CN115907861A

CN115907861A - Advertisement delivery method and advertisement delivery system

Info

Publication number: CN115907861A
Application number: CN202111161721.7A
Authority: CN
Inventors: 杨博文; 朱子晗; 白杨; 杨慧斌
Original assignee: 4Paradigm Beijing Technology Co Ltd
Current assignee: 4Paradigm Beijing Technology Co Ltd
Priority date: 2021-09-30
Filing date: 2021-09-30
Publication date: 2023-04-04

Abstract

Provided are an advertisement delivery method and an advertisement delivery system. The advertisement putting method can comprise the following steps: acquiring delivery target data and environment data of the advertisement; inputting the putting target data into a putting parameter prediction model to obtain putting parameter prediction data; inputting the release parameter prediction data and the environment data into an environment learning model to obtain release target prediction data; inputting the launching target prediction data into a launching parameter optimization model to obtain launching parameter optimization data; and delivering the advertisement according to the delivery parameter optimization data.

Description

Advertisement delivery method and advertisement delivery system

Technical Field

The present disclosure relates to the field of computer technologies, and in particular, to an advertisement delivery method and an advertisement delivery system.

Background

Since the design of the advertisement delivery scheme generally involves prediction and optimization for multiple advertisement delivery targets, it is necessary for operators to have quite abundant advertisement operation experience and excellent business understanding capability and decision-making capability, and thus the requirement on the quality of the operators is high. Some operators may predict an advertisement placement scheme using a machine learning model, but the conventional machine learning model for advertisement placement prediction is less effective.

Disclosure of Invention

Exemplary embodiments of the present disclosure may address, at least in part, the above-mentioned problems.

According to an aspect of the present disclosure, there is provided an advertisement delivery method, including: acquiring delivery target data and environment data of the advertisement; inputting the delivery target data into a delivery parameter prediction model to obtain delivery parameter prediction data; inputting the release parameter prediction data and the environment data into an environment learning model to obtain release target prediction data; inputting the launching target prediction data into a launching parameter optimization model to obtain launching parameter optimization data; and delivering the advertisement according to the delivery parameter optimization data.

Optionally, the delivery parameter prediction model is obtained by training through the following steps: acquiring release parameter sample data and release target sample data of a sample advertisement; determining a return function of the release parameter prediction model by using a reverse reinforcement learning algorithm by respectively taking the release parameter sample data and the release target sample data as input and output of a release parameter prediction model; and adjusting parameters of the input parameter prediction model according to the return function.

Optionally, the environment learning model is obtained by training through the following steps: acquiring environmental sample data of a sample advertisement; inputting the environment sample data and putting parameter prediction training data acquired by using the trained putting parameter prediction model with the putting target sample data as input into an environment learning model to acquire putting target prediction training data; determining a loss function of an environmental learning model according to the putting target prediction training data and the putting target sample data; and adjusting parameters of the environment learning model according to the loss function of the environment learning model.

Optionally, the release parameter optimization model is obtained by training through the following steps: inputting the putting target prediction training data acquired by using the trained environment learning model with the environment sample data and the putting parameter prediction training data as input into an putting parameter optimization model to acquire putting parameter optimization training data; determining a loss function of a release parameter optimization model according to the release parameter optimization training data and the release parameter sample data; and adjusting parameters of the release parameter optimization model according to the loss function of the release parameter optimization model.

Optionally, the delivery parameter prediction data includes a delivery parameter prediction value and a corresponding weight, the delivery parameter optimization data includes a delivery parameter optimization value and a corresponding weight, and/or the delivery parameter prediction training data includes a delivery parameter prediction training value and a corresponding weight.

Optionally, inputting the delivery parameter prediction data and the environmental data to an environmental learning model, comprising: inputting the environmental data and a putting parameter predicted value corresponding to the weight larger than or equal to a preset weight threshold value into an environmental learning model; and/or inputting the environment sample data and the putting parameter prediction training data acquired by using the trained putting parameter prediction model and taking the putting target sample data as input into an environment learning model, wherein the method comprises the following steps: and inputting the environment sample data and a putting parameter prediction training value corresponding to the weight larger than or equal to a preset weight threshold value into an environment learning model.

Optionally, the placement target data includes data related to advertising revenue targets, the environment data includes data related to at least one of an advertising propagation environment, an advertising sponsor, and an auction advertisement, and the placement parameter prediction data and the placement parameter optimization data include data related to advertising placement parameters, respectively.

Optionally, the delivery parameter prediction model, the environment learning model, and the delivery parameter optimization model are updated by performing iterative training using the obtained delivery target data, the environment data, and the delivery parameter optimization data.

