the multidimensional encoded data representing the behavior to obtain the repetitive behavior sequence features of the target object.

[0188] Step 1128: Select multidimensional encoded data representing the same business from the multidimensional encoded data in the encoded data sequence.

[0189] Step 1130: Perform business behavior feature extraction processing on the multidimensional encoded data representing the same business to obtain the business behavior sequence features of the target object.

[0190] Step 1132: Based on the object classification model, classify the target object according to the behavioral sequence characteristics of the target object to obtain the classification result of the target object.

[0191] This application also provides an application scenario in which the above-mentioned object behavior pattern determination method is applied. Specifically, taking the application of the object behavior pattern determination method in a fraud risk control system as an example, the details are as follows: With the rapid innovation and development of mobile payment, payment efficiency has been greatly improved and transaction costs have been reduced. While billions of users use functions such as red envelopes, transfers, and QR code payments for convenient payments every day, they also face increasingly severe risks of telecommunications fraud. Telecommunications fraud, also known as online fraud, refers to fraudulent activities in which fraudsters use social networks and information exchange platforms, employing certain rhetoric and tactics to make victims believe the fraud is genuine and voluntarily transfer funds to the fraudsters, thereby suffering financial losses.

[0192] To combat fraud and purify the payment environment, payment service platforms establish fraud risk control systems to identify and intervene in suspicious fraudulent users and transactions. These systems, designed to protect payment security, utilize big data analytics and modeling to identify, make decisions about, and manage high-risk fraud users and transactions. They primarily include subsystems such as fraud risk identification, strategy configuration, and account management.

[0193] The core of a fraud risk control system is the fraud risk identification subsystem, whose main function is to detect and identify telecommunications fraud and fraud risks between the payer and payee involved in the payment process. The construction of the fraud risk identification subsystem generally involves two steps. The first step is to build an abnormal feature system to identify unusual account behavior. The second step, based on the abnormal feature system, is to train a classification model by constructing positive and negative samples to identify user fraud and fraud risks. The object behavior pattern determination method in this application can be used to identify the abnormal behavior characteristics of each account.

[0194] Specifically, the process by which computer devices identify abnormal behavioral characteristics of each account is as follows: Figure 12 As shown, it mainly includes the following three parts:

[0195] The first part involves the unified encoding of behavioral data sequences. First, computer equipment arranges user payment and operational behaviors in chronological order, resulting in a behavioral data sequence. Each behavioral data point in the sequence is then encoded, yielding multi-dimensional encoded data that matches each behavior. A coded data sequence containing each multi-dimensional encoded data point is then constructed chronologically. The dimensions of the multi-dimensional encoded data include the union of the dimensions of the behavioral description information contained in each behavioral data point. Data preprocessing is then performed to normalize the format of behavioral description information such as time, scenario, channel, and amount for different behaviors, achieving format uniformity for data within the same dimension.

[0196] The second part involves acquiring behavioral sequence feature extraction operators: Based on the uniformly encoded data sequence, time-related, object-related, cumulative-related, and business-related behavioral sequence feature extraction operators are acquired to automatically extract the corresponding behavioral sequence features. In addition to the aforementioned behavioral sequence feature extraction operators, users can also define behavioral sequence feature extraction operators for other dimensions. Based on user behavior analysis, associated account information and profile information can be obtained. Associated account information can be used to implement behavioral sequence feature blackening (i.e., feature transfer, such as marking other accounts under the username of a user as risky when an account is identified as risky), and profile information can be used to establish associations between users and accounts.

[0197] Specifically, the time intervals between the behaviors represented by the various behavioral data in the behavioral data sequence are non-uniform and equidistant. Different behaviors involve different interaction objects, and the information contained in a single behavior is extremely limited. Therefore, defining the relationships between user behaviors is crucial. This application categorizes user behavior patterns into four types: time-related behavior patterns (e.g., rapid inflow and outflow of funds within a short period); object-related behavior patterns (e.g., fund transfers between user A and user B within a certain time period); cumulative behavior patterns (e.g., multiple deposits or withdrawals by user A within a certain time period); and business-related behavior patterns (e.g., multiple payment and refund operations by user A within a certain time period). User behavior patterns are complex and varied. Other more complex behavior patterns can be extracted and analyzed by designing custom behavior sequence feature extraction operators.

