Data processing method and device, electronic equipment, computer readable storage medium and computer program product

Abstract

The invention provides a data processing method and device, electronic equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: constructing a gain function call library based on a gain look-up table, wherein the gain function call library comprises multiple pieces of type information and a gain function corresponding to each piece of type information; determining a target gain function corresponding to the candidate object from a gain function call library based on the type information of the candidate object; based on the target gain function of each candidate object and the feature data of each candidate object, performing gain prediction on each candidate object to obtain gain data of each candidate object; and determining a target object based on the gain data of each candidate object. By means of the method and device, the accuracy and efficiency of determining the target object in the data processing process can be improved.

Description

Technical Field

[0001] The present application relates to computer technology, and in particular to a data processing method, device, electronic device, computer-readable storage medium, and computer program product. Background Art

[0002] In intelligent marketing scenarios, in order to effectively utilize marketing budgets, it is necessary to accurately measure and predict the incremental interventions brought about by intervening on customers to avoid wasting marketing budgets. Therefore, the screening of target customers who need intervention is the key to intelligent marketing algorithms. By building a gain prediction model, it can help accurately determine the customer's sensitivity to intervention measures, thereby optimizing the revenue simulation calculation and delivery strategy, and maximizing the efficiency of intervention promotion. However, the current gain prediction model cannot be directly based on the intervention increment model, and the model building process is complicated, the implementation cost is high, and the prediction accuracy is low. Summary of the invention

[0003] The embodiments of the present application provide a data processing method, device, electronic device, computer-readable storage medium, and computer program product, which can improve the accuracy and efficiency of determining a target object.

[0004] The technical solution of the embodiment of the present application is implemented as follows:

[0005] The present application provides a data processing method, the method comprising:

[0006] Building a gain function call library based on the gain query table, wherein the gain function call library includes a plurality of type information and a gain function corresponding to each type information;

[0007] Based on the type information of the candidate object, determining a target gain function corresponding to the candidate object from the gain function call library;

[0008] Based on the target gain function of each candidate object and the feature data of each candidate object, performing gain prediction on each candidate object to obtain gain data of each candidate object;

[0009] Based on the gain data of each of the candidate objects, a target object is determined.

[0010] The present application provides a data processing device, including:

[0011] A function call library construction module, used to construct a gain function call library based on the gain query table, wherein the gain function call library includes a plurality of type information and a gain function corresponding to each type information;

[0012] An objective function determination module, configured to determine an objective gain function corresponding to the candidate object from the gain function call library based on the type information of the candidate object;

[0013] A gain prediction module, configured to perform gain prediction on each candidate object based on a target gain function of each candidate object and feature data of each candidate object, so as to obtain gain data of each candidate object;

[0014] The object determination module is used to determine the target object based on the gain data of each candidate object.

[0015] An embodiment of the present application provides an electronic device, the electronic device comprising:

[0016] A memory for storing computer executable instructions;

[0017] The processor is used to implement the data processing method provided in the embodiment of the present application when executing the computer executable instructions stored in the memory.

[0018] An embodiment of the present application provides a computer-readable storage medium storing a computer program or computer-executable instructions for implementing the data processing method provided in the embodiment of the present application when executed by a processor.

[0019] An embodiment of the present application provides a computer program product, including a computer program or computer executable instructions. When the computer program or computer executable instructions are executed by a processor, the data processing method provided in the embodiment of the present application is implemented.

[0020] The embodiments of the present application have the following beneficial effects:

[0021] The embodiment of the present application first constructs a gain function call library based on the gain query table. When determining the target object, based on the type information of the candidate object, the target gain function corresponding to the candidate object is determined from the gain function call library, and the gain prediction is performed on each candidate object through the target gain function of each candidate object and the feature data of each candidate object, and the gain data of each candidate object is obtained, thereby determining the target object. In this way, the gain data of each candidate object can be accurately predicted through the target gain function, and the gain data of the candidate object can be determined more simply and efficiently by directly performing gain prediction through the feature data of each candidate object. Therefore, the gain data of the candidate object can be accurately and efficiently determined through the embodiment of the present application, thereby improving the accuracy and efficiency of determining the target object based on the gain data. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a schematic diagram of the architecture of a data processing system 100 provided in an embodiment of the present application;

[0023] Figure 2 is a schematic diagram of the structure of the server 200 provided in an embodiment of the present application;

[0024] Figure 3A It is a schematic diagram of the implementation flow of the data processing method provided in the embodiment of the present application;

[0025] Figure 3B It is a schematic diagram of the implementation process of constructing a gain query table provided in an embodiment of the present application;

[0026] Figure 3C It is a schematic diagram of the implementation process of constructing a gain function call library provided in an embodiment of the present application;

[0027] Figure 3D It is a schematic diagram of the implementation flow of the gain prediction model training provided in the embodiment of the present application;

[0028] Figure 4 It is a schematic diagram of the implementation process of determining the target object provided in an embodiment of the present application. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0030] In the following description, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0031] In the following description, the terms "first\second\third" involved are merely used to distinguish similar objects and do not represent a specific ordering of the objects. It can be understood that "first\second\third" can be interchanged with a specific order or sequence where permitted, so that the embodiments of the present application described here can be implemented in an order other than that illustrated or described here.

[0032] In the embodiments of the present application, the term "module" or "unit" refers to a computer program or a part of a computer program with a predetermined function, and works together with other related parts to achieve a predetermined goal, and can be implemented in whole or in part by using software, hardware (such as processing circuits or memories) or a combination thereof. Similarly, a processor (or multiple processors or memories) can be used to implement one or more modules or units. In addition, each module or unit can be part of an overall module or unit that includes the function of the module or unit.

[0033] Unless otherwise defined, all technical and scientific terms used in the embodiments of the present application have the same meanings as those commonly understood by those skilled in the art. The terms used in the embodiments of the present application are only for the purpose of describing the embodiments of the present application and are not intended to limit the present application.

[0034] The relevant data collection and processing in the embodiments of this application should be strictly in accordance with the requirements of relevant national laws and regulations when applied in examples, obtain the informed consent or separate consent of the personal information subject, and carry out subsequent data use and processing within the scope of authorization of laws and regulations and the personal information subject.

[0035] Before further describing the embodiments of the present application in detail, the nouns and terms involved in the embodiments of the present application are explained. The nouns and terms involved in the embodiments of the present application are subject to the following interpretations.

[0036] 1) Uplift Model: A model used to predict the intervention increment (gain data) caused by the impact of intervention actions on user response behavior.