According to another aspect of the present disclosure, there is provided an advertisement delivery system including: a data acquisition device configured to acquire delivery target data and environment data of an advertisement; a delivery parameter prediction device configured to input the delivery target data to a delivery parameter prediction model to obtain delivery parameter prediction data; a delivery target prediction device configured to input the delivery parameter prediction data and the environment data to an environment learning model to obtain delivery target prediction data; a delivery parameter optimization device configured to input the delivery target prediction data to a delivery parameter optimization model to obtain delivery parameter optimization data; an advertisement delivery device configured to deliver the advertisement according to the delivery parameter optimization data.

Optionally, the delivery parameter prediction model is obtained by training through a first training device, where the first training device includes: the first sample data acquisition device is configured to acquire delivery parameter sample data and delivery target sample data of the sample advertisement; a first function determination device configured to determine a reward function of the release parameter prediction model by using a reverse reinforcement learning algorithm by using the release parameter sample data and the release target sample data as input and output of a release parameter prediction model respectively; a first parameter adjusting device configured to adjust parameters of the delivery parameter prediction model according to the return function.

Optionally, the environment learning model is obtained by training through a second training device, where the second training device includes: the second sample data acquisition device is configured to acquire the environment sample data of the sample advertisement; a delivery target prediction training data acquisition means configured to input the environment sample data and delivery parameter prediction training data acquired by using the trained delivery parameter prediction model with the delivery target sample data as input to an environment learning model to acquire delivery target prediction training data; a second function determination device configured to determine a loss function of an environmental learning model according to the delivery target prediction training data and the delivery target sample data; and a second parameter adjusting device configured to adjust the parameters of the environment learning model according to the loss function of the environment learning model.

Optionally, the release parameter optimization model is obtained by training through a third training device, where the third training device includes: a release parameter optimization training data acquisition means configured to input release target prediction training data acquired by using the trained environment learning model with the environment sample data and the release parameter prediction training data as inputs to a release parameter optimization model to acquire release parameter optimization training data; a third function determination device configured to determine a loss function of the release parameter optimization model according to the release parameter optimization training data and the release parameter sample data; and a third parameter adjusting device configured to adjust parameters of the release parameter optimization model according to the loss function of the release parameter optimization model.

Optionally, the delivery target prediction device is configured to: inputting the environmental data and a launching parameter predicted value corresponding to the weight larger than or equal to a preset weight threshold value into an environmental learning model to obtain launching target predicted data; and/or the putting goal prediction training data acquisition device is configured to: and inputting the environment sample data and the launching parameter prediction training value corresponding to the weight larger than or equal to the preset weight threshold value into an environment learning model to obtain launching target prediction training data.

Optionally, the delivery parameter prediction model, the environment learning model, and the delivery parameter optimization model are updated by performing iterative training using the obtained delivery target data, the obtained environment data, and the obtained delivery parameter optimization data.

According to another aspect of the present disclosure, there is provided a computer-readable storage medium storing instructions that, when executed by at least one computing device, cause the at least one computing device to perform an advertisement delivery method according to the present disclosure.

According to another aspect of the present disclosure, there is provided a system comprising at least one computing device and at least one storage device storing instructions, wherein the instructions, when executed by the at least one computing device, cause the at least one computing device to perform an advertising method according to the present disclosure.

According to the advertisement putting method, the device and the system, at least one of the following beneficial effects can be realized: the method comprises the steps of simultaneously considering traditional advertisement putting variables and environmental variables in the modeling process related to advertisement putting and the application of an advertisement putting model, fully considering multiple factors related to advertisements, training or putting cold start through historical data, continuously and iteratively optimizing the model related to advertisement putting, obtaining the optimal combination of one or more advertisement putting parameters, and guiding putting personnel to carry out advertisement putting according to the optimal putting parameter combination, so that the advertisement conversion rate is improved, the advertiser is helped to save budget, the relevant data of advertisement putting can be tracked for a long time to carry out continuous model updating, the income effect obtained after the prediction scheme is applied to an actual scene is remarkably improved, the short-term income of advertisements is focused, and the full consideration of long-term income such as long-term user retention, repeated purchasing and the like is also increased.

Drawings

These and/or other aspects and advantages of the disclosure will become apparent and more readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:

FIG. 1 is a flow diagram of a method of advertisement delivery, according to an embodiment of the present disclosure;

FIG. 2 is a data flow diagram of an advertisement delivery method according to an embodiment of the present disclosure;

FIG. 3 is a flow chart of a method of training a launch parameter prediction model according to an embodiment of the present disclosure;

FIG. 4 is a flow diagram of a method of training an environmental learning model according to an embodiment of the present disclosure;

FIG. 5 is a flow chart of a training method of a launch parameter optimization model according to an embodiment of the present disclosure;

FIG. 6 is a block diagram of an advertisement delivery system according to an embodiment of the present disclosure;

FIG. 7 is a block diagram of an advertising placement model training system, according to an embodiment of the present disclosure;

fig. 8 is a block diagram of a computing device according to an embodiment of the disclosure.

