Method, device and equipment for analyzing behavior of object group and storage medium

Abstract

The application provides a behavior analysis method and device of an object group, equipment and a storage medium, relates to the technical field of artificial intelligence, and aims to improve the efficiency and accuracy of analyzing the behaviors of objects in different object groups. The method comprises the following steps: obtaining a target object group; the target object group comprises at least two target objects belonging to different candidate object groups; determining each candidate behavior associated with the target object group; determining the information entropy corresponding to each candidate behavior according to the behavior characteristics of each target object in the target object group with respect to each candidate behavior; determining a target behavior from each candidate behavior based on the obtained information entropy; and wherein the difference between the behavior characteristics of the target objects belonging to different candidate object groups with respect to the target behavior meets the difference condition. The method can improve the efficiency and accuracy of determining the target behavior of objects in different object groups.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, device and storage medium for analyzing the behavior of a group of objects. Background Technology

[0002] In the process of analyzing the differences in behavior between objects belonging to different object groups, related technologies generally rely on the personal experience of data analysts to determine the differences in behavior between objects belonging to different object groups. On the one hand, the above analysis process is carried out manually, which is inefficient. On the other hand, the above analysis process relies too much on personal experience, which leads to low accuracy in determining the differences in behavior for objects belonging to different object groups. Summary of the Invention

[0003] This application provides a method, apparatus, device, and storage medium for analyzing the behavior of object groups, which improves the efficiency and accuracy of analyzing the behavior of objects belonging to different object groups.

[0004] The first aspect of this application provides a method for analyzing the behavior of a group of objects, including:

[0005] Obtain a target object group; the target object group includes at least two target objects, and the at least two target objects belong to different candidate object groups;

[0006] Determine each candidate behavior associated with the target object group;

[0007] Based on the behavioral characteristics of each target object in the target object group for each candidate behavior, the information entropy corresponding to each candidate behavior is determined. Each information entropy represents the degree of difference in behavioral characteristics of target objects belonging to different candidate object groups for the corresponding candidate behavior.

[0008] Based on the obtained information entropy, the target behavior is determined from the candidate behaviors; wherein, among the target objects belonging to different candidate object groups, the degree of difference in the behavioral characteristics of the target behavior meets the difference degree condition.

[0009] A second aspect of this application provides a behavior analysis device for a group of objects, comprising:

[0010] A target object determination unit is used to obtain a target object group; the target object group includes at least two target objects, and the at least two target objects belong to different candidate object groups;

[0011] A candidate behavior determination unit is used to determine each candidate behavior associated with the target object group;

[0012] The first analysis unit is used to determine the information entropy corresponding to each candidate behavior based on the behavioral characteristics of each target object in the target object group for each candidate behavior. Each information entropy represents the degree of difference in behavioral characteristics of target objects belonging to different candidate object groups for the corresponding candidate behavior.

[0013] The second analysis unit is used to determine the target behavior from the candidate behaviors based on the obtained information entropy; wherein, the degree of difference in the behavioral characteristics of the target behavior among the target objects belonging to the different candidate object groups meets the difference degree condition.

[0014] In one possible implementation, the first analysis unit is specifically used for:

[0015] Each behavioral feature within the aforementioned behavioral feature range is designated as a corresponding feature segmentation point; or

[0016] Within the range of behavioral features, each behavioral feature with an interval equal to the set behavioral feature step size is determined as a corresponding feature segmentation point.

[0017] In one possible implementation, the first analysis unit is specifically used for:

[0018] For each of the aforementioned feature segmentation points, the following operations are performed:

[0019] Based on one of the feature segmentation points, at least two sub-behavioral feature ranges corresponding to the behavioral feature range are determined.

[0020] Based on the candidate object group to which each target object belongs in each of the at least two sub-behavioral feature ranges, determine the sub-reference information entropy corresponding to each sub-behavioral feature range.

[0021] Based on the sub-reference information entropy corresponding to each of the sub-behavioral feature ranges, the reference information entropy corresponding to the feature segmentation point is determined.

[0022] In one possible implementation, the first analysis unit is specifically used for:

[0023] For each of the aforementioned sub-behavioral feature ranges, the following operations are performed respectively:

[0024] Based on the behavioral characteristics of each target object for the candidate behavior, determine each behavioral characteristic within a sub-behavioral characteristic range within each sub-behavioral characteristic range, and determine the total number of first objects corresponding to the sub-behavioral characteristic range based on the number of target objects corresponding to each determined behavioral characteristic.

[0025] Based on the total number of the first objects and the number of target objects belonging to each candidate object group among the target objects corresponding to each determined behavioral feature, a first estimated value is determined for each candidate object group.

[0026] Based on the first estimated value of each candidate object group, the sub-reference information entropy corresponding to the sub-behavioral feature range is determined.

[0027] In one possible implementation, the first analysis unit is specifically used for:

[0028] For each candidate object group, a first operation is performed to determine the first reference value corresponding to each candidate object group, and based on the determined first reference values, the sub-reference information entropy corresponding to the sub-behavioral feature range is determined.

[0029] The first operation includes:

[0030] Using the first estimated value of one of the candidate object groups as the true number and a preset value as the base, determine the logarithm corresponding to the candidate object group; based on the product of the determined logarithm and the first estimated value of the candidate object group, determine the first reference value corresponding to the candidate object group.

[0031] In one possible implementation, the first analysis unit is specifically used for:

[0032] For each of the sub-behavioral feature ranges, a second operation is performed to obtain a second reference value corresponding to each of the sub-behavioral feature ranges, and based on the determined second reference values, the reference information entropy corresponding to the feature segmentation point is determined.

[0033] The second operation includes:

[0034] Based on the total number of second objects in the target object group and the total number of first objects corresponding to one of the sub-behavioral feature ranges, a second estimated value for the sub-behavioral feature range is determined; and based on the product of the second estimated value and the sub-reference information entropy corresponding to the sub-behavioral feature range, a second reference value corresponding to the sub-behavioral feature range is determined.

[0035] In one possible implementation, the target object determination unit is specifically used for:

[0036] From each candidate object group contained in the different candidate object groups, select the set number of candidate objects that meet the requirements of each candidate object group.

[0037] Based on the selected candidate objects, the target object group is obtained.

[0038] In one possible implementation, the difference between the maximum and minimum values ​​in each of the candidate object groups is less than a difference threshold, or the ratio of the maximum and minimum values ​​is within a preset ratio range.

[0039] In one possible implementation, the second analysis unit is specifically used to determine the target behavior from the candidate behaviors by at least one of the following methods:

[0040] Based on the magnitude of each information entropy, the information entropy is sorted, and the candidate behavior corresponding to the information entropy sorted at the target position is determined as the target behavior;

[0041] The information entropy that is less than the information entropy threshold among the various information entropies is determined, and the candidate behavior corresponding to the determined information entropy is determined as the target behavior.

[0042] A third aspect of this application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described in the first aspect.

[0043] In a fourth aspect, this application provides a computer program product comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method provided in the first aspect.

[0044] In a fifth aspect, this application provides a computer-readable storage medium storing computer instructions that, when executed on a computer, cause the computer to perform the method described in the first aspect.

[0045] Since the embodiments of this application adopt the above-described technical solution, they have at least the following technical effects:

[0046] In this embodiment, the information entropy corresponding to the candidate behavior is determined based on the behavioral characteristics of each target object belonging to different candidate object groups. This information entropy is actually the information entropy of the behavioral characteristics of the target object to which it belongs. This information entropy actually represents the uncertainty (which can be understood as inaccuracy) of distinguishing different target objects to which they belong based on the candidate behavior. That is, in this embodiment, the smaller the information entropy corresponding to a candidate behavior, the lower the uncertainty of distinguishing different target objects to which they belong based on the candidate behavior, and the higher the accuracy of distinguishing different target objects to which they belong based on the candidate behavior. Therefore, the greater the difference in behavioral characteristics of the target objects belonging to the different candidate object groups, the greater the difference in behavioral characteristics of the target objects belonging to the different candidate object groups. Thus, in this embodiment, the target behavior for objects belonging to different candidate object groups can be determined based on the information entropy corresponding to each candidate behavior. Attached Figure Description

[0047] Figure 1 A schematic diagram illustrating an application scenario of a behavior analysis method for object groups provided in an embodiment of this application;

[0048] Figure 2 A schematic diagram of a process for dividing candidate object groups is provided for an embodiment of this application;

[0049] Figure 3 A flowchart illustrating a behavior analysis method for an object group provided in this application embodiment;

[0050] Figure 4 A flowchart for obtaining information entropy corresponding to behavioral features is provided in an embodiment of this application;

[0051] Figure 5 A flowchart for determining the reference information entropy corresponding to each feature segmentation point is provided in an embodiment of this application;

[0052] Figure 6 A schematic diagram illustrating a sub-behavioral feature range corresponding to a behavioral feature range provided in an embodiment of this application;

[0053] Figure 7 A flowchart illustrating a behavior analysis method for an object group provided in an embodiment of this application;

[0054] Figure 8 A structural diagram of a behavior analysis device for an object group provided in an embodiment of this application;

[0055] Figure 9 This is a structural diagram of a computer device provided in an embodiment of this application. Detailed Implementation

[0056] To better understand the technical solutions provided in the embodiments of this application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods. To facilitate a better understanding of the technical solutions of this application by those skilled in the art, some concepts involved in this application will be explained below.

