Advertisement putting strategy optimization method and device, electronic equipment and storage medium

Abstract

The invention discloses an advertisement putting strategy optimization method, and the method comprises the steps: obtaining the putting environment information of an advertisement putting position at a current moment, determining an initial putting strategy according to the putting environment information, and enabling the initial putting strategy to comprise a plurality of advertisement putting strategies; inputting the strategy parameter corresponding to each advertisement putting strategy into an advertisement revenue pre-estimation model for revenue pre-estimation processing to obtain pre-estimated revenue data of each advertisement putting strategy; performing decision analysis based on the estimated income data and the object feature parameters of each advertisement putting strategy, and screening out a target putting strategy from the plurality of advertisement putting strategies; performing resource configuration on the target delivery strategy to obtain a resource configuration result; and under the condition that the resource configuration result meets the preset delivery condition, performing advertisement delivery on the delivery object according to the target delivery strategy. The advertisement revenue estimation is performed on the acquired information based on the advertisement revenue estimation model, the bidding strategy is dynamically adjusted, and the advertisement putting service processing efficiency is improved.

Description

Technical Field

[0001] The present application relates to the technical field of advertisement delivery, and in particular to an optimization method, device, electronic device and storage medium for an advertisement delivery strategy. Background Art

[0002] With the rapid development and widespread application of Internet technology, the advertising business has gradually developed along with the rapid development of the Internet. In advertising, advertising spaces are usually bid through real-time bidding to display advertising services, and the highest bidder obtains the display opportunity of the advertising space. However, due to the short bidding time, which is generally only a few dozen milliseconds, and the huge number of display opportunities, advertisers need to use automatic bidding algorithms to bid automatically.

[0003] In the prior art, the bidding logic of advertising is based on a comprehensive evaluation of advertising plans, exposure of advertising creatives, click-through rates, click costs, conversion costs, and revenue data of the back-end system, and the optimizer manually adjusts the bid. However, the bidding environment of real-time advertising bidding is highly dynamic and changes in real time. The bidding strategy does not take into account the revenue of the back-end conversion of the advertisement. There is a problem of inaccurate determination of the bidding strategy, which leads to a decrease in the advertising yield of advertisers and an inability to stably increase the revenue of advertising delivery in the long term. Therefore, there is an urgent need for an optimization method for advertising delivery strategies to achieve successful bidding to solve the above problems. Summary of the invention

[0004] The present application provides an optimization method, device, electronic device and storage medium for an advertisement delivery strategy. By acquiring delivery environment information of an advertisement delivery position at the current moment and characteristic information of a delivery object, the advertisement revenue is estimated based on the acquired information based on an advertisement revenue estimation model, and the bidding strategy is dynamically adjusted based on the estimated result, thereby solving the problem of inaccurate determination of bidding strategies in real-time bidding for advertisement delivery.

[0005] According to one aspect of the present application, a method for optimizing an advertisement delivery strategy is proposed, the method comprising:

[0006] Acquire the delivery environment information of the current advertisement delivery position, and determine the initial delivery strategy according to the delivery environment information, wherein the initial delivery strategy includes multiple advertisement delivery strategies;

[0007] Obtain the strategy parameters and advertising revenue estimation model corresponding to each advertising delivery strategy;

[0008] Inputting the strategy parameters corresponding to each advertising delivery strategy into the advertising revenue estimation model to perform revenue estimation processing, thereby obtaining estimated revenue data for each advertising delivery strategy;

[0009] Obtaining object characteristic parameters of the delivery object, performing decision analysis based on the estimated revenue data of each advertising delivery strategy and the object characteristic parameters, and selecting a target delivery strategy from the multiple advertising delivery strategies;

[0010] Performing resource allocation on the target delivery strategy to obtain a resource allocation result;

[0011] When the resource configuration result meets the preset delivery condition, advertisement delivery is performed on the delivery object according to the target delivery strategy.

[0012] In a possible implementation, determining the initial delivery strategy according to the delivery environment information includes:

[0013] Acquire an advertisement delivery strategy database, wherein the advertisement delivery strategy database includes preset delivery environment information of advertisement delivery positions and preset advertisement delivery strategies;

[0014] Based on the delivery environment information, the advertisement delivery strategy database is queried to obtain a preset advertisement delivery strategy matching the delivery environment information, and the preset advertisement delivery strategy matching the delivery environment information is determined to be the initial delivery strategy; the advertisement delivery strategy database is constructed based on the preset delivery environment information of the advertisement delivery position and the preset advertisement delivery strategy.

[0015] In a possible implementation, the method for constructing the advertisement delivery strategy database includes:

[0016] Acquire a delivery environment information sample set, wherein the delivery environment information sample set is different preset delivery environment information for advertisement delivery;

[0017] At least one preset advertisement delivery strategy is set based on each preset delivery environment information;

[0018] The advertisement delivery strategy database is constructed based on each preset delivery environment information and the preset advertisement delivery strategy corresponding to each preset delivery environment information.

[0019] In a possible implementation, the decision analysis is performed based on the estimated revenue data of each advertising delivery strategy and the object characteristic parameters to select a target delivery strategy from the multiple advertising delivery strategies, including:

[0020] Performing revenue analysis on the estimated revenue data of each advertising delivery strategy to obtain the estimated revenue of each advertising delivery strategy;

[0021] In the case where the estimated revenue is greater than the preset revenue threshold, determining that the advertisement delivery strategy corresponding to the estimated revenue greater than the preset revenue threshold is an intermediate delivery strategy;

[0022] Performing feature evaluation on the object feature parameters to obtain object feature data;

[0023] A matching analysis is performed on the intermediate delivery strategy and the object feature data, and the intermediate delivery strategy whose matching analysis result satisfies a preset matching condition is determined as the target delivery strategy.

[0024] In a possible implementation, the object characteristic parameters include a mobile device identifier of the object, an object portrait, a behavior characteristic, and an object behavior time.

[0025] The step of performing feature evaluation on the object feature parameters to obtain object feature data includes:

[0026] Calculate the revenue value based on the mobile device identifier, the object portrait, the behavior characteristics and the object behavior time to obtain an estimated revenue value;

[0027] The estimated benefit value is compared with the preset object benefit value to obtain a comparison result, and the object feature data is determined based on the comparison result.

[0028] In a possible implementation, the preset object benefit value includes a first benefit value and a second benefit value, the first benefit value is greater than the second benefit value, the object feature data includes first object feature data, second object feature data and third object feature data,

[0029] The step of comparing the estimated benefit value with the preset object benefit value to obtain a comparison result, and determining the object feature data based on the comparison result includes:

[0030] In a case where the estimated benefit value is greater than or equal to the first benefit value, determining the object feature data to be the first object feature data;

[0031] When the estimated benefit value is less than the first benefit value and greater than or equal to the second benefit value, determining that the object feature data is the second object feature data;

[0032] When the estimated benefit value is less than the second benefit value, the object feature data is determined to be the third object feature data.

