CN117114766A

CN117114766A - Cost control factor determining method, device, equipment and storage medium

Info

Publication number: CN117114766A
Application number: CN202210511707.3A
Authority: CN
Inventors: 李少波
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2022-05-11
Filing date: 2022-05-11
Publication date: 2023-11-24

Abstract

The embodiment of the application provides a method, a device, equipment and a storage medium for determining a cost control factor, which are at least applied to the field of artificial intelligence, wherein the method comprises the following steps: acquiring advertisement data of advertisements to be controlled; inquiring to obtain statistical data characteristics corresponding to the advertisement data from a pre-constructed data inquiry table; wherein the statistical data features are data features of advertisement information which are counted in advance for different advertisements; carrying out data coding processing on the statistical data characteristics to obtain aggregation data and a smoothing coefficient corresponding to the advertisement to be controlled; wherein the aggregate data is data for characterizing attribute information of the advertisement to be controlled; and carrying out smoothing treatment on the aggregated data based on the smoothing coefficient to obtain the cost control factor of the advertisement to be controlled. The application can accurately determine the cost control factor of the advertisement to be controlled.

Description

Cost control factor determining method, device, equipment and storage medium

Technical Field

The embodiment of the application relates to the technical field of Internet, and relates to a cost control factor determining method, device, equipment and storage medium.

Background

In the advertising field, when controlling and adjusting industry independent costs, it is necessary to determine a cost control factor for each advertisement first, so that advertisement costs are accounted and controlled based on the cost control factors. Currently, when determining cost control factors for advertisements, the advertisement initial stage and the advertisement maturation stage are mainly included. In the initial phase of advertising, cost control factors are typically calculated using recent data of similar advertisements in combination with global data; in the advertisement maturation stage, the calculation of the cost control factor is typically done using the data of the advertisement itself in combination with the data of similar advertisements. However, in the related art, similar advertisements of the same type cannot be intelligently identified, and in the calculation of cost control factors of different stages, smoothing coefficients are required, but in the related art, smoothing coefficients applicable to different stages cannot be accurately determined, so that accurate cost control factors cannot be determined in the related art.

Disclosure of Invention

The embodiment of the application provides a cost control factor determining method, device, equipment and storage medium, which are at least applied to the field of artificial intelligence, can accurately identify similar advertisements, and determine smoothing coefficients suitable for each stage, so as to accurately determine the cost control factor.

The technical scheme of the embodiment of the application is realized as follows:

the embodiment of the application provides a cost control factor determining method, which comprises the following steps:

acquiring advertisement data of advertisements to be controlled;

inquiring to obtain statistical data characteristics corresponding to the advertisement data from a pre-constructed data inquiry table; wherein the statistical data features are data features of advertisement information which are counted in advance for different advertisements;

carrying out data coding processing on the statistical data characteristics to obtain aggregation data and a smoothing coefficient corresponding to the advertisement to be controlled; wherein the aggregate data is data for characterizing attribute information of the advertisement to be controlled;

and carrying out smoothing treatment on the aggregated data based on the smoothing coefficient to obtain the cost control factor of the advertisement to be controlled.

The embodiment of the application provides a cost control factor determining device, which comprises:

the acquisition module is used for acquiring advertisement data of the advertisement to be controlled; the query module is used for querying and obtaining statistical data characteristics corresponding to the advertisement data from a pre-constructed data query table; wherein the statistical data features are data features of advertisement information which are counted in advance for different advertisements; the data coding module is used for carrying out data coding processing on the statistical data characteristics to obtain aggregate data and a smooth coefficient corresponding to the advertisement to be controlled; wherein the aggregate data is data for characterizing attribute information of the advertisement to be controlled; and the smoothing processing module is used for carrying out smoothing processing on the aggregated data based on the smoothing coefficient to obtain the cost control factor of the advertisement to be controlled.

The embodiment of the application provides cost control factor determining equipment, which comprises the following components:

a memory for storing executable instructions; and the processor is used for realizing the cost control factor determining method when executing the executable instructions stored in the memory.

