Target user selection method, system, device and storage medium
The method and system enhance advertisement targeting by using prediction models and real-time feedback to improve user selection accuracy, ensuring effective advertisement delivery to engaged users, thus optimizing advertising traffic utilization.
Patent Information
- Application Number
- CN202010657044.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-07-09
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2040-07-09
AI Technical Summary
Existing advertisement targeting methods using DMPs are inaccurate, leading to wasted advertising traffic due to insufficiently precise user selection based on outdated user tags.
A method and system for selecting target users by training prediction models using user matching and feedback, enhancing user selection accuracy through overlapping user identification and model refinement based on real-time feedback.
Improves the accuracy of target user selection, optimizing advertisement delivery by ensuring that advertisements are shown to users most likely to engage with the content, thereby enhancing the efficiency of advertising traffic utilization.
Smart Images

Figure CN113919856B_ABST
Abstract
Description
Background Art
[0002] With the development of Internet technology and the shared bicycle and moped industries, a large amount of user information is stored in the database of shared bicycle service providers. This user information can be used as the object of advertising. When the shared bicycle service provider provides advertising services, the shared bicycle service provider also acts as an advertising pusher and provides advertising services to the merchant side based on the user information stored by itself.
[0003] The traditional way of placing advertisements is through DMP (Data Management Platform), which selects the target users for placing advertisements by matching the user's existing tags. However, such tags usually come from other businesses and are not very accurate for the specific advertisements to be placed. This will cause a waste of advertising traffic. Summary of the invention
[0004] In view of the problems in the prior art, the purpose of the present invention is to provide a target user selection method, system, device and storage medium to improve the accuracy of target user selection when pushing advertisements and improve the utilization efficiency of advertisement delivery traffic.
[0005] The embodiment of the present invention provides a method for selecting a target user, comprising the following steps:
[0006] Receive advertising push demand information from the merchant side;
[0007] Determining the overlapping users of the first user and the second user according to the matching result of the identification information of the first user on the push end and the identification information of the second user on the merchant end;
[0008] Training a first target user prediction model according to the overlapping users;
[0009] The first target user prediction model is used to select a target user from the first users at the push end.
[0010] Optionally, determining the overlapping user of the first user and the second user according to the matching result of the identification information of the first user on the push end and the identification information of the second user on the merchant end includes the following steps:
[0011] Generate a merchant user matching request and push it to the third-party platform, the merchant user matching request includes the push end information and the merchant end information, the third-party platform is configured to obtain the identification information of the first user of the push end and the identification information of the second user of the merchant end according to the merchant user matching request, and perform matching;
[0012] A matching result is obtained from the third-party platform to determine overlapping users of the first user and the second user.
[0013] Optionally, after determining the overlapping users of the first user and the second user, the method further includes determining a target user tag of the overlapping user;
[0014] Training a target user prediction model according to the overlapping users includes the following steps:
[0015] Acquire the feature data of the overlapping users from the user feature database of the push end, and add the feature data to the first training set;
[0016] Determining whether the overlapping user is a positive sample or a negative sample according to the target user label corresponding to the overlapping user, and marking the feature data in the first training set as positive or negative samples;
[0017] The first training set is used to train a first target user prediction model.
[0018] Optionally, selecting a target user from the first user at the push end using the first target user prediction model includes the following steps:
[0019] Acquire feature data of the first user from the user database of the push end and input the feature data into the first target user prediction model to obtain a target probability value of each of the first users output by the first target user prediction model;
[0020] At least one first user with the highest target probability value is selected as the target user.
[0021] Optionally, after selecting at least one first user with the highest target probability value as the target user, the method further includes the following steps:
[0022] After the push end pushes the advertisement information to the user terminal of the target user, feedback information of the pushed user is obtained from the user terminal, and a second target user prediction model is trained according to the feedback information of the pushed user;
[0023] The second target user prediction model is used to select a new target user from the first users at the push end.