[0037] 2) Treatment: Specific actions and strategies taken to influence consumers’ purchasing decisions and behaviors. For example, promotional activities (product discounts, additional gifts, coupons), advertising (broadcasting, online advertising), product recommendations (providing product links, jumping to recommendation pages), etc.

[0038] 3) User response behavior (outcome): consumer demand, purchase, use and feedback on products or services.

[0039] 4) Propensity Score: Propensity data used to assess the probability of a certain event occurring to a user.

[0040] 5) Uplift: It is one of the important indicators to measure marketing effectiveness and results. It represents the improvement in purchasing behavior, brand loyalty, market share, etc. after the user's response behavior changes under the influence of marketing activities or strategies.

[0041] 6) Group tendency data: used to assess the probability of an event occurring in a specific user group.

[0042] 7) Group gain data: It is one of the important indicators to measure the marketing effect and results for a specific user group. It represents the improvement in the purchasing behavior, brand loyalty, market share, etc. of the entire specific user group after the user response behavior changes under the influence of marketing activities or strategies.

[0043] 8) Gain query table: a table or database used to record or query gain data corresponding to different tendency data.

[0044] 9) Gain function call library: provides a series of gain functions for different application scenarios, which are used to call the corresponding gain functions in different scenarios to calculate the corresponding gain data.

[0045] 10) Mode: refers to the value that appears most frequently in a set of data, used to describe the central tendency or central value in a data set.

[0046] In order to better understand the data processing method provided in the embodiments of the present application, the data processing method in the related art and its existing shortcomings are first described.

[0047] In actual marketing scenarios, merchants can reach customers through various channels, such as advertising push, new customer acquisition activities, or targeted coupon distribution, but the costs need to be strictly controlled. Therefore, the core goal of marketing is to maximize the total output under a limited intervention budget, and it is crucial to accurately identify sensitive groups that can be affected by intervention measures, namely target objects. In related technologies, the gain prediction model can be used to predict the causal effect of a certain intervention on an individual's state or behavior. The gain prediction model can be mathematically expressed as the difference between two conditional probabilities, as shown in formula (1):

[0048] p(Y i |X i ,T i =1)-p(Y i |X i ,T i =0) formula (1);

[0049] Among them, i represents the i-th user, Y represents the outcome (such as user clicks, conversions, etc.), X represents user characteristics, and T represents the intervention variable treatment (1 represents intervention, 0 represents no intervention). In related technologies, gain prediction models can be divided into the following three types:

[0050] 1) Two-Model: Two prediction models for user response behavior are established, one using the experimental group data and the other using the control group data. However, the Two-Model model does not directly model the gain data, and may miss some gain data-related signals, resulting in poor performance;

[0051] 2) Single-Model: By building a single model, the intervention action variables are used as input features for training. However, the Single-Model model has large data requirements, high model complexity, and difficult evaluation.

[0052] 3) Direct modeling of gain data: By modifying the existing machine learning algorithm, the gain effect is directly modeled. However, in actual applications, direct modeling of gain data requires a lot of modification and optimization of the machine learning model, which has a high implementation cost.

[0053] Based on the above problems, the embodiments of the present application provide a data processing method, device, electronic device, computer-readable storage medium and computer program product, which can quickly and accurately predict gain, thereby improving the accuracy and efficiency of determining the target object. The following describes an exemplary application of the electronic device provided by the embodiments of the present application. The electronic device provided by the embodiments of the present application can be implemented as various types of terminals such as laptops, tablet computers, desktop computers, set-top boxes, smart phones, smart speakers, smart watches, smart TVs, and vehicle-mounted terminals, and can also be implemented as servers. Below, an exemplary application when the device is implemented as a server will be described.

[0054] See also Figure 1 , Figure 1 It is a schematic diagram of the architecture of the data processing system 100 provided in an embodiment of the present application. The terminal 400 is connected to the server 200 via the network 300. The network 300 may be a wide area network or a local area network, or a combination of the two.

[0055] The terminal 400 is used to display intervention information for the target object on the graphical interface 410, such as marketing services customized according to factors such as the specific market environment, product characteristics, and target object characteristics. The server 200 is used to construct a gain query table based on group tendency data and group gain data, and then construct a gain function call library based on each type of information and the gain function corresponding to each type of information. Then, based on the type information of the candidate object, the target gain function corresponding to the candidate object is determined from the gain function call library. Afterwards, based on the target gain function of each candidate object and the characteristic data of each candidate object, a gain prediction is performed on each candidate object to obtain the gain data of each candidate object, thereby determining the target object from multiple candidate objects based on the gain data of each candidate object, so as to send intervention information to the target object and realize targeted marketing services for the target object.

[0056] In some embodiments, the server 200 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms. The terminal and the server may be directly or indirectly connected via wired or wireless communication, which is not limited in the embodiments of the present application.

[0057] See also Figure 2 , Figure 2 is a schematic diagram of the structure of the server 200 provided in an embodiment of the present application, Figure 2 The server 200 shown includes: at least one processor 210, a memory 230 and at least one network interface 220. The various components in the server 200 are coupled together via a bus system 240. It is understood that the bus system 240 is used to achieve connection and communication between these components. In addition to the data bus, the bus system 240 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, the bus system 240 is not described in detail. Figure 2 Various buses are labeled as bus system 240 .

[0058] The processor 210 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., where the general-purpose processor can be a microprocessor or any conventional processor, etc.

[0059] The memory 230 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state memory, hard disk drives, optical disk drives, etc. The memory 230 may optionally include one or more storage devices that are physically remote from the processor 210.

[0060] The memory 230 includes a volatile memory or a non-volatile memory, and may also include both volatile and non-volatile memories. The non-volatile memory may be a read-only memory (ROM), and the volatile memory may be a random access memory (RAM). The memory 230 described in the embodiment of the present application is intended to include any suitable type of memory.

[0061] In some embodiments, memory 230 can store data to support various operations, examples of which include programs, modules, and data structures, or a subset or superset thereof, as exemplarily described below.

[0062] Operating system 231, including system programs for processing various basic system services and performing hardware-related tasks, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and processing hardware-based tasks;

[0063] The network communication module 232 is used to reach other electronic devices via one or more (wired or wireless) network interfaces 220. Exemplary network interfaces 220 include: Bluetooth, Wireless Authentication (WiFi), and Universal Serial Bus (USB).