[0057] 1) Objects and object groups

[0058] The term "object" generally refers to the identity of a target on the Internet. In the embodiments of this application, the object may include, but is not limited to, an account (also known as a user) on the target platform. The account may include, but is not limited to, at least one of personal accounts and merchant accounts. The target platform may include, but is not limited to, at least one Internet platform such as a game platform, a live streaming platform, a content sharing platform, an e-commerce platform, or a learning platform.

[0059] An object group refers to a collection (also called a group) consisting of at least two objects. The object group will be further described below in the embodiments of this application.

[0060] 2) Information entropy

[0061] The term "information entropy" is a concept borrowed from thermodynamics by Claude Elwood Shannon, the father of information theory. In thermodynamics, thermal entropy is a physical quantity that represents the degree of disorder in the molecular state. Shannon used the concept of information entropy to describe the uncertainty of information sources and explained the relationship between probability and information redundancy in mathematical language. In data mining, information entropy is often used to measure the degree of chaos and separability of feature variables relative to target variables.

[0062] In this embodiment, the information entropy corresponding to the candidate behavior can characterize the degree of uncertainty in distinguishing the candidate object group to which the target object belongs based on the candidate behavior. The smaller the information entropy corresponding to a candidate behavior, the smaller the degree of uncertainty mentioned above. In fact, it characterizes the degree of uncertainty in distinguishing the candidate object group to which different target objects belong based on the candidate behavior. That is, it characterizes the high accuracy of distinguishing the different candidate object groups to which different target objects belong based on the candidate behavior. The greater the difference in behavioral characteristics of the objects included in different candidate object groups regarding the candidate behavior, the greater the difference in behavioral characteristics of the objects in different candidate object groups regarding the candidate behavior. Therefore, in this embodiment, the information entropy corresponding to the behavioral characteristics characterizes the degree of difference in behavioral characteristics of the objects belonging to different candidate object groups regarding the candidate behavior.

[0063] This application relates to Artificial Intelligence (AI), specifically designed based on big data analytics and machine learning (ML) technologies within AI. AI utilizes digital computers or computers-controlled machines to simulate, extend, and expand human intelligence, perceiving the environment, acquiring knowledge, and using that knowledge to achieve optimal results—a theory, method, technology, and application system. In other words, AI is a comprehensive technology within computer science that attempts to understand the essence of intelligence and produce new intelligent machines capable of reacting in a manner similar to human intelligence.

[0064] Artificial intelligence (AI) studies the design principles and implementation methods of various intelligent machines, enabling them to perceive, reason, and make decisions. AI technology mainly includes computer vision, natural language processing, and machine learning or deep learning. With the research and advancement of AI technology, it is being researched and applied in many fields, such as smart homes, intelligent customer service, virtual assistants, smart speakers, intelligent marketing, autonomous driving, robotics, and smart healthcare. It is believed that with further technological development, AI will be applied in even more fields and play an increasingly important role.

[0065] Machine learning is a multidisciplinary field involving probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory, among others. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills, and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to endow computers with intelligence; its applications span all areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, and inductive learning.

[0066] It should be noted that in the embodiments of this application, data related to the acquisition of target groups and the like are involved. When the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0067] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0068] The design concept of this application will be explained below.

[0069] With the rapid development of internet technology, it is common practice to analyze the differentiated behaviors of objects belonging to different groups. Based on these behavioral differences, operational strategies and business plans can be formulated for each group to improve product and operational effectiveness. Currently, this process generally relies on the personal experience of data analysts to analyze relevant data on various behaviors of objects belonging to different groups to identify differentiated behaviors. However, this analysis process is based on human experience, which is inefficient and results in low accuracy in identifying differentiated behaviors. Furthermore, due to the limitations of human experience, the dimensions of the behaviors analyzed in this process are also relatively low.

[0070] In view of this, the inventors have designed a method, apparatus, device, and readable storage medium for analyzing the behavior of objects in different object groups, which improves the efficiency and accuracy of analyzing the differences in behavior between objects in different object groups. In the embodiments of this application, candidate behaviors between objects in different candidate object groups can be analyzed to determine the candidate behaviors whose degree of difference in behavioral characteristics of objects in different candidate object groups meets the degree of difference condition, and these are determined as target behaviors. Specifically, in the embodiments of this application, a target object group can be obtained, each candidate behavior associated with the target object group can be determined, and based on the behavioral characteristics of each target object in the target object group for each candidate behavior, the information entropy corresponding to each candidate behavior can be determined. Based on the obtained information entropy, the target behavior can be determined from each candidate behavior. In this case, the multiple target objects included in the target object group belong to different candidate object groups.

[0071] To better understand the design concept of this application, the following examples illustrate application scenarios in the embodiments of this application. Please refer to... Figure 1 This paper provides an application scenario for a behavior analysis method for an object group, which may include a terminal device 110 and a server 120; the terminal device 110 and the server 120 can communicate via a network, wherein:

[0072] A target client can be installed on the terminal device 110. After an object (including the target object involved in the embodiments of this application and any object among the candidate objects included in each candidate object group) logs into the target client, it can trigger various candidate behaviors. Then, the terminal device 110 can collect the behavioral characteristics of the object for each candidate behavior. The terminal device 110 can save the collected behavioral characteristics locally or send the collected behavioral characteristics to the server 120.

[0073] As one embodiment, the behavior analysis method for object groups provided in this application can be deployed on server 120 for execution. Server 120 obtains a target object group and determines each candidate behavior associated with the target object group. Based on the behavioral characteristics of each target object in the target object group for each candidate behavior, it determines the information entropy corresponding to each candidate behavior. Based on the obtained information entropy, it determines the target behavior from the candidate behaviors. The target objects included in the target object group belong to different candidate object groups. The degree of difference in the behavioral characteristics of the target behavior between the target objects belonging to the different candidate object groups meets the difference degree condition.

[0074] As one embodiment, the behavior analysis method for object groups provided in this application can also be deployed on terminal device 110 for execution. The execution process can be referred to the execution process deployed on service 120, and will not be repeated here.

[0075] The terminal device 110 in this application embodiment (such as, but not limited to, 110-1 or 110-2 shown in the figure) can be a mobile terminal, a fixed terminal, a portable terminal, or a vehicle-mounted terminal, such as a mobile phone, site, unit, device, multimedia computer, multimedia tablet, Internet node, communicator, desktop computer, laptop computer, notebook computer, netbook computer, tablet computer, personal communication system (PCS) device, personal navigation device, personal digital assistant (PDA), audio / video player, digital camera or camcorder, positioning device, television receiver, radio broadcast receiver, e-book device, gaming device, or any combination thereof, including accessories and peripherals of these devices or any combination thereof.

[0076] The server 120 in this embodiment can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or multiple cloud servers (such as, but not limited to, servers 120-1, 120-2 or 120-3 shown in the figure) that provide basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The functions of the server 120 can be implemented by one or more cloud servers, or by one or more cloud server clusters, etc.

[0077] The following section provides a detailed description of the behavior analysis method for object groups in the embodiments of this application. It should be noted that the above application scenarios are shown only to facilitate understanding of the spirit and principles of this application, and the implementation methods of this application are not limited in any way.

[0078] First, the candidate object groups and target object groups involved in the embodiments of this application will be described in detail. In the embodiments of this application, the object group can refer to a set (or group) consisting of at least two objects, the candidate object group is the object group that needs to be analyzed, and the target object group is the object group determined based on different candidate object groups.

[0079] In this embodiment, the method of dividing candidate object groups is not limited. Those skilled in the art can set it according to actual needs. For example, multiple objects can be divided into different candidate object groups according to at least one piece of information such as the attribute characteristics of the objects or the behavior associated with the objects. Several examples are given below:

[0080] Method 1 for dividing objects into groups: Divide objects into different candidate object groups according to their attribute characteristics.