Embodiments of the present application provide a computer program product or computer program comprising executable instructions stored in a computer readable storage medium; the processor of the cost control factor determining device reads the executable instructions from the computer readable storage medium and executes the executable instructions to implement the cost control factor determining method.

The embodiment of the application provides a computer readable storage medium, which stores executable instructions for causing a processor to execute the executable instructions to implement the cost control factor determination method.

The embodiment of the application has the following beneficial effects: and inquiring to obtain statistical data characteristics corresponding to advertisement data of the advertisements to be controlled from a pre-constructed data inquiry table, and carrying out data coding processing on the statistical data characteristics to obtain aggregate data and a smooth coefficient corresponding to the advertisements to be controlled, thereby determining a cost control factor of the advertisements to be controlled based on the aggregate data and the smooth coefficient. Thus, the query is performed based on the pre-constructed data query table, so that the corresponding statistical data can be queried based on any specific attribute parameter of the advertisement to be controlled, and the accurate judgment of the similar advertisement is realized based on the statistical data characteristics corresponding to the statistical data; and the accurate smooth coefficient can be dynamically fitted through data coding processing, so that the cost control factor can be accurately determined.

Drawings

FIG. 1 is a schematic illustration of an advertising process provided by an embodiment of the present application;

FIG. 2 is a schematic diagram of an alternative architecture of a cost control factor determination system provided by an embodiment of the present application;

FIG. 3 is a schematic diagram of a cost control factor determination device according to an embodiment of the present application;

FIG. 4 is a schematic flow chart of an alternative method for determining a cost control factor according to an embodiment of the present application;

FIG. 5 is a schematic flow chart of an alternative advertisement cost control method in an initial stage of advertisement according to an embodiment of the present application;

FIG. 6 is a schematic diagram of an implementation flow of an advertisement initial stage model training method according to an embodiment of the present application;

FIG. 7 is a schematic flow chart of an alternative advertisement cost control method at an advertisement maturation stage according to an embodiment of the present application;

FIG. 8 is a schematic diagram of an implementation flow of an advertisement maturation stage model training method according to an embodiment of the present application;

FIG. 9 is an advertisement bias variation curve of advertisements for commercial A of the brand e-commerce provided by an embodiment of the present application over a day;

FIG. 10 is an advertisement bias variation curve of an advertisement of a commercial B of a brand e-commerce vendor over a day provided by an embodiment of the present application;

FIG. 11 is an advertisement bias variation curve of advertisements of commodity C of a direct nutrient electronic commerce provided by the embodiment of the application in one day;

FIG. 12 is an advertisement bias variation curve of advertisements of commodity D of a direct nutrient electronic commerce over a day provided by an embodiment of the present application;

FIG. 13 is a schematic diagram of an advertisement initial stage cost control factor calculation process according to an embodiment of the present application;

FIG. 14 is a schematic diagram of an advertisement maturation stage cost control factor calculation process according to an embodiment of the present application;

FIG. 15 is an aggregate statistics of similar advertisement data for advertisement maturity stage provided by an embodiment of the present application;

FIG. 16 is a schematic diagram showing the ratio of the estimated conversion count sum to the actual conversion count sum for different commodities according to an embodiment of the present application;

FIG. 17 is a schematic diagram of classifying properties of goods based on different categories according to an embodiment of the present application;

FIG. 18 is a schematic diagram of an advertisement initial stage model according to an embodiment of the present application;

FIG. 19 is a schematic diagram of an advertisement maturity model provided by an embodiment of the present application.

Detailed Description

The present application will be further described in detail with reference to the accompanying drawings, for the purpose of making the objects, technical solutions and advantages of the present application more apparent, and the described embodiments should not be construed as limiting the present application, and all other embodiments obtained by those skilled in the art without making any inventive effort are within the scope of the present application.

