Advertisement putting method and device, storage medium and electronic equipment

Abstract

The invention relates to an advertisement putting method and device, a storage medium and electronic equipment. The method comprises the following steps: acquiring conventional characteristics of each user, inputting the conventional characteristics of each user and a first processing variable into a target multi-task model, and outputting an estimated value of each task of each user under a delivery condition; inputting the conventional features of each user and the second processing variable into a target multi-task model, and outputting an estimated value of each task of each user under the condition of no delivery; according to the delivery target, fusing the pre-estimated value of each task of each user under the delivery condition to obtain a delivery fusion score of each user, and fusing the pre-estimated value of each task of each user under the non-delivery condition to obtain a non-delivery fusion score of each user; and calculating the optimal gain fusion score of each user according to the released fusion score and the non-released fusion score of each user. The technical problem that advertisement putting is difficult to give consideration to multiple targets under dynamic business requirements is solved.

Description

Technical Field

[0001] The present application relates to the field of multimedia intelligent processing, and in particular to a method, device, storage medium and electronic device for advertising delivery. Background Art

[0002] In the field of advertising, advertisers use the RTA (Real-Time Advertising) system to estimate the value of users based on their characteristics and behaviors, and thus decide whether to participate in bidding and bidding strategies. As market competition intensifies, advertisers' business goals are constantly changing. For example, in some periods, they focus on increasing the number of conversions, while in other periods, they focus more on reducing the cost rate. This changing demand places higher demands on the flexibility and responsiveness of the advertising system. Currently, advertising delivery systems usually use multi-task learning models to simultaneously estimate multiple business goals, such as the number of pull-ups, playback time, advertising revenue, and membership revenue. However, due to possible conflicts between business goals, such as increasing the number of conversions may lead to an increase in the cost rate, traditional models are difficult to take into account multiple goals. In addition, it is usually necessary to retrain the model or adjust the data set for different business needs, which increases time and resource costs, and the response speed is limited. Summary of the invention

[0003] The present application provides a method, device, storage medium and electronic device for advertising delivery to solve the technical problem that it is difficult to balance multiple objectives in advertising delivery under dynamic business needs.

[0004] In a first aspect, the present application provides a method for advertising delivery, including: obtaining general features of each user, and inputting the general features of each user and a first processing variable into a target multi-task model, so that the target multi-task model performs a multi-task evaluation of the delivery situation of each user, and outputs an estimated value of each task of each user under the delivery situation, wherein the first processing variable is delivery; inputting the general features of each user and a second processing variable into the target multi-task model, so that the target multi-task model performs a multi-task evaluation of the non-delivery situation of each user, and outputs an estimated value of each task of each user under the non-delivery situation, wherein the second processing variable is non-delivery; according to the delivery target, fusing the estimated value of each task of each user under the delivery situation to obtain a delivery fusion score for each user, and fusing the estimated value of each task of each user under the non-delivery situation to obtain a non-delivery fusion score for each user; according to the delivery fusion score and the non-delivery fusion score of each user, calculating the optimal gain fusion score of each user, wherein the optimal gain fusion score is used to determine the target user for advertising delivery.

[0005] In a second aspect, the present application provides an apparatus for advertising delivery, comprising: a first evaluation module, used to obtain the general features of each user, and input the general features of each user and the first processing variable into a target multi-task model, so that the target multi-task model performs a multi-task evaluation of the delivery situation of each user, and outputs an estimated value of each task of each user under the delivery situation, wherein the first processing variable is delivery; a second evaluation module, used to input the general features of each user and the second processing variable into the target multi-task model, so that the target multi-task model performs a multi-task evaluation of the non-delivery situation of each user, and outputs an estimated value of each task of each user under the non-delivery situation, wherein the second processing variable is non-delivery; a fusion module, used to fuse the estimated values ​​of each task of each user under the delivery situation according to the delivery target, to obtain a delivery fusion score of each user, and to fuse the estimated values ​​of each task of each user under the non-delivery situation, to obtain a non-delivery fusion score of each user; a calculation module, used to calculate the optimal gain fusion score of each user based on the delivery fusion score and the non-delivery fusion score of each user, wherein the optimal gain fusion score is used to determine the target user for advertising delivery.