[0081] In this embodiment of the application, the object includes an account (such as, but not limited to, a personal account or a merchant account), and the attribute features of the object may include, but are not limited to, at least one of the following: account attribute information, account login information, account location information, account preference information, etc.

[0082] When the object is a personal account, the account attribute information may include, but is not limited to, at least one of the following: age, gender, occupation, place of origin, educational background, marital status, and device information of the device used to log in to the personal account. The account login information may include, but is not limited to, at least one of the following: frequency or duration of logging into the personal account, device information of the device used to log in to the personal account, and location of logging into the personal account. The account location information may include, but is not limited to, at least one of the following: the permanent residence of the person corresponding to the account and the location of logging into the personal account. The account preference information may include, but is not limited to, at least one of the following: the personal account's preferred entertainment programs, application software, hobbies, etc.

[0083] When the object is a merchant account, the account attribute information may include, but is not limited to, the merchant type, merchant registration information (such as registration date or registered capital), information on the shareholder of the merchant, information on the services provided by the merchant (such as service type, which may include, but is not limited to, application software development services or consulting services), and information on the goods sold by the merchant (such as, which may include, but is not limited to, the type of goods or the price range of the goods); the account login information may include, but is not limited to, at least one of the following: the time or frequency of the merchant logging into the merchant account, the device information of the device used to log into the merchant account, and the location of the merchant logging into the merchant account; the account location information may include, but is not limited to, the merchant's registered location or the merchant's business location; and the account preference information may include, but is not limited to, at least one of the following: the transaction methods supported by the merchant and the merchant's delivery methods.

[0084] Furthermore, to enhance the flexibility of dividing candidate object groups, in this embodiment of the application, candidate object groups can be divided based on any one or more attribute characteristics of the object. For ease of understanding, a specific example is given here. In this example, the personal account is taken as the above-mentioned object. The candidate object groups are divided according to the age and gender of the personal account. It is possible, but not limited to, dividing the objects corresponding to males under 18 years old, the objects corresponding to females under 18 years old, the objects corresponding to males aged 18-40 years old, the objects corresponding to females aged 18-40 years old, the objects corresponding to males over 40 years old, and the objects corresponding to females over 40 years old into one candidate object group.

[0085] Method 1 for grouping objects: Divide objects into different candidate object groups according to the behaviors associated with them.

[0086] For example, objects that perform the same behavior can be grouped into one candidate object group; for example, objects that use function A can be grouped into candidate object group 1, and objects that use function B can be grouped into candidate object group 2, where using function A and using function B are different behaviors.

[0087] Furthermore, in order to improve the accuracy of analyzing the behavior of objects in different object groups using the target service, the embodiments of this application may also classify objects that perform the same behavior in the target service into a candidate object group. The target service is not limited, and those skilled in the art can set it according to actual needs. For example, the target service may be, but is not limited to, at least one of mobile payment service, information sharing service, live streaming service, game service, etc.

[0088] As one embodiment, this application embodiment can also divide objects into different candidate object groups based on both the object's attribute characteristics and the behaviors associated with the object; for ease of understanding, a specific example is given here, please refer to... Figure 2 In this example, candidate object groups are divided according to the object's gender and the associated behavior. For example, objects corresponding to females using function A are divided into candidate object group 1, objects corresponding to males using function A are divided into candidate object group 2, objects corresponding to females using function B are divided into candidate object group 3, and objects corresponding to females using function B are divided into candidate object group 4. Here, using function A and using function B are different behaviors.

[0089] The following describes the behaviors, candidate behaviors, and target behaviors involved in the embodiments of this application.

[0090] Generally, a behavior refers to an operation or a series of related operations of an object, which can be set by those skilled in the art according to actual needs. In the embodiments of this application, a candidate behavior refers to the behavior associated with the target objects included in the target object group. The candidate behavior may include operations actively performed by the target object or operations passively performed by the object. The target behavior is the behavior in which the degree of difference in behavioral characteristics between objects belonging to different candidate object groups meets the degree of difference condition. That is, if the degree of difference in behavioral characteristics of a certain candidate behavior between objects belonging to different candidate object groups meets the degree of difference condition, then the candidate behavior is determined as the target behavior.

[0091] As one embodiment, to improve the accuracy of analyzing the behavior of objects within different object groups using a target business, the behaviors associated with target objects in a target object group may differ for different target businesses; the behaviors associated with target objects in a target object group may also differ for different scenarios within the same target business; for mobile payment businesses, candidate behaviors may include, but are not limited to, any one or more combinations of electronic resource transfer operations, operations to convert electronic resources into other resources, operations to convert other resources into electronic resources, operations to add other accounts, and operations to delete other accounts; for gaming businesses, candidate behaviors may include, but are not limited to, any one or more combinations of operations to purchase game items, operations to form a team, operations to attack virtual objects, operations to participate in game competitions, and electronic resource transfer operations; for live streaming businesses, candidate behaviors may include, but are not limited to, any one or more combinations of operations to start a live stream, operations to watch a live stream, operations to comment on a live stream, operations to share a live stream, operations to send gifts to the streamer, and operations to purchase goods in a live stream.

[0092] based on Figure 1 The following describes an example of a behavior analysis method for object groups in this application embodiment. Please refer to the following examples. Figure 3 This diagram illustrates a behavior analysis method for an object group designed according to an embodiment of this application, specifically including the following steps:

[0093] Step S301: Obtain a target object group; the target object group includes at least two target objects, and the at least two target objects belong to different candidate object groups.

[0094] As one embodiment, the aforementioned target object group may include two target objects or more than two target objects; the aforementioned different candidate object groups may be two candidate object groups or more than two candidate object groups. In the following embodiments of this application, the aforementioned different candidate object groups are described as M (M is an integer not less than 2) candidate object groups. If this embodiment of the application is used to analyze the target behaviors that are different among the candidate objects contained in the aforementioned M candidate object groups, then in this step, some or all of the candidate objects can be selected from each of the M candidate object groups, and then the selected candidate objects are determined as the aforementioned target objects; the division method of each candidate object group can be referred to the above content, and will not be repeated here.

[0095] Step S302: Determine each candidate behavior associated with the above target object group.

[0096] As one embodiment, this application embodiment can also perform the operation of constructing candidate behaviors for the target objects included in the above-mentioned target object group, and determine each constructed candidate behavior as the candidate behavior associated with the target object group; wherein, the specific content of the candidate behaviors in this application embodiment is not limited, and those skilled in the art can set it according to actual needs, such as referring to the examples of candidate behaviors given above, etc., which will not be repeated here.

[0097] As one embodiment, in order to determine the richness of each candidate behavior associated with the target object group, this embodiment of the application can construct behaviors for each candidate object included in each of the above M candidate object groups, obtain a subset of candidate behaviors associated with each candidate object group, and then determine each candidate behavior included in each of the obtained subsets of candidate behaviors as each candidate behavior associated with the target object group; wherein, a subset of candidate behaviors includes behaviors constructed for the candidate objects included in the corresponding candidate object group, and a subset of candidate behaviors may include at least one candidate behavior.

[0098] As an example, to improve the efficiency and accuracy of determining the candidate behaviors associated with the target object group, this embodiment of the application can also use feature engineering techniques in machine learning to construct behaviors for the target objects contained in the target object group and the candidate objects contained in each of the candidate object groups. Here, feature engineering refers to the task of converting raw data related to the object's behavior into corresponding behavior construction. The raw data can contain detailed information about the candidate objects, such as the object's age, interests, location, average time spent on the target product, and behavioral characteristics of the object for each behavior. The feature engineering in this embodiment of the application can include, but is not limited to, any one or more of the following processing methods: binning, one-hot encoding, hashing trick, embedding, log transformation, scaling, normalization, handling missing values, and feature interaction.

[0099] Step S303: Based on the behavioral characteristics of each target object in the target object group for each candidate behavior, determine the information entropy corresponding to each candidate behavior. Each information entropy represents the degree of difference in behavioral characteristics of target objects belonging to different candidate object groups for the corresponding candidate behavior.

[0100] As an example, in order to improve the accuracy of the information entropy corresponding to each candidate behavior, this embodiment of the application can determine the behavioral feature range of each candidate behavior based on each target object in the above-mentioned target object group and the behavioral features of each candidate behavior; and determine the information entropy corresponding to each behavioral feature based on the reference information entropy of each feature segmentation point corresponding to each behavioral feature range. The specific process will be further explained below.

[0101] Step S304: Based on the obtained information entropy, determine the target behavior from the above candidate behaviors; wherein, among the target objects belonging to the above different candidate object groups, the degree of difference in the behavioral characteristics of the target behavior meets the difference degree condition.

