A positioning method, device and medium for a target population

Abstract

The application discloses a positioning method and device for a target group and a medium, and is suitable for the technical field of data processing. By calling A / B experiments of an experimental group model and a control group model, the experimental group model and the control group model are respectively based on target function triggering users and target function non-triggering users to determine corresponding experimental group probabilities and control group probabilities, and then according to the experimental group probabilities and the control group probabilities, probability values corresponding to respective current users are determined, and current users corresponding to probability values meeting preset conditions are selected from the probability values to realize positioning of the target group. The experimental group model and the control group model based on the division of target function triggering and non-triggering can clearly determine that the positioning of the current target group is based on the target function, the target group positioned based on the target function is more accurate, and resource allocation is performed on the target group determined based on the target function to save resources.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a target group positioning method and device and medium. BACKGROUND

[0002] With the wide application of social application programs, the push function of the social application program is more complex, and the target group positioning of the function is important in the process.

[0003] The existing target group positioning scheme only focuses on the target group, and the accuracy of the target group converted by the diversity of the push function is poor. The positioning process of the target group to which the background push function is pushed is not accurate, and even causes waste of resource allocation.

[0004] Therefore, how to accurately position the target group to reasonably use resources is an urgent problem to be solved by those skilled in the art. SUMMARY

[0005] The purpose of the present application is to provide a target group positioning method, device and medium, which is more accurate in positioning the target group based on the target function, so as to allocate resources to the target group determined by the target function to save resources.

[0006] To solve the above technical problems, the present application provides a target group positioning method, comprising:

[0007] Obtaining user information corresponding to each of the current users, and extracting features from the user information to obtain user feature information, wherein the number of current users is multiple, and the user information includes static information and dynamic behavior information of the current users;

[0008] Calling an experimental group model and a control group model to input the user feature information, and obtaining output parameters of the experimental group model and the control group model as experimental group probability and control group probability, respectively, wherein the experimental group model is determined based on target function triggering, and the control group model is determined based on target function non-triggering;

[0009] Determining the probability value corresponding to each of the current users according to the experimental group probability and the control group probability, and selecting the probability value meeting the preset condition from the probability values to determine the current user corresponding to the probability value meeting the preset condition as the target group.

[0010] Preferably, the user feature information is extracted from the user information, comprising:

[0011] The user information is stored in a user portrait in advance;

[0012] extracting features of the user information based on the user portrait to obtain the user feature information.

[0013] Preferably, the determining process of the experimental group model and the control group model comprises:

[0014] obtaining original user feature information of original users, wherein the number of the original users is multiple and even;

[0015] randomly allocating the original users according to experimental groups and control groups to obtain experimental group users and control group users, wherein the number of the experimental group users and the control group users is the same;

[0016] obtaining initial weight parameters and conversion values corresponding to an initial experimental group model and an initial control group model respectively, wherein the initial weight parameters corresponding to each of the initial experimental group model and the initial control group model are determined by a relationship between the target function and the original user feature information;

[0017] performing loss processing on the original user feature information corresponding to each of the initial weight parameters and the conversion values corresponding to each of the initial experimental group model and the initial control group model to obtain a current target value of the experimental group and a current target value of the control group;

[0018] respectively determining whether the current target value of the experimental group and the current target value of the control group meet respective preset requirements;

[0019] if yes, the establishment of the experimental group model and the control group model is completed;

[0020] if no, the initial weight parameters corresponding to each of the initial experimental group model and the initial control group model are adjusted, and the step of performing loss processing on the original user feature information corresponding to each of the initial weight parameters and the conversion values corresponding to each of the initial experimental group model and the initial control group model is returned to until the current target value of the experimental group and the current target value of the control group meet the respective preset requirements.

[0021] Preferably, the performing loss processing on the original user feature information corresponding to each of the initial weight parameters and the conversion values corresponding to each of the initial experimental group model and the initial control group model to obtain a current target value of the experimental group and a current target value of the control group comprises:

[0022] determining initial experimental group loss values and initial control group model loss values according to relationships between the initial weight parameters corresponding to each of the initial experimental group model and the initial control group model and the original user feature information corresponding to each of the initial experimental group model and the initial control group model;

[0023] determine the final loss value of the experimental group user and the final loss value of the control group user according to the relationship between the respective corresponding conversion value and the initial experimental group loss value and the initial control group model loss value;

[0024] average the final loss value of the experimental group user and the final loss value of the control group user respectively to obtain the current target value of the experimental group and the current target value of the control group.