[0006] As an optional example, the above-mentioned general features include any one or more of basic information, behavior records, interaction preferences, value information, and advertising placement information, and the above-mentioned tasks include any one or more of whether to start, playback time, advertising revenue, membership revenue, and total revenue.

[0007] As an optional example, the above-mentioned device also includes: an acquisition module, used to acquire a training data set before inputting the conventional features and the first processing variables of each of the above-mentioned users into the target multi-task model, wherein the above-mentioned training data set includes training data of multiple training users, and the above-mentioned training data includes conventional features, processing variables and all tasks, wherein the above-mentioned processing variables are the above-mentioned first processing variables or the above-mentioned second processing variables; a training module, used to perform multiple iterative training on the initial multi-task model according to the above-mentioned training data set to obtain the above-mentioned target multi-task model.

[0008] As an optional example, the above-mentioned fusion module includes: a first calculation unit, used to use the Pareto optimal algorithm to calculate the first optimal weight according to the above-mentioned delivery target and the estimated value of each task of each user under the delivery condition, wherein the above-mentioned delivery target includes constraints and optimization objectives; a first processing unit, used to take each user as the current user and perform the following operations on the above-mentioned current user: according to the above-mentioned first optimal weight, weighted fusion is performed on the estimated value of each task of the above-mentioned current user under the delivery condition to obtain the delivery fusion score of the above-mentioned current user.

[0009] As an optional example, the fusion module includes: a first calculation unit, used to use the Pareto optimal algorithm to calculate the second optimal weight according to the delivery target and the estimated value of each task of each user in the case of no delivery, wherein the delivery target includes constraints and optimization objectives; a second processing unit, used for each user as the current user, to perform the following operation on the current user: according to the second optimal weight, weighted fusion is performed on the estimated value of each task of the current user in the case of no delivery to obtain the non-delivery fusion score of the current user.

[0010] As an optional example, the above-mentioned calculation module includes: a third processing unit, used to take each user as the current user, and perform the following operations on the above-mentioned current user: calculate the difference between the above-mentioned current user's delivery fusion score and the above-mentioned current user's non-delivery fusion score; determine the above-mentioned difference as the optimal gain fusion score of the above-mentioned current user.

[0011] As an optional example, the above-mentioned device also includes: a determination module, which is used to determine the user whose optimal gain fusion score is greater than the target threshold as the target user for advertising delivery after calculating the optimal gain fusion score of each user based on the delivery fusion score and non-delivery fusion score of each user.

[0012] In a third aspect, the present application provides a storage medium storing a computer program, wherein the computer program executes the above-mentioned method of advertising delivery when executed by a processor.

[0013] In a fourth aspect, the present application further provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the above-mentioned method of advertising delivery through the computer program.

[0014] The above technical solution provided by the embodiment of the present application has the following advantages compared with the prior art:

[0015] The present application adopts the method of obtaining the conventional features of each user, and inputting the conventional features and the first processing variable of each user into the target multi-task model, so that the target multi-task model performs a multi-task evaluation of the delivery situation of each user, and outputs the estimated value of each task of each user under the delivery situation, wherein the first processing variable is delivery; inputting the conventional features and the second processing variable of each user into the target multi-task model, so that the target multi-task model performs a multi-task evaluation of the non-delivery situation of each user, and outputs the estimated value of each task of each user under the non-delivery situation, wherein the second processing variable is non-delivery; according to the delivery target, the estimated value of each task of each user under the delivery situation is fused to obtain the delivery fusion score of each user, and the estimated value of each task of each user under the non-delivery situation is fused to obtain the non-delivery fusion score of each user; According to the above-mentioned delivery fusion score and non-delivery fusion score of each user, the optimal gain fusion score of each user is calculated, wherein the above-mentioned optimal gain fusion score is used for the method of determining the target users for advertising delivery. In the above-mentioned method, the conventional characteristics of each user are obtained, and they are input into the target multi-task model with the processing variables (delivery or non-delivery), and the multi-task estimated values ​​under delivery and non-delivery conditions are obtained respectively. Then, according to the delivery target, the multi-task estimated values ​​of the user under delivery and non-delivery conditions are fused, and the delivery fusion score and the non-delivery fusion score are calculated respectively. Finally, the optimal gain fusion score of each user is calculated by the delivery fusion score and the non-delivery fusion score, and finally used to screen the target users for advertising delivery, thereby achieving rapid response to dynamic needs, and meeting different delivery targets through fusion score adjustment to improve delivery efficiency, thereby solving the technical problem that it is difficult for advertising delivery to take into account multiple targets under dynamic business needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0018] One or more embodiments are exemplarily described by pictures in the corresponding drawings, and these exemplified descriptions do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0019] Figure 1 is a flow chart of an optional method for advertising delivery according to an embodiment of the present application;