[0102] As one embodiment, in order to improve the accuracy of the determined target behavior, in step S304, the target behavior can be determined from the candidate behaviors using any of the target behavior determination methods 1 to 3:

[0103] Method 1 for determining the target behavior: Based on the magnitude of each of the above information entropies, sort the above information entropies, and determine the candidate behavior corresponding to the information entropy ranked first in the target order as the above target behavior.

[0104] The first target position mentioned above can be one or more sorting positions. That is, in determination method 1, one or more target behaviors can be determined from the candidate behaviors mentioned above. For example, after sorting the information entropies in ascending order, we get {information entropy 1, information entropy 2, information entropy 3, information entropy 4, ..., information entropy 20}. Then, we can determine the candidate behavior corresponding to information entropy 1 or information entropy 2 as the target behavior, but not limited to it. We can also determine the candidate behaviors corresponding to information entropy 1 to information entropy 4 as the target behaviors at the same time.

[0105] Method 2 for determining the target behavior: Determine the information entropy that is less than the information entropy threshold among the above information entropy, and determine the candidate behavior corresponding to the determined information entropy as the above target behavior; wherein, the above information entropy threshold is not limited, and those skilled in the art can set it according to actual needs, such as determining 0.2 or 0.25 times the maximum value among the above information entropy as the above information entropy threshold.

[0106] Target behavior determination method 3: Based on the magnitude of each of the above information entropies, sort the above information entropies, and determine the candidate behavior corresponding to the information entropy ranked first in the target order as the above target behavior; and determine the candidate behavior corresponding to the information entropy less than the information entropy threshold as the above target behavior; for details, please refer to the detailed description of the above target behavior determination method 1 and target behavior determination method 2, which will not be repeated here.

[0107] As one embodiment, in order to improve the flexibility of the determined target behavior, in step S301, candidate objects that meet the set number corresponding to each of the above M candidate object groups can be selected from each candidate object group; based on the obtained candidate objects, the above target object group is obtained; for example, the selected candidate objects can be directly determined as target objects, and the set of determined target objects can be used as the above target object group.

[0108] The method for selecting candidate objects from each candidate object group is not limited. For example, for a candidate object group, a set number of candidate objects corresponding to that candidate object group can be randomly selected from the candidate objects contained in that candidate object group. Alternatively, some object priority selection conditions can be set, and based on these object priority selection conditions, a set number of candidate objects corresponding to that candidate object group can be randomly selected from the candidate objects contained in that candidate object group. The above object priority selection conditions can be set based on the business needs of the target business. For example, if the target business is a game business, considering that the target users aged 10-30 are more likely to use the game business, the above object priority selection conditions can be set to prioritize the target users aged 10-30. If the target business is a mobile payment business, considering that the target users aged 18-40 have funds and are more likely to use the mobile payment business, the above object priority selection conditions can be set to prioritize the target users aged 18-40.

[0109] As one embodiment, to improve the accuracy of analyzing the differences in behavior among different candidate object groups, in step S301, the number of candidate objects selected from different candidate object groups can be set to be within a certain range. For example, the difference between the maximum and minimum values ​​of the set numbers corresponding to each candidate object group can be set to be less than a difference threshold. The method of setting the difference threshold is not limited; those skilled in the art can set it according to actual needs. For example, the difference threshold can be set based on the average number of candidate objects contained in different candidate object groups, such as, but not limited to, setting the average value as the difference threshold. In this embodiment, the difference threshold can also be set based on business experience. The specific value of the difference threshold is also not limited; those skilled in the art can set it according to actual needs. For example, the difference threshold can be set to the range of 0-100. If candidate object group 1 contains 100 candidate objects and candidate object group 2 contains 200 candidate objects, then, but not limited to, 80 candidate objects can be selected from candidate object group 1 and candidate object group 2 respectively, and the selected 160 candidate objects can be determined as the target objects in the target object group.

[0110] As one embodiment, to improve the accuracy of analyzing the differences in behavior among different candidate object groups, step S301 can further set the ratio of the maximum and minimum values ​​among the set numbers corresponding to each candidate object group to be within a preset ratio range. The method of setting the preset ratio range is not limited; those skilled in the art can set it according to actual needs. For example, the preset ratio range can be set based on the ratio of the number of candidate objects contained in different candidate object groups, or it can be set based on business experience. The specific value of the preset ratio range is also not limited. Those skilled in the art can set the ratio according to actual needs, such as setting the preset ratio range to 1.0 to 1.3 or 1.0 to 1.2. If candidate object group 1 contains 200 candidate objects and candidate object group 2 contains 200 candidate objects, and assuming the preset ratio range is 1.0 to 1.2, and 60 candidate objects are selected from candidate object group 1, then 60 can be taken as the maximum value, and 50 to 60 candidate objects can be selected from candidate object group 2. Alternatively, 60 can be taken as the minimum value, and 60 to 72 candidate objects can be selected from candidate object group 2.

[0111] As one example, please refer to Figure 4 To improve the accuracy of the information entropy corresponding to each candidate behavior, in step S303, in this embodiment of the application, the following steps S401-S404 can be performed for each of the above-mentioned behavioral features to obtain the information entropy corresponding to each behavioral feature:

[0112] Step S401: Based on the behavioral characteristics of each target object in the target object group for one of the candidate behaviors, determine the range of behavioral characteristics of the candidate behavior.

[0113] As one embodiment, the behavioral features of the aforementioned candidate behavior can be, but are not limited to, discrete values. Therefore, the set of behavioral features of each target object in the target object group for the aforementioned candidate behavior can be determined as the range of behavioral features for the aforementioned candidate behavior. This range can be a set of deduplicated behavioral features; alternatively, the range of behavioral features for the aforementioned candidate behavior can be determined based on the maximum and minimum values ​​among the behavioral features of each target object, such as defining the interval between the maximum and minimum values ​​as the range of behavioral features for the aforementioned candidate behavior. For ease of understanding, a specific example is given here. In this example, electronic resource transfer operation is taken as one of the above-mentioned candidate behaviors, and the transferred electronic resource value is taken as the above-mentioned behavior feature. The target object group includes target object 1 to target object 10, and the electronic resource values ​​transferred by target object 1 to target object 10 for the above-mentioned electronic resource transfer operation are 1, 3, 2, 3, 4, 5, 8, 1, 7, and 7, respectively. Then, {1, 2, 3, 4, 5, 7, 8} is determined as the behavior feature range of the above-mentioned electronic resource transfer operation. Alternatively, the interval [1, 8] can be determined as the behavior feature range of the above-mentioned electronic resource transfer operation.

[0114] As one embodiment, the behavioral feature of the aforementioned candidate behavior can also be a continuous value. Therefore, the maximum and minimum values ​​among the continuous values ​​of the behavioral feature of each target object in the target object group for the aforementioned candidate behavior can be determined as the behavioral feature range of the aforementioned candidate behavior. For example, the interval combining the maximum and minimum values ​​can be determined as the behavioral feature range of the aforementioned candidate behavior. For ease of understanding, a specific example is given here. In this example, the operation of reading an e-book is taken as the aforementioned candidate behavior, and the continuous page range of the candidate object reading the e-book is taken as the behavioral feature. The target object group includes target object 1 to target object 5, and the page ranges of target object 1 to target object 10 for the aforementioned e-book reading operation are 2-10, 3-17, 6-18, 20-40, and 13-21, respectively. It can be determined that the maximum value in the page range is 40 and the minimum value is 2. Therefore, the interval [2, 40] can be set as the behavioral feature range of the aforementioned e-book reading operation.

[0115] It should be noted that the above-described method for determining the behavioral feature range of a candidate behavior is merely an example. Those skilled in the art can flexibly set the specific method for determining the behavioral feature range of each candidate behavior according to actual business needs.

[0116] Step S402: Within the range of the above-mentioned behavioral features, determine at least one feature segmentation point.

[0117] As an example, in order to determine the accuracy and flexibility of feature segmentation points, each behavioral feature in the above-mentioned behavioral feature range can be determined as a corresponding feature segmentation point. For example, when a behavioral feature range is {1, 2, 3, 4, 5, 7, 8}, then 1, 2, 3, 4, 5, 7, and 8 can be determined as a feature segmentation point. When a behavioral feature range is [1, 10] and the values ​​in this range can only be integers, then 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10 can be determined as a feature segmentation point.