[0025] Preferably, the model of the experimental group model and the control group model is the same, at least including one or more of a deep learning model, a machine learning model, a binary classification model and an artificial intelligence model.

[0026] Preferably, the determination of the respective corresponding probability value of the current user according to the experimental group probability and the control group probability comprises:

[0027] The experimental group probability and the control group probability are subtracted to obtain the respective corresponding probability value of the current user.

[0028] Preferably, the selection of the probability value meeting the preset condition in each probability value to determine the current user corresponding to the probability value meeting the preset condition as the target population comprises:

[0029] When the preset condition is that the first N probability values all exceed the preset probability, the determination process of the target population comprises:

[0030] Sort each probability value from large to small;

[0031] Determine whether the first N probability values after sorting all exceed the preset probability;

[0032] If yes, the current user corresponding to the first N probability values is the target population;

[0033] Or, when the preset condition is whether the number of the first N probability values reaches a preset number, the determination process of the target population comprises:

[0034] Sort each probability value from large to small;

[0035] Determine whether the number of the first N probability values after sorting reaches the preset number;

[0036] If yes, the current user corresponding to the first N probability values is the target population.

[0037] To solve the above technical problems, the application also provides a target population positioning device, comprising:

[0038] An acquisition module is configured to acquire user information corresponding to each of current users, and to perform feature extraction on the user information to obtain user feature information, wherein the number of the current users is multiple, and the user information includes static information and dynamic behavior information of the current users;

[0039] An invocation module is configured to invoke an experimental group model and a control group model respectively to input the user feature information, and to obtain output parameters of the experimental group model and the control group model as experimental group probability and control group probability respectively, wherein the experimental group model is determined based on target function triggering, and the control group model is determined based on target function non-triggering;

[0040] A determination module is configured to determine probability values corresponding to each of the current users according to the experimental group probability and the control group probability, and to select a probability value meeting a preset condition from the probability values to determine the current user corresponding to the probability value meeting the preset condition as a target population.

[0041] To solve the above technical problems, the application further provides a positioning device for a target population, comprising:

[0042] A memory is configured to store a computer program;

[0043] A processor is configured to implement the steps of the positioning method for a target population when the computer program is executed.

[0044] To solve the above technical problems, the application further provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the steps of the positioning method for a target population.

[0045] The positioning method for a target population provided by the application is based on A / B experiment of an experimental group model and a control group model, the experimental group model and the control group model are respectively based on target function triggering users and target function non-triggering users to determine corresponding experimental group probability and control group probability, and then the probability values corresponding to each of the current users are determined according to the experimental group probability and the control group probability, and the current user corresponding to the probability value meeting the preset condition is selected from the probability values to realize the positioning of the target population. The experimental group model and the control group model based on the target function triggering and non-triggering can clearly determine that the positioning of the current target population is based on the target function, and the target population positioned based on the target function is more accurate, so that the target population determined based on the target function can be allocated resources to save resources.

[0046] In addition, the application further provides a positioning device for a target population and a medium, which have the same beneficial effects as the positioning method for a target population. BRIEF DESCRIPTION OF DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. Obviously, the drawings described in the following embodiments are only some of the embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort based on these drawings.

[0048] Figure 1 The flow chart of the target population positioning method provided by the embodiments of the present application;

[0049] Figure 2 The schematic diagram of the training and prediction of the experimental group model and the control group model provided by the embodiments of the present application;

[0050] Figure 3 The structural diagram of the target population positioning device provided by the embodiments of the present application;

[0051] Figure 4 The structural diagram of another target population positioning device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0052] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without any creative effort fall within the protection scope of the present application.

[0053] The core of the present application is to provide a target population positioning method, device and medium. The target population positioned based on this target function is more accurate, so as to save resources by allocating resources to the target population determined by the target function.

[0054] In order to make the personnel in the technical field better understand the present application, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0055] It should be noted that the social software program is designed to help people find real friendship in the virtual world, and each function of the social software has its own audience, so it is extremely important to target the target audience using the function. The existing target audience positioning is to push all target functions to all users, and some of them have very low conversion rates, and some have high conversion rates. Even if the target function is not pushed to the user with high conversion rate, the target function will also be targeted and converted. For example, whether an advertisement should be displayed to a user, if the user can be targeted and converted for any function push, the advertisement of the function does not need to be pushed to the user. If the user's conversion rate is very low before watching the advertisement, and the conversion rate is improved after watching the advertisement, the advertisement of the function needs to be pushed to the user in this case. The common model of the target personnel positioning is response modeling, which does not distinguish between natural conversion users, and some users can be converted even if the advertisement function is not used. The prior art pushes all target functions to all users, which essentially leads to waste of some push resources and does not accurately allocate resources. Therefore, the target audience positioning method provided by the present application can target users with low conversion rates before not pushing the target function and high conversion rates after pushing the target function as target users, and push the target function to save push resources.