[0033] In a possible implementation, the intermediate strategy includes a first strategy, a second strategy, and a third strategy, and the object feature data includes first object feature data, second object feature data, and third object feature data.

[0034] The performing matching analysis on the intermediate delivery strategy and the object feature data, and determining the intermediate delivery strategy whose matching analysis result satisfies a preset matching condition as the target delivery strategy, includes:

[0035] In a case where the object characteristic data is the first object characteristic data, determining the first strategy as the target delivery strategy;

[0036] In a case where the object characteristic data is the second object characteristic data, determining the second strategy as the target delivery strategy;

[0037] In a case where the object characteristic data is the third object characteristic data, the third strategy is determined to be the target delivery strategy.

[0038] In a possible implementation, the training method of the advertising revenue estimation model includes:

[0039] Acquire the initial advertising revenue estimation model, where the advertising revenue estimation model is constructed based on a reinforcement learning model;

[0040] Inputting the strategy parameters of the sample advertising delivery strategy into the initial advertising revenue estimation model to perform revenue estimation processing to obtain estimated revenue data corresponding to the sample advertising delivery strategy;

[0041] Matching the revenue label data of the sample advertising delivery strategy with the estimated revenue data of the corresponding sample advertising delivery strategy to obtain a matching result;

[0042] In the case where the matching result does not satisfy the preset condition, determining difference information between the revenue label data of the sample advertisement delivery strategy and the estimated revenue data corresponding to the sample advertisement delivery strategy;

[0043] The network parameters of the initial advertising revenue estimation model are updated based on the difference information until the matching result satisfies the preset condition, and the initial advertising revenue estimation model corresponding to the matching result that satisfies the preset condition is determined as the advertising revenue estimation model.

[0044] On the other hand, a device for optimizing an advertisement delivery strategy is provided, the device comprising:

[0045] A current environment information acquisition module is used to obtain the delivery environment information of the current advertisement delivery position, and determine an initial delivery strategy according to the delivery environment information, wherein the initial delivery strategy includes multiple advertisement delivery strategies;

[0046] A model acquisition module is used to obtain the strategy parameters and advertising revenue estimation model corresponding to each advertising delivery strategy;

[0047] A revenue estimation processing module, used for inputting the strategy parameters corresponding to each advertising delivery strategy into the advertising revenue estimation model to perform revenue estimation processing, and obtain the estimated revenue data of each advertising delivery strategy;

[0048] A decision analysis module, used to obtain object characteristic parameters of the delivery object, perform decision analysis based on the estimated revenue data of each advertising delivery strategy and the object characteristic parameters, and select a target delivery strategy from the multiple advertising delivery strategies;

[0049] A resource configuration module, used to perform resource configuration on the target delivery strategy and obtain a resource configuration result;

[0050] The advertisement delivery module is used to deliver advertisements to the delivery object according to the target delivery strategy when the resource configuration result meets the preset delivery conditions.

[0051] In a possible implementation, the current environment information acquisition module includes an initial delivery strategy determination module, and the initial delivery strategy determination module is used to:

[0052] Acquire an advertisement delivery strategy database, wherein the advertisement delivery strategy database includes preset delivery environment information of advertisement delivery positions and preset advertisement delivery strategies;

[0053] Based on the delivery environment information, the advertisement delivery strategy database is queried to obtain a preset advertisement delivery strategy matching the delivery environment information, and the preset advertisement delivery strategy matching the delivery environment information is determined to be the initial delivery strategy; the advertisement delivery strategy database is constructed based on the preset delivery environment information of the advertisement delivery position and the preset advertisement delivery strategy.

[0054] In a possible implementation, the current environment information acquisition module includes an advertisement delivery strategy database construction unit, and the advertisement delivery strategy database construction unit is used to:

[0055] Acquire a delivery environment information sample set, wherein the delivery environment information sample set is different preset delivery environment information for advertisement delivery;

[0056] Setting at least one preset advertisement delivery strategy based on each preset delivery environment information;

[0057] The advertisement delivery strategy database is constructed based on each preset delivery environment information and the preset advertisement delivery strategy corresponding to each preset delivery environment information.

[0058] In a possible implementation, the decision analysis module is used to:

[0059] Performing revenue analysis on the estimated revenue data of each advertising delivery strategy to obtain the estimated revenue of each advertising delivery strategy;

[0060] In the case where the estimated revenue is greater than the preset revenue threshold, determining that the advertisement delivery strategy corresponding to the estimated revenue greater than the preset revenue threshold is an intermediate delivery strategy;

[0061] Performing feature evaluation on the object feature parameters to obtain object feature data;

[0062] A matching analysis is performed on the intermediate delivery strategy and the object feature data, and the intermediate delivery strategy whose matching analysis result satisfies a preset matching condition is determined as the target delivery strategy.

[0063] In a possible implementation, the object feature parameters include a mobile device identifier, an object portrait, a behavior feature, and an object behavior time of the object, and the decision analysis module includes a feature evaluation unit, and the feature evaluation unit is used to:

[0064] Calculate the revenue value based on the mobile device identifier, the object portrait, the behavior characteristics and the object behavior time to obtain an estimated revenue value;

[0065] The estimated benefit value is compared with the preset object benefit value to obtain a comparison result, and the object feature data is determined based on the comparison result.

[0066] In a possible implementation, the preset object benefit value includes a first benefit value and a second benefit value, the first benefit value is greater than the second benefit value, the object feature data includes first object feature data, second object feature data and third object feature data, and the decision analysis module includes a comparison unit, and the comparison unit is used to:

[0067] In a case where the estimated benefit value is greater than or equal to the first benefit value, determining the object feature data to be the first object feature data;

[0068] When the estimated benefit value is less than the first benefit value and greater than or equal to the second benefit value, determining that the object feature data is the second object feature data;

[0069] When the estimated benefit value is less than the second benefit value, the object feature data is determined to be the third object feature data.

[0070] In a possible implementation, the intermediate strategy includes a first strategy, a second strategy, and a third strategy, the object feature data includes first object feature data, second object feature data, and third object feature data, and the decision analysis module includes a matching analysis unit, and the matching analysis unit is used to:

[0071] In a case where the object characteristic data is the first object characteristic data, determining the first strategy as the target delivery strategy;

[0072] In a case where the object characteristic data is the second object characteristic data, determining the second strategy as the target delivery strategy;

[0073] In a case where the object characteristic data is the third object characteristic data, the third strategy is determined to be the target delivery strategy.

[0074] In a possible implementation, the revenue estimation processing module includes an advertisement revenue estimation model training unit, and the advertisement revenue estimation model training unit is used to:

[0075] Acquire the initial advertising revenue estimation model, where the advertising revenue estimation model is constructed based on a reinforcement learning model;

[0076] Inputting the strategy parameters of the sample advertising delivery strategy into the initial advertising revenue estimation model to perform revenue estimation processing to obtain estimated revenue data corresponding to the sample advertising delivery strategy;

[0077] Matching the revenue label data of the sample advertising delivery strategy with the estimated revenue data of the corresponding sample advertising delivery strategy to obtain a matching result;

[0078] In the case where the matching result does not satisfy the preset condition, determining difference information between the revenue label data of the sample advertisement delivery strategy and the estimated revenue data corresponding to the sample advertisement delivery strategy;

[0079] The network parameters of the initial advertising revenue estimation model are updated based on the difference information until the matching result satisfies the preset condition, and the initial advertising revenue estimation model corresponding to the matching result that satisfies the preset condition is determined as the advertising revenue estimation model.