[0118] As an embodiment, to ensure the accuracy and flexibility of determining feature segmentation points, this embodiment further defines each behavioral feature within the aforementioned behavioral feature range, spaced at intervals equal to a set behavioral feature step size, as a corresponding feature segmentation point. The defined behavior is not overly limited; those skilled in the art can set it according to actual needs. For example, if a behavioral feature range is {1, 2, 3, 4, 5, 6} and the set behavioral feature step size is 2, then 1, 3, and 5 can be defined as a feature segmentation point, but is not limited to this; 2, 4, and 6 can also be defined as a feature segmentation point. If a behavioral feature range is [1-50] and the set behavioral feature step size is 10, then 1, 11, 21, 31, and 41 can be defined as a feature segmentation point, but is not limited to this; 10, 20, 30, 40, and 50 can also be defined as a feature segmentation point, and 5, 15, 25, 35, and 45 can also be defined as a feature segmentation point. Other cases can be deduced similarly, and no further explanation is provided here.

[0119] Step S403: Determine the reference information entropy corresponding to each feature segmentation point.

[0120] As one embodiment, each feature segmentation point in step S403 is actually the feature segmentation point determined in step S402. In this embodiment, one feature segmentation point can divide the above-mentioned behavioral feature range into at least two sub-behavioral feature ranges. Therefore, in this embodiment, for one of the feature segmentation points, the sub-reference information entropy corresponding to each of the at least two sub-behavioral feature ranges obtained based on the one feature segmentation point can be determined, and then the reference information entropy corresponding to the one feature segmentation point can be determined based on the determined sub-reference information entropy. In this embodiment, the specific method of determining the reference information entropy corresponding to each feature segmentation point will be further explained below.

[0121] Step S404: Among the obtained reference information entropies, the one that meets the set value condition is determined as the information entropy corresponding to the above-mentioned behavioral feature.

[0122] As one embodiment, the minimum value among the obtained reference information entropies can be taken as the reference information entropy that meets the set value conditions. Therefore, the minimum value among the obtained reference information entropies is determined as the information entropy corresponding to the above-mentioned behavioral feature. That is, the reference information entropies corresponding to each of the above-mentioned feature segmentation points are respectively denoted as F1, F2, ..., F n (where n is the identifier of the feature segmentation point), then "F1, F2, ..., F n The minimum value in the above is determined as the information entropy corresponding to one of the aforementioned behavioral features.

[0123] As one embodiment, the obtained reference information entropies can be sorted according to their magnitudes, and the reference information entropy ranked in the second target position can be determined as the information entropy corresponding to the aforementioned behavioral feature. For example, after sorting the obtained reference information entropies in ascending order, the reference information entropy ranked in the first, second, or third position can be determined as the information entropy corresponding to the aforementioned behavioral feature, etc.

[0124] As one example, please refer to Figure 5 In this embodiment of the application, in step S403 above, the following steps S501-S504 can be performed on each of the above feature segmentation points to improve the accuracy of the reference information entropy corresponding to each determined feature segmentation point.

[0125] Step S501: Based on one of the feature segmentation points mentioned above, determine at least two sub-behavioral feature ranges corresponding to the aforementioned behavioral feature range.

[0126] As one embodiment, for a single feature segmentation point, the aforementioned behavioral feature range can be divided into two corresponding sub-behavioral feature ranges; please refer to [link to relevant documentation]. Figure 6For example, if the behavioral feature range is {1, 2, 3, 4, 5, 7, 8} and the feature split point is 3, then "less than 3" (i.e., "<3") and "greater than or equal to 3" (">=3") can be defined as two sub-behavioral feature ranges corresponding to {1, 2, 3, 4, 5, 7, 8}. Similarly, "less than or equal to 3" (i.e., "<=3") and "greater than 3" (i.e., ">3") can also be defined as two sub-behavioral feature ranges corresponding to {1, 2, 3, 4, 5, 7, 8}. {1, 2} and {3, 4, 5, 7, 8} are defined as the two sub-behavioral feature ranges corresponding to {1, 2, 3, 4, 5, 7, 8}. Alternatively, {1, 2, 3} and {4, 5, 7, 8} can also be defined as the two sub-behavioral feature ranges corresponding to {1, 2, 3, 4, 5, 7, 8}. For the continuous value type behavioral feature range [N1, N2], the two corresponding sub-behavioral feature ranges can be determined in the same way as described above. N1 is less than N2, and N1 and N2 can be values ​​of any type.

[0127] As one example, for a single feature segmentation point, the aforementioned behavioral feature range can be divided into three corresponding sub-behavioral feature ranges. Please refer to [link to relevant documentation]. Figure 6 When the feature segmentation point is 3, "less than 3" (i.e., "<3"), "equal to 3" (i.e., "=3"), and "greater than 3" (i.e., ">3") can be defined as the three sub-behavioral feature ranges corresponding to {1, 2, 3, 4, 5, 7, 8}. Similarly, {1, 2}, {3}, and {4, 5, 7, 8} can be defined as the three sub-behavioral feature ranges corresponding to {1, 2, 3, 4, 5, 7, 8}. For the continuous value type behavior feature range [N1, N2], the corresponding three sub-behavioral feature ranges can be determined in the same way as above. N1 is less than N2, and N1 and N2 can be values ​​of any type.

[0128] Step S502: Based on the candidate object group to which each target object belongs in each of the above at least two sub-behavioral feature ranges, determine the sub-reference information entropy corresponding to each of the above sub-behavioral feature ranges.

[0129] As one embodiment, in step S502, this embodiment of the application may, but is not limited to, perform the following steps S5021 to S5023 for each of the above-mentioned sub-behavioral feature ranges, in order to determine the sub-reference information entropy corresponding to each sub-behavioral feature range.

[0130] Step S5021: Based on the behavioral characteristics of each target object for the candidate behavior, determine each behavioral characteristic within a sub-behavioral characteristic range within the range of sub-behavioral characteristics, and determine the total number of first objects corresponding to the sub-behavioral characteristic range based on the number of target objects corresponding to each determined behavioral characteristic.

[0131] As one embodiment, the number of target objects corresponding to each determined behavioral feature can be, but is not limited to, being determined as the total number of first objects corresponding to the aforementioned sub-behavioral feature range. For example, if the target object group includes target objects 1 to 10, and the behavioral features of target objects 1 to 10 for the aforementioned candidate behavior are behavioral features 1 to 10 respectively, and assuming that behavioral features 1, 5, and 8 are within the aforementioned sub-behavioral range, then since behavioral features 1, 5, and 8 correspond to target objects 1, 5, and 8 respectively, the total number of first objects corresponding to the aforementioned sub-behavioral feature range can be determined to be 3.

[0132] Step S5022: Based on the total number of the first objects and the number of target objects belonging to each candidate object group among the target objects corresponding to each behavioral feature, determine the first estimated value of each candidate object group.

[0133] As one embodiment, the first estimate for each candidate object group can be determined based on, but is not limited to, the following formula (1):

[0134]

[0135] In formula (1), m is the identifier of the candidate object group in each of the above candidate object groups; P1 m is the first estimated value corresponding to the candidate object group identified as m; i is the identifier of the sub-behavioral feature range, K i K represents the total number of the first objects corresponding to the sub-behavioral feature range denoted as i; m This refers to the number of target objects belonging to the candidate object group identified as m among the target objects corresponding to each behavioral feature within the aforementioned sub-behavioral feature range. For ease of understanding, a specific example is given above: the target object group includes target objects 1 to 10, and the behavioral features of target objects 1 to 10 for the aforementioned candidate behavior are behavioral features 1 to 10, respectively. Assuming that behavioral features 1, 5, and 8 are within the aforementioned sub-behavioral range, target object 1 belongs to candidate object group 1, and target objects 5 and 8 belong to candidate object group 2, then the total number K of the first objects corresponding to the aforementioned sub-behavioral feature range is 3, and the first estimated value of candidate object group 1 is... First estimate of candidate group 2

[0136] Step S5023: Based on the first estimated value of each of the above candidate object groups, determine the sub-reference information entropy corresponding to the above sub-behavioral feature range.

[0137] As one embodiment, in step S5023, a first operation can be performed for each of the above-mentioned candidate object groups to determine the first reference value corresponding to each of the above-mentioned candidate object groups, and based on the determined first reference values, to determine the sub-reference information entropy corresponding to the above-mentioned sub-behavioral feature range; wherein, the above-mentioned first operation includes: using the first estimated value of one of the above-mentioned candidate object groups as the truth and a preset value as the base, to determine the logarithm corresponding to the above-mentioned candidate object group; and based on the product of the determined logarithm and the first estimated value of the above-mentioned candidate object group, to determine the first reference value corresponding to the above-mentioned candidate object group.