[0020] Figure 2 is a specific implementation flow chart of an optional advertising delivery method according to an embodiment of the present application;

[0021] Figure 3 is a schematic structural diagram of an optional device for advertising delivery according to an embodiment of the present application;

[0022] Figure 4 is a schematic diagram of an optional electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0024] The disclosure below provides many different embodiments or examples to realize the different structures of the present application. In order to simplify the disclosure of the present application, the parts and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the present application. In addition, the present application can repeat reference numbers and / or letters in different examples. This repetition is for the purpose of simplification and clarity, and does not itself indicate the relationship between the various embodiments and / or settings discussed.

[0025] According to a first aspect of an embodiment of the present application, a method for advertising delivery is provided. Optionally, as follows: Figure 1 As shown, the above method includes:

[0026] S102, obtaining the conventional features of each user, and inputting the conventional features of each user and the first processing variable into the target multi-task model, so that the target multi-task model performs a multi-task evaluation of the delivery situation of each user, and outputs an estimated value of each task of each user under the delivery situation, wherein the first processing variable is the delivery;

[0027] S104, inputting the conventional features and the second processing variable of each user into the target multi-task model, so that the target multi-task model performs a multi-task evaluation on each user in the case of no delivery, and outputs an estimated value of each task of each user in the case of no delivery, wherein the second processing variable is no delivery;

[0028] S106, according to the delivery target, the estimated values ​​of each task of each user in the delivery situation are integrated to obtain the delivery integrated score of each user, and the estimated values ​​of each task of each user in the non-delivery situation are integrated to obtain the non-delivery integrated score of each user;

[0029] S108, calculating the optimal gain fusion score of each user according to the delivery fusion score and the non-delivery fusion score of each user, wherein the optimal gain fusion score is used to determine the target user for advertisement delivery.

[0030] Optionally, in this embodiment, if Figure 2 As shown in the specific implementation flowchart, the general characteristics of each user are first collected, including basic information (such as age, gender, occupation, etc.), behavior records (such as recent behavior data, historical behavior data, etc.), interaction preferences (such as advertising click-through rate, content type preference, etc.), value information (such as consumption amount, active days, etc.), advertising delivery information (number of advertising competitions, bidding success rate, exposure frequency, etc.). These characteristics serve as the basic information for model input and comprehensively describe the user's behavior and attributes. The general characteristics of each user and the first processing variable (delivery) are used as input and passed to the target multi-task model. The model evaluates the multi-task goals under delivery and outputs the estimated values ​​of the delivered multi-tasks, including whether to start, playback time, advertising revenue, membership revenue and other tasks. Once again, the general characteristics of each user and the second processing variable (not delivered) are used as input and passed to the target multi-task model. The model evaluates the multi-task goals under non-delivery and outputs similar multi-task estimates for non-delivery. According to the preset delivery target, the estimated value of each task under delivery is integrated to calculate the delivery integration score of each user. The integration process is based on the Pareto optimal algorithm, and different weight combinations are used to comprehensively evaluate the multi-task targets. Similarly, the estimated value of each task under non-delivery is integrated to calculate the non-delivery integration score of each user. The delivery target is the specific business goal or effect that the advertiser hopes to achieve during the advertising process, which is usually set according to the purpose of the advertising campaign. The delivery target can be to maximize the number of conversions under the premise of meeting the cost rate standard, to maximize the conversion revenue under the premise of meeting the cost rate standard, to maximize the total revenue, etc. Based on the delivery integration score and the non-delivery integration score, the optimal gain integration score of each user is calculated. The integration score represents the relative gain of the user value under delivery and non-delivery conditions, and is used to measure the optimization effect of advertising delivery on the user conversion target. According to the optimal gain integration score, high-value users are selected as the target users for advertising delivery. By setting a threshold limit, user groups can be accurately selected to improve the effect and efficiency of advertising delivery.