[0056] Figure 1 A flowchart of a target audience positioning method provided by an embodiment of the present application is shown in Figure 1 The method comprises the following steps:

[0057] S11: Obtain respective user information of the current users, and perform feature extraction on the user information to obtain user feature information;

[0058] The number of current users is multiple, and the user information includes static information and dynamic behavior information of the current users.

[0059] It can be understood that the number of current users is multiple, and the present embodiment determines the final target audience among the multiple current users. The user information includes static information and dynamic behavior information of the current users. The static information refers to information that does not change every day or changes at intervals on the social platform, such as basic information, interest information, etc. The basic information and interest information are selected and completed when registering.

[0060] Dynamic behavioral information refers to the activities users engage in using online tools to search for, select, absorb, utilize, exchange, and publish information, driven by their information needs and motivations. In this embodiment, the information generated during interactions on social platforms is dynamically changing. For example, browsing social platform pages and topics daily, actively engaging in discussions on social platforms, chatting with other users, participating in topics, and actions such as liking, commenting, and forwarding all constitute user behavioral information.

[0061] Based on user information, feature extraction is performed to obtain corresponding user feature information. This embodiment does not limit the feature extraction method; it can be based on one or more features, a specific feature extraction algorithm, or a user profile. As one embodiment, the feature extraction process extracts features based on the characteristics of the target function being pushed, in order to achieve targeted positioning of the target audience.

[0062] S12: Call the experimental group model and the control group model respectively to input user feature information, and obtain the output parameters of the experimental group model and the control group model as the experimental group probability and the control group probability respectively;

[0063] The experimental group model was determined based on the target function trigger, while the control group model was determined based on the target function not being triggered.

[0064] The experimental and control models use the same model, hence the designation as the experimental and control groups. The only difference lies in the transformed samples within the models. Both groups aim for transformation, but the transformation process depends on whether the target function in each model is triggered. For the experimental group model, the target function is triggered, while for the control group model, it is not. Both models transform positive samples but not negative samples.

[0065] For example, the experimental group model triggers user engagement when the ad push function is enabled, while the control group model does not trigger user engagement when the ad push function is disabled. The corresponding target function can be a single function or a combination of multiple functions; there is no limitation here, and it can be set according to the actual situation.

[0066] During the model prediction process, the user's feature information is input into both the experimental group model and the control group model to obtain their respective output parameters as the experimental group probability and the control group probability. Since there are multiple users, the experimental group probability and control group probability are calculated separately for each user.

[0067] S13: Determine the respective probability values of the current users according to the experimental group probability and the control group probability, and select the probability value meeting the preset condition in the probability values to determine the current user corresponding to the probability value meeting the preset condition as the target population.

[0068] After determining the respective experimental group probability and control group probability of the plurality of users, the final probability value is determined according to the two probabilities. After the plurality of users obtain the respective probability values, the probability value meeting the preset condition is selected in the plurality of probability values to determine the current user corresponding to the probability value meeting the preset condition as the target population of the target function push.

[0069] It should be noted that the preset condition can be based on the relationship between the probability value and the threshold, or based on the top users corresponding to the probability value after the sorting of the probability value, which is not limited here and can be set according to the actual situation.

[0070] The positioning method of the target population provided by the embodiment of the application determines the respective experimental group probability and control group probability of the target function triggered user and the target function not triggered user by calling the A / B experiment of the experimental group model and the control group model, and then determines the respective probability values of the current users according to the experimental group probability and the control group probability, and selects the current user corresponding to the probability value meeting the preset condition in the probability values to realize the positioning of the target population. The experimental group model and the control group model for the target function triggered and not triggered can clearly determine that the positioning of the current target population is based on the target function, and the target population positioned based on the target function is more accurate, so that the target population determined for the target function can be allocated resources to save resources.

[0071] On the basis of the above embodiment, the user feature information is obtained by performing feature extraction on the user information in step S11, including:

[0072] The user information is stored in the user portrait in advance;

[0073] The user feature information is obtained by performing feature extraction on the user information based on the user portrait.

[0074] Specifically, the user portrait needs to process a large amount of logs, and the content contained is not completely fixed. The features concerned are different according to different industries and products. The user portrait will contain population attributes and behavior features. The population attributes mainly refer to the age, gender, province and city where the user is located, education level, marital status, fertility, industry and occupation where the user works, etc. The behavior features mainly include activity and loyalty indicators. In addition to the above common features, different types of websites extract user portraits with different emphases.