[0138] As one embodiment, the sub-reference information entropy corresponding to the above-mentioned sub-behavioral feature range can be determined based on the following formula (2);

[0139]

[0140] In formula (2), j is the identifier of the candidate behavior; i is the identifier of the sub-behavioral feature range of the behavioral feature range of the candidate behavior identified as j; Ej i P1 is the sub-reference information entropy corresponding to the sub-behavioral feature range identified as i within the sub-behavioral feature range corresponding to the behavioral feature range of candidate behavior identified as j; m is the identifier of the candidate object group in each of the above candidate object groups, and M is the total number of each of the above candidate object groups; m q is the first estimated value corresponding to the candidate object group identified as m; q is the above preset value. Those skilled in the art can set q according to actual needs, such as setting q to 2, 10 or e, etc., but not limited to it.

[0141] Step S503: Based on the sub-reference information entropy corresponding to each of the above sub-behavioral feature ranges, determine the reference information entropy corresponding to the above feature segmentation point.

[0142] As one embodiment, in step S503, a second operation can be performed for each of the above-mentioned sub-behavioral feature ranges to obtain a second reference value corresponding to each of the above-mentioned sub-behavioral feature ranges, and a reference information entropy corresponding to the above-mentioned feature segmentation point can be determined based on the determined second reference values; wherein, the above-mentioned second operation includes: determining a second estimated value of the above-mentioned sub-behavioral feature range based on the second total number of target objects contained in the above-mentioned target object group and the first total number of objects corresponding to one of the above-mentioned sub-behavioral feature ranges; and determining a second reference value corresponding to the above-mentioned sub-behavioral feature range based on the product of the above-mentioned second estimated value and the sub-reference information entropy corresponding to the above-mentioned sub-behavioral feature range.

[0143] As one embodiment, the reference information entropy corresponding to the above feature segmentation points can be determined based on the following formula (3);

[0144]

[0145] In formula (3), j is the identifier of the candidate behavior; i is the identifier of the sub-behavioral feature range of the behavioral feature range of the candidate behavior identified as j; Ej i It is the sub-reference information entropy corresponding to the sub-behavioral feature range identified as i within the sub-behavioral feature range corresponding to the behavioral feature range of candidate behavior identified as j, where Ej i It can be obtained, but is not limited to, through the principle of Formula 1 above; P2 i It is the second estimated value corresponding to the sub-behavioral feature range identified as i within the sub-behavioral feature range corresponding to the behavioral feature range of the candidate behavior identified as j; n is the identifier of the feature segmentation point among at least one feature segmentation point of the behavioral feature range of the candidate behavior identified as j; N is the total number of feature segmentation points among the above at least one feature segmentation point; F n It is the reference information entropy corresponding to the feature segmentation point identified as n.

[0146] As one embodiment, the second estimate of the range of each sub-behavioral feature can be determined based on, but is not limited to, the following formula (4):

[0147]

[0148] In formula (4), i is the identifier of the sub-behavioral feature range, and K i is the total number of the first objects corresponding to the sub-behavioral feature range represented by i; K is the total number of target objects contained in the target object group (i.e., the total number of the second objects mentioned above).

[0149] To facilitate understanding of the object group behavior analysis method provided in the embodiments of this application, the following specific example is provided in the embodiments of this application:

[0150] In this example, the behavior between objects in candidate object group A and candidate object group B is analyzed. Candidate object group A contains accounts that use the transfer function in mobile payment services, while candidate object group B contains accounts that use the payment receiving function in mobile payment services. Using the transfer function can be understood as sending electronic resources to other accounts outside of itself; using the payment receiving function can be understood as receiving electronic resources from other accounts outside of itself.

[0151] Please see Figure 7 Candidate object group A can contain objects A1 to Aw (w is an integer greater than 1, and considering that the number of candidate objects in a candidate object group in a real business scenario will not be too small, the value of w is generally not too small). Object group B can contain objects B1 to Bs (s is an integer greater than 1, and considering that the number of candidate objects in a candidate object group in a real business scenario will not be too small, the value of s is generally not too small). In this example, we take selecting 5 candidate objects from each of candidate object group A and candidate object group B as the target objects. Select objects A1 to A5, and then select objects B1 to B5 from candidate object group B as target objects. The resulting target object set is {object A1, object A2, object A3, object A4, object A5, object B1, object B2, object B3, object B4, object B5}. Record the group label of each target object. This group label is used to represent the candidate object group to which the target object belongs. Those skilled in the art can set the corresponding group label according to actual needs. In this example, group label "1" represents that the target object belongs to candidate object group A, and group label "0" represents that the target object belongs to candidate object group B.

[0152] Furthermore, in this example, feature engineering techniques in machine learning are used to construct candidate behaviors for a group of target objects. Specifically, the relevant data of the behaviors of each target object in the target object group can be processed by binning, one-hot encoding feature hashing, nesting, logarithmic scaling, feature scaling, standardization, handling missing values, and feature interaction to obtain each candidate behavior associated with the target object group. Here, candidate behaviors U1 to U2 are used as examples to illustrate the candidate behaviors associated with the target object group. Based on the relevant content of step S303 above, the information entropy H{U1} to H{U5} corresponding to each of candidate behaviors U1 to U2 is determined.

[0153] After obtaining H{U1} to H{U5}, sort them in ascending order. Please continue to the next step. Figure 7 Here, assuming the sorting result is H{U3}, H{U1}, H{U5}, H{U4}, H{U2}, then correspondingly, the degree of difference between the objects in candidate object group A and the objects in candidate object group B for candidate behavior U3, candidate behavior U1, candidate behavior U5, candidate behavior U4, and candidate behavior U2 decreases sequentially; furthermore, but not limited to, candidate behavior U3 and candidate behavior U1 can be determined as the target behavior.

[0154] The following example illustrates the process of determining the information entropy H{U1} to H{U5} corresponding to candidate behaviors U1 to U2, using electron transfer operation as candidate behavior U1 and the electron resource value transferred by the electron transfer operation as the behavioral feature of candidate behavior U1. The process of determining the information entropy corresponding to candidate object U1 is explained as follows: Assume that the electron resource values ​​of objects A1, A2, A3, A4, A5, B1, B2, B3, B4, and B5 for candidate behavior U1 are respectively 2. Given the sequence {2, 3, 4, 5}, we can determine that the behavioral feature range of candidate behavior U1 is {2, 3, 4, 5}. Further, within {2, 3, 4, 5}, we determine the feature segmentation points as "2", "3", "4", and "5". Then, we further determine the reference information entropy corresponding to each feature segmentation point "2", "3", "4", and "5" as F2, F3, F4, and F5, respectively. The following section uses the determination of the reference information entropy corresponding to feature segmentation point "3" as an example to illustrate the process of obtaining F2, F3, F4, and F5:

[0155] In this example, feature segmentation point "3" divides the behavioral feature range of candidate behavior U1 from {2, 3, 4, 5} into two sub-behavioral feature ranges: "U1<3" (indicating that the electronic resource value of candidate behavior U1 is less than 3) and "U1>=3" (indicating that the electronic resource value of candidate behavior U1 is less than 3). Then, an information table as shown in Table 1 can be created for feature segmentation point "3".

[0156] Table 1:

[0157]

[0158] Based on the information in Table 1 above, determine the sub-reference information entropy EUI corresponding to the sub-behavioral feature range "U1<3". U1<3 The process is as follows:

[0159] Based on the above formula (1), the first estimated value of candidate object group A can be determined. The first estimate of candidate group B is Where K UI<3 It is the total number of the first objects corresponding to "U1<3"; then, based on the above formula (2), the entropy of the sub-reference information corresponding to "U1<3" is determined as follows.

[0160] Furthermore, based on a similar approach, the sub-reference information entropy corresponding to the sub-feature behavior range "U1>=3" is determined as follows:

[0161] In determining the sub-reference information entropy EUI corresponding to "U1<3" U1<3 The sub-reference information entropy corresponding to "U1>=3" is EUI. U1≥3 Then, a second estimate of "U1<3" can be determined based on the above formula (4). And the second estimate of U1>=3 is determined based on formula (4).

[0162] Furthermore, based on the above formula (3), the reference information entropy corresponding to feature segmentation point "3" can be determined as follows:

[0163] The method provided in this application embodiment, on the one hand, determines the information entropy corresponding to the candidate behavior based on the behavioral characteristics of each target object belonging to different candidate object groups for the aforementioned candidate behavior. The smaller the information entropy corresponding to a candidate behavior, the higher the accuracy of distinguishing different target objects belonging to different candidate object groups based on the candidate behavior. Thus, the greater the difference in behavioral characteristics of objects belonging to different candidate object groups for the candidate behavior, the more accurate the method can be. Therefore, in this application embodiment, the target behavior for objects belonging to different candidate object groups can be determined based on the information entropy corresponding to each candidate behavior. On the other hand, this application embodiment improves the accuracy and dimensionality of the determined behavioral characteristics associated with the target object group through feature engineering technology in machine learning, thereby enabling the analysis of differentiated target behaviors among objects included in different candidate object groups at a deeper level.