[0198] The third part involves user behavior pattern extraction and training. Computer devices extract time-related, object-related, cumulative, and business-related behavioral sequence features using behavioral sequence feature extraction operators, and use these features as input to a classifier to enhance the ability to identify user risks.

[0199] In the above embodiments, behavioral data is uniformly encoded, and behavioral sequence features of target users in different dimensions are extracted based on behavioral sequence feature extraction operators. The behavioral patterns of target objects are analyzed, and positive and negative samples are constructed by further judging whether the behavioral sequence features are abnormal behavioral features. A classification model is then trained to identify user fraud and fraud risks.

[0200] To examine the advantages of this application's solution compared to the traditional method of manually mining user risk tags, a comparison was made between the efficiency of behavioral sequence feature extraction and the modeling effect during the classification model's modeling process. The comparison results are shown in Table 1 below.

[0201] Table 1: Comparison of Efficiency and Modeling Results of Behavioral Sequence Feature Extraction during Modeling

[0202]

[0203] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0204] Based on the same inventive concept, this application also provides an object behavior pattern determination apparatus for implementing the object behavior pattern determination method described above. The solution provided by this apparatus is similar to the implementation scheme described in the above method; therefore, the specific limitations in one or more object behavior pattern determination apparatus embodiments provided below can be found in the limitations of the object behavior pattern determination method described above, and will not be repeated here.

[0205] In one embodiment, such as Figure 13 As shown, an object behavior pattern determination device 1300 is provided, including: a data acquisition module 1302, a data encoding module 1304, a data mapping module 1306, and a behavior pattern determination module 1308, wherein:

[0206] The data acquisition module 1302 is used to acquire behavioral data of each behavior generated by the target object, wherein the behavioral data contains behavioral description information of at least one dimension.

[0207] The data encoding module 1304 is used to encode each of the behavioral data separately to obtain multidimensional encoded data that matches each of the behaviors; the dimensions of the multidimensional encoded data include the union of the dimensions of the behavioral description information contained in each of the behavioral data.

[0208] The data mapping module 1306 is used to perform data mapping processing on the encoded data sequence composed of the multidimensional encoded data in at least one dimension to obtain behavioral sequence features in the at least one dimension.

[0209] The behavior pattern determination module 1308 is used to determine the behavior pattern of the target object based on the behavior sequence features under the at least one dimension.

[0210] In one embodiment, the object behavior pattern determination device further includes an arrangement information determination module, which is used to determine the dimensional union of the dimensions to which each of the behavior description data belongs based on the behavior description information contained in each of the behavior data; and to arrange the dimensions in the dimensional union to determine the arrangement information of each dimension.

[0211] The data encoding module is further configured to encode the behavior description information contained in each behavior data to obtain the encoded characters of the behavior description information; and to arrange the encoded characters based on the arrangement information of the dimension to which the behavior description information belongs to obtain the multidimensional encoded data of the behavior data.

[0212] In one embodiment, the data encoding module is further configured to arrange the encoded characters based on the arrangement information of the dimension to which the behavior description information belongs, to obtain initial multidimensional encoded data containing partial encoding; and to fill the dimensions of the initial multidimensional encoded data with target data to obtain updated multidimensional encoded data.

[0213] In one embodiment, the data acquisition module can also be used to acquire the original behavior data of each behavior generated by the target object; and to perform format normalization processing on the behavior description information of the same dimension in each of the original behavior data to obtain the behavior data of each behavior.

[0214] In one embodiment, the multidimensional encoded data in the encoded data sequence is arranged in chronological order according to the occurrence of the behavior represented by the multidimensional encoded data; the data mapping module is further configured to perform data mapping processing on the encoded data sequence in at least one dimension according to the arrangement order of each multidimensional encoded data in the encoded data sequence, to obtain behavioral sequence features in at least one dimension.

[0215] In one embodiment, the object behavior pattern determination device further includes a sorting module, used to sort the multidimensional encoded data according to the chronological order of the occurrence of the behaviors represented by the multidimensional encoded data, to obtain an encoded data sequence.

[0216] In one embodiment, the data acquisition module is further configured to acquire a sequence of behavioral data of the target object; the behavioral data in the sequence of behavioral data is used to characterize the behavior generated by the target object, and each behavioral data is sorted according to the occurrence time of the behavior; the sorting position of the multidimensional encoded data in the sequence of encoded data is the same as the sorting position of the behavioral data matched by the multidimensional encoded data in the sequence of behavioral data.

