A user screening method and device, electronic equipment and storage medium

Abstract

The application discloses a user screening method and device, electronic equipment and storage medium. The method comprises the following steps: obtaining a user portrait of each target user, and screening a candidate user from the target user according to the user portrait; obtaining a push value and an intention probability of a specified category of each candidate user; the push value and the intention probability are obtained by analyzing the user portrait; and screening a user for pushing information under the specified category from the candidate user according to the push value and the intention probability. After the candidate user is screened from the target user, the push value and the intention probability of the candidate user are obtained, and the candidate user is further screened according to the push value and the intention probability, so that the push precision of the crowd is improved, and the coverage and effectiveness of the push are improved. Moreover, the intention probability of the user is for the category to which the information to be pushed belongs, so that the final user screened is more targeted, and the push precision of the crowd can be further improved.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and more specifically to a user screening method, apparatus, electronic device, and storage medium. Background Technology

[0002] Currently, the similarity group expansion scheme calculates the similarity between candidate users (users other than seed users) and the seed user set, and selects a certain number of candidate users as the similarity group of the seed user set in descending order of similarity.

[0003] However, the screening accuracy of similar groups determined by this similarity method is not high, thus reducing the effectiveness of information push. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of the prior art by providing a user screening method, apparatus, electronic device, and storage medium, which is achieved through the following technical solutions.

[0005] A first aspect of the present invention provides a user screening method, the method comprising:

[0006] Obtain user profiles for each target user, and filter candidate users from each target user based on the user profiles;

[0007] The push value and intention probability for each candidate user are obtained; both the push value and intention probability are obtained by analyzing user profiles.

[0008] Based on the push value and intent probability, users are selected from the candidate users to push information under the specified category.

[0009] In some embodiments of this application, the step of filtering candidate users from various target users based on user profiles includes:

[0010] For each target user, the probability of pushing a request is predicted based on the user profile of the target user; candidate users are then selected from each target user based on the probability of pushing a request.

[0011] In some embodiments of this application, predicting the push probability based on the target user's user profile includes:

[0012] Based on the user profile of the target user, extract user features of a first specified type; input the user features into a trained first classification model, so that the first classification model can predict and output the push probability of the target user based on the user features.

[0013] In some embodiments of this application, the training process for the first classification model includes:

[0014] Based on the user profiles of users who have placed orders, extract user features of a first specified type as positive sample features; based on the user profiles of users who have not placed orders but whose activity level is higher than a first threshold, extract user features of a first specified type as negative sample features; use the positive sample features and the negative sample features to train the constructed first classification model until convergence.

[0015] In some embodiments of this application, the step of filtering candidate users from various target users based on push probability includes:

[0016] Select target users whose probability of receiving push notifications is greater than the second threshold from among the target users as candidate users.

[0017] In some embodiments of this application, obtaining the push value of each candidate user includes:

[0018] For each candidate user, a second specified type of user feature is extracted based on the user profile of the candidate user; the second specified type of user feature is input into different regression models, so that each regression model predicts the sub-push value of the candidate user based on the user feature; the push value of the candidate user is determined based on the sub-push value output by each regression model.

[0019] In some embodiments of this application, the training process for each regression model includes:

[0020] Based on the user profiles of users who have placed orders, extract user features of a second specified type as positive sample features, and obtain the average transaction amount of users who have placed orders as the sample label corresponding to the positive sample features; based on the user profiles of users who have not placed orders, extract user features of a second specified type as negative sample features, and use the first preset value as the sample label corresponding to the negative sample features; use the positive sample features and their corresponding sample labels, and the negative sample features and their corresponding sample labels to train each regression model separately until convergence.

[0021] In some embodiments of this application, after training each regression model separately using positive sample features and their corresponding sample labels, and negative sample features and their corresponding sample labels, the method further includes:

[0022] The negative sample features of users who did not place an order are input into each trained regression model, so that each trained regression model can predict the sub-push value of users who did not place an order based on the negative sample features. The push value of users who did not place an order is determined based on the sub-push value output by each regression model, and the sample label corresponding to the negative sample features is updated using the push value. Each trained regression model is retrained using the negative sample features and their corresponding updated sample labels until convergence.