[0024] Optionally, the input of the second target user prediction model is the user's feature data, and the output is the user's target probability value;
[0025] The step of training the second target user prediction model according to the feedback information of the pushed user comprises the following steps:
[0026] Acquire the feature data of the pushed user from the user feature database of the push end, and add the feature data to the second training set;
[0027] Adding positive and negative sample labels to the feature data in the second training set according to the feedback information of the pushed user;
[0028] The second training set is used to train a second target user prediction model.
[0029] Optionally, adding positive and negative sample labels to the feature data in the second training set according to the feedback information of the pushed user includes the following steps:
[0030] According to the positive feedback classification rule of the merchant end, it is determined whether the pushed user feedback information belongs to positive feedback or negative feedback, and the corresponding feature data in the second training set are marked as positive and negative samples.
[0031] Optionally, after training the second target user prediction model according to the feedback information of the pushed user, the method further includes the following steps:
[0032] Selecting a first target user and a second target user from the first user at the push end by respectively using the first target user prediction model and the second target user prediction model;
[0033] After the push end pushes the advertisement to the first target user and the second target user respectively, obtaining feedback information of the pushed user from the user terminal;
[0034] Determining a first prediction effect of the first target user prediction model according to the feedback information of the first target user, and determining a second prediction effect of the second target user prediction model according to the feedback information of the second target user;
[0035] The target user prediction model corresponding to the better prediction result is selected as the model used in the subsequent selection of new target users.
[0036] The embodiment of the present invention further provides a target user selection system, which is used to implement the target user selection method. The system includes:
[0037] The demand receiving module is used to receive the advertisement push demand information from the merchant side;
[0038] A user matching module, used to determine the overlapping users of the first user and the second user according to the matching result of the identification information of the first user at the push end and the identification information of the second user at the merchant end;
[0039] A model training module, used for training a first target user prediction model according to the overlapping users;
[0040] The user selection module is used to select a target user from the first users at the push end by using the first target user prediction model.
[0041] The embodiment of the present invention further provides a target user selection device, including:
[0042] processor;
[0043] a memory storing executable instructions of the processor;
[0044] Wherein, the processor is configured to perform the steps of the target user selection method by executing the executable instructions.
[0045] The embodiment of the present invention further provides a computer-readable storage medium for storing a program, wherein the program implements the steps of the target user selection method when executed.
[0046] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure.
[0047] The present invention provides a more efficient target user selection method, system, device and storage medium, which finds seed users based on user matching between a merchant end and a push end, and then expands the population based on the seed users using a target user prediction model to obtain target users. Since the target users have a high similarity with the seed users, the accuracy of target user selection during advertisement push is improved, thereby improving the utilization efficiency of advertisement delivery traffic. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] Other features, objectives and advantages of the present invention will become more apparent from a reading of the detailed description of non-limiting embodiments made with reference to the following accompanying drawings.
[0049] Figure 1 is a flow chart of a target user selection method according to an embodiment of the present invention;
[0050] Figure 2 is a schematic diagram of the interaction between the push end and the merchant end according to an embodiment of the present invention;
[0051] Figure 3 is a schematic diagram of implementing model switching according to an embodiment of the present invention;
[0052] Figure 4 is a schematic diagram of the structure of a target user selection system according to an embodiment of the present invention;
[0053] Figure 5 is a schematic diagram of the structure of a target user selection device according to an embodiment of the present invention;
[0054] Figure 6It is a schematic diagram of the structure of a computer-readable storage medium according to an embodiment of the present invention. DETAILED DESCRIPTION
[0055] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the disclosure will be more comprehensive and complete and to fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
[0056] In addition, the accompanying drawings are only schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices.
[0057] like Figure 1 As shown, an embodiment of the present invention provides a target user selection method, comprising the following steps:
[0058] S100: receiving advertisement push demand information from a merchant end, where the merchant end is the merchant that needs to publish advertisements;
[0059] S200: Determine the overlapping user of the first user and the second user according to the matching result of the identification information of the first user on the push end and the identification information of the second user on the merchant end;
[0060] Here, the push end is the party that provides the advertising push service, and can push advertisements to its existing users based on the advertising push demand information of the merchant end;
[0061] S300: training a first target user prediction model according to the overlapping users;
[0062] S400: Select a target user from the first users at the push end using the first target user prediction model.