[0217] In one embodiment, the data mapping processing dimension includes a time dimension; the data mapping module further includes a time dimension data mapping unit for acquiring time dimension operators; the time dimension operators characterize the data mapping relationship in the time dimension; the duration condition matched by the time dimension operators is determined; continuous multidimensional encoded data whose time data meets the duration condition is determined from the encoded data sequence; based on the time dimension operators, data mapping processing is performed on the continuous multidimensional encoded data to obtain the behavioral sequence features of the target object in the time dimension.

[0218] In one embodiment, the time dimension data mapping unit is further configured to determine the time window matched by the time dimension operator; based on the dynamically sliding time window in the encoded data sequence, each multidimensional encoded data in the same time window is regarded as continuous multidimensional encoded data.

[0219] In one embodiment, the time-dimensional data mapping unit is further configured to extract behavioral temporal features from continuous multidimensional encoded data within the same time window based on the time-dimensional operator, thereby obtaining continuous behavioral sequence features of the target object.

[0220] In one embodiment, the data mapping processing dimension includes an object dimension; the data mapping module further includes an object dimension data mapping unit for obtaining object dimension operators; the object dimension operators characterize the data mapping relationship of the object dimension; the object selection conditions matched by the object dimension operators are determined; multidimensional encoded data containing object information and satisfying the object selection conditions are filtered from the encoded data sequence; based on the object dimension operators, data mapping processing is performed on the filtered multidimensional encoded data to obtain the behavioral sequence features of the target object in the object dimension.

[0221] In one embodiment, the object selection condition is the object that performs the interactive behavior; the object dimension data mapping unit is further configured to obtain the target encoded characters representing the interactive object in the multidimensional encoded data, divide the multidimensional encoded data with the same target encoded characters into the same group to obtain a multidimensional encoded data group; for each multidimensional encoded data group, the interactive behavior feature is extracted to obtain the interactive behavior sequence feature of the target object.

[0222] In one embodiment, the object dimension data mapping unit is further configured to sort the multidimensional encoded data with the same target encoded character according to the occurrence time of the behavior represented by the multidimensional encoded data to obtain a multidimensional encoded data group.

[0223] In one embodiment, the data mapping processing dimension includes a cumulative behavior dimension; the data mapping module further includes a cumulative behavior dimension data mapping unit, used to determine multidimensional encoded data representing the same type of behavior from the multidimensional encoded data in the encoded data sequence; for each type of behavior, the multidimensional encoded data representing the behavior is used to extract repeated behavior features to obtain the repeated behavior sequence features of the target object.

[0224] In one embodiment, the data mapping processing dimension includes a business dimension; the data mapping module further includes a business dimension data mapping unit, used to filter out multi-dimensional coded data representing the same business from the multi-dimensional coded data in the coded data sequence; and to perform business behavior feature extraction processing on the multi-dimensional coded data representing the same business to obtain the business behavior sequence features of the target object.

[0225] In one embodiment, the object behavior pattern determination device further includes a classification module, used to acquire an object classification model trained based on behavior sequence feature sample data, wherein the behavior sequence feature sample data carries object category labels; and based on the object classification model, classify the target object according to the behavior sequence features of the target object to obtain the classification result of the target object.

[0226] In one embodiment, the target object includes a target account; the object classification model is a binary classification model that classifies accounts into risky accounts or non-risky accounts; the object behavior pattern determination device further includes an associated account marking module, used to obtain associated accounts that are the same as the subject to which the target account belongs when the classification result of the target account is a risky account; and to mark the behavioral sequence features of the associated accounts as risky.

[0227] In one embodiment, the data acquisition module is further configured to acquire behavioral data of each behavior generated by the target object in each time period;

[0228] The data mapping module is further configured to determine the encoded data sequence consisting of multi-dimensional encoded data corresponding to behaviors within the same time period; perform data mapping processing on the encoded data sequence of each time period according to at least one dimension to obtain the behavior sequence features of each time period; and determine the behavior pattern of the target object based on the behavior sequence features of each time period.