[0023] In some embodiments of this application, obtaining the push value of each candidate user includes:

[0024] For each candidate user, if the candidate user places an order, the average transaction amount of the candidate user is used as the push value; if the candidate user does not place an order, a second preset value is used as the push value.

[0025] In some embodiments of this application, obtaining the intention probability of each candidate user for a specified category includes:

[0026] For each candidate user, user features of a first specified type are extracted based on the user profile of the candidate user; the user features of the first specified type are input into the second classification model corresponding to the specified category, so that the second classification model can predict the probability of the candidate user's intention to the specified category based on the user features.

[0027] In some embodiments of this application, the training process for the second classification model includes:

[0028] Get users who have placed orders under the specified category, and extract user features of a first specified type as positive sample features based on the user profiles of the users who have placed orders; extract user features of the first specified type as negative sample features based on the user profiles of users who have not placed orders; use the positive sample features and the negative sample features to train the constructed second classification model until convergence.

[0029] In some embodiments of this application, the step of selecting users from candidate users for pushing information under the specified category based on the push value and intention probability includes:

[0030] Select candidate users whose push value is higher than the third threshold and whose intention probability is higher than the fourth threshold from the candidate users, and use them as users to push information under the specified category.

[0031] A second aspect of the present invention provides a user screening device, the device comprising:

[0032] The first filtering module is used to obtain the user profile of each target user and filter candidate users from each target user based on the user profile.

[0033] The acquisition module is used to acquire the push value and the probability of intent for a specified category for each candidate user; both the push value and the probability of intent are obtained by analyzing user profiles.

[0034] The second filtering module is used to filter users from the candidate users for pushing information under the specified category based on the push value and intention probability.

[0035] A third aspect of the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, performs the steps of the method described in the first aspect above.

[0036] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in the first aspect above.

[0037] Based on the user screening method and apparatus described in the first and second aspects above, the present invention has at least the following beneficial effects or advantages:

[0038] After selecting candidate users from the target audience, the system acquires their push value and intent probability for a specific category. Further filtering based on these factors improves the accuracy of push notifications, thereby increasing coverage and effectiveness. Furthermore, since the user's intent probability is tied to the category of the information to be pushed, the final selected users are more targeted, further enhancing the accuracy of push notifications. Attached Figure Description

[0039] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0040] Figure 1 This is a flowchart illustrating an embodiment of a user screening method according to an exemplary embodiment of the present invention;

[0041] Figure 2 According to the present invention Figure 1 The embodiment shown illustrates a schematic diagram of a candidate user push value prediction process.

[0042] Figure 3 According to the present invention Figure 1 The illustrated embodiment presents a schematic diagram of a candidate user's intention probability prediction process for a specified category.

[0043] Figure 4 This is a schematic diagram illustrating the structure of a user screening device according to an exemplary embodiment of the present invention;

[0044] Figure 5 This is a schematic diagram of the hardware structure of an electronic device according to an exemplary embodiment of the present invention;

[0045] Figure 6This is a schematic diagram illustrating the structure of a storage medium according to an exemplary embodiment of the present invention. Detailed Implementation

[0046] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention as detailed in the appended claims.

[0047] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The singular forms “a,” “the,” and “the” used in this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0048] It should be understood that although the terms first, second, third, etc., may be used in this invention to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first information may also be referred to as second information without departing from the scope of this invention, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0049] The quality of existing similarity-based audience segments requires that these users be online and information be pushed to them. The quality of these audience segments is then evaluated by acquiring the behavior of these users. However, this results in losses for the information pusher due to the low accuracy of the filtering.

[0050] To address the aforementioned technical issues, this application proposes a user screening method, which involves obtaining user profiles for each target user, screening candidate users from each target user based on the user profiles, obtaining the push value and intention probability for each candidate user for a specified category, wherein the push value and intention probability are obtained by analyzing the user profiles, and further screening users for pushing information under the specified category from the candidate users based on the push value and intention probability.