[0080] On the other hand, an electronic device is provided, including a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the optimization method of the advertising delivery strategy in any of the above aspects.

[0081] On the other hand, a computer-readable storage medium is provided, in which at least one instruction or at least one program is stored, and the at least one instruction or the at least one program is loaded and executed by a processor to implement a method for optimizing an advertising delivery strategy as described in any of the above aspects.

[0082] On the other hand, a computer program product or a computer program is provided, the computer program product or the computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the electronic device executes the optimization method of the advertising delivery strategy of any aspect above.

[0083] The embodiment of the present application obtains the delivery environment information of the current advertising delivery position, and determines the initial delivery strategy according to the delivery environment information, wherein the initial delivery strategy includes multiple advertising delivery strategies; obtains the strategy parameters and the advertising revenue estimation model corresponding to each advertising delivery strategy; inputs the strategy parameters corresponding to each advertising delivery strategy into the advertising revenue estimation model for revenue estimation processing, and obtains the estimated revenue data of each advertising delivery strategy; obtains the object characteristic parameters of the delivery object, performs decision analysis based on the estimated revenue data of each advertising delivery strategy and the object characteristic parameters, and selects the target delivery strategy from the multiple advertising delivery strategies; performs resource allocation for the target delivery strategy to obtain the resource allocation result; and when the resource allocation result meets the preset delivery condition, performs advertising delivery on the delivery object according to the target delivery strategy. By obtaining the delivery environment information of the current advertising delivery position and the characteristic information of the delivery object, the advertising revenue is estimated based on the advertising revenue estimation model for the acquired information, and the bidding strategy is dynamically adjusted based on the estimated result, the processing efficiency of the advertising delivery business is improved, the problem of inaccurate determination of the bidding strategy in the real-time bidding of advertising delivery is solved, the success rate of the real-time bidding of advertising delivery is improved, and the yield rate of advertising delivery is also improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0085] Figure 1 It is a flowchart of an optimization method for an advertisement delivery strategy provided in an embodiment of the present application;

[0086] Figure 2 It is a flow chart of a method for constructing an advertisement delivery strategy database of an advertisement delivery strategy optimization method provided in an embodiment of the present application;

[0087] Figure 3 It is a flowchart of a training method for an advertising revenue estimation model of an optimization method for advertising delivery strategy provided in an embodiment of the present application;

[0088] Figure 4 It is a flowchart of determining a target delivery strategy of the method for optimizing the advertising delivery strategy provided in an embodiment of the present application;

[0089] Figure 5 It is a structural block diagram of an optimization device for advertising delivery strategy provided in an embodiment of the present application. DETAILED DESCRIPTION

[0090] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only 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.

[0091] With the rapid development and widespread application of Internet technology, the advertising business has gradually developed along with the rapid development of the Internet. In advertising, advertising spaces are usually bid through real-time bidding to display advertising services, and the highest bidder obtains the display opportunity of the advertising space. However, due to the short bidding time, which is generally only a few dozen milliseconds, and the huge number of display opportunities, advertisers need to use automatic bidding algorithms to bid automatically.

[0092] The common bidding logic in the advertising industry is to conduct a comprehensive evaluation based on the advertising plan, exposure of the advertising creative, click-through rate, click cost, conversion cost, and revenue data of the back-end system, and the optimizer will manually adjust the bid. This method relies on manual labor and has several very obvious disadvantages. First, because the back-end revenue data includes advertising revenue (real-time / offline) and order revenue (real-time), there is no algorithm estimation for offline revenue, and it is impossible to adjust the advertising bid according to the ROI (Return on Investment) indicator of the advertising plan. Second, because the effect of the back-end conversion of the advertisement is not taken into account, it is impossible to automatically close the loss-making plan in time, or to amplify the high-revenue plan in time, and it is impossible to stably increase the revenue of advertising in the long term. The bidding strategy for advertising is to successfully bid for real-time display opportunities for advertisements, so as to maximize the revenue of advertisements.

[0093] The embodiment of the present application provides an optimization method for advertising delivery strategy, which obtains the delivery environment information of the current advertising delivery position and the characteristic information of the delivery object, estimates the advertising revenue based on the acquired information based on the advertising revenue estimation model, and dynamically adjusts the bidding strategy based on the estimation result, thereby improving the processing efficiency of advertising delivery business and solving the problem of inaccurate determination of bidding strategy in real-time bidding for advertising delivery. Figure 1 The method for optimizing the advertising delivery strategy includes steps S101 to S1011.

[0094] In step S101, the delivery environment information of the current advertisement delivery position is obtained, and an initial delivery strategy is determined according to the delivery environment information. The initial delivery strategy includes multiple advertisement delivery strategies.

[0095] In a possible implementation manner, the delivery environment information represents the market environment for advertisement delivery.

[0096] In a possible implementation, in the field of advertising, the advertising delivery system is regarded as an intelligent agent and the advertising market environment is regarded as an environment. The goal of advertising delivery is to maximize advertising revenue by selecting appropriate advertising strategies within a limited budget.

[0097] In a possible implementation, the placement environment information of the advertisement placement includes target audience, geographic location, media format, and competition environment.

[0098] In a possible implementation, the target audience is the type of users covered by the advertising placement, and the type of users is determined based on information such as the user's age, gender, occupation, interests and hobbies; the geographical location is the area where the advertising placement is placed; the media forms include billboards, light boxes, LED screens, and online media; and the competitive environment refers to whether there are other advertisements with high similarity near the placement environment of the advertising placement.

[0099] In a possible implementation, determining the initial delivery strategy according to the delivery environment information includes:

[0100] Acquire an advertisement delivery strategy database, wherein the advertisement delivery strategy database includes preset delivery environment information of advertisement delivery positions and preset advertisement delivery strategies;

[0101] Based on the delivery environment information, the advertisement delivery strategy database is queried to obtain a preset advertisement delivery strategy matching the delivery environment information, and the preset advertisement delivery strategy matching the delivery environment information is determined to be the initial delivery strategy; the advertisement delivery strategy database is constructed based on the preset delivery environment information of the advertisement delivery position and the preset advertisement delivery strategy.

[0102] In a possible implementation manner, an initial delivery strategy is determined according to the delivery environment information, and the initial delivery strategy is at least one of preset advertisement delivery strategies.

[0103] In a possible implementation, the preset advertising delivery strategy includes the materials for advertising delivery, that is, the manner in which the advertising content is presented to the user, specifically, images, videos, and texts, etc., and may also be short videos and audios.