Detailed Description

Advertisement delivery involves various complex factors, and advertisement delivery optimization belongs to a complex multi-objective optimization problem, and is much more complex and more difficult to solve compared with a traditional single-objective optimization problem. In addition to traditional ad placement parameters (e.g., ad spots, materials, keywords, bids, slots, budgets, demographics, etc.), the effectiveness of ad placement may also be affected by uncertainty factors such as bid ad placement, new media hotwords, accidental star effects, etc.

Because advertisement delivery belongs to the problem of multi-objective optimization, the traditional method needs operators to have strong business understanding capability and decision-making capability to generate relatively good advertisement delivery effect. The traditional advertisement putting model only considers basic and single putting parameters, but the influence of environment variables on putting results in the real putting process is large, and the consideration of the environment variables is often ignored or the environment variables are difficult to optimize the advertisement putting strategy. Moreover, the traditional advertisement putting model is difficult to realize continuous learning iteration, only focuses on short-term benefits, and does not well consider long-term benefits such as long-term user retention, repeated purchase and the like.

The advertisement putting technical scheme provided by the invention can simultaneously take the traditional advertisement putting variable and the environmental variable into account in the modeling process and the model application process related to advertisement putting, remarkably improves the income effect obtained after the prediction scheme is applied to the actual scene, not only focuses on the short-term income of the advertisement, but also can increase the full consideration of the long-term income such as long-term user retention, repeated purchasing and the like.

The technical scheme of the invention is described in the following with reference to the accompanying drawings. The following description with reference to the accompanying drawings is provided to assist in a comprehensive understanding of the embodiments of the disclosure as defined by the claims and their equivalents. Various specific details are included to aid understanding, but these are merely to be considered exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present disclosure. In addition, descriptions of well-known functions and constructions are omitted for clarity and conciseness.

It should be noted that "at least one of the items" appearing in the present disclosure means a case where three types of juxtaposition including "any one of the items", "a combination of any two or more of the items", and "all of the items" are included. For example, "including at least one of the first unit and the second unit" includes the following three cases in parallel: (1) includes a first unit; (2) comprises a second unit; and (3) comprises a first unit and a second unit. For another example, "at least one of the first step and the second step is performed", which means that the following three cases are juxtaposed: (1) executing the step one; (2) executing the step two; and (3) executing the step one and the step two. In this disclosure, the term "and/or" includes any one of the associated listed items and any combination of any two or more.

Fig. 1 is a flow chart of an advertisement delivery method according to an embodiment of the present disclosure. Fig. 2 is a data flow diagram of an advertisement delivery method according to an embodiment of the present disclosure.

Referring to fig. 1 and 2, in step S11, delivery target data and environment data of an advertisement are acquired. According to embodiments of the present disclosure, placement targeting data may include data related to advertising revenue targeting, and environmental data may include data related to at least one of an advertising dissemination environment, an advertising sponsor, and an auction advertisement. For example, the placement targeting data may include data related to revenue targets (e.g., short term revenue targets, long term revenue targets) that the advertisement operator desires to achieve, such as advertisement click-through rates, advertisement conversion rates, advertisement collection amounts, collection amounts of products corresponding to the advertisements, buyback rates of products corresponding to the advertisements, gross Merchandis Volume (GMV) of products corresponding to the advertisements, customer order promotion amounts, new user introduction amounts, old user buyback rates, affiliate promotion amounts, and so forth. For example, environmental data may include data related to external influencing factors for ad placement (e.g., competitive ad placement, large disk contingencies for ad placement platforms, off-site marketing campaigns, hot news, star effects for ad promoters, hot search keywords, etc.). The delivery target data and the environmental data of the advertisement can be acquired through various advertisement delivery and operation channels. In an embodiment of the present disclosure, environmental data of an advertisement may be acquired in real time and changes in the environmental data may be monitored.

In step S12, the delivery target data may be input to the delivery parameter prediction model to obtain delivery parameter prediction data. The placement parameter prediction data may include data relating to ad placement parameters (e.g., ad resource slots, product bids corresponding to ads, ad placement content, etc.). According to embodiments of the present disclosure, the placement parameter prediction data may include a placement parameter prediction value and a corresponding weight.

In the embodiment of the present disclosure, the delivery parameter prediction model may be trained by the method shown in fig. 3. Fig. 3 is a flow chart of a method of training a launch parameter prediction model according to an embodiment of the present disclosure. As shown in fig. 3, in step S31, sample data of delivery parameters and sample data of delivery targets of a sample advertisement are acquired.