[0229] In one embodiment, the target object includes a target account; the object behavior pattern determination device further includes a matching module, used to profile the user of the target account based on the behavior pattern of the target account, and obtain user profile data; when the user profile data fails to match the authentication object of the target account, the user is marked based on the user profile data to mark the behavior pattern of the target account as a user.

[0230] The modules in the aforementioned object behavior pattern determination device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can invoke and execute the operations corresponding to each module.

[0231] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 14 As shown, this computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores behavioral data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When executed by the processor, the computer program implements a method for determining object behavior patterns.

[0232] Those skilled in the art will understand that Figure 14 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0233] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0234] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0235] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0236] 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, data stored, data displayed, 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 the relevant data shall comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0237] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0238] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0239] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for determining object behavior patterns, characterized in that, The method includes: Acquire behavioral data for each action generated by the target object, wherein the behavioral data contains behavioral description information in at least one dimension; the target object is a user or an account used by a user. Based on the behavioral description information contained in each behavioral data, determine the dimensional union of the dimensions to which each behavioral description data belongs; Each of the aforementioned behavioral data is encoded to obtain multidimensional encoded data matching each of the aforementioned behaviors; the dimension of the multidimensional encoded data is the dimension contained in the union of the dimensions. The encoded data sequence composed of the multidimensional encoded data is subjected to data mapping processing in at least one dimension to obtain behavioral sequence features in the at least one dimension; the dimension of the data mapping processing includes at least one of time dimension, object dimension, behavior accumulation dimension and business dimension; Based on the behavioral sequence features under the at least one dimension, the behavioral pattern of the target object is determined; When the target object is a target account, a user profile is created based on the behavior pattern of the target account to obtain user profile data; the user profile data is used to determine the association between the user of the target account and the registrant of the target account. When the user profile data fails to match the authentication object of the target account, the user tag is applied to the behavior pattern of the target account based on the user profile data.

2. The method according to claim 1, characterized in that, The method further includes: Arrange the dimensions in the union of the given dimensions to determine the arrangement information for each dimension; The step of encoding each of the aforementioned behavioral data to obtain multidimensional encoded data matching each of the aforementioned behaviors includes: For each of the aforementioned behavioral data, the behavioral description information contained in the behavioral data is encoded to obtain the encoded characters of the behavioral description information; Based on the arrangement information of the dimension to which the behavior description information belongs, the encoded characters are arranged to obtain the multidimensional encoded data of the behavior data.

3. The method according to claim 2, characterized in that, The arrangement of the coded characters based on the arrangement information of the dimension to which the behavior description information belongs, to obtain the multidimensional coded data of the behavior data, includes: Based on the arrangement information of the dimension to which the behavior description information belongs, the encoded characters are arranged to obtain initial multidimensional encoded data containing partial encoding; The dimensions of the initial multidimensional encoded data that are missing encoded characters are filled with target data to obtain updated multidimensional encoded data.

4. The method according to any one of claims 1 to 3, characterized in that, The multidimensional encoded data in the encoded data sequence are arranged in chronological order according to the occurrence of the behavior represented by the multidimensional encoded data; The step of performing at least one dimension data mapping processing on the encoded data sequence composed of each of the multidimensional encoded data to obtain behavioral sequence features under the at least one dimension includes: According to the arrangement order of the multidimensional encoded data in the encoded data sequence, the encoded data sequence is subjected to data mapping processing in at least one dimension to obtain behavioral sequence features in at least one dimension.

5. The method according to claim 4, characterized in that, The acquisition of behavioral data for each behavior generated by the target object includes: Obtain a sequence of behavioral data of a target object; the behavioral data in the sequence is used to characterize the behavior generated by the target object, and each behavioral data is sorted according to the occurrence time of the behavior; The sorting position of the multidimensional encoded data in the encoded data sequence is the same as the sorting position of the behavioral data matched by the multidimensional encoded data in the behavioral data sequence.

6. The method according to claim 4, characterized in that, The dimensions of the data mapping process include the time dimension; The step of performing data mapping processing on the encoded data sequence according to the arrangement order of the multidimensional encoded data in the encoded data sequence to obtain behavioral sequence features in at least one dimension includes: Obtain the time dimension operator; the time dimension operator represents the data mapping relationship in the time dimension; Determine the duration condition matched by the time dimension operator; From the encoded data sequence, determine continuous multidimensional encoded data whose time data meets the duration condition; Based on the time dimension operator, the continuous multidimensional encoded data is processed by data mapping to obtain the behavioral sequence features of the target object in the time dimension.