[0104] In a possible implementation manner, a preset advertisement delivery strategy corresponding to the delivery environment information is acquired from the advertisement delivery strategy database according to the delivery environment information.

[0105] In this implementation, a preset advertising delivery strategy is obtained from a pre-built advertising delivery strategy database as an initial delivery strategy according to the delivery environment information, and the delivery strategy is scoped, thereby reducing the determination scope of the advertising delivery strategy and improving the efficiency of the advertising delivery strategy.

[0106] In a possible implementation, the method for constructing the advertisement delivery strategy database includes:

[0107] Acquire a delivery environment information sample set, wherein the delivery environment information sample set is different preset delivery environment information for advertisement delivery;

[0108] At least one preset advertisement delivery strategy is set based on each preset delivery environment information;

[0109] The advertisement delivery strategy database is constructed based on each preset delivery environment information and the preset advertisement delivery strategy corresponding to each preset delivery environment information.

[0110] In one possible implementation, Figure 2 As shown, the method for constructing the advertisement delivery strategy database includes steps S201 to S205:

[0111] In step S201, a delivery environment information sample set is obtained, where the delivery environment information sample set is different preset delivery environment information for advertisement delivery;

[0112] In step S203, at least one preset advertisement delivery strategy is set based on each preset delivery environment information;

[0113] In step S205, the advertisement delivery strategy database is constructed based on each preset delivery environment information and the preset advertisement delivery strategy corresponding to each preset delivery environment information.

[0114] In a possible implementation manner, before acquiring the advertisement delivery strategy database, the advertisement delivery strategy database is constructed.

[0115] In a possible implementation, the advertisement delivery strategy database is constructed based on preset delivery environment information of advertisement delivery positions and preset advertisement delivery strategies, and each preset delivery environment information corresponds to at least one preset advertisement delivery strategy.

[0116] In a possible implementation, materials that can be displayed in the advertising placement position are determined based on the target audience, geographic location, media format and competitive environment of the preset placement environment information, and then a preset advertising placement strategy corresponding to the preset placement environment information is determined.

[0117] In this implementation, the advertising delivery strategy database is constructed based on preset delivery environment information and preset advertising delivery strategies corresponding to the preset delivery environment information, which facilitates the acquisition of multiple preset advertising delivery strategies corresponding to the delivery environment information of the advertising delivery position at the current moment, thereby improving the accuracy of determining the advertising delivery strategy.

[0118] In step S103, the strategy parameters and the advertising revenue estimation model corresponding to each advertising delivery strategy are obtained.

[0119] In one possible implementation, before obtaining the advertising revenue estimation model, the initial advertising revenue estimation model is trained based on sample advertising delivery strategies and revenue label data of the sample advertising delivery strategies to determine the advertising revenue estimation model, and the initial advertising revenue estimation model is constructed based on a reinforcement learning model.

[0120] In one possible implementation, reinforcement learning is a method that allows an agent to learn the optimal behavior strategy through interaction with the environment. The agent performs actions in a given state, and the environment gives rewards based on the actions. The agent's goal is to maximize the long-term cumulative rewards. Specifically, the advertising delivery system is the agent; the market environment for advertising delivery is the environment; the target users, advertising positions and other information for advertising delivery are the state, that is, the description of the environment at a certain moment; the action is the operation that the agent can take in a certain state, and the strategy for advertising delivery is the action; the reward is the feedback obtained by the agent after taking action, and the revenue from advertising delivery is the reward.

[0121] In one possible implementation, the initial advertising revenue estimation model is built based on the Deep Q-Network (DQN). DQN is a deep reinforcement learning algorithm that combines deep learning and Q-Learning to solve the problem of large state space by approximating the Q function through a neural network. DQN introduces experience replay and target network to improve training stability and efficiency. DQN's Q-Learning is a model-free reinforcement learning algorithm that predicts the expected return of taking a certain action in a certain state by learning a state-action value function (Q function). DQN uses two neural networks with the same structure but different parameters: one for predicting Q values ​​(main network) and the other for calculating target Q values ​​(target network). The parameters of the target network are updated regularly, which helps to reduce instability during training. Specifically, the revenue from advertising is the Q value.

[0122] In a possible implementation, each advertising plan can control the budget of each audience group, the advertising delivery time and evaluate the delivery effect of each audience by setting delivery goals, user targeting, promotion budget, promotion cycle, bidding, etc. An advertising plan may contain one or more advertising creatives.

[0123] In one possible implementation, advertising creativity is the smallest unit of delivery. Advertising creativity is a level in the delivery process of an advertising platform. It is mainly used to encapsulate scattered material capabilities at this level to produce advertising materials for actual delivery.

[0124] In one possible implementation, the strategy parameters corresponding to each advertising delivery strategy include data such as exposure, clicks, costs, click costs, and conversion costs of the advertising plans corresponding to each advertising delivery strategy, and data such as exposure, clicks, costs, click costs, and conversion costs of the advertising creatives corresponding to each advertising delivery strategy.

[0125] In a possible implementation, there are multiple sample advertising delivery strategies, which are obtained based on the server connecting to the advertising delivery platform.

[0126] In a possible implementation, the policy parameters of the sample advertisement delivery policy include target user information and advertisement delivery position information.

[0127] In one possible implementation, the target user information is the user's behavior data, such as registration conversion, order placement, order payment, advertising exposure, advertising clicks, and other data, as well as the channel ID, link ID, advertising plan, advertising creativity and other information carried by the data.

[0128] In one possible implementation, the advertising placement information is the data such as the advertising plan, exposure, clicks, costs, click costs and conversion costs of advertising creatives obtained by the server when connecting to the advertising delivery platform, as well as the data carrying channel ID, link ID, advertising plan, advertising creatives and other information.

[0129] In a possible implementation, the revenue tag data of the sample advertising delivery strategy is the revenue after the sample advertising delivery strategy is delivered. Specifically, the revenue includes real-time order revenue and offline order revenue, the real-time order revenue is the revenue on the day of advertising delivery, and the offline order revenue is the revenue after the day of advertising delivery.

[0130] In a possible implementation, the training method of the advertising revenue estimation model includes:

[0131] Acquire the initial advertising revenue estimation model, where the advertising revenue estimation model is constructed based on a reinforcement learning model;

[0132] Inputting the strategy parameters of the sample advertising delivery strategy into the initial advertising revenue estimation model to perform revenue estimation processing to obtain estimated revenue data corresponding to the sample advertising delivery strategy;

[0133] Matching the revenue label data of the sample advertising delivery strategy with the estimated revenue data of the corresponding sample advertising delivery strategy to obtain a matching result;

[0134] In the case where the matching result does not satisfy the preset condition, determining difference information between the revenue label data of the sample advertisement delivery strategy and the estimated revenue data corresponding to the sample advertisement delivery strategy;

[0135] The network parameters of the initial advertising revenue estimation model are updated based on the difference information until the matching result satisfies the preset condition, and the initial advertising revenue estimation model corresponding to the matching result that satisfies the preset condition is determined as the advertising revenue estimation model.