According to an embodiment of the present disclosure, the sample data of the serving parameters of the sample advertisement may include data related to historical serving operations of the sample advertisement (e.g., advertisement resource slots, product prices corresponding to the advertisement, advertisement serving content, etc.), and the sample data of serving objectives may include data related to revenue objectives of the sample advertisement. For example, the placement target sample data may include data related to revenue targets (e.g., short-term revenue targets, long-term revenue targets) desired by an operator of the sample advertisement, such as advertisement attention, advertisement click-through rate, advertisement conversion rate, purchase rate of a product corresponding to the advertisement, deal amount GMV (Gross merchandis Volume) of a product corresponding to the advertisement, customer price promotion amount, new user introduction amount, old user repurchase rate, membership promotion amount, and the like.

In step S32, the release parameter sample data and the release target sample data are respectively used as input and output of the release parameter prediction model, and a reward function of the release parameter prediction model is determined by using a reverse reinforcement learning algorithm. In step S33, parameters of the projected parameter prediction model are adjusted according to the reward function.

By training the delivery parameter prediction model by using a reverse reinforcement learning algorithm, delivery parameter sample data of the sample advertisement can be analyzed, and delivery parameter prediction training data containing delivery variable weights are obtained. For example, the placement parameter prediction training data may include placement parameter prediction training values and corresponding weights. When the prediction is carried out by using the trained putting parameter prediction model, putting parameter prediction data containing putting variable weights can be obtained, so that putting parameter combination suggestions can be automatically given.

Referring again to fig. 1, in step S13, the above-mentioned delivery parameter prediction data and environment data may be input to an environment learning model to obtain delivery target prediction data.

In an embodiment of the present disclosure, the environmental learning model may be trained by the method shown in fig. 4. Fig. 4 is a flow diagram of a method of training an environmental learning model according to an embodiment of the present disclosure.

In step S41, environmental sample data of the sample advertisement may be acquired. The environment sample data may include sample data related to at least one of a sample advertisement dissemination environment, a sample advertisement speaker, and an auction advertisement. For example, environmental sample data may include sample data related to external influencing factors of sample advertising impressions (e.g., auction advertising impressions, large disk contingencies of advertising platforms, off-site marketing campaigns, hot news, star effects of advertising speakers, hot search keywords, etc.).

In step S42, the environmental sample data and the putting parameter prediction training data obtained by using the trained putting parameter prediction model with the putting target sample data as input may be input to the environmental learning model to obtain the putting target prediction training data. In embodiments of the present disclosure, the placement parameter prediction training data may include placement parameter prediction training values and corresponding weights. For example, a ranking algorithm may be used to screen the delivery parameter prediction training data with relatively high weight from the delivery parameter prediction training data, and the screened delivery parameter prediction training data may be input to the environmental learning model. Optionally, a ranking algorithm and a preset prior rule may be used to obtain a combination of the launching parameter prediction training data with weights ranked from large to small from the launching parameter prediction training data, and then the combination of the launching parameter prediction training data with relatively high weights is preferentially input to the environment learning model. Alternatively, the environmental learning model may be input with environmental sample data, and a delivery parameter prediction training value corresponding to a weight greater than or equal to a predetermined weight threshold.

In step S43, a loss function of the environmental learning model may be determined according to the delivery target prediction training data and the delivery target sample data. In step S44, parameters of the environmental learning model may be adjusted according to a loss function of the environmental learning model. By introducing the environmental data in the training process of the environmental learning model, the influence of environmental factors in the advertisement putting process on the advertisement putting income and effect can be simulated.

According to the embodiment of the disclosure, the environment learning model can be trained by using machine learning algorithms such as a reinforcement learning algorithm, a GBDT (Gradient Boosting Decision Tree) algorithm, an XGBOOST (Extreme Gradient Boosting algorithm) algorithm and the like.

Since advertisement placement is heavily influenced by external environmental factors (e.g., auction placement factors, promotional campaigns, off-site campaigns, etc.), these external environmental factors often cannot be recorded and perceived. Environmental data serving as hidden variables can be introduced and utilized through the environmental learning model, environmental factors are considered in model training, and fitting with a real advertisement putting environment is greatly improved.

The environment learning model obtained by training according to the method does not need to depend on a large amount of feedback data with fine granularity excessively, and is favorable for well fitting the real world. Under the condition that strong regularity and mechanistic hypothesis exist in a prediction target, an environment learning model obtained by training according to the method can be combined with a priori rule and a machine learning algorithm to construct a virtual environment which is very similar to a real environment, and the model is continuously iterated in the virtual environment through numerous delivery experiments to promote the environment learning model and other related models to achieve the optimal effect.

According to the embodiment of the disclosure, the delivery parameter prediction data and the environment data may be input to a trained environment learning model to obtain delivery target prediction data. Alternatively, environmental data, projected parameter predictions corresponding to weights greater than or equal to a predetermined weight threshold may be input to the environmental learning model. The environment learning model obtained by training in the method shown in fig. 4 can improve the simulation accuracy of the advertisement delivery environment.