[0064] In some embodiments, the device provided in the embodiments of the present application can be implemented in software. Figure 2 The data processing device 233 stored in the memory 230 is shown, which can be software in the form of a program and a plug-in, etc., and includes the following software modules: a function call library construction module 2331, a target function determination module 2332, a gain prediction module 2333 and an object determination module 2334. These modules are logical, and therefore can be arbitrarily combined or further split according to the functions implemented. The functions of each module will be described below.

[0065] In other embodiments, the device provided in the embodiments of the present application can be implemented in hardware. As an example, the device provided in the embodiments of the present application can be a processor in the form of a hardware decoding processor, which is programmed to execute the data processing method provided in the embodiments of the present application. For example, the processor in the form of a hardware decoding processor can adopt one or more application-specific integrated circuits (Application Specific Integ rated Circuit, ASIC), digital signal processors (Digital Signal Processor, DSP), programmable logic devices (Programmable Logic Device, PLD), complex programmable logic devices (Complex Programmable Logic Device, CPLD), field programmable gate arrays (Field-Progr ammable Gate Array, FPGA) or other electronic components.

[0066] The data processing method provided in the embodiment of the present application will be explained in combination with the exemplary application and implementation of the server provided in the embodiment of the present application.

[0067] The following describes the data processing method provided in the embodiment of the present application. As mentioned above, the electronic device that implements the data processing method in the embodiment of the present application may be a server, so the execution subject of each step will not be repeatedly described below.

[0068] It should be noted that in the data processing examples below, the object is consumption data as an example. Based on their understanding of the following text, those skilled in the art can apply the data processing method provided in the embodiments of the present application to data processing including other types of objects.

[0069] See also Figure 3A , Figure 3A is a schematic diagram of the implementation flow of the data processing method provided in the embodiment of the present application, which will be combined with Figure 3A The steps shown are explained.

[0070] In step 101, a gain function call library is constructed based on a gain query table.

[0071] In the embodiment of the present application, the gain function call library includes multiple types of information and a gain function corresponding to each type of information.

[0072] Here, the gain query table includes group tendency data and group gain data corresponding to multiple object sets, and the gain query table includes group tendency data and group gain data corresponding to multiple object sets of the same type of information. Each object set includes multiple sample objects, and the group tendency data is used to estimate the probability of a certain behavior occurring as a whole among the multiple sample objects included in this object set. For example, the group tendency data can characterize the probability that the sample objects in this object set purchase a specified product, wherein the higher the group tendency data, the greater the possibility that this object set will purchase the specified product. The group gain data can also be expressed as a group intervention increment, which is used to estimate the positive impact or benefits of the intervention behavior on the multiple sample objects included in this object set as a whole. For example, the group gain data can refer to the increase in product sales by providing product recommendation services to this object set.

[0073] In some embodiments, the gain function call library includes gain functions corresponding to different types of information, and the gain function is determined according to the group tendency data and group gain data corresponding to the multiple object sets included in the gain query table. When a plurality of gain query tables corresponding to different types of information are obtained, a gain function can be constructed according to the gain query table of each type of information, and then a gain function call library is constructed according to the plurality of gain functions.

[0074] In some embodiments, before executing the above step 101, a gain query table may be constructed first, see Figure 3B , a gain query table can be constructed through steps 201 to 204, as described in detail below:

[0075] In step 201, feature data of a plurality of sample objects corresponding to preset type information is obtained.

[0076] Here, the preset type information may be one of the types of information such as long-tail type, balance type, potential type, high-value type, loyal type, and churn type. Feature data may include: basic information, consumer behavior information (purchase frequency, promotion sensitivity, consumption amount, purchase channel), preference information (service requirements, special preferences), interaction history information (interaction with related products, click links, open emails, leave messages, etc.), etc. For example, basic information, consumer behavior information, preference information, and interaction history information of multiple sample objects of potential types are obtained.

[0077] In step 202, a plurality of sample objects are divided into a plurality of object sets.

[0078] Here, the sample object can be predicted based on the target tendency prediction model and the characteristic data of the sample object to obtain the tendency data of the sample object; then the sample objects are sorted in descending order based on the tendency data of the sample object to obtain the sorted sample objects; finally, the sorted sample objects are divided to obtain multiple object sets. Among them, the target tendency prediction model is a model trained in advance for predicting tendency data. The characteristic data of the sample object is input into the target tendency prediction model, and the tendency data corresponding to the sample object can be output by performing tendency prediction through the target tendency prediction model.

[0079] Among them, the sorted sample objects can be equally divided into segments by the Propensity Score Matching (PSM) technology to form N homogeneous groups as multiple object sets, each of which consists of sample objects that have been intervened and sample objects that have not been intervened. Among them, the size of each object set can be determined by the propensity score interval or the number of quantiles. For example, there are 6 sample objects 1, 2, 3, 4, 5, 6, and the corresponding propensity data are: 0, 0.03, 0.01, 0.04, 0.07, 0.09. Then the sorted sample objects are: 6, 5, 4, 2, 3, 1. By dividing the propensity data interval [0, 0.05), [0.05, 0.1), the sample objects 6, 5, 4, 2 are one object set, and the sample objects 3, 1 are another object set. Alternatively, the partitioning process is performed using the quantile numbers 3 and 6, so that the first three sample objects 6, 5, and 4 of the sorted sample objects are one object set, and the third to sixth sample objects 2, 3, and 1 of the sorted sample objects are another object set.

[0080] In step 203, based on the feature data of the sample objects included in each object set, the group tendency data and the group gain data of each object set are determined.

[0081] Here, firstly, for each object set, the group tendency data of the object set is determined based on the tendency data of the sample objects included in the object set; then, the sample objects are divided into a first object subset and a second object subset, and a first value of a preset feature corresponding to the first object subset is determined based on the feature data of the first object subset, and a second value of a preset feature corresponding to the second object subset is determined based on the feature data of the second object subset; finally, the group gain data of the object set is determined based on the first value and the second value. The first object subset includes sample objects that have received intervention information, and the second object subset includes sample objects that have not received intervention information.

[0082] Among them, the average value or mode of the tendency data of multiple sample objects can be used as the group tendency data of the object set. For example, the object set includes four sample objects, and the corresponding tendency data are: 0.02, 0.02, 0.04, and 0.08. It can be calculated that the average value of the tendency data of multiple sample objects is 0.04, so 0.04 can be used as the group tendency data of the object set. Alternatively, the mode of the tendency data of multiple sample objects, 0.02, is used as the group tendency data of the object set. Specifically, according to the actual application scenario, the data that can better reflect the overall tendency of the object set can be used as the group tendency data.