[0075] Content-based media or reading websites, search engines or general navigation websites, often extract the user's interest in browsing content features, such as sports, entertainment, food, finance, tourism, real estate, and cars. The user portrait of social networking sites can extract the user's social network, and find the user group with close relationship and the star node playing the role of opinion leader in the community. The user portrait of e-commerce shopping websites generally extracts the user's online shopping interest and consumption capacity. Online shopping interest mainly refers to the user's category preference when shopping online, such as clothing, luggage, home, baby, washing and food. Consumption capacity refers to the user's purchasing power. If it is done in detail, the actual consumption level and the psychological consumption level of each category can be distinguished, and the feature latitude can be established. In addition, the user's environmental attributes can also be added, such as current time, access location LBS features, local weather, and holiday conditions.

[0076] The extraction of user features, that is, the production process of user portrait, can be roughly divided into the following steps:

[0077] 1. User modeling, which refers to determining the extracted user feature dimensions and the data sources needed.

[0078] 2. Data collection, through data collection tools such as log collection system (Flume) or script program, the data needed is stored in a distributed system infrastructure (Hadoop) cluster.

[0079] 3. Data cleaning, the process of data cleaning is usually located in the Hadoop cluster, and it is also possible to perform data cleaning at the same time as data collection. The main work of this step is to extract the fields of the collected data from various sources and in disarray to obtain the target features of interest.

[0080] 4. Model training, some features may not be directly obtained from data cleaning, such as user's interest in content or user's consumption level, which can be learned and predicted through the known features collected.

[0081] 5. Attribute prediction, using the trained model and the user's known features to predict the user's unknown features.

[0082] 6. Data merging, merging the features extracted by the user through various data sources, and giving a certain credibility.

[0083] 7. Data distribution, for the combined result data, distribute to each platform such as precision marketing, personalized recommendation, customer relationship management (CRM), etc., to provide data support. Of course, for a specific website or application (App), there are special user dimensions that need to be further refined to provide more accurate personalized services and content to users.

[0084] The user portrait provided by the embodiment is related to the target function, and the corresponding extracted user feature information is also related to the target function.

[0085] The user feature information obtained by the feature extraction according to the user information provided by the embodiment is relatively comprehensive through the user portrait, and the user's features can be deeply studied so that more accurate target users can be matched subsequently.

[0086] On the basis of the above embodiment, the determination process of the experimental group model and the control group model in step S12 includes:

[0087] Obtain the original user feature information of the original user, wherein the number of original users is multiple and even;

[0088] Randomly allocate the original users according to the experimental group and the control group to obtain the experimental group users and the control group users, wherein the number of the experimental group users and the control group users is the same;

[0089] Obtain the initial weight parameters corresponding to the initial experimental group model and the initial control group model respectively and the conversion values corresponding to the initial experimental group model and the initial control group model respectively, wherein the initial weight parameters corresponding to the initial experimental group model and the initial control group model respectively are determined by the relationship between the target function and the original user feature information;

[0090] According to the respective conversion values, the respective initial weight parameters and the respective original user feature information, the current target value of the experimental group and the current target value of the control group are obtained by loss processing;

[0091] Determine whether the current target value of the experimental group and the current target value of the control group meet the respective preset requirements respectively;

[0092] If it is satisfied, the establishment of the experimental group model and the control group model is completed;

[0093] If it is not satisfied, the respective initial weight parameters are adjusted, and the step of obtaining the current target value of the experimental group and the current target value of the control group by loss processing according to the respective conversion values, the respective initial weight parameters and the respective original user feature information is returned until the current target value of the experimental group and the current target value of the control group meet the respective preset requirements.

[0094] Specifically, in the model training process, the original users are even, which are randomly assigned to the experimental group and the control group, for example, 100 users, the experimental group and the control group are randomly assigned to 50 users respectively, the original user feature information of the original users can be feature information extracted by user portrait, or can be obtained based on other feature extraction algorithms, in order to keep consistent with the feature extraction of the above embodiment, the user portrait is used for feature extraction to obtain the original user feature information.

[0095] Taking the determination of the experimental group model as an example, for 50 users, a plurality of feature information is set in the model training process, and the plurality of feature information respectively corresponds to respective initial weight parameters and conversion values. It should be noted that the sum of each initial weight parameter is not necessarily 1, and each weight parameter has its own value corresponding to 0-1 in general. The corresponding conversion value can be 1 or 0. The initial weight parameter is determined by the relationship between the target function and the original user feature information. For example, the advertising push function has a higher degree of association with a certain feature, so the initial weight parameter corresponding to the gender feature will be higher than that of other features.