[0063] In the target user selection method of this embodiment, firstly, the advertisement push demand information sent by the merchant end is received in step S100, and then the overlapping users are found according to the user matching of the merchant end and the push end in step S200. The overlapping users determined here meet two conditions: they belong to the first user of the push end, that is, they are the optional objects of advertisement push; and they belong to the second user of the merchant end. Here, the second user of the merchant end is the first user selected by the merchant end to match its business, that is, the seed user that the merchant end considers to be more valuable.
[0064] Considering that the number of overlapping users is limited, it is often unable to meet the merchant's advertising traffic demand. Therefore, the target user prediction model is further trained according to the seed user in step S300, and the target user is selected from the first user on the push end based on the target user prediction model in step S400, thereby achieving the population expansion of the seed user. Since the target user obtained in step S400 has a high similarity with the seed user, it can be ensured that the advertisement is pushed to the user who best matches the merchant's advertising push demand, which improves the accuracy of the target user selection during advertisement push, thereby improving the utilization efficiency of the advertisement traffic.
[0065] In this embodiment, in step S100, the advertisement push demand information sent by the merchant terminal may include merchant terminal information (such as merchant terminal name, merchant ID, etc.), advertisement content, advertisement push requirements (such as traffic requirements, time requirements, etc.), etc. After receiving the advertisement push demand information, it is determined that a target user matching the merchant terminal needs to be selected.
[0066] In this embodiment, in order to protect the user's data privacy security, in step S200, matching the identification information of the first user on the push end with the identification information of the second user on the merchant end needs to be implemented on a third-party platform. For the present invention, it is necessary to send a matching request to the third-party platform and obtain a matching result.
[0067] Specifically, step S200 includes the following steps:
[0068] Generate a merchant user matching request, and push the merchant user matching request to the third-party platform, the merchant user matching request including push end information (such as push end ID, push end name, etc.) and merchant end information (such as merchant end ID, merchant name, etc.);
[0069] A matching result is obtained from the third-party platform to determine overlapping users of the first user and the second user.
[0070] Among them, the third-party platform is configured to obtain the identification information of the first user of the push end (pre-stored in the third-party platform or obtained from the user feature database of the push end) and the identification information of the second user of the merchant end (pre-stored in the third-party platform or obtained from the merchant end) after receiving the merchant user matching request, and perform matching.
[0071] For example, the third-party platform can use encrypted mobile phone numbers to match the first user group in the push end with the second user group in the merchant end, thereby obtaining overlapping users in the two user groups. Among them, the identification information of the first user may include the name, ID, mobile phone number, ID card number and other personal information of the first user, and the identification information of the second user may include the name, ID, mobile phone number, ID card number and other personal information of the second user. The third-party platform used here is a platform that is recognized by both the users on the merchant end and the users on the push end and can obtain user information. It can achieve matching of overlapping users under the premise of fully protecting the privacy of user data.
[0072] like Figure 2 As shown, the interactive relationship between the merchant end H100, the push end H200, the third-party platform H300 and the user terminal H400 is shown. The target user selection method in the present invention can be implemented on the server of the push end H200. After selecting the target user, the push end H200 pushes the advertisement to the user terminal H400 of the target user. In other alternative embodiments, the target user selection method can also be implemented on other servers, which can communicate with the push end H200. After selecting the target user, the target user information is sent to the push end H200, and the push end H200 pushes the advertisement to the user terminal H400 of the target user. The user terminal H400 refers to the terminal device used by the user, including but not limited to mobile phones, tablet computers, laptops, etc. The user terminal H400 can see the pushed advertisements on the APP or webpage provided by the push end H200, and the push end H200 can further obtain the user's operation data on the user terminal H400, such as the click data of the advertisement in the user terminal H400, the product purchase data, etc.