[0164] This technical solution, when used in a product operation insight system targeting mobile payment services, analyzes different user groups (i.e., the aforementioned candidate groups, with users being the aforementioned objects). Based on the analyzed target behaviors, corresponding operational strategies are set for users belonging to different user groups, which can improve the effectiveness by more than 50%, significantly enhancing the refined operation and growth of users.

[0165] Please refer to Figure 8 Based on the same inventive concept, embodiments of this application provide a behavior analysis device for a group of objects, comprising:

[0166] The target object determination unit 8001 is used to obtain a target object group; the target object group includes at least two target objects, and the at least two target objects belong to different candidate object groups.

[0167] The candidate behavior determination unit 8002 is used to determine each candidate behavior associated with the above-mentioned target object group;

[0168] The first analysis unit 8003 is used to determine the information entropy corresponding to each of the candidate behaviors based on the behavioral characteristics of each target object in the target object group for each candidate behavior. Each information entropy represents the degree of difference in behavioral characteristics of target objects belonging to different candidate object groups for the corresponding candidate behaviors.

[0169] The second analysis unit 8004 is used to determine the target behavior from the candidate behaviors based on the obtained information entropy; wherein, the degree of difference in the behavioral characteristics of the target behavior among the target objects belonging to the different candidate object groups meets the difference degree condition.

[0170] As one embodiment, the first analysis unit 8003 is specifically used for:

[0171] For each of the above behavioral characteristics, the following operations shall be performed:

[0172] Based on the behavioral characteristics of each target object in the above target object group for one of the above candidate behaviors, the range of behavioral characteristics of the above candidate behavior is determined.

[0173] Within the scope of the aforementioned behavioral characteristics, at least one feature segmentation point is determined;

[0174] Determine the reference information entropy corresponding to each feature segmentation point;

[0175] The reference information entropy that meets the set value conditions among the obtained reference information entropies is determined as the information entropy corresponding to the aforementioned behavioral feature.

[0176] As one embodiment, the first analysis unit 8003 is specifically used for:

[0177] Each behavioral feature within the aforementioned range is designated as a corresponding feature segmentation point; or

[0178] Within the aforementioned range of behavioral features, each behavioral feature with an interval equal to the set behavioral feature step size is defined as a corresponding feature segmentation point.

[0179] As one embodiment, the first analysis unit 8003 is specifically used for:

[0180] For each of the above feature segmentation points, the following operations are performed:

[0181] Based on one of the above feature segmentation points, at least two sub-behavioral feature ranges corresponding to the above behavioral feature range are determined.

[0182] Based on the candidate object group to which each target object belongs in each of the above at least two sub-behavioral feature ranges, determine the sub-reference information entropy corresponding to each of the above sub-behavioral feature ranges.

[0183] Based on the sub-reference information entropy corresponding to each of the above sub-behavioral feature ranges, the reference information entropy corresponding to the above feature segmentation point is determined.

[0184] As one embodiment, the first analysis unit 8003 is specifically used for:

[0185] For each of the above sub-behavioral feature ranges, the following operations are performed respectively:

[0186] Based on the behavioral characteristics of each target object for the above candidate behavior, determine each behavioral characteristic within a sub-behavioral characteristic range within the range of each sub-behavioral characteristic, and determine the total number of first objects corresponding to the above sub-behavioral characteristic range based on the number of target objects corresponding to each determined behavioral characteristic.

[0187] Based on the total number of the first objects mentioned above, and the number of target objects belonging to each candidate object group among the target objects corresponding to each determined behavioral feature, the first estimated value of each candidate object group is determined respectively.

[0188] Based on the first estimated value of each of the above candidate object groups, the sub-reference information entropy corresponding to the above sub-behavioral feature range is determined.

[0189] As one embodiment, the first analysis unit 8003 is specifically used for:

[0190] For each of the above candidate object groups, the first operation is performed to determine the first reference value corresponding to each of the above candidate object groups, and based on the determined first reference values, the sub-reference information entropy corresponding to the above sub-behavioral feature range is determined.

[0191] The first operation mentioned above includes:

[0192] Using the first estimated value of one of the candidate object groups as the true number and the preset value as the base, determine the logarithm corresponding to the candidate object group; based on the product of the determined logarithm and the first estimated value of the candidate object group, determine the first reference value corresponding to the candidate object group.

[0193] As one embodiment, the first analysis unit 8003 is specifically used for:

[0194] For each of the above sub-behavioral feature ranges, the second operation is performed to obtain the second reference value corresponding to each of the above sub-behavioral feature ranges, and based on the determined second reference values, the reference information entropy corresponding to the above feature segmentation point is determined.

[0195] The second operation includes: determining a second estimated value for a sub-behavioral feature range based on the total number of second objects in the target object group and the total number of first objects corresponding to a sub-behavioral feature range; and determining a second reference value corresponding to the sub-behavioral feature range based on the product of the second estimated value and the sub-reference information entropy corresponding to the sub-behavioral feature range.

[0196] As one embodiment, the target object determination unit 8001 is specifically used for:

[0197] From each of the candidate object groups contained in the above different candidate object groups, select the set number of candidate objects that meet the requirements of each candidate object group.

[0198] Based on the selected candidate objects, the above-mentioned target object group is obtained.

[0199] As one embodiment, in the set number corresponding to each candidate object group, the difference between the maximum and minimum values ​​is less than the difference threshold, or the ratio of the maximum and minimum values ​​is within a preset ratio range.

[0200] As one embodiment, the second analysis unit 8004 is specifically used to determine the target behavior from the candidate behaviors described above by at least one of the following methods:

[0201] Based on the magnitude of each of the above information entropies, the above information entropies are sorted, and the candidate behaviors corresponding to the information entropies sorted at the target order are determined as the above target behaviors.

[0202] The information entropy that is less than the information entropy threshold among the above information entropy values ​​is identified, and the candidate behavior corresponding to the identified information entropy is determined as the above target behavior.

[0203] As one example, Figure 8The device described above can be used to implement any of the behavior analysis methods for object groups discussed earlier.

[0204] Based on the same inventive concept as the above-described method embodiments, this application also provides a computer device. This computer device can be used for data processing based on pushed content. In one embodiment, the computer device can be a server, such as... Figure 1 The server 120 is shown. In this embodiment, the structure of the computer device can be as follows: Figure 9 As shown, it includes a memory 901, a communication module 903, and one or more processors 902.

[0205] The memory 901 is used to store computer programs executed by the processor 902. The memory 901 may mainly include a program storage area and a data storage area. The program storage area may store the operating system and programs required to run instant messaging functions, etc.; the data storage area may store various instant messaging information and operation instruction sets, etc.

[0206] Memory 901 may be volatile memory, such as random-access memory (RAM); memory 901 may also be non-volatile memory, such as read-only memory, flash memory, hard disk drive (HDD), or solid-state drive (SSD); or memory 901 may be any other medium capable of carrying or storing desired program code having the form of instructions or data structures and accessible by a computer, but is not limited thereto. Memory 901 may be a combination of the above-described memories.

[0207] Processor 902 may include one or more central processing units (CPUs) or digital processing units, etc. Processor 902 is used to implement the behavior analysis method for the aforementioned object group when calling computer programs stored in memory 901.

[0208] The communication module 903 is used to communicate with terminal devices and other servers.

[0209] This application embodiment does not limit the specific connection medium between the memory 901, communication module 903, and processor 902 described above. This application embodiment... Figure 9 The memory 901 and the processor 902 are connected via a bus 904, which is in... Figure 9The connections between other components are shown in thick lines and are for illustrative purposes only, not as limiting information. The 904 bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, Figure 9 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0210] The memory 901 stores a computer storage medium containing computer-executable instructions for implementing the account feature extraction method of this application embodiment. The processor 902 is used to execute the behavior analysis method for the aforementioned object group.

[0211] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0212] Alternatively, if the integrated unit described above is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0213] Based on the same technical concept, embodiments of this application also provide a computer-readable storage medium storing computer instructions that, when executed on a computer, cause the computer to perform the behavior analysis method for object groups as described above.