[0083] The sample objects include sample objects that have received intervention information and sample objects that have not received intervention information. The sample objects that have received intervention information can be understood as objects affected by the intervention action, such as users who have received recommendation information about a specified product; the sample objects that have not received intervention information can be understood as objects that have not been affected by the intervention action, such as users who have not received recommendation information about a specified product. The sample objects that have received intervention information are divided into a first object subset, and the sample objects that have not received intervention information are divided into a second object subset, so that the sample objects are divided into a first object subset and a second object subset. The preset feature can be a certain type of feature data in the feature data, such as the number of purchases. The first value of the preset feature corresponding to the first object subset represents the number of purchases of the product by the object in the first object subset when the intervention information is received, such as 3; the second value of the preset feature corresponding to the second object subset represents the number of purchases of the product by the object in the second object subset when the intervention information is not received, such as 1. The group gain data of the object set can be the difference between the first value and the second value, for example, the difference 2 is obtained by subtracting the first value 3 from the second value 1, and then the difference 2 is used as the group gain data of the object set.

[0084] In step 204, a gain query table corresponding to the preset type information is constructed based on the group tendency data and the group gain data of each object set.

[0085] Here, the gain query table includes group tendency data and group gain data, wherein the group tendency data and the group gain data correspond one to one. For example, a blank table including two columns can be first established, and then the group tendency data can be sequentially imported into one of the columns, and then the group gain data corresponding to each group tendency data can be sequentially imported into the other column to construct a gain query table. There is a corresponding relationship between the group tendency data and the group gain data in the same row.

[0086] In the embodiment of the present application, characteristic data of multiple sample objects corresponding to preset type information are obtained, multiple sample objects are divided and processed to obtain multiple object sets, and then for each object set, the group tendency data and group gain data of the object set are determined based on the characteristic data of the sample objects included in the object set, and finally a gain query table corresponding to the preset type information is constructed based on the group tendency data and group gain data of each object set. In this way, the gain data corresponding to sample objects with different tendency data can be quickly determined according to the gain query table, and the specific performance of sample objects with different types of information when receiving intervention information and not receiving intervention information can be determined, so as to more accurately and comprehensively understand the gain data of sample objects.

[0087] It should be noted that the above steps 201 to 204 are described by taking a certain type of information as the preset type information as an example. For multiple sample objects corresponding to other types of information, a gain query table corresponding to other types of information can also be constructed through similar steps from steps 201 to 204.

[0088] In some embodiments, see Figure 3C Step 101 "building a gain function call library based on the gain query table" can be implemented through steps 1011 to 1013, which are described in detail below:

[0089] In step 1011, each type of information and the corresponding gain query table are obtained.

[0090] Here, different types of information involve objects with different characteristics and behaviors, and can be classified and divided according to different standards to obtain multiple types of information, such as long-tail type, balance type, potential type, high-value type, loyal type, churn type, etc. Among them, one type of information corresponds to one gain query table, and multiple types of information correspond to multiple gain query tables.

[0091] In step 1012, for each gain lookup table, a gain function between the group tendency data and the group gain data is determined.

[0092] In some embodiments, for each gain query table, a mapping relationship between group tendency data and group gain data is determined, and the mapping relationship is used to establish a corresponding relationship between group tendency data and group gain data. Nonlinear function fitting can be used to establish the mapping relationship between group tendency data and group gain data. When the number of gain query tables is large, a deep neural network can also be used to learn the mapping relationship, so as to determine the gain function between group tendency data and group gain data.

[0093] Here, the gain function is used to determine the group gain data based on the group tendency data, and can describe the influence of the group tendency data on the group gain data. According to the mapping relationship, the corresponding relationship between the group tendency data and the group gain data can be determined, and then the group gain data is used as the output and the group tendency data is used as the input to construct the corresponding gain function. For example, there is a linear relationship between the group tendency data and the group gain data, and a suitable linear mathematical function can be selected to construct the gain function. At the same time, the parameters in the gain function can be fitted through data analysis or a machine learning algorithm to complete the construction of the gain function.

[0094] In step 1013, a gain function call library is constructed based on each type information and the gain function corresponding to each type information.

[0095] Here, one type of information corresponds to one gain function, and different gain functions can be fitted based on each type of information, thereby constructing a gain function call library. When performing gain prediction, the corresponding gain function can be selected from the gain function call library according to the type information for gain prediction.

[0096] In the embodiment of the present application, a gain query table corresponding to a plurality of type information is obtained, a mapping relationship between group tendency data and group gain data is determined for each gain query table, a gain function between group tendency data and group gain data is determined based on the mapping relationship, and finally a gain function call library is constructed based on each type information and the gain function corresponding to each type information. In this way, the corresponding gain function is fitted according to different type information, and when performing gain prediction, the corresponding gain function is selected in the gain function call library according to the type information for gain prediction, which can improve the accuracy of gain prediction.

[0097] Continue to see Figure 3A , and the description continues with step 101 above.

[0098] In step 102, based on the type information of the candidate object, a target gain function corresponding to the candidate object is determined from a gain function call library.

[0099] The target gain function is used to characterize the mapping relationship between the tendency data and gain data of the candidate object corresponding to the type information.

[0100] Here, the candidate objects are multiple objects of the same type of information, and it is necessary to determine the object that needs to be intervened, that is, the target object, from the multiple candidate objects. The feature data of the candidate objects is used to describe and distinguish the key features (features related to the current marketing scenario) and information of different objects, so as to better understand the users and formulate personalized marketing strategies. For example, the feature data of the candidate objects may include: basic information, consumption behavior information (purchase frequency, promotion sensitivity, consumption amount, purchase channel), preference information (service requirements, special preferences), interaction history information (interaction with related products, click connection, open email, message feedback, etc.) and other information. The type information includes one type of information from multiple types such as long tail type, balance type, potential type, high value type, loyal type, and loss type. For example, if the type information of multiple candidate objects is a long tail type, the gain function corresponding to the long tail type is determined from the gain function call library as the target gain function. Among them, the target gain function can characterize the mapping relationship between the group tendency data and the group gain data of the object set of the long tail type, so as to characterize the mapping relationship between the tendency data and the gain data of multiple candidate objects.

[0101] In step 103, based on the target gain function of each candidate object and the feature data of each candidate object, gain prediction is performed on each candidate object to obtain gain data of each candidate object.