[0051] The technical effects that can be achieved based on the above-described technical solution are:

[0052] After selecting candidate users from the target audience, the system acquires their push value and intent probability for a specific category. Further filtering based on these factors improves the accuracy of push notifications, thereby increasing coverage and effectiveness. Furthermore, since the user's intent probability is tied to the category of the information to be pushed, the final selected users are more targeted, further enhancing the accuracy of push notifications.

[0053] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0054] Example 1:

[0055] Figure 1 This is a flowchart illustrating an embodiment of a user filtering method according to an exemplary embodiment of the present invention. This user filtering method can be applied to any network-connected electronic device, such as... Figure 1 As shown, this user filtering method includes the following steps:

[0056] Step 101: Obtain user profiles for each target user and filter candidate users from each target user based on the user profiles.

[0057] In this embodiment, the target user refers to the users selected by the information pusher for expansion. The user profile is obtained by the backend server of the APP application installed on the device through ETL (Extract, Transform, Load) processing based on the user's natural attribute information and social attribute information.

[0058] Among them, user natural attribute information refers to the personal information filled in by users when logging into or registering for the APP. User natural attribute information includes information such as gender, age, income, place of origin, and region of residence.

[0059] User social attribute information is obtained by parsing real-time received user behavior information. This information includes app login activity, preferred categories, purchasing power, and order amount. The user behavior information includes actions such as browsing, clicking, saving, liking, and purchasing generated by the user after logging into the app.

[0060] It should be noted that since the behavioral information and / or personal information that users generate in the APP will change over time, the backend server can dynamically update the user profile at regular intervals so that the user profile can better reflect the user's current characteristics.

[0061] In one optional specific embodiment, the screening process for candidate users can be carried out by predicting the push probability based on the user profile of each target user, and then screening candidate users from each target user based on the push probability.

[0062] Among them, the push probability represents the value of the information push to the target user; the higher the push probability, the higher the value of the information push.

[0063] Based on this, when filtering candidate users according to the push probability, target users whose push probability is greater than the second threshold can be selected from each target user as candidate users.

[0064] Optionally, for the prediction process of push probability, user features of a first specified type can be extracted based on the user profile of the target user, and the user features can be input into a trained first classification model, so that the first classification model can predict the push probability of the target user based on the user features and output it, thereby transforming the problem of predicting the similarity between users into a binary classification problem and expanding the audience coverage of the information.

[0065] The first specified type of user feature can be a set of multiple features or a single feature, reflecting user characteristics such as purchasing power, activity level, gender, and age. To make the first classification model output probability values ​​instead of classification results, a sigmoid function is added to the output layer of the first classification model to convert the classification results into probability outputs.

[0066] It is worth noting that before applying the first classification model, it is necessary to pre-train the first classification model. For the training process of the first classification model, firstly, user features of the first specified type are extracted as positive sample features based on the user profile of users who have placed orders, and user features of the first specified type are extracted as negative sample features based on the user profile of users who have not placed orders and whose activity level is higher than the first threshold. Then, the constructed first classification model is trained using the positive sample features and negative sample features until convergence.

[0067] Among them, the positive sample of users who have placed orders refers to users who have made purchases, while the negative sample refers to users who frequently log in to the APP but have not made any purchases.

[0068] Furthermore, before training, it is necessary to set positive sample labels for positive sample features, such as setting them to "1" to represent positive samples, and set negative sample labels for negative sample features, such as setting them to "0" to represent negative samples.

[0069] Those skilled in the art will understand that this application does not specifically limit the structure of the first classification model for predicting push probability, and can be implemented using any existing classification model structure, such as Random Forest (RF), Gradient Boosting Decision Tree (GBDT), and XGBoost.

[0070] Step 102: Obtain the push value and intention probability for each candidate user for a specified category. The push value and intention probability are obtained by analyzing user profiles.