[0214] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0215] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for analyzing the behavior of a group of objects, characterized in that, include: Obtain the target object group; The target object group includes at least two target objects, and the at least two target objects belong to different candidate object groups; Determine each candidate behavior associated with the target object group; Based on the behavioral characteristics of each target object in the target object group for each candidate behavior, the information entropy corresponding to each candidate behavior is determined. Each information entropy represents the degree of difference in behavioral characteristics of target objects belonging to different candidate object groups for the corresponding candidate behavior. Based on the obtained information entropy, the target behavior is determined from the candidate behaviors, and different operational strategies are set for the different candidate object groups; wherein, the degree of difference in the behavioral characteristics of the target behavior among the target objects belonging to the different candidate object groups meets the difference degree condition; Specifically, based on the behavioral characteristics of each target object in the target object group for each candidate behavior, the information entropy corresponding to each candidate behavior is determined, and the following operations are performed for each behavioral characteristic: Based on the behavioral characteristics of each target object in the target object group for one of the candidate behaviors, the range of behavioral characteristics of the candidate behavior is determined. For the continuous behavioral feature range, at least two values ​​that satisfy the segmentation point selection conditions are selected from the behavioral feature range as feature segmentation points; for the offline behavioral feature range, at least two values ​​are selected as feature segmentation points at preset intervals. For at least one feature segmentation point within the range of the aforementioned behavioral features, the following operations are performed respectively: Based on at least two sub-behavioral feature ranges corresponding to the behavioral feature range determined by the feature segmentation points, and the candidate object group to which each target object belongs corresponding to each sub-behavioral feature range, the sub-reference information entropy corresponding to each of the at least two sub-behavioral feature ranges is determined. Based on the sub-reference information entropy corresponding to each of the at least two sub-behavioral feature ranges, according to the formula Determine the reference information entropy corresponding to the feature segmentation point; where j is the identifier of the candidate behavior; i is the identifier of the sub-behavioral feature range of the behavioral feature range of the candidate behavior identified as j; It is the sub-reference information entropy corresponding to the sub-behavioral feature range of the candidate behavior identified as j, and the sub-reference information entropy corresponding to the sub-behavioral feature range of the candidate behavior identified as j. It is the second estimated value corresponding to the sub-behavioral feature range identified as i within the sub-behavioral feature range corresponding to the behavioral feature range of the candidate behavior identified as j; the second estimated value is determined based on the total number of second objects of the target objects included in the target object group and the total number of first objects corresponding to one sub-behavioral feature range in each sub-behavioral feature range; n is the identifier of the feature segmentation point among at least two feature segmentation points of the behavioral feature range of the candidate behavior identified as j, and N is the total number of feature segmentation points among the at least two feature segmentation points; It is the reference information entropy corresponding to the feature segmentation point identified as n; The reference information entropy that meets the set value conditions among the reference information entropies corresponding to each of the at least one feature segmentation point is determined as the information entropy corresponding to the behavioral feature.

2. The method as described in claim 1, characterized in that, At least one feature segmentation point within the range of the behavioral features is determined using the following method: Each behavioral feature within the aforementioned behavioral feature range is designated as a corresponding feature segmentation point; or Within the range of behavioral features, each behavioral feature with an interval equal to the set behavioral feature step size is determined as a corresponding feature segmentation point.

3. The method as described in claim 1, characterized in that, The step of determining the sub-reference information entropy corresponding to each of the at least two sub-behavioral feature ranges based on the behavioral feature range determined by the feature segmentation points, and the candidate object group to which each target object belongs for each sub-behavioral feature range, includes: For each sub-behavioral feature range, the following operations are performed: Based on the behavioral characteristics of each target object for the candidate behavior, determine each behavioral characteristic within a sub-behavioral characteristic range within each sub-behavioral characteristic range, and determine the total number of first objects corresponding to the sub-behavioral characteristic range based on the number of target objects corresponding to each determined behavioral characteristic. Based on the total number of the first objects and the number of target objects belonging to each candidate object group among the target objects corresponding to each determined behavioral feature, a first estimated value is determined for each candidate object group. Based on the first estimated value of each candidate object group, the sub-reference information entropy corresponding to the sub-behavioral feature range is determined.

4. The method as described in claim 3, characterized in that, The step of determining the sub-reference information entropy corresponding to the sub-behavioral feature range based on the first estimated value of each of the candidate object groups includes: For each candidate object group, a first operation is performed to determine the first reference value corresponding to each candidate object group, and based on the determined first reference values, the sub-reference information entropy corresponding to the sub-behavioral feature range is determined. The first operation includes: Using the first estimated value of one of the candidate object groups as the true number and a preset value as the base, determine the logarithm corresponding to the candidate object group; based on the product of the determined logarithm and the first estimated value of the candidate object group, determine the first reference value corresponding to the candidate object group.

5. The method as described in claim 1, characterized in that, The acquisition of the target object group includes: From each candidate object group contained in the different candidate object groups, select the set number of candidate objects that meet the requirements of each candidate object group. Based on the selected candidate objects, the target object group is obtained.

6. The method as described in claim 5, characterized in that, In each candidate object group, the difference between the maximum and minimum values ​​is less than the difference threshold, or the ratio of the maximum and minimum values ​​is within a preset ratio range.

7. The method according to any one of claims 1-6, characterized in that, The determination of the target behavior from the candidate behaviors based on the obtained information entropy includes at least one of the following methods: Based on the magnitude of each information entropy, the information entropy is sorted, and the candidate behavior corresponding to the information entropy sorted at the target position is determined as the target behavior; The information entropy that is less than the information entropy threshold among the various information entropies is determined, and the candidate behavior corresponding to the determined information entropy is determined as the target behavior.

8. A behavior analysis device for a group of objects, characterized in that, include: Target object determination unit, used to obtain a group of target objects; The target object group includes at least two target objects, and the at least two target objects belong to different candidate object groups; A candidate behavior determination unit is used to determine each candidate behavior associated with the target object group; The first analysis unit is used to determine the information entropy corresponding to each candidate behavior based on the behavioral characteristics of each target object in the target object group for each candidate behavior. Each information entropy represents the degree of difference in behavioral characteristics of target objects belonging to different candidate object groups for the corresponding candidate behavior. The second analysis unit is used to determine the target behavior from the candidate behaviors based on the obtained information entropy, so as to set different operational strategies for the different candidate object groups; wherein, the degree of difference in the behavioral characteristics of the target behavior among the target objects belonging to the different candidate object groups meets the difference degree condition. The first analysis unit is specifically used for: Based on the behavioral characteristics of each target object in the target object group for one of the candidate behaviors, the range of behavioral characteristics of the candidate behavior is determined. For the continuous behavioral feature range, at least two values ​​that satisfy the segmentation point selection conditions are selected from the behavioral feature range as feature segmentation points; for the offline behavioral feature range, at least two values ​​are selected as feature segmentation points at preset intervals. For at least one feature segmentation point within the range of the aforementioned behavioral features, the following operations are performed respectively: Based on at least two sub-behavioral feature ranges corresponding to the behavioral feature range determined by the feature segmentation points, and the candidate object group to which each target object belongs corresponding to each sub-behavioral feature range, the sub-reference information entropy corresponding to each of the at least two sub-behavioral feature ranges is determined. Based on the sub-reference information entropy corresponding to each of the at least two sub-behavioral feature ranges, according to the formula Determine the reference information entropy corresponding to the feature segmentation point; where j is the identifier of the candidate behavior; i is the identifier of the sub-behavioral feature range of the behavioral feature range of the candidate behavior identified as j; It is the sub-reference information entropy corresponding to the sub-behavioral feature range of the candidate behavior identified as j, and the sub-reference information entropy corresponding to the sub-behavioral feature range of the candidate behavior identified as j. It is the second estimated value corresponding to the sub-behavioral feature range identified as i within the sub-behavioral feature range corresponding to the behavioral feature range of the candidate behavior identified as j; the second estimated value is determined based on the total number of second objects of the target objects included in the target object group and the total number of first objects corresponding to one sub-behavioral feature range in each sub-behavioral feature range; n is the identifier of the feature segmentation point among at least two feature segmentation points of the behavioral feature range of the candidate behavior identified as j, and N is the total number of feature segmentation points among the at least two feature segmentation points; It is the reference information entropy corresponding to the feature segmentation point identified as n; The reference information entropy that meets the set value conditions among the reference information entropies corresponding to each of the at least one feature segmentation point is determined as the information entropy corresponding to the behavioral feature.

9. A computer program product, characterized in that, The computer program product includes computer instructions stored in a computer-readable storage medium, a processor of a computer device reading the computer instructions from the computer-readable storage medium, and the processor executing the computer instructions to cause the computer device to perform the method of any one of claims 1-7.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method of any one of claims 1-7.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed on a computer, cause the computer to perform the method as described in any one of claims 1-7.

Images