[0031] Optionally, in this embodiment, by estimating and integrating the two situations of delivery and non-delivery, the impact of advertising delivery on user behavior is accurately measured to help advertisers quickly identify high-value users. The unified multi-task model supports flexible applications in multi-target scenarios at the same time, improving the practicality and delivery efficiency of the model.

[0032] As an optional example, conventional features include any one or more of basic information, behavior records, interaction preferences, value information, and advertising placement information, and tasks include any one or more of whether to start, playback duration, advertising revenue, membership revenue, and total revenue.

[0033] Optionally, in this embodiment, conventional features are data that describe user behavior, attributes, and their interactions with advertisements, and may include the following categories: basic information: describes the user's static characteristics, which may include age, gender, occupation, region, etc.; behavior records: describes the user's dynamic behavior data, reflecting the user's recent activities, which may include the frequency of visits, viewing content categories, shopping records, device usage time, etc. in the past 7 days or 30 days; interaction preferences: describes the user's preferences for advertisements, content, or products, which may include category preferences for clicking on advertisements, types of content viewed, keywords searched, etc.; value information: describes the user's potential contribution value to the platform, which may include historical consumption amounts, ad click-through rates, paid membership status, estimated lifetime value, etc.; advertising delivery information: describes the user's historical behavior and responses in advertising delivery, which may include whether the ad has been clicked, post-click behavior (such as purchase or registration), number of ad exposures, ad conversion rate, etc.

[0034] Optionally, in this embodiment, the task is the target variable that the model needs to predict and optimize, which may include the following categories: whether to start: whether the user starts the application or performs related operations after the advertisement is placed; playback time: the cumulative time the user watches the content; advertising revenue: the user's direct contribution to the revenue from advertising; membership revenue: membership subscription revenue or renewal revenue brought by advertising; total revenue: the overall revenue brought by the user after the advertisement is placed, including the sum of advertising revenue and membership revenue.

[0035] Optionally, in this embodiment, by combining multiple features and tasks, the effects of advertising delivery in different dimensions can be comprehensively evaluated, thereby achieving a more accurate and efficient advertising delivery strategy.

[0036] As an optional example, before inputting the conventional features and the first processing variable of each user into the target multi-task model, the method further includes:

[0037] Obtaining a training data set, wherein the training data set includes training data of multiple training users, the training data includes conventional features, processing variables, and all tasks, wherein the processing variable is a first processing variable or a second processing variable;

[0038] The initial multi-task model is trained iteratively multiple times according to the training data set to obtain the target multi-task model.

[0039] Optionally, in this embodiment, in order to improve the accuracy of the delivery strategy in the evaluation of the effect of advertising, it is first necessary to collect and process user data, and use a multi-task model to predict the effect of advertising. The training data set is the basis for building a multi-task model. The data set contains the historical behavior and attribute information of multiple users, and covers the following key data: conventional features, processing variables (delivery or non-delivery) and all tasks. Using the above training data set, a multi-task model is preliminarily trained. The model learns information of different dimensions such as user behavior patterns, advertising response preferences, and advertising revenue prediction through multiple iterations, and gradually optimizes the model parameters. The multi-task model can explore the potential correlation between different advertising effects by integrating multiple task outputs, thereby improving the prediction accuracy of the effect of advertising. During the training process, the model will model the conventional features and processing variables according to the user data, and learn the law of user behavior changes before and after the advertising is delivered. Through repeated iterative training processes, a target multi-task model is finally obtained, which can simultaneously handle the prediction of multiple advertising delivery tasks, such as: the user's advertising revenue and member income estimation under delivery and non-delivery conditions.

[0040] Optionally, in this embodiment, the target multi-task model is a model obtained by combining a gain model and a multi-task model. The multi-task model can use an MMOE (Multi-Task Multi-Output Expert, a multi-task, multi-output model based on an expert mechanism) model, a PLE (Pooling-based Layer-wise Enhancement Mode, a deep learning model based on a pooling mechanism) model, etc., and the gain model can use an S-Learner (Single Model, single model method) model, a DragonNet (multi-task natural language processing model based on deep learning and pre-training) model, etc.