7. The method according to claim 6, characterized in that, The determination of the duration condition matched by the time dimension operator includes: Determine the time window matched by the time dimension operator; The step of determining continuous multidimensional encoded data from the encoded data sequence that meets the duration condition includes: Based on the dynamically sliding time window in the encoded data sequence, each multidimensional encoded data in the same time window is regarded as continuous multidimensional encoded data.

8. The method according to claim 7, characterized in that, The step of performing data mapping processing on the continuous multidimensional encoded data based on the time dimension operator to obtain the behavioral sequence features of the target object in the time dimension includes: Based on the time dimension operator, behavioral temporal features are extracted from continuous multidimensional encoded data within the same time window to obtain continuous behavioral sequence features of the target object.

9. The method according to claim 4, characterized in that, The dimensions of the data mapping process include the object dimension; The step of performing data mapping processing on the encoded data sequence according to the arrangement order of the multidimensional encoded data in the encoded data sequence to obtain behavioral sequence features in at least one dimension includes: Obtain the object dimension operator; the object dimension operator represents the data mapping relationship of the object dimension; Determine the object selection criteria matched by the object dimension operator; From the encoded data sequence, select multidimensional encoded data that contains object information and whose object information satisfies the object selection criteria; Based on the object dimension operator, the selected multidimensional encoded data is processed by data mapping to obtain the behavioral sequence features of the target object in the object dimension.

10. The method according to claim 9, characterized in that, The object selection criteria are objects that have engaged in interactive behavior; the method further includes: Obtain the target encoded character representing the interactive object from the multidimensional encoded data, and divide the multidimensional encoded data with the same target encoded character into the same group to obtain the multidimensional encoded data group; The step of performing data mapping processing on the selected multidimensional encoded data based on the object dimension operator to obtain the behavioral sequence features of the target object in the object dimension includes: For each multidimensional encoded data group, interactive behavior features are extracted from the multidimensional encoded data group to obtain the interactive behavior sequence features of the target object.

11. The method according to claim 10, characterized in that, The step of dividing multidimensional encoded data with the same target encoded character into the same group to obtain a multidimensional encoded data group includes: The multidimensional encoded data with the same target encoded character are sorted according to the occurrence time of the behavior represented by the multidimensional encoded data to obtain a multidimensional encoded data group.

12. The method according to claim 4, characterized in that, The dimensions of the data mapping process include the cumulative behavior dimension; The step of performing data mapping processing on the encoded data sequence according to the arrangement order of the multidimensional encoded data in the encoded data sequence to obtain behavioral sequence features in at least one dimension includes: From the multidimensional encoded data in the encoded data sequence, determine the multidimensional encoded data that represents the same type of behavior; For each type of behavior, repeating behavior features are extracted from the multidimensional encoded data representing the behavior to obtain the repeating behavior sequence features of the target object.

13. The method according to claim 4, characterized in that, The dimensions of the data mapping process include business dimensions; The step of performing data mapping processing on the encoded data sequence according to the arrangement order of the multidimensional encoded data in the encoded data sequence to obtain behavioral sequence features in at least one dimension includes: From the multidimensional encoded data in the encoded data sequence, multidimensional encoded data representing the same service are selected; The business behavior feature extraction process is performed on the multidimensional encoded data representing the same business to obtain the business behavior sequence features of the target object.

14. The method according to claim 4, characterized in that, The method further includes: Obtain an object classification model trained based on behavioral sequence feature sample data, wherein the behavioral sequence feature sample data carries object category labels; Based on the object classification model, the target object is classified according to its behavioral sequence characteristics to obtain the classification result of the target object.

15. The method according to claim 14, characterized in that, The target object includes the target account; the object classification model is a binary classification model that classifies accounts into risky accounts or non-risky accounts; The method further includes: When the target account is classified as a risky account, obtain the associated account that is the same as the entity to which the target account belongs; Risk labeling is performed on the behavioral sequence characteristics of the associated accounts.

16. The method according to claim 1, characterized in that, The acquisition of behavioral data for each behavior generated by the target object includes: Obtain the raw behavioral data of each action produced by the target object; The behavior description information of the same dimension in each of the original behavior data is normalized to obtain the behavior data of each behavior.