[0073] The following takes the merchant end as the bank end and the push end as the shared bicycle service provider as an example to specifically explain the implementation methods of steps S100 and S200. First, corresponding to step S100, the bank end pushes the advertising push demand information of the latest business to the push end. Corresponding to step S200, the push end generates a merchant user matching request based on the bank end information and its own information, and pushes the merchant user matching request to the third-party platform. The third-party platform performs a database collision with the bank user data of the bank end (corresponding to the second user) and the riding user data of the push end (corresponding to the first user). Here, the bank user data of the bank end is preferably the bank seed user pre-screened by the bank end. The third-party platform can obtain overlapping users who are both bank seed users and riding users.
[0074] In this embodiment, the second users of the merchant end may include not only seed users with a relatively high degree of matching, but also non-seed users, which are used for the selection of positive and negative samples when training the model later. Specifically, the identification information of the second user also includes the target user tag of the second user, and the target user tag can indicate whether the second user is a seed user of the merchant end. If it is a seed user, the target user tag is set to a first value, and if it is a non-seed user, the target user tag is set to a second value. This target user tag is pre-added by the merchant end to the identification information of the second user. After determining the overlapping users of the first user and the second user, it also includes determining the target user tag of the overlapping users, which can be obtained from a third-party platform.
[0075] The step S300: training a target user prediction model according to the overlapping users, comprises the following steps:
[0076] The feature data of the overlapping users is obtained from the user feature database of the push end, and added to the first training set; the specific content of the user's feature data here is defined according to the service model of the push end. For example, when the push end is a shared bicycle service provider, the user's feature data may include the user's basic information and riding attributes. The basic information may include the user's age, gender, region, etc. The riding attributes may include the user's riding time attributes, riding duration attributes, riding location attributes, etc.; for another example, when the push end is a travel platform, the user's feature data may include the user's basic information and travel attributes. The travel attributes may include the user's travel location, travel time, travel group type, etc.;
[0077] According to the target user label corresponding to the overlapping user, determine whether the overlapping user belongs to a positive sample or a negative sample, and mark the feature data in the first training set as positive and negative samples. Specifically, if the target user label corresponding to a user is a first value, the corresponding feature data is used as a positive sample; if the target user label corresponding to a user is a second value, the corresponding feature data is used as a negative sample;
[0078] The first training set is used to train the first target user prediction model, for example, a gradient descent method or other training methods may be used to obtain a trained first target user prediction model, which can be used to expand the number of target users in the first user group in step S400.
[0079] The first target user prediction model may be a binary classification model trained based on a machine learning tree model, such as a LightGBM (a gradient boosting framework based on a decision tree algorithm) model, a decision tree model, a random forest model, etc. In other alternative implementations, the first target user prediction model may also adopt other types of machine learning models that can achieve user classification, and is not limited to the model types listed here.
[0080] In this embodiment, the input of the first target user prediction model is the user's feature data, and the output is the user's target probability value. In step S400, the target user can be selected from the first users according to the user's target probability value output by the first target user prediction model.
[0081] Specifically, in step S400, selecting a target user from the first user at the push end using the first target user prediction model includes the following steps:
[0082] Acquire feature data of all first users from the user database of the push end and input the feature data into the first target user prediction model to obtain a target probability value of each first user, where the target probability value indicates the probability score of a user being a target user and can be a value between 0 and 1;
[0083] A target user is selected according to the target probability value of each of the first users, and at least one first user with the highest target probability value is selected as the target user. For example, the first users can be sorted from high to low according to the target probability value, and the number of target users n is determined according to the advertising push requirements of the merchant side, and n first users with the highest target probability values are selected as the target users.
[0084] Furthermore, in this embodiment, after the first batch of target users is obtained according to the first target user prediction model, it is possible to effectively find the target users that match the merchant end among the first users of the push end during the cold start. After the push end pushes the advertisement to the target user, the feedback information of the target user can be obtained from the user terminal. Here, the feedback information is whether the user clicks on the advertisement, whether the user accepts the product in the advertisement, etc. After receiving the feedback information of the target user, the first target user prediction model can be further optimized and corrected according to the feedback information of the target user, or the target user prediction model can be trained based on the feedback information of the user to whom the advertisement has been pushed, so that a better target user can be obtained based on the real scene information.