[0136] In one possible implementation, Figure 3 As shown, the training method of the advertising revenue estimation model includes steps S301 to S309:

[0137] In step S301, the initial advertising revenue estimation model is obtained, and the advertising revenue estimation model is constructed based on a reinforcement learning model;

[0138] In step S303, the strategy parameters of the sample advertisement delivery strategy are input into the initial advertisement revenue estimation model for revenue estimation processing to obtain estimated revenue data corresponding to the sample advertisement delivery strategy;

[0139] In step S305, the revenue label data of the sample advertisement delivery strategy is matched with the estimated revenue data of the corresponding sample advertisement delivery strategy to obtain a matching result;

[0140] In step S307, when the matching result does not meet the preset condition, determining the difference information between the revenue label data of the sample advertisement delivery strategy and the estimated revenue data corresponding to the sample advertisement delivery strategy;

[0141] In step S309, the network parameters of the initial advertising revenue estimation model are updated based on the difference information until the matching result satisfies the preset condition, and the initial advertising revenue estimation model corresponding to the matching result that satisfies the preset condition is determined as the advertising revenue estimation model.

[0142] In one possible implementation, the server connects to the advertising delivery platform to obtain data such as advertising plan, advertising creative exposure, clicks, costs, click costs and conversion costs; at the same time, through real-time big data calculations, real-time user behavior data is obtained, such as registration conversion, order placement, order payment, advertising exposure, advertising clicks and other data; these data are aligned according to the advertising plan and advertising creative dimensions to obtain the strategy parameters of the sample advertising delivery strategy.

[0143] In a possible implementation, the revenue after the sample advertisement delivery strategy is delivered is obtained by the server as the revenue label data of the sample advertisement delivery strategy.

[0144] In one possible implementation, the strategy parameters of the sample advertising delivery strategy are input into the initial advertising revenue estimation model for revenue estimation processing to obtain estimated revenue data corresponding to the sample advertising delivery strategy; the revenue label data of the sample advertising delivery strategy and the estimated revenue data of the corresponding sample advertising delivery strategy are matched to obtain a matching result; if the matching result does not meet a preset condition, the difference information between the revenue label data of the sample advertising delivery strategy and the estimated revenue data of the corresponding sample advertising delivery strategy is determined; based on the difference information, the network parameters of the initial advertising revenue estimation model are updated until the matching result meets the preset condition, and the initial advertising revenue estimation model corresponding to the matching result that meets the preset condition is determined to be the advertising revenue estimation model.

[0145] In one possible implementation, when the difference information between the revenue label data of the sample advertising delivery strategy and the estimated revenue data of the corresponding sample advertising delivery strategy is less than a preset difference value, it is determined that the matching result meets the preset condition so that the advertising revenue estimation model can accurately predict the revenue of the advertising delivery strategy.

[0146] In a possible implementation, historical data of advertisement delivery strategies generated in real-time bidding are collected, and when the historical data meets a preset data volume condition, the advertisement revenue estimation model is optimized based on the historical data.

[0147] In this implementation, the advertising revenue estimation model based on reinforcement learning can find advertising plans more flexibly and reasonably, and determine the advertising delivery strategy based on the estimated ROI of the advertising plan and advertising creativity, as well as the user's in-app behavior characteristics, so as to facilitate the subsequent automatic adjustment of bids based on the estimated revenue of the advertising delivery strategy, thereby improving the return on advertising revenue.

[0148] In step S105, the strategy parameters corresponding to each advertisement delivery strategy are input into the advertisement revenue estimation model for revenue estimation processing to obtain estimated revenue data for each advertisement delivery strategy.

[0149] In a possible implementation, the strategy parameters corresponding to each advertisement delivery strategy are input into the advertisement revenue estimation model to obtain the output of the advertisement revenue estimation model as the estimated revenue data of each advertisement delivery strategy.

[0150] In step S107, object characteristic parameters of the delivery object are obtained, and decision analysis is performed based on the estimated revenue data of each advertisement delivery strategy and the object characteristic parameters, so as to select a target delivery strategy from the multiple advertisement delivery strategies.

[0151] In a possible implementation, the decision analysis is performed based on the estimated revenue data of each advertising delivery strategy and the object characteristic parameters to select a target delivery strategy from the multiple advertising delivery strategies, including:

[0152] Performing revenue analysis on the estimated revenue data of each advertising delivery strategy to obtain the estimated revenue of each advertising delivery strategy;

[0153] In the case where the estimated revenue is greater than the preset revenue threshold, determining that the advertisement delivery strategy corresponding to the estimated revenue greater than the preset revenue threshold is an intermediate delivery strategy;

[0154] Performing feature evaluation on the object feature parameters to obtain object feature data;

[0155] A matching analysis is performed on the intermediate delivery strategy and the object feature data, and the intermediate delivery strategy whose matching analysis result satisfies a preset matching condition is determined as the target delivery strategy.

[0156] In a possible implementation, the object characteristic parameters include a mobile device identifier of the object, an object portrait, a behavior characteristic, and an object behavior time.

[0157] The step of performing feature evaluation on the object feature parameters to obtain object feature data includes:

[0158] Calculate the revenue value based on the mobile device identifier, the object portrait, the behavior characteristics and the object behavior time to obtain an estimated revenue value;

[0159] The estimated benefit value is compared with the preset object benefit value to obtain a comparison result, and the object feature data is determined based on the comparison result.

[0160] In a possible implementation, the preset object benefit value includes a first benefit value and a second benefit value, the first benefit value is greater than the second benefit value, the object feature data includes first object feature data, second object feature data and third object feature data,

[0161] The step of comparing the estimated benefit value with the preset object benefit value to obtain a comparison result, and determining the object feature data based on the comparison result includes:

[0162] In a case where the estimated benefit value is greater than or equal to the first benefit value, determining the object feature data to be the first object feature data;

[0163] When the estimated benefit value is less than the first benefit value and greater than or equal to the second benefit value, determining that the object feature data is the second object feature data;

[0164] When the estimated benefit value is less than the second benefit value, the object feature data is determined to be the third object feature data.

[0165] In a possible implementation, the intermediate strategy includes a first strategy, a second strategy, and a third strategy, and the object feature data includes first object feature data, second object feature data, and third object feature data.

[0166] The performing matching analysis on the intermediate delivery strategy and the object feature data, and determining the intermediate delivery strategy whose matching analysis result satisfies a preset matching condition as the target delivery strategy, includes:

[0167] In a case where the object characteristic data is the first object characteristic data, determining the first strategy as the target delivery strategy;

[0168] In a case where the object characteristic data is the second object characteristic data, determining the second strategy as the target delivery strategy;

[0169] In a case where the object characteristic data is the third object characteristic data, the third strategy is determined to be the target delivery strategy.