[0071] The designated category refers to the category to which the information to be pushed belongs. In the field of cosmetic medicine, the category may include double eyelid surgery, eyebrow tattooing, eyebrow lifting, etc.

[0072] Since both the push value and the probability of intent for a specific category are obtained through user profile analysis, the backend server can analyze and obtain these values ​​in advance and dynamically update them at regular intervals. Therefore, when in use, the device can directly obtain the push value and probability of intent for candidate users.

[0073] Those skilled in the art will understand that, since user profiles change over time, in order to ensure the real-time nature of push notification value and intent probability, the current user profile can be analyzed again at the time of use.

[0074] In one optional specific embodiment, the value of a user's push notification is closely related to their purchasing power. Therefore, the push notification value can be predicted directly based on the order amount feature in the user profile. That is, for each candidate user, if the candidate user places an order, the average transaction amount of the candidate user is used as the push notification value. If the candidate user does not place an order, the second preset value is used as the push notification value. This prediction method is simple and easy to operate.

[0075] The second preset value is a default value set for users who have not placed an order.

[0076] It should be noted that the above-described method for predicting push value only considers one characteristic of the user. In order to improve the accuracy of push value prediction, other characteristics of the user also need to be considered for prediction. For the specific prediction process, please refer to the relevant description of the following embodiments, which will not be detailed here.

[0077] Furthermore, the process of predicting the probability of intent for a specified category can be found in the relevant description of the following embodiments, which will not be detailed here.

[0078] Step 103: Based on the push value and intention probability, select users from the candidate users to push information under the specified category.

[0079] Among them, users selected from the candidate users belong to the extended audience for the specified category. By pushing information under the specified category to the extended audience, the click-through rate and conversion rate of the information can be improved.

[0080] In one optional specific embodiment, candidate users whose push value is higher than a third threshold and whose intention probability is higher than a fourth threshold can be selected from the candidate users as users to push information under the specified category.

[0081] This completes the above. Figure 1 The screening process shown involves first identifying candidate users from the target user pool. Then, it obtains the push value and intent probability of each candidate user for a specified category. Based on these factors, the candidate users are further filtered to improve the accuracy of push notifications, thereby increasing the coverage and effectiveness of the push. Furthermore, since the user's intent probability is related to the category of the information to be pushed, the final selected users are more targeted, further enhancing the accuracy of push notifications.

[0082] Example 2:

[0083] Figure 2 According to the present invention Figure 1 The illustrated embodiment presents a schematic diagram of a candidate user push value prediction process. Figure 1 Based on the illustrated embodiments, as Figure 2 As shown, the process of predicting the push value of candidate users includes the following steps:

[0084] Step 201: Extract user features of the second specified type based on the user profiles of candidate users.

[0085] In this embodiment, the second specified type is a subset of the first specified type, and the user features of the second specified type are of higher importance than the user features of other types in the first specified type.

[0086] Optionally, before executing step 201, the user features of the first specified type can be sorted by importance, and then a certain number of user features ranked at the top can be selected as the optimal feature subset, i.e., user features of the second specified type.

[0087] Those skilled in the art will understand that the process of ranking feature importance can be implemented using relevant technologies, and this application does not impose specific limitations on it. For example, the XGBoost embedded feature selection method can be used to rank the importance.

[0088] Step 202: Input the user features of the second specified type into different regression models, so that each regression model can predict the sub-push value of the candidate user based on the user features.

[0089] In this embodiment, in order to improve the accuracy of the push value, multiple different regression models are used to perform regression predictions, and the push value of candidate users is obtained by combining the prediction results of multiple different regression models.

[0090] Before executing step 202, for the training process of each regression model, user features of a second specified type can be extracted as positive sample features based on the user profile of users who have placed orders, and the average transaction amount of users who have placed orders can be obtained as the sample label corresponding to the positive sample features. At the same time, user features of a second specified type can be extracted as negative sample features based on the user profile of users who have not placed orders, and the first preset value can be used as the sample label corresponding to the negative sample features. Then, each regression model is trained separately using the positive sample features and their corresponding sample labels, and the negative sample features and their corresponding sample labels, until convergence.