17. The method according to claim 1, characterized in that, The method further includes: The multidimensional encoded data is sorted according to the chronological order of the behaviors represented by the multidimensional encoded data to obtain an encoded data sequence.

18. The method according to any one of claims 1 to 3, 5 to 17, characterized in that, The acquisition of behavioral data for each behavior generated by the target object includes: Obtain behavioral data for each action of the target object within each time period; The step of performing at least one dimension data mapping processing on the encoded data sequence composed of each of the multidimensional encoded data to obtain behavioral sequence features under the at least one dimension includes: Determine the sequence of coded data consisting of multidimensional coded data corresponding to behaviors within the same time period; Data mapping processing is performed on the encoded data sequence of each time period according to at least one dimension to obtain the behavioral sequence features of each time period. The behavioral pattern of the target object is determined based on the behavioral sequence characteristics of each time period.

19. An object behavior pattern determination device, characterized in that, The device includes: The data acquisition module is used to acquire behavioral data of each behavior generated by the target object, wherein the behavioral data includes behavioral description information of at least one dimension; the target object is a user or an account used by a user. The permutation information determination module is used to determine the dimensional union of the dimensions to which each behavioral description data belongs, based on the behavioral description information contained in each behavioral data. The data encoding module is used to encode each of the behavioral data separately to obtain multidimensional encoded data that matches each of the behaviors; the dimension of the multidimensional encoded data is the dimension contained in the union of the dimensions. The data mapping module is used to perform data mapping processing on the encoded data sequence composed of the multidimensional encoded data in at least one dimension to obtain behavioral sequence features under the at least one dimension; the dimension of the data mapping processing includes at least one of time dimension, object dimension, behavior accumulation dimension and business dimension; A behavior pattern determination module is used to determine the behavior pattern of the target object based on the behavior sequence features under the at least one dimension. When the target object is a target account, a user profile is created based on the behavior pattern of the target account to obtain user profile data; the user profile data is used to determine the association between the user of the target account and the registrant of the target account. When the user profile data fails to match the authentication object of the target account, the user tag is applied to the behavior pattern of the target account based on the user profile data.

20. The apparatus according to claim 19, characterized in that, The arrangement information determination module is used to arrange the dimensions in the union of the dimensions and determine the arrangement information of each dimension. The data encoding module is further configured to encode the behavior description information contained in each behavior data to obtain the encoded characters of the behavior description information; and to arrange the encoded characters based on the arrangement information of the dimension to which the behavior description information belongs to obtain the multidimensional encoded data of the behavior data.

21. The apparatus according to claim 20, characterized in that, The data encoding module is further configured to arrange the encoded characters based on the arrangement information of the dimension to which the behavior description information belongs, to obtain initial multidimensional encoded data containing partial encoding; and to fill the dimensions with missing encoded characters in the initial multidimensional encoded data with target data to obtain updated multidimensional encoded data.

22. The apparatus according to claim 19, characterized in that, The data acquisition module can also be used to acquire the original behavior data of each behavior generated by the target object; and to perform format normalization processing on the behavior description information of the same dimension in each of the original behavior data to obtain the behavior data of each behavior.

23. The apparatus according to any one of claims 19 to 22, characterized in that, The multidimensional encoded data in the encoded data sequence are arranged in chronological order according to the occurrence of the behaviors represented by the multidimensional encoded data; the data mapping module is further configured to perform data mapping processing on the encoded data sequence in at least one dimension according to the arrangement order of the multidimensional encoded data in the encoded data sequence, so as to obtain behavioral sequence features in at least one dimension.

24. The apparatus according to claim 19, characterized in that, The object behavior pattern determination device further includes a sorting module, which sorts the multidimensional encoded data according to the chronological order of the occurrence of the behaviors represented by the multidimensional encoded data to obtain an encoded data sequence.

25. The apparatus according to claim 19, characterized in that, The data acquisition module is further configured to acquire a sequence of behavioral data of the target object; the behavioral data in the sequence of behavioral data is used to characterize the behavior generated by the target object, and each behavioral data is sorted according to the occurrence time of the behavior; The sorting position of the multidimensional encoded data in the encoded data sequence is the same as the sorting position of the behavioral data matched by the multidimensional encoded data in the behavioral data sequence.