[0170] In one possible implementation, Figure 4 As shown, the decision analysis is performed based on the estimated revenue data of each advertising delivery strategy and the object characteristic parameters to select a target delivery strategy from the multiple advertising delivery strategies, including steps S401 to S407:

[0171] In step S401, the estimated revenue data of each advertising delivery strategy is analyzed to obtain the estimated revenue of each advertising delivery strategy;

[0172] In step S403, when the estimated revenue is greater than the preset revenue threshold, determining the advertisement delivery strategy corresponding to the estimated revenue greater than the preset revenue threshold as the intermediate delivery strategy;

[0173] In step S405, feature evaluation is performed on the object feature parameters to obtain object feature data;

[0174] In step S407, a matching analysis is performed on the intermediate delivery strategy and the object feature data, and the intermediate delivery strategy whose matching analysis result satisfies a preset matching condition is determined as the target delivery strategy.

[0175] In a possible implementation, an object characteristic parameter of the delivery object is obtained, and if the object characteristic parameter indicates that the delivery object is a user in the basic blacklist, the bidding is directly abandoned.

[0176] In a possible implementation, the object feature data is the calculated user estimated revenue score of the object. Specifically, the user estimated revenue score is calculated by combining the user's oaidmd5 value (an identifier recognized by a mobile device), the offline calculated user portrait, the user's click behavior, browsing history and other behavioral features, with time features and other parameters to calculate the user's estimated revenue score rankscore.

[0177] rankscore=estimatedCPC*pv+w*estimatedecpm+price,

[0178] in,

[0179] Estimated CPC is the estimated cost per click. CPC (Cost Per Click) represents a model in which advertisers pay fees to advertising platforms based on the number of times users click on their ads.

[0180] pv: pv (page views) refers to page views in advertising, that is, the number of visits to a website page;

[0181] Estimated ecpm: the estimated advertising revenue per thousand impressions;

[0182] price: estimated order revenue;

[0183] w: is the ecpm weight coefficient, which is configured by the operator.

[0184] In one possible implementation, the PV value reflects the popularity of a website. For example, large portal websites such as Sina and NetEase have over 100 million visitors every day. PV directly reflects the exposure of the advertisement and the attention of the audience. Advertisers usually evaluate the effectiveness of advertising based on PV data, and adjust advertising strategies to improve the attractiveness and click-through rate of the advertisement.

[0185] In one possible implementation, eCPM (effective cost per mille) refers to the advertising revenue that can be obtained for every thousand impressions. The unit of display can be a web page, an advertising unit, or even a single advertisement.

[0186] In a possible implementation, a matching analysis is performed on the intermediate delivery strategy and the object feature data, and the intermediate delivery strategy whose matching analysis result meets the preset matching condition is determined as the target delivery strategy, that is, the target delivery strategy is determined according to the user's estimated revenue rank score and traffic grouping. The higher the user's estimated revenue score, the greater the probability of obtaining revenue from advertising delivery, and the greater the probability of obtaining the highest revenue.

[0187] In a possible implementation, users are divided into high-value users, medium-value users, and low-value users according to their estimated revenue scores, wherein the user estimated revenue scores of high-value users are greater than or equal to the first revenue value, the user estimated revenue scores of medium-value users are less than the first revenue value and greater than or equal to the second revenue value, and the user estimated revenue scores of low-value users are less than the second revenue value.

[0188] In a possible implementation, the target delivery strategy for matching high-value users is the first strategy, the target delivery strategy for matching medium-value users is the second strategy, and the target delivery strategy for matching low-value users is the third strategy.

[0189] In this implementation, the estimated revenue data of each advertising delivery strategy is analyzed to obtain the estimated revenue of each advertising delivery strategy; when the estimated revenue is greater than a preset revenue threshold, the advertising delivery strategy corresponding to the estimated revenue greater than the preset revenue threshold is determined to be the intermediate delivery strategy; the object feature parameters are evaluated to obtain object feature data; the intermediate delivery strategy and the object feature data are matched and analyzed, and the intermediate delivery strategy whose matching analysis results meet the preset matching conditions is determined as the target delivery strategy. Based on the estimated revenue data and the object feature parameters of the delivery object, the advertising delivery strategy for the delivery object is determined, which improves the accuracy of advertising delivery and thereby improves the return on advertising delivery. It also reduces labor costs and improves the efficiency of real-time bidding.

[0190] In step S109, resources are configured for the target delivery strategy to obtain a resource configuration result.

[0191] In a possible implementation, the bidding configuration information of the intelligent strategy library is read. Through the bidding intelligent strategy configured by the advertising delivery platform, the intelligent agent module can obtain this configuration by accessing the API interface provided by the advertising delivery platform, load it into the memory, and conduct user bidding according to the bidding logic of the intelligent strategy.

[0192] In a possible implementation, an advertisement bidding is conducted based on the target delivery strategy. If the bidding is successful, the bidding strategy, plan ID and other information are recorded, subsequent conversion behaviors are tracked, and real-time ROI (reward) is calculated.

[0193] In a possible implementation, resources are configured for the target delivery strategy to obtain a resource configuration result, that is, the budget for the target delivery strategy is increased to increase the probability of successful bidding for the target delivery strategy.

[0194] In step S1011, when the resource configuration result meets the preset delivery condition, advertisement delivery is performed on the delivery object according to the target delivery strategy.

[0195] In a possible implementation, when the resource configuration result meets the preset delivery condition, it is determined that the bidding is successful, and the advertisement is delivered to the delivery object according to the target delivery strategy.

[0196] In one possible implementation, after a bid is successful, the user's behavior data after the bid, such as in-app ad clicks and member purchase orders, is combined with the cost, budget, clicks and other data of the advertising plan to re-optimize the model and calculate the ROI of the plan; the advertising plans are sorted from high to low according to the calculated ROI scores of the plans.

[0197] In a possible implementation, the advertising plan with the ROI score is encapsulated into a return structure and returned to the advertising delivery end. After the subsequent advertising delivery system obtains the advertising plan object with the score, it will adjust the advertising plan budget according to the score from high to low, put the budget into the plan with high ROI, and shut down the advertising plan with low ROI.

[0198] In a possible implementation, in the environment of real-time bidding for advertising, the flowmode field of the traffic grouping brought by the advertising plan is split into two AB sub-queues, namely the intelligent adjustment candidate advertising queue and the manual strategy candidate advertising queue, and a label is added to each click behavior of the advertising plan. Flowmode = A means that the user comes from the intelligent strategy, that is, the target delivery strategy under the optimization method of the advertising delivery strategy; flowmode = B means that the user comes from the manual strategy.

[0199] In a possible implementation, for each bid, the subsequent conversion revenue when flowmode=A is recorded, and the subsequent conversion revenue when flowmode=B is recorded, and the revenue brought by the intelligent strategy is verified by comparing the subsequent conversion differences.