[0091] For users who did not place an order, since the order transaction amount is 0, a default value (i.e., the first preset value) can be set for them as their sample label.

[0092] Optionally, considering the complexity of prediction and device performance, three different regression models can be used for regression prediction.

[0093] Specifically, these three different regression models can be based on three base learners: Random Forest (RF), Gradient Boosting Decision Tree (GBDT), and XGBoost.

[0094] It should be noted that, in order to improve the accuracy and stability of the model and avoid overfitting, after completing the first training as described above, the training samples can be predicted and the prediction results can be used as training samples for the second training to train each regression model again.

[0095] That is, after training each regression model using positive sample features and their corresponding sample labels, and negative sample features and their corresponding sample labels, the negative sample features of users who have not placed orders are input into each trained regression model. Each trained regression model predicts the sub-push value of users who have not placed orders based on the negative sample features. Then, the push value of users who have not placed orders is determined based on the sub-push value output by each regression model, and the sample labels corresponding to the negative sample features are updated using the push value. Then, each trained regression model is retrained using the negative sample features and their updated sample labels until convergence.

[0096] Step 203: Determine the push value of candidate users based on the sub-push value output by each regression model.

[0097] Optionally, the values ​​of each sub-push can be weighted and summed to obtain the push value of the candidate user.

[0098] This completes the above. Figure 2 The prediction process shown improves the accuracy of push value measurement by using multiple regression models to predict the sub-push values ​​of each user, thus better measuring user value. Furthermore, the second specified type of user feature input to each regression model includes different dimensional features of the user; by comprehensively considering these different dimensional features when predicting push value, accuracy can be further improved.

[0099] Example 3:

[0100] Figure 3 According to the present invention Figure 1 The illustrated embodiment presents a schematic diagram of a candidate user's intention probability prediction process for a specified category. Figures 1 to 2 Based on the illustrated embodiments, as Figure 3 As shown, the intention probability prediction process for a specified category includes the following steps:

[0101] Step 301: Extract user features of the first specified type based on the user profile of the candidate users.

[0102] The extraction process for user features of the first specified type can be found in the relevant description of step 101 above, and will not be repeated here.

[0103] Step 302: Input the user features of the first specified type into the second classification model corresponding to the specified category, so that the second classification model can predict the probability of the candidate user's intention to the specified category based on the user features.

[0104] In this embodiment, in order to ensure the accuracy of the intention probability for a specified category, each category has a dedicated second classification model for prediction.

[0105] It should be noted that the second classification model can adopt the same structure as the first classification model. Although both the first and second classification models are used to predict probabilities, they are trained using different training samples, so their prediction targets are different.

[0106] Optionally, the training process for the second classification model includes: acquiring users who have placed orders under a specified category, extracting user features of a first specified type as positive sample features based on the user profiles of the acquired users who have placed orders, and extracting user features of the first specified type as negative sample features based on the user profiles of users who have not placed orders, and then using the positive sample features and the negative sample features to train the constructed second classification model until convergence.

[0107] Before training, it is also necessary to set positive sample labels for positive sample features, such as setting "1" to represent a positive sample, and set negative sample labels for negative sample features, such as setting "0" to represent a negative sample.

[0108] Therefore, the positive samples for training the first classification model come from users who have placed orders across all categories, while the positive samples for training the second classification model come from users who have placed orders within a specified category. Furthermore, the negative samples for training the first classification model come from relatively active users who have not placed orders, while the negative samples for training the second classification model only need to come from users who have not placed orders. Thus, the intention probability predicted by the second classification model is more targeted, as it targets the probability of a user's purchase intention for a specified category.

[0109] This completes the above. Figure 3 The prediction process shown involves setting up a dedicated second classification model for each category and using the second classification model corresponding to the specified category to predict the probability of candidate users' intention for the specified category. Since the second classification model is highly targeted, the accuracy of the probability of intention for the specified category can be guaranteed.