[0096] The conversion value of the present embodiment is an actual value, and the current target value, that is, the predicted value, is obtained by loss processing according to the actual value, the initial weight parameter and the corresponding original user feature information. The preset requirement can be that the difference between the predicted value and the actual value is less than a threshold value, or only whether the predicted value is less than a certain threshold value, etc. If it is satisfied, it means that the current initial weight parameter setting can meet the establishment of the model. If it is not satisfied, adjust the specific value of each initial weight parameter, and return to calculate the current target value until the preset requirement is met.

[0097] The establishment of the control group model is also the same. The difference between the two models is only whether the target function is triggered or not, and the rest is the same.

[0098] On the basis of the above embodiment, as an embodiment, the current target value of the corresponding experimental group and the current target value of the control group are obtained by loss processing according to the respective corresponding conversion values, the respective corresponding initial weight parameters and the respective corresponding original user feature information, including:

[0099] The initial experimental group loss value and the initial control group model loss value are respectively determined according to the relationship between the respective corresponding initial weight parameters and the respective corresponding original user feature information;

[0100] The final loss value of the experimental group user and the final loss value of the control group user are respectively determined according to the relationship between the initial experimental group loss value and the initial control group model loss value corresponding to the respective corresponding conversion values.

[0101] The final loss value of the experimental group user and the final loss value of the control group user are averaged respectively to obtain the current target value of the experimental group and the current target value of the control group.

[0102] Specifically, taking the experimental group model as an example, for example, the initial weight parameters are w1, w2 and w3, the corresponding features are A, B and C, and the determination function of the initial experimental loss value is f(w1*A+w2*B+w3*C). According to the relationship between the conversion value and the initial experimental loss value, the final loss value of the experimental group user is determined, the conversion value is Y, and the final loss value is Y-f(w1*A+w2*B+w3*C). In combination with the above example, 50 users in the experimental group need to determine their respective final loss values, and the 50 final loss values are averaged to obtain the current target value of the experimental group. Similarly, the control group model is the same, and will not be repeated here.

[0103] On the basis of the above embodiment, the model of the experimental group model and the control group model is the same, and at least includes one or more of a deep learning model, a machine learning model, a binary classification model and an artificial intelligence model.

[0104] It can be understood that the experimental group model and the control group model can be based on an existing mathematical model, or a plurality of models can be combined, or other models can be set, which are not limited here, and can also be a classical binary classification model, such as: logistic regression, tree model, etc., which will not be repeated here.

[0105] The determination process of the experimental group model and the control group model provided in this embodiment determines the final weight parameter through loss processing, trains based on user information, and improves the model accuracy of the experimental group and the control group.

[0106] On the basis of the above embodiment, as an embodiment, the respective probability values of the current users are determined according to the experimental group probability and the control group probability, including:

[0107] The experimental group probability and the control group probability are subtracted to obtain the respective probability values of the current users.

[0108] It can be understood that the experimental group probability and the control group probability are subtracted, which is to subtract a small value from a large value to ensure that the probability value is positive. As long as the experimental group probability and the control group probability can be obtained as the final probability value, which is positive, the requirement for which one as the minuend and the subtrahend is not required.

[0109] In addition, as an embodiment, the probability values meeting the preset condition are selected from the probability values to determine the current users corresponding to the probability values meeting the preset condition as the target population, including:

[0110] In a preset condition that the first N probability values exceed a preset probability, the process of determining the target group includes:

[0111] The probability values are sorted from large to small.

[0112] It is judged whether the first N probability values after sorting exceed the preset probability.

[0113] If yes, the current users corresponding to the first N probability values are taken as the target group.

[0114] Alternatively, in a preset condition that the number of the first N probability values reaches a preset number, the process of determining the target group includes:

[0115] The probability values are sorted from large to small.

[0116] It is judged whether the number of the first N probability values after sorting reaches the preset number.

[0117] If yes, the current users corresponding to the first N probability values are taken as the target group.

[0118] It can be understood that the preset condition can be that the users corresponding to the probability values exceeding the preset probability are taken as the target group, or the target group can be selected according to the preset number after sorting, both of which can be used. However, the prerequisite is to sort the probability values of the users, which can be sorted from large to small or from small to large, and is not limited herein. In this embodiment, the order from large to small is used, and if the order from small to large is used, the last probability value is selected from the front, and the preset number is selected from the front of the last one.