[0041] Optionally, in this embodiment, by combining conventional features with tasks, the model can more accurately predict the multi-dimensional effects after advertising, such as whether the user starts, the playback time, advertising revenue, etc., optimize the advertising delivery strategy from different dimensions, and improve the accuracy and efficiency of advertising. Multiple delivery strategy predictions can be obtained through one model training, without the need to frequently change the model structure, saving model adjustment time and labor costs. The multi-task model combines the prediction results of user conventional features and processing variables to help advertisers more effectively identify potential target users, thereby improving the conversion rate and value of advertising.

[0042] As an optional example, according to the delivery target, the estimated value of each task of each user under the delivery condition is integrated to obtain the delivery integration score of each user, including:

[0043] Using the Pareto optimal algorithm, a first optimal weight is calculated according to the delivery target and the estimated value of each task of each user under the delivery condition, wherein the delivery target includes constraint conditions and optimization targets;

[0044] Take each user as the current user and perform the following operations on the current user:

[0045] According to the first optimal weight, the estimated value of each task of the current user under the delivery condition is weightedly fused to obtain the delivery fusion score of the current user.

[0046] Optionally, in this embodiment, a Pareto optimal algorithm is used to calculate the delivery fusion score. This algorithm aims to find a set of solutions so that no solution can significantly improve a certain goal without compromising other goals. That is to say, in multi-objective optimization, certain specific combinations of solutions cannot be superior. This algorithm will find the optimal weight based on the delivery goal and the estimated value of each user's task to ensure that the trade-off between different tasks best meets the target constraints. The constraint condition can be the cost rate, such as what conditions the cost rate satisfies. According to business needs, the optimization goal can be to maximize the number of people pulled up, maximize the exchange income, maximize the total income, etc. Taking the multi-task estimated value of each user in the delivery situation as input, combined with the delivery target constraints and optimization goals, the Pareto optimal algorithm will perform weight calculation to obtain the first optimal weight. The first optimal weight is used to adjust the influence of each task estimate value to ensure the optimal trade-off under the multi-task model. According to the calculated first optimal weight, the estimated value of each task of each user in the delivery situation is weighted and fused by the following formula to obtain the delivery fusion score, which is used to measure the comprehensive value of the current user in the delivery situation:

[0047] K1=∑ i w i *P i ;

[0048] Among them, K1 is the integration score, w i is the first optimal weight of the task estimate, P i is the estimated value of each task.

[0049] Optionally, in this embodiment, the Pareto optimal algorithm can be used to comprehensively consider the multi-task estimated values ​​and delivery targets, and accurately evaluate the value of users in advertising delivery. By using weight calculation, it is ensured that the estimated values ​​of different tasks are optimally weighed according to business goals, thereby improving the overall effect of advertising delivery. Because model training and Pareto optimal algorithm are divided into two stages, each time the advertiser's needs change, there is no need to change the model training stage, only the Pareto optimal algorithm hyperparameters need to be changed to achieve the update of the final fusion score.

[0050] As an optional example, the estimated values ​​of each task of each user without delivery are fused to obtain the fused score of each user without delivery, including:

[0051] The Pareto optimal algorithm is used to calculate the second optimal weight according to the delivery target and the estimated value of each task of each user without delivery, wherein the delivery target includes constraint conditions and optimization targets;

[0052] Take each user as the current user and perform the following operations on the current user:

[0053] According to the second optimal weight, the estimated value of each task of the current user in the case of no delivery is weightedly fused to obtain the non-delivery fusion score of the current user.

[0054] Optionally, in this embodiment, the Pareto optimal algorithm is used to calculate the non-delivery fusion score in the same way as the calculation process of the delivery fusion score. The multi-task estimated value of each user in the non-delivery case is used as input. Combined with the delivery target constraint and the optimization target, the Pareto optimal algorithm will perform weight calculation to obtain the second optimal weight. The second optimal weight is used to adjust the influence of each task estimate to ensure the optimal trade-off under the multi-task model. According to the calculated second optimal weight, the estimated value of each task of each user in the non-delivery case is weighted and fused to obtain the non-delivery fusion score, which is used to measure the comprehensive value of the current user in the non-delivery case.

[0055] As an optional example, according to the delivery fusion score and non-delivery fusion score of each user, the optimal gain fusion score of each user is calculated to include:

[0056] Take each user as the current user and perform the following operations on the current user:

[0057] Calculate the difference between the current user's delivery integration score and the current user's non-delivery integration score;

[0058] The difference is determined as the optimal gain fusion score for the current user.