[0102] Here, each of the multiple candidate objects has corresponding feature data, and gain prediction is performed using the target gain function of each candidate object and the feature data of each candidate object to obtain the gain data of each candidate object.

[0103] In some embodiments, there are two implementations of step 103. The first implementation is to perform gain prediction based on the target gain function and feature data. When implementing, the target tendency prediction model can be used to first perform tendency prediction on the feature data of each candidate object to obtain tendency data corresponding to each candidate object; then, for the tendency data corresponding to each candidate object, gain prediction is performed on each candidate object based on the target gain function to obtain the gain data of the candidate object.

[0104] Here, the target tendency prediction model is used to predict the tendency data of the candidate object. The characteristic data of the candidate object is input into the target tendency prediction model. The tendency data corresponding to the object can be output by the target tendency prediction model. Then, the tendency data of the candidate object is used as input to calculate through the target gain function to obtain the gain data of the candidate object. In this way, the tendency data of the candidate object is first determined by the target tendency prediction model, and then the tendency data of the candidate object is used as the input of the target gain function for calculation. The output calculation result is the gain data of the candidate object. The accuracy of the prediction gain result is improved by the target tendency prediction model and the target gain function.

[0105] In some embodiments, the target tendency prediction model can be obtained through the following process: obtaining feature data of multiple sample objects corresponding to each type of information, and constructing a tendency training data set based on the feature data of multiple sample objects corresponding to the type information; then, performing tendency prediction training on the initial tendency prediction model based on the tendency training data set to obtain multiple candidate tendency prediction models; finally, based on the tendency prediction accuracy of multiple candidate tendency prediction models, determining the target tendency prediction model from multiple candidate tendency prediction models.

[0106] Here, the sample object can be a randomly selected individual or group, including sample objects of various types of information. The characteristic data of the sample object is used to describe and distinguish the key characteristics and information of different users, and may include: basic information, consumption behavior information (purchase frequency, promotion sensitivity, consumption amount, purchase channel), preference information (service demand, special preferences), interaction history information (interaction with related products, click links, open emails, message feedback, etc.) and other information. The initial tendency prediction model is a basic tendency prediction model obtained by simple model construction, which has low performance and generalization ability. The initial tendency prediction model can be different types of models, such as linear regression model, logistic regression model, etc. Based on the tendency training data set, the characteristic data of multiple sample objects are input into the initial tendency prediction model, so that the initial tendency prediction model performs tendency prediction based on the characteristic data, obtains the predicted tendency value corresponding to each sample object, and then determines the loss value through the preset loss function, the predicted tendency value and the real tendency data of multiple sample objects, and back-propagates the loss value to the initial tendency prediction model. Based on the loss value, the parameters of the initial tendency prediction model are adjusted by the gradient descent algorithm until the training end condition is reached, and multiple candidate tendency prediction models are obtained. Then, the actual tendency data of multiple sample objects are compared with the tendency data output by multiple candidate tendency prediction models to determine the tendency prediction accuracy of multiple candidate tendency prediction models. Finally, the candidate tendency prediction model with the highest tendency prediction accuracy is used as the target tendency prediction model, or the candidate tendency prediction model with the highest tendency prediction accuracy and reaching a preset accuracy is used as the target tendency prediction model. In this way, through the characteristic data of multiple sample objects corresponding to various types of information, the tendency prediction training of multiple candidate tendency prediction models is carried out, so as to determine the target tendency prediction model, which can ensure that the tendency prediction accuracy of the target tendency prediction model is higher.

[0107] Another implementation of step 103 is to perform gain prediction using only feature data. During the implementation, a trained gain prediction model can be directly obtained, and gain prediction is performed on each candidate object based on the feature data using the trained gain prediction model to obtain the gain data of the candidate object.

[0108] Here, the trained gain prediction model is a model based on gain data and feature data, and is used to predict the gain data of the candidate object. The gain data of the candidate object is input into the trained gain prediction model, and the gain prediction is performed through the trained gain prediction model to obtain the gain data corresponding to the object. In this way, the feature data is directly input into the trained gain prediction model to obtain the gain data, which can more conveniently and quickly determine the gain data of the candidate object.

[0109] In some embodiments, see Figure 3D, the trained gain prediction model can be obtained through steps 301 to 304, as described in detail below:

[0110] In step 301, feature data and target tendency prediction models of multiple sample objects corresponding to multiple types of information are obtained, and tendency prediction is performed on the feature data of the multiple sample objects based on the target tendency prediction model to obtain tendency data of the multiple sample objects.

[0111] Here, the sample object can be a randomly selected individual or group, including sample objects of various types of information. The characteristic data of the sample object is used to describe and distinguish the key characteristics and information of different users, and may include: basic information, consumption behavior information (purchase frequency, number of purchases, promotion sensitivity, consumption amount, purchase channel), preference information (service needs, special preferences), interaction history information (interaction with related products, click links, open emails, leave messages, feedback, etc.) and other information. The tendency data of the sample object is used to characterize the probability of the sample object performing a certain behavior, such as the probability of the sample object purchasing a specified product. The characteristic data of multiple sample objects are respectively input into the target tendency prediction model for tendency prediction, and the corresponding tendency data of multiple sample objects are obtained.

[0112] In step 302, for each type of information, a gain function corresponding to the type information is determined, and based on the gain function corresponding to the type information and the tendency data of the multiple sample objects corresponding to the type information, gain data of the multiple sample objects are determined.

[0113] Here, since the obtained multiple sample objects include multiple types of information, it is necessary to determine the corresponding gain function for each type of information. Gain prediction is performed on multiple sample objects corresponding to this type according to the gain function corresponding to each type of information. Specifically, for each sample object among the multiple sample objects corresponding to this type, the tendency data of the sample object is used as the input of the gain function corresponding to this type to calculate and obtain the gain data of this sample object. When the gain prediction for each sample object is completed, the gain data of multiple sample objects can be obtained.

[0114] In step 303, a gain training data set is constructed based on the feature data of the plurality of sample objects and the gain data of the plurality of sample objects.

[0115] Here, the gain training data set refers to a data set used to train a gain prediction model, which includes: feature data of multiple sample objects and gain data of multiple sample objects.

[0116] In step 304, gain prediction training is performed on the initial gain prediction model based on the gain training data set to obtain a trained gain prediction model.