26. The apparatus according to claim 23, characterized in that, The data mapping processing includes a time dimension; the data mapping module further includes a time dimension data mapping unit for acquiring time dimension operators; the time dimension operators represent the data mapping relationship of the time dimension; the duration conditions matched by the time dimension operators are determined; and continuous multidimensional encoded data that meet the duration conditions are determined from the encoded data sequence. Based on the time dimension operator, the continuous multidimensional encoded data is processed by data mapping to obtain the behavioral sequence features of the target object in the time dimension.

27. The apparatus according to claim 26, characterized in that, The time dimension data mapping unit is also used to determine the time window matched by the time dimension operator; based on the time window that slides dynamically in the encoded data sequence, the multidimensional encoded data in the same time window are regarded as continuous multidimensional encoded data.

28. The apparatus according to claim 27, characterized in that, The time-dimensional data mapping unit is also used to extract behavioral temporal features from continuous multidimensional encoded data within the same time window based on the time-dimensional operator, thereby obtaining continuous behavioral sequence features of the target object.

29. The apparatus according to claim 23, characterized in that, The data mapping process includes an object dimension; the data mapping module further includes an object dimension data mapping unit for obtaining object dimension operators; the object dimension operators represent the data mapping relationship of the object dimension; the object selection conditions matched by the object dimension operators are determined; and multidimensional encoded data containing object information and satisfying the object selection conditions are filtered from the encoded data sequence. Based on the object dimension operator, the selected multidimensional encoded data is processed by data mapping to obtain the behavioral sequence features of the target object in the object dimension.

30. The apparatus according to claim 29, characterized in that, The object selection condition is the object that has an interactive behavior; the object dimension data mapping unit is also used to obtain the target encoded character representing the interactive object in the multidimensional encoded data, and divide the multidimensional encoded data with the same target encoded character into the same group to obtain the multidimensional encoded data group; For each multidimensional encoded data group, interactive behavior features are extracted from the multidimensional encoded data group to obtain the interactive behavior sequence features of the target object.

31. The apparatus according to claim 30, characterized in that, The object dimension data mapping unit is further configured to sort the multidimensional encoded data with the same target encoded characters according to the occurrence time of the behavior represented by the multidimensional encoded data to obtain a multidimensional encoded data group.

32. The apparatus according to claim 23, characterized in that, The data mapping process includes a cumulative behavior dimension; the data mapping module further includes a cumulative behavior dimension data mapping unit, used to determine multidimensional encoded data representing the same type of behavior from the multidimensional encoded data in the encoded data sequence. For each type of behavior, repeating behavior features are extracted from the multidimensional encoded data representing the behavior to obtain the repeating behavior sequence features of the target object.

33. The apparatus according to claim 23, characterized in that, The data mapping process includes business dimensions; the data mapping module further includes a business dimension data mapping unit, used to filter out multi-dimensional coded data representing the same business from the multi-dimensional coded data in the coded data sequence. The business behavior feature extraction process is performed on the multidimensional encoded data representing the same business to obtain the business behavior sequence features of the target object.

34. The apparatus according to claim 23, characterized in that, The object behavior pattern determination device further includes a classification module for acquiring an object classification model trained based on behavior sequence feature sample data, wherein the behavior sequence feature sample data carries object category labels. Based on the object classification model, the target object is classified according to its behavioral sequence characteristics to obtain the classification result of the target object.

35. The apparatus according to claim 34, characterized in that, The target object includes the target account; the object classification model is a binary classification model that classifies accounts into risky accounts or non-risky accounts; the object behavior pattern determination device further includes an associated account marking module, which is used to obtain associated accounts that are the same as the subject to which the target account belongs when the classification result of the target account is a risky account; Risk labeling is performed on the behavioral sequence characteristics of the associated accounts.

36. The apparatus according to claims 19 to 22, 24 to 35, characterized in that, The data acquisition module is also used to acquire behavioral data of each behavior generated by the target object in each time period; The data mapping module is also used to determine the encoded data sequence consisting of multi-dimensional encoded data corresponding to the behavior within the same time period; and to perform data mapping processing on the encoded data sequence of each time period according to at least one dimension to obtain the behavior sequence features of each time period. The behavioral pattern of the target object is determined based on the behavioral sequence characteristics of each time period.

37. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 18.

38. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 18.

39. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the method according to any one of claims 1 to 18.

Images