[0059] Optionally, in this embodiment, the optimal gain fusion score is obtained by subtracting the non-delivery fusion score from the delivery fusion score. The fusion score represents the relative gain of user value between delivery and non-delivery, and is used to measure the optimization effect of advertising delivery on user conversion goals.

[0060] As an optional example, after calculating the optimal gain fusion score of each user according to the delivery fusion score and the non-delivery fusion score of each user, the method further includes:

[0061] The users whose optimal gain fusion score is greater than the target threshold are determined as the target users for advertisement delivery.

[0062] Optionally, in this embodiment, high-value users are screened as target users for advertising based on the optimal gain fusion score. Specifically, a target threshold can be set to accurately select user groups and improve the effectiveness and efficiency of advertising. If a user is determined to be a target user, all advertisements will participate in the competition.

[0063] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0064] According to another aspect of the embodiment of the present application, a device for delivering advertisements is also provided. Figure 3 As shown, including:

[0065] The first evaluation module 302 is used to obtain the general features of each user, and input the general features of each user and the first processing variable into the target multi-task model, so that the target multi-task model performs a multi-task evaluation of the delivery situation of each user, and outputs an estimated value of each task of each user under the delivery situation, wherein the first processing variable is the delivery;

[0066] The second evaluation module 304 is used to input the conventional features and the second processing variable of each user into the target multi-task model, so that the target multi-task model performs a multi-task evaluation on each user in the case of no delivery, and outputs an estimated value of each task of each user in the case of no delivery, wherein the second processing variable is no delivery;

[0067] Fusion module 306, used for fusing the estimated values ​​of each task of each user in the delivery situation according to the delivery target to obtain the delivery fusion score of each user, and fusing the estimated values ​​of each task of each user in the non-delivery situation to obtain the non-delivery fusion score of each user;

[0068] The calculation module 308 is used to calculate the optimal gain fusion score of each user according to the delivery fusion score and the non-delivery fusion score of each user, wherein the optimal gain fusion score is used to determine the target user for advertisement delivery.

[0069] It should be noted that the first evaluation module 302 in this embodiment can be used to execute step S102 in the embodiment of the present application, the second evaluation module 304 in this embodiment can be used to execute step S104 in the embodiment of the present application, the fusion module 306 in this embodiment can be used to execute step S106 in the embodiment of the present application, and the calculation module 308 in this embodiment can be used to execute step S108 in the embodiment of the present application.

[0070] As an optional example, conventional features include any one or more of basic information, behavior records, interaction preferences, value information, and advertising placement information, and tasks include any one or more of whether to start, playback duration, advertising revenue, membership revenue, and total revenue.

[0071] As an optional example, the above device further includes:

[0072] an acquisition module, used for acquiring a training data set before inputting the conventional features and the first processing variable of each user into the target multi-task model, wherein the training data set includes training data of a plurality of training users, the training data includes conventional features, processing variables and all tasks, wherein the processing variable is the first processing variable or the second processing variable;

[0073] The training module is used to perform multiple iterative training on the initial multi-task model according to the training data set to obtain the target multi-task model.

[0074] As an optional example, the fusion module includes:

[0075] A first calculation unit is used to calculate a first optimal weight using a Pareto optimal algorithm according to a delivery target and an estimated value of each task of each user under the delivery condition, wherein the delivery target includes a constraint condition and an optimization target;

[0076] The first processing unit is used to take each user as the current user and perform the following operations on the current user:

[0077] According to the first optimal weight, the estimated value of each task of the current user under the delivery condition is weightedly fused to obtain the delivery fusion score of the current user.

[0078] As an optional example, the fusion module includes:

[0079] A second calculation unit is used to calculate a second optimal weight using a Pareto optimal algorithm according to a delivery target and an estimated value of each task of each user without delivery, wherein the delivery target includes a constraint condition and an optimization target;

[0080] The second processing unit is used to take each user as the current user and perform the following operations on the current user:

[0081] According to the second optimal weight, the estimated value of each task of the current user in the case of no delivery is weightedly fused to obtain the non-delivery fusion score of the current user.