[0117] Here, the initial gain prediction model is a basic gain prediction model obtained through simple model construction, which has low performance and generalization ability. Based on the gain training data set, the feature data of multiple sample objects are used as input data, and the gain data of multiple sample objects are used as output labels. The initial gain prediction model is trained for gain prediction. During the training process, the parameters of the model are continuously adjusted and optimized according to the output data of the gain prediction, so as to obtain a trained gain prediction model. Among them, the trained gain prediction model has better performance and higher gain prediction accuracy because it has been fully trained and tuned.

[0118] In an embodiment of the present application, feature data and a target tendency prediction model of multiple sample objects corresponding to multiple types of information are obtained, and the characteristic data of the multiple sample objects are predicted by the target tendency prediction model to obtain the tendency data of the multiple sample objects. Then, a gain function corresponding to each type of information is determined, and the gain data of the multiple sample objects is determined based on the gain function and the tendency data. Finally, a gain training data set is constructed based on the feature data of the multiple sample objects and the gain data of the multiple sample objects, and the initial gain prediction model is trained for gain prediction to obtain a trained gain prediction model. In this way, the gain data of sample objects of multiple types of information are obtained using the gain function, and then modeling is directly performed based on the feature data and the gain data, thereby improving the efficiency of gain prediction modeling and the accuracy of gain prediction.

[0119] Continue to see Figure 3A , and the description continues with step 103 above.

[0120] In step 104, a target object is determined based on the gain data of each candidate object.

[0121] Here, the target object is the object that is about to receive the intervention information for intervention. Based on the gain data of each candidate object, the target gain data within the preset gain range can be determined, and the candidate object corresponding to the target gain data can be determined as the target object. Among them, the preset gain range can be understood as the expected gain range of the product or service set in advance when conducting marketing or sales activities. The preset gain range can be formulated according to factors such as market demand, target customer group, competition situation and business goals. For example, if the purchasing power of the target customer group is strong and the corresponding gain data is generally high, the upper and lower limits of the preset gain range can be adjusted higher; or, if the current market demand for the specified product is weak and the generated gain data is not high, the upper and lower limits of the preset gain range can be lowered.

[0122] In an embodiment of the present application, a gain query table is first generated by the group tendency data and group gain data corresponding to a plurality of object sets, and then a gain function call library is constructed according to the gain query table. When a target object is determined from a plurality of candidate objects, a target gain function corresponding to the type information of the plurality of candidate objects is determined from the gain function call library, and the mapping relationship between the group tendency data and the group gain data of the object set corresponding to the type information of the plurality of candidate objects can be more accurately determined by the target gain function. Afterwards, the candidate objects are subjected to gain prediction by at least one of the target gain function and the feature data to obtain the gain data of the candidate objects, and then the target object is determined from the plurality of candidate objects. In this way, the gain data of the plurality of candidate objects can be accurately predicted by the target gain function, and the gain data of the candidate objects can be determined more simply and efficiently by directly performing gain prediction by the feature data of the candidate objects. Therefore, the gain data of the candidate objects can be accurately and efficiently determined by the embodiment of the present application, thereby improving the accuracy and efficiency of determining the target object according to the gain data.

[0123] The data processing method provided in the embodiment of the present application can be applied to marketing activities, and can help merchants estimate the effects of specific activities, such as advertising, promotional activities, or social media marketing, through gain prediction, so as to optimize marketing budgets and improve marketing efficiency. Or in some platforms, gain prediction can be used to recommend products or content that may be most attractive to users, and personalized recommendation services can be provided based on user historical behaviors and preferences.

[0124] The data processing method provided in the embodiment of the present application can also be applied to the financial field, helping users evaluate the expected returns of different investment projects through gain prediction, and assisting investors in deciding how to allocate funds. By providing predictions of future gains, users can be helped to optimize the decision-making process and improve the accuracy and effectiveness of decisions.

[0125] The following is an explanation of an exemplary application of the embodiments of the present application in a practical application scenario.

[0126] The present application embodiment provides a data processing method for determining target objects in a marketing scenario. Figure 4 , is a schematic diagram of the implementation process of determining the target object provided in the embodiment of the present application, which is described in detail below:

[0127] The first stage generates a gain query table, which specifically includes steps 401 and 402, which are described in detail below:

[0128] Step 401: perform tendency prediction and classification processing on sample objects.

[0129] Here, we first need to determine the target tendency prediction model. Collect feature data of multiple sample objects corresponding to various types of information from the big data analysis platform (feature data may include: user basic information characteristics, APP usage, similar marketing product purchase data, etc.) to build a tendency training data set, and then use the tendency training data set to train multiple initial tendency prediction models to obtain multiple candidate tendency prediction models, and then select the model with the best tendency prediction performance from the multiple candidate tendency prediction models as the target tendency prediction model.

[0130] Then, in the tendency training data set, characteristic data of multiple sample objects corresponding to preset type information (such as long-tail users) are obtained, and the target tendency prediction model is used to perform tendency prediction on the characteristic data of multiple sample objects corresponding to the preset type information to obtain the tendency data of the sample objects.

[0131] Finally, according to the propensity data of the sample objects, the sample objects are arranged in descending order, and then the propensity score matching technology is used to divide the sorted sample objects into equal intervals to form N homogeneous groups as multiple object sets. The size of each object set can be determined by the propensity data interval (such as [0, 0.05), [0.05, 0.1), etc.) or the number of quantiles (such as 50, 100, etc.).

[0132] Step 402: construct a gain lookup table.

[0133] Here, each object set is composed of intervened sample objects and unintervented sample objects. Therefore, the group tendency data of each object set can be determined according to the tendency data of the sample objects in each object set (for example, the average value, mode, etc. of the tendency data of multiple sample objects are used as the group tendency data), and the group gain data of each object set can be determined according to the gain data of the sample objects in each object set (for example, the click-through rate gain, purchase rate gain, etc. in the characteristic data of the sample objects are determined as the group gain data), and then a gain query table is constructed based on the group tendency data and the group gain data.

[0134] In the first stage, the target tendency prediction model obtained can accurately predict the tendency data of users with different characteristics, has a certain generalization ability, and can ensure the accuracy of tendency prediction in different situations.

[0135] The second stage includes step 403, which is described in detail below:

[0136] Step 403: construct a gain function.

[0137] Here, according to the generated gain query table, a nonlinear function is used to fit the mapping relationship between the group tendency data and the group gain data. When the magnitude of N is large enough, a deep neural network can also be used to learn the mapping relationship, thereby improving the robustness and generalization ability of the gain prediction model. Among them, different gain functions can be fitted for object sets of different types of information, and different gain functions can be constructed as a gain function call library. Therefore, in actual use, the corresponding gain function can be selected according to the actual use scenario to improve the accuracy of gain prediction.