[0119] The determination of the target group provided in the embodiment of the application realizes the positioning of the target group, and the target group positioned based on the target function is more accurate.

[0120] Figure 2 An experimental group model and a control group model provided in the embodiment of the application are shown in a schematic diagram of training and prediction of the experimental group model and the control group model, as shown in the figure, model one is an experimental group model reaching users for a target function, which is divided into conversion and non-conversion, model two is a control group model not reaching users for a target function, which is divided into conversion and non-conversion, and the model training stage includes: Figure 2

[0121] 1. First, the users are randomly grouped, and A / B experiment is performed, wherein the control group reaches users for a target function, and the experimental group does not reach users for a target function.

[0122] ​2、Firstly, the users in the control group are modeled, the target is conversion, if conversion is positive sample, if no conversion is negative sample, the features include static information and behavior data of the users. Then the users in the experimental group are modeled, the target is also conversion, if conversion is positive sample, if no conversion is negative sample, the features include static information and behavior data of the users.

[0123] Model prediction stage:

[0124] 1、After a new user comes, the model trained by the experimental group data and the model trained by the control group data are requested respectively, and two probabilities are obtained.

[0125] 2、Subtract the control group model probability from the experimental group model probability to obtain the final probability, and sort the probability from large to small, that is, the target users of the specific function can be selected.

[0126] For the training and prediction schematic diagram of the experimental group model and the control group model provided by the application, please refer to the above method embodiment, and the application will not be repeated here, which has the same beneficial effects as the above target population positioning method.

[0127] The above detailed description of the target population positioning method corresponds to each embodiment, and on this basis, the application also discloses a target population positioning device corresponding to the above method, Figure 3 A structure diagram of a target population positioning device provided by an embodiment of the application. As shown in Figure 3 The target population positioning device comprises:

[0128] The acquisition module 11 is configured to acquire user information corresponding to each of the current users, and extract user feature information according to the user information, wherein the number of the current users is multiple, and the user information includes static information and dynamic behavior information of the current users.

[0129] The calling module 12 is configured to call the experimental group model and the control group model respectively to input the user feature information, and acquire output parameters of the experimental group model and the control group model as an experimental group probability and a control group probability respectively, wherein the experimental group model is determined based on target function triggering, and the control group model is determined based on target function non-triggering.

[0130] The determination module 13 is configured to determine a probability value corresponding to each of the current users according to the experimental group probability and the control group probability, and select a probability value meeting a preset condition in the probability values to determine the current user corresponding to the probability value meeting the preset condition as the target population.

[0131] Since the embodiments of the device part correspond to the above embodiments, the embodiments of the device part will be described with reference to the above method embodiments, and will not be repeated here.

[0132] For the positioning device of a target population provided by the present application, please refer to the above-mentioned method embodiments, and the present application will not be repeated here, which has the same beneficial effects as the above-mentioned positioning method of a target population.

[0133] Figure 4 The structure diagram of another positioning device of a target population provided by the present application is shown in FIG. 2, which comprises: Figure 4

[0134] The memory 21 is used for storing a computer program.

[0135] The processor 22 is used for executing the computer program to realize the steps of the positioning method of a target population.

[0136] The positioning device of a target population provided by the present application can include but is not limited to a tablet computer, a notebook computer or a desktop computer, etc.

[0137] The processor 22 can include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 22 can be implemented in at least one of a hardware form of a digital signal processor (DSP), a field-programmable gate array (FPGA), a programmable logic array (PLA). The processor 22 can also include a main processor and a coprocessor. The main processor is a processor for processing data in an awake state, also known as a central processing unit (CPU). The coprocessor is a low-power processor for processing data in a standby state. In some embodiments, the processor 22 can be integrated with a graphics processor (GPU) that is responsible for rendering and drawing the content required to be displayed on the display screen. In some embodiments, the processor 22 can also include an artificial intelligence (AI) processor that is used to process machine learning-related computing operations.

[0138] ​The memory 21 can include one or more computer-readable storage media. The memory 21 can also include high-speed random access memory and non-volatile, computer-readable storage media such as one or more magnetic disk storage devices, flash memory devices. In this embodiment, the memory 21 is used to store at least the following computer program 211, wherein the computer program is loaded and executed by the processor 22, and can realize the steps of the positioning method of the target population disclosed in any of the above embodiments. In addition, the resources stored in the memory 21 can also include an operating system 212 and data 213, etc., and the storage mode can be temporary storage or permanent storage. The operating system 212 can include Windows, Unix, Linux, etc. The data 213 can include but is not limited to the data involved in the positioning method of the target population, etc.