Referring again to fig. 1, in step S14, the delivery target prediction data is input to a delivery parameter optimization model to obtain delivery parameter optimization data. For example, placement parameter optimization data may include data related to optimized advertisement placement parameters.

In the embodiment of the present disclosure, the release parameter optimization model may be trained by the method shown in fig. 5. Fig. 5 is a flow chart of a training method of a launch parameter optimization model according to an embodiment of the present disclosure.

As shown in fig. 5, in step S51, the putting target prediction training data obtained by using the trained environment learning model with the environment sample data and the putting parameter prediction training data as inputs may be input to the putting parameter optimization model to obtain the putting parameter optimization training data. For example, the placement parameter optimization training data may include optimized data related to placement operations of sample advertisements.

In step S52, a loss function of the release parameter optimization model is determined according to the release parameter optimization training data and the release parameter sample data. In step S53, parameters of the release parameter optimization model are adjusted according to the loss function of the release parameter optimization model. According to the embodiment of the disclosure, the release parameter optimization model can be trained by using machine learning algorithms such as a reinforcement learning algorithm, a GBDT (Gradient Boosting Decision Tree) algorithm, an XGBOOST (Extreme Gradient Boosting algorithm) and the like.

The delivery parameter prediction model, the environment learning model and the delivery parameter optimization model obtained by training through the training method can construct a highly simulated virtual environment by means of the environment learning model and carry out a large number of iterative training fitting, so that an infinite delivery mode can be tried in the virtual environment to optimize the delivery parameter combination for actual advertisement delivery.

Referring again to fig. 1, in step S15, advertisements are delivered according to the delivery parameter optimization data. According to the delivery parameter prediction model, the environment learning model and the delivery parameter optimization model, delivery parameter optimization data can be obtained to guide actual advertisement delivery.

A placement parameter prediction model, an environmental learning model, and a placement parameter optimization model according to embodiments of the present disclosure may be updated by iterative training using acquired placement objective data, environmental data, and placement parameter optimization data. In the iterative training process, the delivery target data, the environmental data and the delivery parameter optimization data acquired in the advertisement delivery process can be respectively used as delivery target sample data, environmental sample data and delivery parameter sample data in the corresponding model training process, so as to carry out iterative training on the delivery parameter prediction model, the environmental learning model and the delivery parameter optimization model. Therefore, the delivery target data, the environmental data and the delivery parameter optimization data in each advertisement delivery process can be fed back and fall back to the iterative training of the delivery parameter prediction model, the environmental learning model and the delivery parameter optimization model, so that the delivery parameter prediction model, the environmental learning model and the delivery parameter optimization model can be continuously updated, the advertisement user image is updated, and preparation is made for long-term user asset deposition.

The advertisement delivery method according to an embodiment of the present disclosure is described above with reference to fig. 1 to 5, but the present disclosure is not limited thereto. An advertising system and an advertising model training system according to an embodiment of the present disclosure are described below in conjunction with fig. 6.

Fig. 6 is a block diagram of an advertisement delivery system 60 according to an embodiment of the present disclosure. Fig. 7 is a block diagram of an advertising model training system 70 according to an embodiment of the present disclosure.

As shown in fig. 6, the advertisement delivery system 60 may include a data acquisition device 601 configured to acquire delivery target data and environment data of an advertisement. The data acquisition device 601 may acquire the delivery target data and the environment data of the advertisement by the data acquisition method described above.

The advertisement delivery system 60 may include a delivery parameter prediction device 602 configured to input delivery targeting data to a delivery parameter prediction model to obtain delivery parameter prediction data.

The advertisement delivery system 60 may comprise a delivery target prediction means 603 configured to input said delivery parameter prediction data and said environment data to an environment learning model to obtain delivery target prediction data.

The advertisement delivery system 60 may comprise a delivery parameter optimization device 604 configured to input the delivery target prediction data to a delivery parameter optimization model to obtain delivery parameter optimization data.

Advertisement delivery system 60 may comprise an advertisement delivery device 605 configured to deliver the advertisement according to the delivery parameter optimization data.

How to obtain the delivery parameter prediction model, the environmental learning model and the delivery parameter optimization model through the training of the advertisement delivery model training system 70 is described below with reference to fig. 7.

As shown in fig. 7, according to the embodiment of the present disclosure, the placement parameter prediction model is trained by the first training device 701 in the advertisement placement model training system 70. The first training device 701 includes: a first sample data acquiring device 701A configured to acquire sample data of a delivery parameter and sample data of a delivery target of a sample advertisement; a first function determining device 701B configured to determine a reward function of the release parameter prediction model by using a reverse reinforcement learning algorithm by taking the release parameter sample data and the release target sample data as input and output of a release parameter prediction model, respectively; a first parameter adjusting device 701C configured to adjust parameters of the delivery parameter prediction model according to the reward function.