[0085] like Figure 3 Specifically, after the step S400: selecting at least one first user with the highest target probability value as the target user, the following steps are further included:
[0086] The push end is configured to push the advertisement information to the target user selected in step S400;
[0087] S510: after the push end pushes the advertisement information to the user terminal of the target user, the target user selected in step S400 is regarded as a pushed user, feedback information of the pushed user is obtained from the user terminal, and a second target user prediction model is trained according to the feedback information of the pushed user;
[0088] S520: Select a new target user from the first user at the push end using the second target user prediction model, where the second target user prediction model is a target user prediction model trained based on real delivery data.
[0089] In this embodiment, the input of the second target user prediction model is the user's feature data, and the output is the user's target probability value. In the step S520, the second target user prediction model is used to select a new target user from the first user at the push end, that is, the feature data of the first user at the push end is input into the second target user prediction model, the target probability value of each user output by the second target user prediction model is obtained, and a new target user is selected from high to low according to the target probability value as the target user for the next round of advertising delivery.
[0090] In this embodiment, the step S510: training the second target user prediction model according to the feedback information of the pushed user includes the following steps:
[0091] Acquire the feature data of the pushed user from the user feature database of the push end, and add them to the second training set. If the feature data of the pushed user is less, the second training set may further include the feature data of the overlapping users obtained in step S200, and perform model optimization training through the added feature data of the pushed user. If the feature data of the pushed user is sufficient, the second training set may only include the feature data of the pushed user, and retrain the model completely based on the real delivery scenario data;
[0092] According to the feedback information of the pushed user, positive and negative sample labels are added to the feature data in the second training set. For example, if the feedback information of a user is that the advertisement content has been accepted, the feature data of the user is a positive sample, and if the feedback information of a user is that the advertisement content has not been accepted, the feature data of the user is a negative sample;
[0093] The second target user prediction model is trained using the second training set. The second target user prediction model may use the same type of model as the first target user prediction model, such as the aforementioned LightGBM, decision tree, random forest and other binary classification models. The second target user prediction model may be an optimized target user prediction model trained on the basis of the first target user prediction model using the second training set, or may be a target user prediction model retrained using the second training set.
[0094] Further, adding positive and negative sample labels to the feature data in the second training set according to the feedback information of the pushed user includes the following steps:
[0095] According to the positive feedback classification rule of the merchant end, it is determined whether the pushed user feedback information belongs to positive feedback or negative feedback, and the corresponding feature data in the second training set are marked as positive and negative samples.
[0096] Here, the positive feedback classification rule of the merchant end is a judgment rule for identifying whether the user's feedback information is positive feedback or negative feedback. For example, for a merchant end, it defines that a user clicks on an advertisement as accepting an advertisement. When a user's click operation is received, the user's feedback information is regarded as positive feedback, and the user's feature data is regarded as a positive sample. When a user's click operation is not received, the user's feedback information is regarded as negative feedback, and the user's feature data is regarded as a negative sample. For another example, for another merchant end, it defines that a user purchases an advertised product as accepting an advertisement. Only when the user's purchase data is received, the user's feedback information is regarded as positive feedback, and the user's feature data is regarded as a positive sample.
[0097] like Figure 3As shown, in this embodiment, the time of model switching can be further controlled, that is, the second target user prediction model is switched only when the prediction effect of the second target user prediction model is better than that of the first target user prediction model, otherwise the first target user prediction model is continued to be used. The comparison of the prediction effect can be carried out by advertising traffic segmentation, that is, the advertising traffic is divided into two parts, one part selects the first target user through the first target user prediction model, and the other part selects the second target user through the second target user prediction model.