[0139] In some embodiments, the positioning device of the target population can further include a display screen 23, an input / output interface 24, a communication interface 25, a power supply 26 and a communication bus 27.

[0140] Those skilled in the art can understand that the structure shown in the above embodiments does not constitute a limitation on the positioning device of the target population, and can include more or fewer components than those shown in the drawings. Figure 4 The processor 22 realizes the positioning method of the target population provided in any of the above embodiments by calling the instructions stored in the memory 21.

[0141] The processor 22 realizes the positioning method of the target population provided in any of the above embodiments by calling the instructions stored in the memory 21.

[0142] For the positioning device of the target population provided by the present application, please refer to the above method embodiments, and the present application will not be described here again, which has the same beneficial effects as the above positioning method of the target population.

[0143] Further, the present application also provides a computer-readable storage medium, and the computer-readable storage medium stores a computer program, and the computer program is executed by the processor 22 to realize the steps of the positioning method of the target population.

[0144] It can be understood that if the method in the above embodiment is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part of the prior art that contributes to the present application or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and performs all or part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (Read-Only Memory, ROM), a random access memory (Random Access Memory, RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0145] For the computer readable storage medium provided by the present application, please refer to the above method embodiment, and the present application will not be repeated here. It has the same beneficial effects as the positioning method of the target population described above.

[0146] The above describes in detail the target population positioning method, target population positioning device and medium provided by the present application. The embodiments in the specification are described in a progressive manner, and each embodiment mainly describes the difference from other embodiments. The same and similar parts of each embodiment can be referred to. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the related parts can be referred to the method part. It should be pointed out that for ordinary skilled in the art, without departing from the principle of the present application, the present application can be improved and modified, and these improvements and modifications also fall within the protection scope of the claims of the present application.

[0147] It should be further noted that in the present specification, the relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment. Without more limitations, the element defined by the statement "including a" does not exclude the presence of other identical elements in the process, method, article or equipment including the element.

Claims

1. A method of locating a target population, characterized by, The method comprises the following steps: obtaining user information corresponding to each of the current users, and performing feature extraction on the user information to obtain user feature information, wherein the number of the current users is multiple, and the user information comprises static information and dynamic behavior information of the current users; calling an experimental group model and a control group model respectively to input the user feature information, and obtaining output parameters of the experimental group model and the control group model as experimental group probability and control group probability respectively, wherein the experimental group model is determined based on target function triggering, and the control group model is determined based on target function non-triggering; determining probability values corresponding to each of the current users according to the experimental group probability and the control group probability, and selecting a probability value meeting a preset condition from the probability values to determine the current user corresponding to the probability value meeting the preset condition as a target population; correspondingly, the determination process of the experimental group model and the control group model comprises the following steps: obtaining original user feature information of original users, wherein the number of the original users is multiple and even; randomly allocating the original users into experimental group users and control group users according to an experimental group and a control group, wherein the number of the experimental group users is the same as that of the control group users; obtaining initial weight parameters and conversion values corresponding to an initial experimental group model and an initial control group model respectively, wherein the initial weight parameters are determined by the relationship between the target function and the original user feature information; performing loss processing on the conversion values, the initial weight parameters and the original user feature information to obtain current target values of the experimental group and the control group respectively; determining whether the current target values of the experimental group and the control group meet preset requirements respectively; if yes, the establishment of the experimental group model and the control group model is completed; if no, the initial weight parameters are adjusted, and the step of performing loss processing on the conversion values, the initial weight parameters and the original user feature information to obtain current target values of the experimental group and the control group is returned until the current target values of the experimental group and the control group meet the preset requirements; correspondingly, the step of performing loss processing on the conversion values, the initial weight parameters and the original user feature information to obtain current target values of the experimental group and the control group comprises the following steps: determining initial experimental group model loss values and initial control group model loss values according to the relationship between the initial weight parameters and the original user feature information respectively; determining final loss values of the experimental group users and the control group users according to the relationship between the conversion values and the initial experimental group model loss values and the initial control group model loss values respectively. The final loss value of the experimental group user and the final loss value of the control group user are respectively averaged to obtain a current target value of the experimental group and a current target value of the control group.

2. The method of targeting according to claim 1, wherein, The feature extraction based on the user information comprises: The user information is stored in a user portrait in advance; The feature extraction of the user information based on the user portrait comprises:

3. The method of targeting according to claim 2, wherein, The model of the experimental group and the model of the control group are the same, and at least one or more of a deep learning model, a machine learning model, a binary classification model and an artificial intelligence model.