According to the embodiment of the present disclosure, the environmental learning model is trained by the second training device 702 in the advertisement delivery model training system 70. The second training device 702 comprises: a second sample data acquiring means 702A configured to acquire environmental sample data of the sample advertisement; a putting target prediction training data acquisition means 702B configured to input the environment sample data and putting parameter prediction training data acquired by using the trained putting parameter prediction model with the putting target sample data as input to an environment learning model to acquire putting target prediction training data; a second function determining device 702C configured to determine a loss function of the environmental learning model according to the delivery target prediction training data and the delivery target sample data; a second parameter adjusting device 702D configured to adjust the parameters of the environment learning model according to the loss function of the environment learning model.

According to an embodiment of the present disclosure, the delivery target prediction training data obtaining means 702B may be configured to: and inputting the environmental sample data and the putting parameter prediction training value corresponding to the weight greater than or equal to the preset weight threshold value into the environmental learning model to obtain putting target prediction training data. Optionally, the placement target predicting device 603 may be configured to: and inputting the environmental data and the launching parameter predicted value corresponding to the weight larger than or equal to the preset weight threshold value into the trained environmental learning model to obtain launching target predicted data.

According to the embodiment of the present disclosure, the release parameter optimization model is obtained by training through the third training device 703. The third training device 703 includes: a launch parameter optimization training data acquisition means 703A configured to input launch target prediction training data acquired by using the trained environment learning model with the environment sample data and the launch parameter prediction training data as inputs to a launch parameter optimization model to acquire launch parameter optimization training data; a third function determining device 703B configured to determine a loss function of the release parameter optimization model according to the release parameter optimization training data and the release parameter sample data; a third parameter adjusting device 703C configured to adjust parameters of the release parameter optimization model according to a loss function of the release parameter optimization model.

According to an embodiment of the present disclosure, the delivery parameter prediction data may include a delivery parameter prediction value and a corresponding weight, and the delivery parameter optimization data includes a delivery parameter optimization value and a corresponding weight. The placement parameter prediction training data may include placement parameter prediction training values and corresponding weights.

According to embodiments of the present disclosure, placement targeting data may include data related to advertising revenue targeting. The environmental data may include data relating to at least one of an advertisement dissemination environment, an advertisement speaker, and an auction advertisement. The placement parameter prediction data and the placement parameter optimization data may each include data related to ad placement parameters.

According to an embodiment of the present disclosure, the delivery parameter prediction model, the environmental learning model, and the delivery parameter optimization model may be updated by iterative training using the obtained delivery objective data, the environmental data, and the delivery parameter optimization data. In the iterative training process, the delivery target data, the environmental data and the delivery parameter optimization data acquired in the advertisement delivery process can be respectively used as delivery target sample data, environmental sample data and delivery parameter sample data in the corresponding model training process, so as to carry out iterative training on the delivery parameter prediction model, the environmental learning model and the delivery parameter optimization model.

The functions and operations of the components of the advertisement delivery system 60 and the advertisement delivery model training system 70 may be understood with reference to the advertisement delivery method and the model training method described in conjunction with fig. 1 to 5, and will not be described herein again for brevity.

According to the advertisement putting method, the system and the related device thereof, the traditional advertisement putting variable and the environmental variable can be simultaneously considered in the modeling process related to advertisement putting and the application of the advertisement putting model, multiple factors related to the advertisement are fully considered, the model related to advertisement putting is continuously iteratively optimized through historical data training or putting cold start, the optimal combination of one or more advertisement putting parameters is obtained, and the putting personnel is guided to carry out advertisement putting according to the optimal putting parameter combination, so that the advertisement conversion rate is improved, and the advertiser is helped to save budget. In addition, the advertisement delivery technical scheme provided by the invention can track relevant data of advertisement delivery for a long time so as to carry out continuous model updating, remarkably improve the income effect obtained after the prediction scheme is applied to an actual scene, not only focus on the short-term income of the advertisement, but also increase the full consideration of long-term income such as long-term user retention, repurchase and the like.

Further, the advertisement delivery method according to the present disclosure may also be implemented in a computing device. Fig. 8 is a block diagram of computing device 8, according to an embodiment of the present disclosure.

Referring to fig. 8, a computing device 8 according to an embodiment of the present disclosure may include a memory 81 and a processor 82, a computer program 83 stored on the memory 81, when the computer program 83 is executed by the processor 82, an advertisement delivery method according to an embodiment of the present disclosure may be implemented.

In an embodiment of the present disclosure, when the computer program 83 is executed by the processor 82, the operations of the advertisement delivery method described with reference to fig. 1 to 5 may be implemented: acquiring delivery target data and environment data of the advertisement; inputting the delivery target data into a delivery parameter prediction model to obtain delivery parameter prediction data; inputting the release parameter prediction data and the environment data into an environment learning model to obtain release target prediction data; inputting the delivery target prediction data into a delivery parameter optimization model to obtain delivery parameter optimization data; and releasing the advertisement according to release parameter optimization data. The functions and operations of the various components in computing device 8 may be understood with reference to the advertisement delivery methods and model training methods described in conjunction with fig. 1-5, which are not repeated herein for brevity.