[0138] In the second stage, since different partitioning processes will result in object sets of different sizes, small-scale object sets can better reflect individual performance but may cause group uplift distortion. Large-scale object sets can obtain more reliable uplift but may deviate from individual performance. Therefore, by tending to the data interval and the point interval, a more suitable object set can be determined, which can accurately reflect individual performance while avoiding deviation from individual performance.

[0139] The third stage: screening of target objects, specifically including steps 404 to 406, which are described in detail below:

[0140] Step 404: predict the tendency of the candidate object.

[0141] Here, firstly, the target tendency prediction model trained in the first stage is used to predict the tendency of multiple candidate objects to obtain the tendency data of the candidate objects. The tendency data of the candidate objects can be annotated on the ID information of the candidate objects to obtain a list of candidate objects with tendency annotations.

[0142] Step 405: perform gain prediction on the candidate object.

[0143] Here, the tendency data is input into the target gain function to perform gain prediction on the candidate object, thereby obtaining the gain data of the candidate object. The gain data of the candidate object can be marked on the ID information of the candidate object to obtain a list of candidate objects with gain marking.

[0144] Step 406, determine the target object.

[0145] Here, according to the gain data and business requirements of the candidate objects, the objects whose gain data are within the preset gain range are determined as target objects. Then, the business indicators are improved by marketing the target objects.

[0146] In the third stage, the group tendency data is determined by selecting the average value, mode or other statistical methods to improve the accuracy of constructing the gain query table. At the same time, the gain function is determined by the mapping relationship between the group tendency data and the group gain data to improve the accuracy of the gain data prediction.

[0147] In an embodiment of the present application, a target tendency prediction model is used to predict and divide the sample objects, and then a gain function is constructed based on the group tendency data and the group gain data. In an embodiment of the present application, the tendency data can be used as an intermediate result to model the gain data, which mainly involves the training of the tendency data prediction model. Therefore, only conventional prediction model training techniques need to be used, and there are no excessive requirements for data scale, distribution, and model structure. In practical applications, the gain data of multiple object sets can be added to the gain query table according to demand. Therefore, the specific gain data of the user can be obtained by querying the gain query table, thereby improving the rationality and reliability of the target object acquisition. In addition, the embodiment of the present application can also use the gain function to obtain the gain data of a large number of users, and then directly model the gain data based on the feature data to obtain the gain data prediction model, which solves the problem that the gain data in the related art cannot be directly modeled.

[0148] The following further describes an exemplary structure of the data processing device 233 provided in the embodiment of the present application implemented as a software module. In some embodiments, Figure 2 As shown, the software modules stored in the data processing device 233 of the memory 230 may include:

[0149] A function call library construction module 2331 is used to construct a gain function call library based on the gain query table, wherein the gain function call library includes a plurality of type information and a gain function corresponding to each type information;

[0150] The target function determination module 2332 is used to determine the target gain function corresponding to the candidate object from the gain function call library based on the type information of the candidate object.

[0151] The gain prediction module 2333 is used to perform gain prediction on each candidate object based on the target gain function of each candidate object and the feature data of each candidate object to obtain the gain data of each candidate object.

[0152] The object determination module 2334 is used to determine the target object based on the gain data of each candidate object.

[0153] In some embodiments, the data processing device 233 further includes:

[0154] A first acquisition module, used to acquire feature data of multiple sample objects corresponding to preset type information;

[0155] A division module, used for performing division processing on the plurality of sample objects to obtain a plurality of object sets;

[0156] A first determination module, configured to determine group tendency data and group gain data of each of the object sets based on feature data of sample objects included in each of the object sets;

[0157] The first construction module is used to construct a gain query table corresponding to the preset type information based on the group tendency data and the group gain data of each object set.

[0158] In some embodiments, the partitioning module is further used to:

[0159] Predicting the tendency of the sample object based on the target tendency prediction model and the characteristic data of the sample object to obtain the tendency data of the sample object;

[0160] Based on the tendency data of the sample objects, sorting the sample objects in descending order to obtain sorted sample objects;

[0161] The sorted sample objects are divided to obtain the multiple object sets.

[0162] In some embodiments, the data processing device 233 further includes:

[0163] A second acquisition module is used to acquire feature data of multiple sample objects corresponding to each type of information, and to construct a tendency training data set based on the feature data of the multiple sample objects corresponding to the type of information;

[0164] A training module, used for performing tendency prediction training on the initial tendency prediction model based on the tendency training data set to obtain multiple candidate tendency prediction models;

[0165] The second determination module is used to determine a target tendency prediction model from the multiple candidate tendency prediction models based on the tendency prediction accuracy rates of the multiple candidate tendency prediction models.

[0166] In some embodiments, the first determining module is further configured to:

[0167] For each of the object sets, determining group tendency data of the object set based on tendency data of sample objects included in the object set;

[0168] Dividing the sample objects into a first object subset and a second object subset, the first object subset includes the sample objects that have received the intervention information, and the second object subset includes the sample objects that have not received the intervention information;

[0169] Determine a first value of a preset feature corresponding to the first object subset based on the feature data of the first object subset, and determine a second value of the preset feature corresponding to the second object subset based on the feature data of the second object subset;

[0170] Based on the first value and the second value, population gain data for the set of objects is determined.

[0171] In some embodiments, the function call library construction module 2331 is further used to:

[0172] Obtain the gain query table corresponding to each type of information;

[0173] For each of the gain lookup tables, determining a gain function between the group tendency data and the group gain data;

[0174] The gain function calling library is constructed based on each piece of the type information and the gain function corresponding to each piece of the type information.

[0175] In some embodiments, the gain prediction module 2333 is further used to:

[0176] Using the target tendency prediction model to perform tendency prediction on the feature data of each candidate object to obtain tendency data corresponding to each candidate object;

[0177] For the tendency data corresponding to each candidate object, gain prediction is performed on each candidate object based on the target gain function to obtain gain data of each candidate object.

[0178] In some embodiments, the object determination module 2334 is further configured to:

[0179] Based on the gain data of each candidate object, determining target gain data within a preset gain range;

[0180] The candidate object corresponding to the target gain data is determined as the target object.

[0181] The embodiment of the present application provides a computer program product, which includes a computer program or a computer executable instruction, and the computer program or the computer executable instruction is stored in a computer-readable storage medium. The processor of the electronic device reads the computer executable instruction from the computer-readable storage medium, and the processor executes the computer executable instruction, so that the electronic device executes the data processing method described above in the embodiment of the present application.