[0110] Corresponding to the aforementioned embodiments of the user screening method, the present invention also provides embodiments of a user screening device.

[0111] Figure 4 This is a schematic diagram illustrating the structure of a user filtering device according to an exemplary embodiment of the present invention. The device is used to execute the user filtering method provided in any of the above embodiments, such as... Figure 4 As shown, the user screening device includes:

[0112] The first filtering module 410 is used to obtain the user profile of each target user and filter candidate users from each target user based on the user profile.

[0113] The acquisition module 420 is used to acquire the push value and the probability of intent for a specified category for each candidate user; the push value and the probability of intent are both obtained by analyzing user profiles;

[0114] The second filtering module 430 is used to filter users from the candidate users for pushing information under the specified category based on the push value and intention probability.

[0115] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0116] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the present invention according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0117] The present invention also provides an electronic device corresponding to the user screening method provided in the foregoing embodiments, for executing the above-described user screening method.

[0118] Figure 5 This is a hardware structure diagram of an electronic device according to an exemplary embodiment of the present invention. The electronic device includes: a communication interface 601, a processor 602, a memory 603, and a bus 604; wherein the communication interface 601, the processor 602, and the memory 603 communicate with each other through the bus 604. The processor 602 can execute the user selection method described above by reading and executing machine-executable instructions corresponding to the control logic of the user selection method in the memory 603. The specific content of the method is described in the above embodiment and will not be repeated here.

[0119] The memory 603 mentioned in this invention can be any electronic, magnetic, optical, or other physical storage device, and can contain stored information such as executable instructions, data, etc. Specifically, the memory 603 can be RAM (Random Access Memory), flash memory, storage drive (such as hard disk drive), any type of storage disk (such as optical disc, DVD, etc.), or similar storage media, or combinations thereof. Communication between this system network element and at least one other network element is achieved through at least one communication interface 601 (which can be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc., can be used.

[0120] Bus 604 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. The memory 603 is used to store programs, and the processor 602 executes the programs after receiving execution instructions.

[0121] Processor 602 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 602 or by instructions in software form. The processor 602 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application-Specific Integrated Circuit (ASIC), an On-Premises Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor.

[0122] The electronic device provided in this application embodiment and the user screening method provided in this application embodiment are based on the same inventive concept and have the same beneficial effects as the methods they adopt, operate or implement.

[0123] This application also provides a computer-readable storage medium corresponding to the user screening method provided in the foregoing embodiments. Please refer to... Figure 6 As shown, the computer-readable storage medium is an optical disc 30, on which a computer program (i.e., a program product) is stored. When the computer program is run by a processor, it executes the user screening method provided in any of the foregoing embodiments.

[0124] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical and magnetic storage media, which will not be elaborated here.

[0125] The computer-readable storage medium provided in the above embodiments of this application and the user screening method provided in the embodiments of this application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the applications stored therein.

[0126] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the following claims.

[0127] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0128] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A user screening method, characterized in that, The method includes: Obtain user profiles for each target user, and filter candidate users from each target user based on the user profiles; The push value and intention probability for each candidate user are obtained; both the push value and intention probability are obtained by analyzing user profiles. The process of obtaining the push value of each candidate user and the probability of intent for a specified category includes: For each candidate user, extract user features of a first specified type based on the user profile of the candidate user; Based on the user characteristics of the first specified type, determine the intention probability for the specified category; The user features of the second specified type are respectively input into different regression models, so that each regression model predicts the sub-push value of the candidate user based on the user features of the second specified type; the push value of the candidate user is determined based on the sub-push value output by each regression model; the user features of the second specified type are a subset of a specified number of user features of the first specified type whose importance is ranked first; Based on the push value and intent probability, users are selected from the candidate users to push information under the specified category.

2. The method according to claim 1, characterized in that, The step of filtering candidate users from various target users based on user profiles includes: For each target user, predict the probability of push notifications based on the user profile of the target user; Candidate users are selected from each target user group based on the probability of being pushed to them.