The computing device illustrated in fig. 8 is only one example and should not impose any limitations on the functionality or scope of use of embodiments of the disclosure.

Exemplary embodiments of the present disclosure may also be implemented as a computing device that may include a storage component having stored therein a set of computer-executable instructions that, when executed by a processor, perform a method of advertising according to exemplary embodiments of the present disclosure.

In particular, computing devices may be deployed in servers or clients, as well as on node devices in a distributed network environment. Further, the computing device may be a PC computer, tablet device, personal digital assistant, smart phone, web application, or other device capable of executing the set of instructions.

The computing device need not be a single computing device, but can be any device or collection of circuits capable of executing the instructions (or sets of instructions) described above, individually or in combination. The computing device may also be part of an integrated control system or system manager, or may be configured as a portable electronic device that interfaces with local or remote (e.g., via wireless transmission).

In a computing device, a processor may include a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), a programmable logic device, a special purpose processor system, a microcontroller, or a microprocessor. By way of example, and not limitation, processors may also include analog processors, digital processors, microprocessors, multi-core processors, processor arrays, network processors, and the like.

Some of the operations described in the advertisement delivery method according to the exemplary embodiments of the present disclosure may be implemented by software, some of the operations may be implemented by hardware, and furthermore, the operations may be implemented by a combination of hardware and software.

The processor may execute instructions or code stored in one of the storage components, which may also store data. The instructions and data may also be transmitted or received over a network via a network interface device, which may employ any known transmission protocol.

The memory component may be integral to the processor, e.g., having RAM or flash memory disposed within an integrated circuit microprocessor or the like. Further, the storage component may comprise a stand-alone device, such as an external disk drive, storage array, or any other storage device usable by a database system. The storage component and the processor may be operatively coupled or may communicate with each other, such as through an I/O port, a network connection, etc., such that the processor can read files stored in the storage component.

In addition, the computing device may also include a video display (such as a liquid crystal display) and a user interaction interface (such as a keyboard, mouse, touch input device, etc.). All components of the computing device may be connected to each other via a bus and/or a network.

The advertisement delivery method according to the exemplary embodiment of the present disclosure may be described as various interconnected or coupled functional blocks or functional diagrams. However, these functional blocks or functional diagrams may be equally integrated into a single logic device or operated on by non-exact boundaries.

The various elements or devices in the advertising system 60 shown in fig. 6 and the advertising model training system 70 shown in fig. 7 may be configured as software, hardware, firmware, or any combination thereof that perform particular functions. For example, each unit or device may correspond to an application-specific integrated circuit, to pure software code, or to a module combining software and hardware. Furthermore, one or more functions implemented by each unit or device may also be uniformly executed by components in a physical entity device (e.g., a processor, a client, a server, or the like).

Further, the advertisement delivery method described with reference to fig. 1 to 5 may be implemented by a program (or instructions) recorded on a computer-readable storage medium. For example, according to an exemplary embodiment of the present disclosure, a computer-readable storage medium storing instructions may be provided, wherein the instructions, when executed by at least one computing device, cause the at least one computing device to perform an advertising method according to the present disclosure.

The computer program in the computer-readable storage medium may be executed in an environment deployed in a computer device such as a client, a host, a proxy device, a server, and the like, and it should be noted that the computer program may also be used to perform additional steps other than the above steps or perform more specific processing when the above steps are performed, and the content of the additional steps and the further processing is already mentioned in the description of the related method with reference to fig. 1 to 5, and therefore will not be described again in order to avoid repetition.

It should be noted that each unit or device in the advertisement delivery system 60 shown in fig. 6 and the advertisement delivery model training system 70 shown in fig. 7 may completely depend on the execution of the computer program to realize the corresponding functions, that is, the functional architecture of the computer program in each unit or device corresponds to each step, so that the whole system is called by a special software package (e.g., lib library) to realize the corresponding functions.

On the other hand, each unit or device in fig. 6 and 7 may also be implemented by hardware, software, firmware, middleware, microcode, or any combination thereof. When implemented in software, firmware, middleware or microcode, the program code or code segments to perform the corresponding operations may be stored in a computer-readable medium such as a storage medium, so that a processor may perform the corresponding operations by reading and executing the corresponding program code or code segments.

Accordingly, an advertisement delivery method according to an exemplary embodiment of the present disclosure may be implemented by a system including at least one computing device and at least one storage device storing instructions.

According to an exemplary embodiment of the present disclosure, at least one computing device is a computing device for performing an advertisement delivery method according to an exemplary embodiment of the present disclosure, the storage device having stored therein a set of computer-executable instructions that, when executed by the at least one computing device, perform an advertisement delivery method according to the present disclosure.