[0182] The present application embodiment provides a computer-readable storage medium, in which computer executable instructions or computer programs are stored. When the computer executable instructions or computer programs are executed by a processor, the processor will execute the data processing method provided by the present application embodiment, for example, Figure 3A The data processing method is shown.

[0183] In some embodiments, the computer-readable storage medium may be a memory such as RAM, ROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or may be various devices including one or any combination of the above memories.

[0184] In some embodiments, computer executable instructions may be in the form of a program, software, software module, script or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a stand-alone program or as a module, component, subroutine or other unit suitable for use in a computing environment.

[0185] As an example, computer-executable instructions may, but need not, correspond to a file in a file system, may be stored as part of a file that stores other programs or data, such as in one or more scripts in a HyperText Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files storing one or more modules, subroutines, or code portions).

[0186] As an example, computer executable instructions may be deployed to be executed on one electronic device, or on multiple electronic devices located at one site, or on multiple electronic devices distributed at multiple sites and interconnected by a communication network.

[0187] In summary, the target tendency prediction model is used in the embodiment of the present application to predict the tendency and divide the sample objects, and then a gain function is constructed according to the group tendency data and the group gain data. The gain data can be accurately determined through the gain function, thereby improving the rationality and reliability of the target object acquisition. The tendency data can also be used as an intermediate result to model the gain data. Only conventional prediction model training technology is needed, and there are no excessive requirements for the data scale, distribution and model structure. In addition, the embodiment of the present application can also use the gain function to obtain the gain data of a large number of sample objects, and then directly model the gain data based on the feature data to obtain the gain data prediction model, which solves the problem that the gain data in the related art cannot be directly modeled. .

[0188] The above is only an embodiment of the present application and is not intended to limit the protection scope of the present application. Any modifications, equivalent substitutions and improvements made within the spirit and scope of the present application are included in the protection scope of the present application.

Claims

1. A data processing method, characterized in that: The method comprises: Building a gain function call library based on the gain query table, wherein the gain function call library includes a plurality of type information and a gain function corresponding to each type information; Based on the type information of the candidate object, determining a target gain function corresponding to the candidate object from the gain function call library; Based on the target gain function of each candidate object and the feature data of each candidate object, performing gain prediction on each candidate object to obtain gain data of each candidate object; Based on the gain data of each of the candidate objects, a target object is determined.

2. The method according to claim 1, characterized in that: The method further comprises: Acquire feature data of multiple sample objects corresponding to preset type information; Performing division processing on the multiple sample objects to obtain multiple object sets; Determining group tendency data and group gain data of each of the object sets based on feature data of the sample objects included in each of the object sets; Based on the group tendency data and the group gain data of each of the object sets, a gain query table corresponding to the preset type information is constructed.

3. The method according to claim 2, characterized in that The dividing process of the plurality of sample objects to obtain a plurality of object sets includes: Predicting the tendency of the sample object based on the target tendency prediction model and the characteristic data of the sample object to obtain the tendency data of the sample object; Based on the tendency data of the sample objects, sorting the sample objects in descending order to obtain sorted sample objects; The sorted sample objects are divided to obtain the multiple object sets.

4. The method according to claim 2, characterized in that: The method further comprises: Acquire feature data of multiple sample objects corresponding to each type of information, and construct a tendency training data set based on the feature data of the multiple sample objects corresponding to the type of information; Performing tendency prediction training on the initial tendency prediction model based on the tendency training data set to obtain multiple candidate tendency prediction models; Based on the tendency prediction accuracy rates of the plurality of candidate tendency prediction models, a target tendency prediction model is determined from the plurality of candidate tendency prediction models.

5. The method according to claim 2, characterized in that: The determining of the group tendency data and the group gain data of each of the object sets based on the feature data of the sample objects included in each of the object sets comprises: For each of the object sets, determining group tendency data of the object set based on tendency data of sample objects included in the object set; Dividing the sample objects into a first object subset and a second object subset, the first object subset includes the sample objects that have received the intervention information, and the second object subset includes the sample objects that have not received the intervention information; Determine a first value of a preset feature corresponding to the first object subset based on the feature data of the first object subset, and determine a second value of the preset feature corresponding to the second object subset based on the feature data of the second object subset; Based on the first value and the second value, population gain data for the set of objects is determined.

6. The method according to claim 1, characterized in that The step of constructing a gain function call library based on the gain query table includes: Obtain the gain query table corresponding to each type of information; For each of the gain lookup tables, determining a gain function between the group tendency data and the group gain data; The gain function calling library is constructed based on each piece of the type information and the gain function corresponding to each piece of the type information.

7. The method according to claim 1, characterized in that The step of performing gain prediction on each candidate object based on the target gain function of each candidate object and the feature data of each candidate object to obtain the gain data of each candidate object includes: Using the target tendency prediction model to perform tendency prediction on the feature data of each candidate object to obtain tendency data corresponding to each candidate object; For the tendency data corresponding to each candidate object, gain prediction is performed on each candidate object based on the target gain function to obtain gain data of each candidate object.

8. The method according to claim 1, characterized in that: The step of determining a target object based on the gain data of each candidate object comprises: Based on the gain data of each candidate object, determining target gain data within a preset gain range; The candidate object corresponding to the target gain data is determined as the target object.

9. A data processing device, characterized in that: The device comprises: A function call library construction module, used to construct a gain function call library based on the gain query table, wherein the gain function call library includes a plurality of type information and a gain function corresponding to each type information; An objective function determination module, configured to determine an objective gain function corresponding to the candidate object from the gain function call library based on the type information of the candidate object; A gain prediction module, configured to perform gain prediction on each candidate object based on a target gain function of each candidate object and feature data of each candidate object, so as to obtain gain data of each candidate object; The object determination module is used to determine the target object based on the gain data of each candidate object.

10. An electronic device, characterized in that: The electronic device comprises: A memory for storing computer executable instructions; A processor, configured to implement the method according to any one of claims 1 to 8 when executing the computer executable instructions stored in the memory.

11. A computer-readable storage medium storing computer-executable instructions or a computer program, characterized in that: When the computer executable instructions or computer programs are executed by a processor, the method according to any one of claims 1 to 8 is implemented.

12. A computer program product comprising computer executable instructions or a computer program, characterized in that: When the computer executable instructions or computer programs are executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images