3. The method according to claim 2, characterized in that, The step of predicting the push probability based on the target user's user profile includes: Extract user features of a first specified type based on the user profile of the target user; The user features are input into a trained first classification model, which then predicts and outputs the probability of pushing a target user based on the user features.

4. The method according to claim 3, characterized in that, The training process for the first classification model includes: Extract user features of the first specified type as positive sample features based on the user profile of users who have placed orders; Based on the user profiles of users who have not placed an order and whose activity level is higher than the first threshold, user features of the first specified type are extracted as negative sample features. The first classification model is trained using the positive and negative sample features until it converges.

5. The method according to claim 2, characterized in that, The step of filtering candidate users from each target user group based on the push probability includes: Select target users whose probability of receiving push notifications is greater than the second threshold from among the target users as candidate users.

6. The method according to claim 1, characterized in that, The training process for each regression model includes: Based on the user profile of users who have placed orders, extract user features of the second specified type as positive sample features, and obtain the average transaction amount of users who have placed orders as the sample label corresponding to the positive sample features; Based on the user profile of users who have not placed an order, extract the second specified type of user features as negative sample features, and use the first preset value as the sample label corresponding to the negative sample features; Each regression model is trained separately using positive sample features and their corresponding sample labels, and negative sample features and their corresponding sample labels, until convergence.

7. The method according to claim 6, characterized in that, After training each regression model separately using positive sample features and their corresponding sample labels, and negative sample features and their corresponding sample labels, the method further includes: The negative sample features of users who did not place an order are input into each trained regression model, so that each trained regression model can predict the sub-push value of users who did not place an order based on the negative sample features. The push value of the user who did not place an order is determined based on the sub-push value output by each regression model, and the sample label corresponding to the negative sample feature is updated using the push value. Each trained regression model is retrained using the negative sample features and their corresponding updated sample labels until convergence.

8. The method according to claim 1, characterized in that, The process of obtaining the push value for each candidate user includes: For each candidate user, if the candidate user places an order, the average transaction amount of the candidate user is used as the push value. If the candidate user does not place an order, the second preset value will be used as the push value.

9. The method according to claim 1, characterized in that, The step of obtaining the probability of each candidate user's intention for a specified category also includes: The user features of the first specified type are input into the second classification model corresponding to the specified category, so that the second classification model can predict the probability of the candidate user's intention to the specified category based on the user features.

10. The method according to claim 9, characterized in that, The training process for the second classification model includes: Get users who have placed orders under the specified category, and extract user features of the first specified type as positive sample features based on the user profiles of the users who have placed orders. Based on the user profiles of users who have not placed orders, extract user features of the first specified type as negative sample features; The constructed second classification model is trained using the positive sample features and the negative sample features until convergence.

11. The method according to claim 1, characterized in that, The step of selecting users from candidate users to push information under the specified category based on the push value and intention probability includes: Select candidate users whose push value is higher than the third threshold and whose intention probability is higher than the fourth threshold from the candidate users, and use them as users to push information under the specified category.

12. A user screening device, characterized in that, The device includes: The first filtering module is used to obtain the user profile of each target user and filter candidate users from each target user based on the user profile. The acquisition module is used to acquire the push value and the probability of intent for a specified category for each candidate user; both the push value and the probability of intent are obtained by analyzing user profiles. The acquisition module, when acquiring the push value and intention probability for each candidate user and a specified category, specifically performs the following steps: for each candidate user, extracting user features of a first specified type based on the candidate user's user profile; determining the intention probability for the specified category based on the user features of the first specified type; inputting user features of a second specified type into different regression models, so that each regression model predicts the sub-push value of the candidate user based on the user features of the second specified type; determining the push value of the candidate user based on the sub-push value output by each regression model; the user features of the second specified type are a subset consisting of a specified number of user features ranked first in importance among the user features of the first specified type. The second filtering module is used to filter users from the candidate users for pushing information under the specified category based on the push value and intention probability.

13. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method as described in any one of claims 1-11.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by a processor, it implements the steps of the method as described in any one of claims 1-11.

Images