The control logic or functions performed by various components or controllers in a system, device, etc. may be represented by flowcharts or the like in one or more of the figures. These figures provide representative control strategies and/or logic that may be implemented using one or more processing strategies (e.g., event-driven, interrupt-driven, multi-tasking, multi-threading, and so forth). As such, various steps or functions illustrated may be performed in the sequence illustrated, in parallel, or in some cases omitted. Although not always explicitly illustrated, one of ordinary skill in the art will recognize that one or more of the illustrated steps or functions may be repeatedly performed depending on the particular processing strategy being used.

While various exemplary embodiments of the present disclosure have been described above, it should be understood that the above description is exemplary only, and not exhaustive, and that the present disclosure is not limited to the disclosed exemplary embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the disclosure. Therefore, the protection scope of the present disclosure should be subject to the scope of the claims.

Claims

1. An advertisement delivery method, comprising:

acquiring delivery target data and environment data of the advertisement;

inputting the putting target data into a putting parameter prediction model to obtain putting parameter prediction data;

inputting the release parameter prediction data and the environment data into an environment learning model to obtain release target prediction data;

inputting the launching target prediction data into a launching parameter optimization model to obtain launching parameter optimization data;

and delivering the advertisement according to the delivery parameter optimization data.

2. The advertisement delivery method according to claim 1, wherein the delivery parameter prediction model is obtained by training:

acquiring release parameter sample data and release target sample data of a sample advertisement;

determining a return function of the release parameter prediction model by using a reverse reinforcement learning algorithm by respectively taking the release parameter sample data and the release target sample data as input and output of a release parameter prediction model;

and adjusting parameters of the input parameter prediction model according to the return function.

3. The advertisement delivery method according to claim 2, wherein the environment learning model is trained by the following steps:

acquiring environmental sample data of a sample advertisement;

inputting the environment sample data and putting parameter prediction training data acquired by using the trained putting parameter prediction model with the putting target sample data as input into an environment learning model to acquire putting target prediction training data;

determining a loss function of an environmental learning model according to the putting target prediction training data and the putting target sample data;

and adjusting parameters of the environment learning model according to the loss function of the environment learning model.

4. The advertisement delivery method according to claim 3, wherein the delivery parameter optimization model is trained by the following steps:

inputting the putting target prediction training data acquired by using the trained environment learning model by taking the environment sample data and the putting parameter prediction training data as input into a putting parameter optimization model to acquire putting parameter optimization training data;

determining a loss function of a release parameter optimization model according to the release parameter optimization training data and the release parameter sample data;

and adjusting parameters of the release parameter optimization model according to the loss function of the release parameter optimization model.

5. The advertisement delivery method according to claim 3 or 4,

the delivery parameter prediction data comprises delivery parameter prediction values and corresponding weights, the delivery parameter optimization data comprises delivery parameter optimization values and corresponding weights, and/or

The putting parameter prediction training data comprises putting parameter prediction training values and corresponding weights.

6. The method of claim 5, wherein inputting the placement parameter prediction data and the environmental data to an environmental learning model comprises: inputting the environmental data and a putting parameter predicted value corresponding to the weight larger than or equal to a preset weight threshold value into an environmental learning model; and/or

Inputting the environment sample data and the putting parameter prediction training data acquired by using the trained putting parameter prediction model and taking the putting target sample data as input into an environment learning model, wherein the method comprises the following steps: and inputting the environment sample data and a putting parameter prediction training value corresponding to the weight larger than or equal to a preset weight threshold value into an environment learning model.

7. The advertisement delivery method according to any one of claims 1 to 4,

the placement targeting data includes data relating to advertising revenue targeting,

the environmental data includes data relating to at least one of an advertisement dissemination environment, an advertisement speaker and an auction advertisement,

the delivery parameter prediction data and the delivery parameter optimization data each include data related to advertisement delivery parameters.

8. An advertisement delivery system, comprising:

a data acquisition device configured to acquire delivery target data and environment data of an advertisement;

a delivery parameter prediction device configured to input the delivery target data to a delivery parameter prediction model to obtain delivery parameter prediction data;

a delivery target prediction device configured to input the delivery parameter prediction data and the environment data to an environment learning model to obtain delivery target prediction data;

a delivery parameter optimization device configured to input the delivery target prediction data to a delivery parameter optimization model to obtain delivery parameter optimization data;

an advertisement delivery device configured to deliver the advertisement according to the delivery parameter optimization data.

9. A computer-readable storage medium storing instructions that, when executed by at least one computing device, cause the at least one computing device to perform the method of advertisement delivery of any of claims 1-8.

10. A system comprising at least one computing device and at least one storage device storing instructions that, when executed by the at least one computing device, cause the at least one computing device to perform the advertisement delivery method of any of claims 1-8.