[0200] Figure 5 The schematic diagram of the structure of an optimization device 500 for an advertisement delivery strategy provided by an embodiment of the present application is shown. The device has the function of implementing the optimization method for advertisement delivery strategy in the above method embodiment. The function can be implemented by hardware or by hardware executing corresponding software. Figure 5 As shown, the device may include:

[0201] The current environment information acquisition module 501 is used to obtain the delivery environment information of the current advertisement delivery position, and determine the initial delivery strategy according to the delivery environment information, wherein the initial delivery strategy includes multiple advertisement delivery strategies;

[0202] Model acquisition module 502, used to acquire strategy parameters and advertising revenue estimation models corresponding to each advertising delivery strategy;

[0203] The revenue estimation processing module 503 is used to input the strategy parameters corresponding to each advertising delivery strategy into the advertising revenue estimation model to perform revenue estimation processing, and obtain the estimated revenue data of each advertising delivery strategy;

[0204] A decision analysis module 504 is used to obtain object characteristic parameters of the delivery object, perform decision analysis based on the estimated revenue data of each advertising delivery strategy and the object characteristic parameters, and select a target delivery strategy from the multiple advertising delivery strategies;

[0205] The resource configuration module 505 is used to perform resource configuration on the target delivery strategy to obtain a resource configuration result;

[0206] The advertisement delivery module 506 is used to deliver advertisements to the delivery object according to the target delivery strategy when the resource configuration result meets the preset delivery condition.

[0207] In a possible implementation, the current environment information acquisition module 501 includes an initial delivery strategy determination module, and the initial delivery strategy determination module is used to:

[0208] Acquire an advertisement delivery strategy database, wherein the advertisement delivery strategy database includes preset delivery environment information of advertisement delivery positions and preset advertisement delivery strategies;

[0209] Based on the delivery environment information, the advertisement delivery strategy database is queried to obtain a preset advertisement delivery strategy matching the delivery environment information, and the preset advertisement delivery strategy matching the delivery environment information is determined to be the initial delivery strategy; the advertisement delivery strategy database is constructed based on the preset delivery environment information of the advertisement delivery position and the preset advertisement delivery strategy.

[0210] In a possible implementation, the current environment information acquisition module 501 includes an advertisement delivery strategy database construction unit, and the advertisement delivery strategy database construction unit is used to:

[0211] Acquire a delivery environment information sample set, wherein the delivery environment information sample set is different preset delivery environment information for advertisement delivery;

[0212] Setting at least one preset advertisement delivery strategy based on each preset delivery environment information;

[0213] The advertisement delivery strategy database is constructed based on each preset delivery environment information and the preset advertisement delivery strategy corresponding to each preset delivery environment information.

[0214] In a possible implementation, the decision analysis module 504 is used to:

[0215] Performing revenue analysis on the estimated revenue data of each advertising delivery strategy to obtain the estimated revenue of each advertising delivery strategy;

[0216] In the case where the estimated revenue is greater than the preset revenue threshold, determining that the advertisement delivery strategy corresponding to the estimated revenue greater than the preset revenue threshold is an intermediate delivery strategy;

[0217] Performing feature evaluation on the object feature parameters to obtain object feature data;

[0218] A matching analysis is performed on the intermediate delivery strategy and the object feature data, and the intermediate delivery strategy whose matching analysis result satisfies a preset matching condition is determined as the target delivery strategy.

[0219] In a possible implementation, the object feature parameters include a mobile device identifier, an object portrait, a behavior feature, and an object behavior time of the object, and the decision analysis module 504 includes a feature evaluation unit, and the feature evaluation unit is used to:

[0220] Calculate the revenue value based on the mobile device identifier, the object portrait, the behavior characteristics and the object behavior time to obtain an estimated revenue value;

[0221] The estimated benefit value is compared with the preset object benefit value to obtain a comparison result, and the object feature data is determined based on the comparison result.

[0222] In a possible implementation, the preset object benefit value includes a first benefit value and a second benefit value, the first benefit value is greater than the second benefit value, the object feature data includes first object feature data, second object feature data and third object feature data, and the decision analysis module 504 includes a comparison unit, and the comparison unit is used to:

[0223] In a case where the estimated benefit value is greater than or equal to the first benefit value, determining the object feature data to be the first object feature data;

[0224] When the estimated benefit value is less than the first benefit value and greater than or equal to the second benefit value, determining that the object feature data is the second object feature data;

[0225] When the estimated benefit value is less than the second benefit value, the object feature data is determined to be the third object feature data.

[0226] In a possible implementation, the intermediate strategy includes a first strategy, a second strategy, and a third strategy, the object feature data includes first object feature data, second object feature data, and third object feature data, and the decision analysis module 504 includes a matching analysis unit, and the matching analysis unit is used to:

[0227] In a case where the object characteristic data is the first object characteristic data, determining the first strategy as the target delivery strategy;

[0228] In a case where the object characteristic data is the second object characteristic data, determining the second strategy as the target delivery strategy;

[0229] In a case where the object characteristic data is the third object characteristic data, the third strategy is determined to be the target delivery strategy.

[0230] In a possible implementation, the revenue estimation processing module 503 includes an advertisement revenue estimation model training unit, and the advertisement revenue estimation model training unit is used to:

[0231] Acquire the initial advertising revenue estimation model, where the advertising revenue estimation model is constructed based on a reinforcement learning model;

[0232] Inputting the strategy parameters of the sample advertising delivery strategy into the initial advertising revenue estimation model to perform revenue estimation processing to obtain estimated revenue data corresponding to the sample advertising delivery strategy;

[0233] Matching the revenue label data of the sample advertising delivery strategy with the estimated revenue data of the corresponding sample advertising delivery strategy to obtain a matching result;

[0234] In the case where the matching result does not satisfy the preset condition, determining difference information between the revenue label data of the sample advertisement delivery strategy and the estimated revenue data corresponding to the sample advertisement delivery strategy;

[0235] The network parameters of the initial advertising revenue estimation model are updated based on the difference information until the matching result satisfies the preset condition, and the initial advertising revenue estimation model corresponding to the matching result that satisfies the preset condition is determined as the advertising revenue estimation model.

[0236] It should be noted that the device provided in the above embodiment, when implementing its functions, is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0237] An embodiment of the present application provides an electronic device, which includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement an optimization method for any one of the advertising delivery strategies provided in the above method embodiments.

[0238] The memory can be used to store software programs and modules. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, application programs required for functions, etc.; the data storage area can store data created according to the use of the device, etc. In addition, the memory can include a high-speed random access memory and can also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices. Accordingly, the memory can also include a memory controller to provide the processor with access to the memory.

[0239] An embodiment of the present application also provides a computer-readable storage medium, which can be set in an electronic device to store at least one instruction or at least one program related to a method for optimizing an advertising delivery strategy. The at least one instruction or the at least one program is loaded and executed by the processor to implement any one of the methods for optimizing an advertising delivery strategy provided in the above method embodiments.

[0240] Optionally, in this embodiment, the above-mentioned storage medium may include but is not limited to: a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and other media that can store program codes.