[0082] As an optional example, the calculation module includes:

[0083] The third processing unit is used to take each user as the current user and perform the following operations on the current user:

[0084] Calculate the difference between the current user's delivery integration score and the current user's non-delivery integration score;

[0085] The difference is determined as the optimal gain fusion score for the current user.

[0086] As an optional example, the above device further includes:

[0087] The determination module is used to determine the user whose optimal gain fusion score is greater than the target threshold as the target user for advertisement delivery after calculating the optimal gain fusion score of each user according to the delivery fusion score and non-delivery fusion score of each user.

[0088] For other examples of this embodiment, please refer to the above examples and will not be repeated here.

[0089] Figure 4 is a schematic diagram of an optional electronic device according to an embodiment of the present application, such as Figure 4 As shown, it includes a processor 402, a communication interface 404, a memory 406 and a communication bus 408, wherein the processor 402, the communication interface 404 and the memory 406 communicate with each other through the communication bus 408, wherein,

[0090] Memory 406, used to store computer programs;

[0091] The processor 402 is used to implement the following steps when executing the computer program stored in the memory 406:

[0092] Obtaining the conventional features of each user, and inputting the conventional features of each user and the first processing variable into the target multi-task model, so that the target multi-task model performs a multi-task evaluation of the delivery situation of each user, and outputs an estimated value of each task of each user under the delivery situation, wherein the first processing variable is the delivery;

[0093] Inputting the conventional features and the second processing variable of each user into the target multi-task model, so that the target multi-task model performs a multi-task evaluation on each user in the case of no delivery, and outputs an estimated value of each task of each user in the case of no delivery, wherein the second processing variable is no delivery;

[0094] According to the delivery target, the estimated values ​​of each task of each user in the delivery situation are integrated to obtain the delivery integrated score of each user, and the estimated values ​​of each task of each user in the non-delivery situation are integrated to obtain the non-delivery integrated score of each user;

[0095] The optimal gain fusion score of each user is calculated based on the delivery fusion score and the non-delivery fusion score of each user, wherein the optimal gain fusion score is used to determine the target user for advertisement delivery.

[0096] Optionally, in this embodiment, the communication bus may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The communication bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 The communication interface is used for communication between the above electronic device and other devices.

[0097] The memory may include RAM, or may include non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0098] As an example, the memory 406 may include, but is not limited to, the first evaluation module 302, the second evaluation module 304, the fusion module 306, and the calculation module 308 in the apparatus for advertising delivery. In addition, it may also include, but is not limited to, other module units in the apparatus for advertising delivery, which will not be described in detail in this example.

[0099] The above-mentioned processor can be a general-purpose processor, which can include but not be limited to: CPU (Central Processing Unit), NP (Network Processor), etc.; it can also be DSP (Digital Signal Processing), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0100] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments, and this embodiment will not be described in detail here.

[0101] It can be understood by those skilled in the art that Figure 4 The structure shown is for illustration only. The device for implementing the above-mentioned advertising delivery method may be a terminal device, which may be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a PDA, a mobile Internet device (Mobile Internet Devices, MID), a PAD, and other terminal devices. Figure 4 The structure of the electronic device is not limited. For example, the electronic device may also include Figure 4 More or fewer components (such as network interfaces, display devices, etc.) shown in, or having Figure 4 Different configurations shown.

[0102] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, which can include: a flash drive, ROM, RAM, a magnetic disk or an optical disk, etc.

[0103] According to another aspect of the embodiments of the present application, a computer-readable storage medium is provided, in which a computer program is stored, wherein the computer program executes the steps in the above-mentioned method for advertisement delivery when executed by a processor.

[0104] Optionally, in this embodiment, a person of ordinary skill in the art may understand that all or part of the steps in the various methods of the above embodiments may be completed by instructing hardware related to the terminal device through a program, and the program may be stored in a computer-readable storage medium, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, etc.

[0105] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0106] If the integrated units in the above embodiments are implemented in the form of software functional units and sold or used as independent products, they can be stored in the above computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling one or more computer devices (which may be personal computers, servers, or network devices, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application.