[0098] Specifically, after the step S510: training the second target user prediction model according to the feedback information of the pushed user, the following steps are also included:
[0099] S511: Selecting a first target user and a second target user from the first user at the push end by using the first target user prediction model and the second target user prediction model respectively;
[0100] The push end is configured to distribute the advertisement traffic between the first target user and the second target user. For example, to push advertisements to 1,000 users in total, 800 first target users are selected using the first target user prediction model, and 200 second target users are selected using the second target user prediction model.
[0101] S512: After the push end pushes the advertisement to the first target user and the second target user respectively, feedback information of the pushed user is obtained from the user terminal, where the pushed user refers to the first target user and the second target user selected in step S511;
[0102] S513: Determine a first prediction effect of the first target user prediction model according to the feedback information of the first target user, and determine a second prediction effect of the second target user prediction model according to the feedback information of the second target user;
[0103] S514: Compare the first prediction effect and the second prediction effect;
[0104] The first prediction effect can be characterized by the ratio of the number of positive feedback users among the first target users to the total number of the first target users, and the second prediction effect can be characterized by the ratio of the number of positive feedback users among the second target users to the total number of the second target users.
[0105] If the first prediction effect is better than the second prediction effect, then continue with step S400, that is, continue to select the first target user prediction model to select a subsequent new target user from the first user, and after the push end pushes the advertisement to the new target user, then continue with step S510 to obtain the latest feedback information and optimize and train the second target user prediction model;
[0106] If the second prediction effect is better than the first prediction effect, continue with step S520, that is, select the second target user prediction model to select a subsequent new target user from the first user, and after the push end pushes the advertisement to the new target user, continue with step S510 to obtain the latest feedback information and continuously optimize and train the second target user prediction model.
[0107] In other alternative embodiments, if the second prediction effect is better than the first prediction effect, the traffic can also be gradually switched to the second target user prediction model, that is, the proportion of the second target users among all target users can be increased. For example, if a total of 1,000 users are pushed advertisements, 200 first target users are selected using the first target user prediction model, and 800 second target users are selected using the second target user prediction model. After the second prediction effect reaches a preset standard (for example, the ratio of the number of second target users with positive feedback to the total number of second target users is higher than a preset threshold), all traffic can be switched to the second target user prediction model.
[0108] like Figure 4 As shown, an embodiment of the present invention further provides a target user selection system for implementing the target user selection method, the system comprising:
[0109] The demand receiving module M100 is used to receive the advertisement push demand information from the merchant side;
[0110] A user matching module M200, configured to determine the overlapping users of the first user and the second user according to the matching result of the identification information of the first user at the push end and the identification information of the second user at the merchant end;
[0111] A model training module M300 is used to train a first target user prediction model according to the overlapping users;
[0112] The user selection module M400 is used to select a target user from the first users at the push end by using the first target user prediction model.
[0113] In the target user selection system of this embodiment, the demand receiving module M100 first receives the advertising push demand information sent by the merchant side, and then the user matching module M200 finds the overlapping users according to the user matching of the merchant side and the push side. The overlapping users are the seed users that best meet the merchant side's needs. Then, the model training module M300 trains the target user prediction model according to the seed users, and the user selection module M400 selects the target user from the first user at the push side based on the target user prediction model, thereby achieving the population expansion of the seed user. Since the target user obtained by the user selection module M400 has a high similarity with the seed user, it can ensure that the advertisement is pushed to the user who best matches the merchant side's advertisement push demand, which improves the accuracy of the target user selection during advertisement push, thereby improving the utilization efficiency of the advertisement delivery traffic.
[0114] In the target user selection system of the present invention, the functions of each module can be implemented by adopting the specific implementation methods of each step in the above-mentioned target user selection method. Specifically, the demand receiving module M100 can adopt the specific implementation method of the above-mentioned step S100 to receive the advertising push demand information, the user matching module M200 can adopt the specific implementation method of the above-mentioned step S200 to obtain the overlapping users, the model training module M300 can adopt the specific implementation method of the above-mentioned step S300 to obtain the target user prediction model, and the user selection module M400 can adopt the specific implementation method of the above-mentioned step S400 to select the target user.