[0241] It should be noted that the above-mentioned sequence of the embodiments of the present application is for description only and does not represent the advantages and disadvantages of the embodiments. The above-mentioned specific embodiments of this specification are described. Other embodiments are within the scope of the attached claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0242] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0243] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0244] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for optimizing an advertising delivery strategy, characterized in that: The method comprises: Acquire the delivery environment information of the current advertisement delivery position, and determine the initial delivery strategy according to the delivery environment information, wherein the initial delivery strategy includes multiple advertisement delivery strategies; Obtain the strategy parameters and advertising revenue estimation model corresponding to each advertising delivery strategy; Inputting the strategy parameters corresponding to each advertising delivery strategy into the advertising revenue estimation model to perform revenue estimation processing, thereby obtaining estimated revenue data for each advertising delivery strategy; Obtaining object characteristic parameters of the delivery object, performing decision analysis based on the estimated revenue data of each advertising delivery strategy and the object characteristic parameters, and selecting a target delivery strategy from the multiple advertising delivery strategies; Performing resource allocation on the target delivery strategy to obtain a resource allocation result; When the resource configuration result meets the preset delivery condition, advertisement delivery is performed on the delivery object according to the target delivery strategy.

2. The method for optimizing the advertising delivery strategy according to claim 1, characterized in that: The determining of the initial delivery strategy according to the delivery environment information includes: Acquire an advertisement delivery strategy database, wherein the advertisement delivery strategy database includes preset delivery environment information of advertisement delivery positions and preset advertisement delivery strategies; Based on the delivery environment information, the advertisement delivery strategy database is queried to obtain a preset advertisement delivery strategy matching the delivery environment information, and the preset advertisement delivery strategy matching the delivery environment information is determined to be the initial delivery strategy; the advertisement delivery strategy database is constructed based on the preset delivery environment information of the advertisement delivery position and the preset advertisement delivery strategy.

3. The method for optimizing the advertising delivery strategy according to claim 2, characterized in that: The method for constructing the advertisement delivery strategy database includes: Acquire a delivery environment information sample set, where the delivery environment information sample set is different preset delivery environment information for advertisement delivery; Setting at least one preset advertisement delivery strategy based on each preset delivery environment information; The advertisement delivery strategy database is constructed based on each preset delivery environment information and the preset advertisement delivery strategy corresponding to each preset delivery environment information.

4. The method for optimizing the advertising delivery strategy according to claim 1, characterized in that: The decision analysis is performed based on the estimated revenue data of each advertising delivery strategy and the object characteristic parameters to select a target delivery strategy from the multiple advertising delivery strategies, including: Performing revenue analysis on the estimated revenue data of each advertising delivery strategy to obtain the estimated revenue of each advertising delivery strategy; In the case where the estimated revenue is greater than the preset revenue threshold, determining that the advertisement delivery strategy corresponding to the estimated revenue greater than the preset revenue threshold is an intermediate delivery strategy; Performing feature evaluation on the object feature parameters to obtain object feature data; A matching analysis is performed on the intermediate delivery strategy and the object feature data, and the intermediate delivery strategy whose matching analysis result satisfies a preset matching condition is determined as the target delivery strategy.

5. The method for optimizing the advertising delivery strategy according to claim 4, characterized in that: The object characteristic parameters include the object's mobile device identifier, object portrait, behavior characteristics and object behavior time. The step of performing feature evaluation on the object feature parameters to obtain object feature data includes: Calculate the revenue value based on the mobile device identifier, the object portrait, the behavior characteristics and the object behavior time to obtain an estimated revenue value; The estimated benefit value is compared with the preset object benefit value to obtain a comparison result, and the object feature data is determined based on the comparison result.

6. The method for optimizing the advertising delivery strategy according to claim 5, characterized in that: The preset object benefit value includes a first benefit value and a second benefit value, the first benefit value is greater than the second benefit value, the object feature data includes first object feature data, second object feature data and third object feature data, The step of comparing the estimated benefit value with the preset object benefit value to obtain a comparison result, and determining the object feature data based on the comparison result includes: In a case where the estimated benefit value is greater than or equal to the first benefit value, determining the object feature data to be the first object feature data; When the estimated benefit value is less than the first benefit value and greater than or equal to the second benefit value, determining that the object feature data is the second object feature data; When the estimated benefit value is less than the second benefit value, the object feature data is determined to be the third object feature data.

7. The method for optimizing the advertising delivery strategy according to claim 4, characterized in that: The intermediate strategy includes a first strategy, a second strategy and a third strategy, and the object feature data includes a first object feature data, a second object feature data and a third object feature data. The performing matching analysis on the intermediate delivery strategy and the object feature data, and determining the intermediate delivery strategy whose matching analysis result satisfies a preset matching condition as the target delivery strategy, includes: In a case where the object characteristic data is the first object characteristic data, determining the first strategy as the target delivery strategy; In a case where the object characteristic data is the second object characteristic data, determining the second strategy as the target delivery strategy; In a case where the object characteristic data is the third object characteristic data, the third strategy is determined to be the target delivery strategy.

8. The method for optimizing the advertising delivery strategy according to claim 1, characterized in that: The training method of the advertising revenue estimation model includes: Acquire the initial advertising revenue estimation model, where the advertising revenue estimation model is constructed based on a reinforcement learning model; Inputting the strategy parameters of the sample advertising delivery strategy into the initial advertising revenue estimation model to perform revenue estimation processing to obtain estimated revenue data corresponding to the sample advertising delivery strategy; Matching the revenue label data of the sample advertising delivery strategy with the estimated revenue data of the corresponding sample advertising delivery strategy to obtain a matching result; In the case where the matching result does not satisfy the preset condition, determining difference information between the revenue label data of the sample advertisement delivery strategy and the estimated revenue data corresponding to the sample advertisement delivery strategy; The network parameters of the initial advertising revenue estimation model are updated based on the difference information until the matching result satisfies the preset condition, and the initial advertising revenue estimation model corresponding to the matching result that satisfies the preset condition is determined as the advertising revenue estimation model.

9. An optimization device for advertising delivery strategy, characterized in that: The device comprises: A current environment information acquisition module is used to obtain the delivery environment information of the current advertisement delivery position, and determine an initial delivery strategy according to the delivery environment information, wherein the initial delivery strategy includes multiple advertisement delivery strategies; A model acquisition module is used to obtain the strategy parameters and advertising revenue estimation model corresponding to each advertising delivery strategy; A revenue estimation processing module, used for inputting the strategy parameters corresponding to each advertising delivery strategy into the advertising revenue estimation model to perform revenue estimation processing, and obtain the estimated revenue data of each advertising delivery strategy; A decision analysis module, used to obtain object characteristic parameters of the delivery object, perform decision analysis based on the estimated revenue data of each advertising delivery strategy and the object characteristic parameters, and select a target delivery strategy from the multiple advertising delivery strategies; A resource configuration module, used to perform resource configuration on the target delivery strategy and obtain a resource configuration result; The advertisement delivery module is used to deliver advertisements to the delivery object according to the target delivery strategy when the resource configuration result meets the preset delivery conditions.

10. An electronic device, characterized in that: It comprises a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor to implement the optimization method of the advertising delivery strategy according to any one of claims 1 to 8.

Images