[0107] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0108] In the several embodiments provided in the present application, it should be understood that the disclosed client can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0109] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0110] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0111] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for placing an advertisement, characterized in that: include: Obtaining conventional features of each user, and inputting the conventional features of each user and a first processing variable into a target multi-task model, so that the target multi-task model performs a multi-task evaluation of the delivery situation of each user, and outputs an estimated value of each task of each user under the delivery situation, wherein the first processing variable is delivery; Inputting the conventional features and the second processing variable of each user into the target multi-task model, so that the target multi-task model performs a multi-task evaluation on each user under the condition of no delivery, and outputs an estimated value of each task of each user under the condition of no delivery, wherein the second processing variable is no delivery; According to the delivery target, the estimated values ​​of each task of each user in the delivery situation are integrated to obtain the delivery integrated score of each user, and the estimated values ​​of each task of each user in the non-delivery situation are integrated to obtain the non-delivery integrated score of each user; The optimal gain fusion score of each user is calculated according to the delivery fusion score and the non-delivery fusion score of each user, wherein the optimal gain fusion score is used to determine the target user for advertisement delivery.

2. The method according to claim 1, characterized in that The conventional features include any one or more of basic information, behavior records, interaction preferences, value information, and advertising information; the tasks include any one or more of whether to start, playback duration, advertising revenue, membership revenue, and total revenue.

3. The method according to claim 1, characterized in that Before inputting the conventional features of each user and the first processing variable into the target multi-task model, the method further includes: Acquire a training data set, wherein the training data set includes training data of multiple training users, and the training data includes conventional features, processing variables, and all tasks, wherein the processing variable is the first processing variable or the second processing variable; The initial multi-task model is iteratively trained multiple times according to the training data set to obtain the target multi-task model.

4. The method according to claim 1, characterized in that: According to the delivery target, the estimated value of each task of each user under the delivery condition is integrated to obtain the delivery integration score of each user, which includes: Using a Pareto optimal algorithm, a first optimal weight is calculated according to the delivery target and the estimated value of each task of each user under the delivery condition, wherein the delivery target includes a constraint condition and an optimization target; Take each user as the current user and perform the following operations on the current user: According to the first optimal weight, the estimated value of each task of the current user in the delivery situation is weightedly fused to obtain the delivery fusion score of the current user.

5. The method according to claim 1, characterized in that The step of fusing the estimated value of each task of each user without delivery to obtain the non-delivery fusion score of each user includes: Using a Pareto optimal algorithm, a second optimal weight is calculated according to the delivery target and the estimated value of each task of each user without delivery, wherein the delivery target includes a constraint condition and an optimization target; Take each user as the current user and perform the following operations on the current user: According to the second optimal weight, weighted fusion is performed on the estimated value of each task of the current user in the case of no delivery to obtain a no-delivery fusion score of the current user.

6. The method according to claim 1, characterized in that The calculating the optimal gain fusion score of each user according to the delivery fusion score and the non-delivery fusion score of each user comprises: Take each user as the current user and perform the following operations on the current user: Calculate the difference between the delivery integration score of the current user and the non-delivery integration score of the current user; The difference is determined as the optimal gain fusion score of the current user.

7. The method according to claim 1, characterized in that After calculating the optimal gain fusion score of each user according to the delivery fusion score and the non-delivery fusion score of each user, the method further includes: The user whose optimal gain fusion score is greater than the target threshold is determined as the target user for advertisement delivery.

8. An advertisement delivery device, characterized in that: include: A first evaluation module is used to obtain the general features of each user, and input the general features of each user and a first processing variable into a target multi-task model, so that the target multi-task model performs a multi-task evaluation of the delivery situation of each user, and outputs an estimated value of each task of each user under the delivery situation, wherein the first processing variable is delivery; A second evaluation module is used to input the conventional features and the second processing variable of each user into the target multi-task model, so that the target multi-task model performs a multi-task evaluation on each user under the condition of no delivery, and outputs an estimated value of each task of each user under the condition of no delivery, wherein the second processing variable is no delivery; A fusion module is used to fuse the estimated values ​​of each task of each user in the delivery situation according to the delivery target to obtain the delivery fusion score of each user, and to fuse the estimated values ​​of each task of each user in the non-delivery situation to obtain the non-delivery fusion score of each user; The calculation module is used to calculate the optimal gain fusion score of each user according to the delivery fusion score and the non-delivery fusion score of each user, wherein the optimal gain fusion score is used to determine the target user for advertisement delivery.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is executed.

10. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to execute the method according to any one of claims 1 to 7 through the computer program.

Images