[0115] Furthermore, in this embodiment, the model training module M300 can also be used to obtain feedback information of the pushed user from the user terminal after the push end pushes the advertising information to the user terminal of the target user, and train a second target user prediction model based on the feedback information of the pushed user. In this embodiment, the input of the second target user prediction model is the user's feature data, and the output is the user's target probability value.
[0116] The user selection module M400 can also be used to select a target user from the first user at the push end using the second target user prediction model.
[0117] Furthermore, the user selection module M400 can also use the method of the above steps S511 to S514 to compare the prediction effects of the first target user prediction model and the second target user prediction model, so as to determine the traffic segmentation of the first target user prediction model and the second target user prediction model, and when the second prediction effect of the second target user prediction model reaches a preset standard, the second target user prediction model is fully used to select the target user from the first user at the push end.
[0118] An embodiment of the present invention also provides a target user selection device, comprising a processor; a memory, in which executable instructions of the processor are stored; wherein the processor is configured to execute the steps of the target user selection method by executing the executable instructions.
[0119] It will be appreciated by those skilled in the art that various aspects of the present invention may be implemented as systems, methods or program products. Therefore, various aspects of the present invention may be specifically implemented in the following forms, namely: complete hardware implementation, complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, which may be collectively referred to herein as "circuits", "modules" or "platforms".
[0120] Refer to the following Figure 5 The electronic device 600 according to this embodiment of the present invention is described. Figure 5 The electronic device 600 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.
[0121] like Figure 5 As shown, the electronic device 600 is in the form of a general computing device. The components of the electronic device 600 may include, but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different system components (including the storage unit 620 and the processing unit 610), a display unit 640, etc.
[0122] The storage unit stores a program code, which can be executed by the processing unit 610, so that the processing unit 610 performs the steps according to various exemplary embodiments of the present invention described in the target user selection method section of this specification. For example, the processing unit 610 can perform the following steps: Figure 1 Follow the steps shown in .
[0123] The storage unit 620 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 6201 and / or a cache memory unit 6202 , and may further include a read-only memory unit (ROM) 6203 .
[0124] The storage unit 620 may also include a program / utility 6204 having a set (at least one) of program modules 6205, such program modules 6205 including but not limited to: an operating system, one or more application programs, other program modules and program data, each of which or some combination may include the implementation of a network environment.
[0125] Bus 630 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.
[0126] The electronic device 600 may also communicate with one or more external devices 700 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 600, and / or communicate with any device that enables the electronic device 600 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 650. Furthermore, the electronic device 600 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 660. The network adapter 660 may communicate with other modules of the electronic device 600 via a bus 630. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.
[0127] The embodiment of the present invention further provides a computer-readable storage medium for storing a program, which implements the steps of the target user selection method when executed. In some possible implementations, various aspects of the present invention may also be implemented in the form of a program product, which includes a program code, and when the program product is executed on a terminal device, the program code is used to cause the terminal device to execute the steps of various exemplary embodiments of the present invention described in the above target user selection method section of this specification.
[0128] refer to Figure 6 As shown, a program product 800 for implementing the above method according to an embodiment of the present invention is described, which can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be executed on a terminal device, such as a personal computer. However, the program product of the present invention is not limited thereto. In this document, a readable storage medium can be any tangible medium containing or storing a program, which can be used by or in combination with an instruction execution system, an apparatus or a device.
[0129] The program product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.
[0130] The computer readable storage medium may include a data signal propagated in a baseband or as part of a carrier wave, wherein a readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The readable storage medium may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by an instruction execution system, an apparatus, or a device or used in combination with it. The program code contained on the readable storage medium may be transmitted with any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the above.
[0131] Program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages such as Java, C++, etc., and conventional procedural programming languages such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).
[0132] In summary, the present invention provides a more efficient target user selection method, system, device and storage medium, which finds seed users based on user matching between the merchant end and the push end, and then expands the population based on the seed users using a target user prediction model to obtain the target users. Since the target users have a high similarity with the seed users, the accuracy of target user selection during advertisement push is improved, thereby improving the utilization efficiency of advertisement delivery traffic.