4. The method of targeting according to claim 3, wherein, The determination of the probability value corresponding to each of the current users based on the experimental group probability and the control group probability comprises: The experimental group probability and the control group probability are subtracted to obtain the probability value corresponding to each of the current users.

5. The method of targeting according to claim 4, wherein, The current user corresponding to the probability value meeting the preset condition is determined as the target population, which comprises: If the preset condition is that the first N probability values all exceed a preset probability, the determination process of the target population comprises: The probability values are sorted from large to small; It is judged whether the first N probability values after sorting all exceed the preset probability; If yes, the current users corresponding to the first N probability values are taken as the target population; Or, if the preset condition is whether the number of the first N probability values reaches a preset number, the determination process of the target population comprises: The probability values are sorted from large to small; It is judged whether the number of the first N probability values after sorting reaches the preset number; If yes, the current users corresponding to the first N probability values are taken as the target population.

6. A targeting device for a target population, characterized in that, It comprises: An acquisition module is configured to acquire user information corresponding to each of current users, and to perform feature extraction based on the user information to obtain user feature information, wherein the number of the current users is multiple, and the user information comprises static information and dynamic behavior information of the current users; A calling module is configured to call an experimental group model and a control group model respectively to input the user feature information, and to acquire output parameters of the experimental group model and the control group model as an experimental group probability and a control group probability respectively, wherein the experimental group model is determined based on a target function trigger, and the control group model is determined based on that the target function does not trigger; A determination module is configured to determine a probability value corresponding to each of the current users based on the experimental group probability and the control group probability, and to select a probability value meeting a preset condition from the probability values to determine the current user corresponding to the probability value meeting the preset condition as a target population. Correspondingly, the determination process of the experimental group model and the control group model comprises: Original user feature information of original users is acquired, wherein the number of the original users is multiple and even; The original users are randomly allocated based on an experimental group and a control group to obtain experimental group users and control group users, wherein the number of the experimental group users and the control group users is the same; obtaining initial weight parameters corresponding to each of the initial experiment group model and the initial control group model respectively and conversion values corresponding to each of the initial experiment group model and the initial control group model respectively, wherein the initial weight parameters corresponding to each of the initial experiment group model and the initial control group model respectively are determined by a relationship between the target function and the original user feature information; performing loss processing on the conversion values corresponding to each of the initial experiment group model and the initial control group model respectively, the initial weight parameters corresponding to each of the initial experiment group model and the initial control group model respectively and the original user feature information corresponding to each of the initial experiment group model and the initial control group model respectively to obtain a current target value of the experiment group and a current target value of the control group; respectively judging whether the current target value of the experiment group and the current target value of the control group meet respective preset requirements; if yes, completing establishment of the experiment group model and the control group model; if no, adjusting the initial weight parameters corresponding to each of the initial experiment group model and the initial control group model respectively and returning to the step of performing loss processing on the conversion values corresponding to each of the initial experiment group model and the initial control group model respectively, the initial weight parameters corresponding to each of the initial experiment group model and the initial control group model respectively and the original user feature information corresponding to each of the initial experiment group model and the initial control group model respectively to obtain a current target value of the experiment group and a current target value of the control group until the current target value of the experiment group and the current target value of the control group meet the respective preset requirements; correspondingly, the step of performing loss processing on the conversion values corresponding to each of the initial experiment group model and the initial control group model respectively, the initial weight parameters corresponding to each of the initial experiment group model and the initial control group model respectively and the original user feature information corresponding to each of the initial experiment group model and the initial control group model respectively to obtain a current target value of the experiment group and a current target value of the control group comprises: determining initial experiment group model loss values and initial control group model loss values respectively according to a relationship between the initial weight parameters corresponding to each of the initial experiment group model and the initial control group model respectively and the original user feature information corresponding to each of the initial experiment group model and the initial control group model respectively; determining final loss values of the experiment group users and final loss values of the control group users respectively according to a relationship between the conversion values corresponding to each of the initial experiment group model and the initial control group model respectively and the initial experiment group model loss values and the initial control group model loss values; respectively performing average processing on the final loss values of the experiment group users and the final loss values of the control group users to obtain the current target value of the experiment group and the current target value of the control group.

7. A targeting device for a target population, characterized in that, comprise: a memory for storing a computer program; a processor for implementing steps of the positioning method of the target population according to any one of claims 1 to 5 when executing the computer program.

8. A computer-readable storage medium, characterized in that, The computer program is stored on the computer readable storage medium and is executed by the processor to implement steps of the positioning method of the target population according to any one of claims 1 to 5.

Images