[0133] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.
Claims
1. A target user selection method, characterized in that: The steps include: Receive advertising push demand information from the merchant side; Determine the overlapping users of the first user and the second user according to the matching result of the identification information of the first user on the push end and the identification information of the second user on the merchant end; wherein the matching result is obtained from a third-party platform, and the third-party platform is configured to match according to the identification information of the first user on the push end and the identification information of the second user on the merchant end; Training a first target user prediction model according to the overlapping users; The first target user prediction model is used to select a target user from the first users at the push end, including: obtaining feature data of the first user from a user database at the push end and inputting the feature data into the first target user prediction model to obtain a target probability value for each of the first users output by the first target user prediction model; and selecting at least one first user with the highest target probability value as the target user.
2. The target user selection method according to claim 1, characterized in that: After determining the overlapping users of the first user and the second user, the method further includes determining a target user tag of the overlapping user; Training a target user prediction model according to the overlapping users includes the following steps: Acquire the feature data of the overlapping users from the user feature database of the push end, and add the feature data to the first training set; Determining whether the overlapping user is a positive sample or a negative sample according to the target user label corresponding to the overlapping user, and marking the feature data in the first training set as positive or negative samples; The first training set is used to train a first target user prediction model.
3. The target user selection method according to claim 1, characterized in that: After selecting at least one first user with the highest target probability value as the target user, the following steps are also included: After the push end pushes the advertisement information to the user terminal of the target user, feedback information of the pushed user is obtained from the user terminal, and a second target user prediction model is trained according to the feedback information of the pushed user; The second target user prediction model is used to select a new target user from the first users at the push end.
4. The target user selection method according to claim 3, characterized in that: The input of the second target user prediction model is the user's feature data, and the output is the user's target probability value; The step of training the second target user prediction model according to the feedback information of the pushed user comprises the following steps: Acquire the feature data of the pushed user from the user feature database of the push end, and add the feature data to the second training set; Adding positive and negative sample labels to the feature data in the second training set according to the feedback information of the pushed user; The second training set is used to train a second target user prediction model.
5. The target user selection method according to claim 4, characterized in that: Adding positive and negative sample labels to the feature data in the second training set according to the feedback information of the pushed user includes the following steps: According to the positive feedback classification rule of the merchant end, it is determined whether the pushed user feedback information belongs to positive feedback or negative feedback, and the corresponding feature data in the second training set are marked as positive and negative samples.
6. The target user selection method according to claim 4, characterized in that: After the second target user prediction model is trained according to the feedback information of the pushed user, the following steps are also included: Selecting a first target user and a second target user from the first user at the push end by respectively using the first target user prediction model and the second target user prediction model; After the push end pushes the advertisement to the first target user and the second target user respectively, obtaining feedback information of the pushed user from the user terminal; Determining a first prediction effect of the first target user prediction model according to the feedback information of the first target user, and determining a second prediction effect of the second target user prediction model according to the feedback information of the second target user; The target user prediction model corresponding to the better prediction result is selected as the model used in the subsequent selection of new target users.
7. A target user selection system, used to implement the target user selection method according to any one of claims 1 to 6, characterized in that: The system comprises: The demand receiving module is used to receive the advertisement push demand information from the merchant side; A user matching module, used to determine the overlapping users of the first user and the second user according to the matching result of the identification information of the first user at the push end and the identification information of the second user at the merchant end; A model training module, used for training a first target user prediction model according to the overlapping users; The user selection module is used to select a target user from the first users at the push end by using the first target user prediction model.
8. A target user selection device, characterized in that: include: processor; a memory storing executable instructions of the processor; The processor is configured to execute the steps of the target user selection method according to any one of claims 1 to 6 by executing the executable instructions.
9. A computer-readable storage medium for storing a program, characterized in that: When the program is executed, the steps of the target user selection method according to any one of claims 1 to 6 are implemented.