In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is to be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with one another without conflict. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which embodiments of this application belong. The terminology used in the embodiments of the application is for the purpose of describing embodiments of the application only and is not intended to be limiting of the application.

Before explaining the cost control factor determination method according to the embodiment of the present application, the technical terms related to the embodiment of the present application are explained first.

(1) OCPA advertisement: is a special advertisement bidding mode. FIG. 1 is a schematic diagram of an advertisement process according to an embodiment of the present application, and as shown in FIG. 1, the entire advertisement process is summarized into four stages: exposure, clicking, conversion, payment. The charging point is what way the platform side charges, if the charging point is exposed, the platform side charges the advertiser with the exposure of the advertisement; if the charging point is clicked, the platform side charges the advertiser for the click quantity of the advertisement; and so on. The bid point is in what way the advertiser is bidding, if the bid point is exposing, the advertiser is priced with the exposure of the advertisement; if the bid point is on a click, the advertiser bids with the click-through amount of the advertisement. If both the billing point and the bid point are focused on exposure, then such an advertisement is a thousand browsing Cost (CPM) advertisement; if both the billing point and the bid point are focused on clicks, such an advertisement is an average Click-through (CPC) advertisement; if both the billing point and the bid point are focused on conversion, then such an advertisement is a Cost Per Action (CPA) advertisement. CPA advertising is naturally good, but many platforms cannot take this form. For example, when the conversion data of an advertisement platform depends on the feedback of the advertiser, if the CPA mode is adopted, the advertiser has a very powerful power to not transmit or to transmit less conversion numbers. In this case, it is necessary to optimize the CPA (OCPA, optimized CPA) in a more specific manner. OCPA is a special mode, charging points are separated from bidding points, charging points are clicked, bidding points are converted, OCPA is a bidding mode of bidding advertisements, and the optimization targets of the advertisements are directly bid. The method has the advantages that the data advantage is utilized to help advertisers to estimate the advertisement conversion rate (pCVR) in conversion bidding, and charging of a platform is guaranteed.

(2) pCVR: the estimated conversion rate of an advertisement generally refers to the click conversion rate of the advertisement, that is, the probability of obtaining conversion after the advertisement is clicked.

(3) pCVR calibration: this is a post-adjustment to the output of the pCVR predictive model. Targeted adjustments may be made in a particular industry, for example, pCVR calibration for a camping elet.

(4) Direct nutrient electricity quotient: the direct-camping electronic commerce is also called as a second-class electronic commerce, and the shopping trip of the user of the direct-camping electronic commerce is approximately as follows: reading certain media platforms, finding information flow advertisements, clicking on advertisements, reading landing pages, filling in a receiving address order/discarding purchase jumps to step one, continuing to read certain media platforms, waiting for quick delivery to check out payment. The whole process is relatively simple, and the commodity is actively pushed to the user of the media platform by the information flow advertisement instead of being actively sent to the electronic commerce platform by the user to search for the commodity.

(5) Industry factor: in the advertisement algorithm system, the industry factor plays a role of readjusting estimated thousand times of display profits (eCPM, effective Cost Per Mille) of advertisements of a specific industry. The readjustment strategy can be to adjust the pCVR predictive value in a specific industry based on a certain industry; or the specific crowd effect enhancement can be performed according to specific crowd in a certain industry. On the direct-camping E-commerce, the industry factors mainly comprise direct-camping E-commerce pCVR compensation factors and direct-camping E-commerce high-conversion crowd enhancement factors.

(6) Advertisement initial stage: also referred to as an insufficient advertisement conversion phase, in the initial advertisement phase, the absolute value of the difference between the advertisement bias of the advertisement and the global bias of the advertisement throughout the day is greater than or equal to a preset threshold. The initial stage of the advertisement in the embodiment of the application is the underexposure stage of the advertisement when the advertisement just begins to be exposed in one day. At this time, the conversion number of the advertisement < =2 or the consumption of the advertisement < =2, the target CPA value (target_cpa).

(7) Advertisement maturation stage: also referred to as the full advertisement conversion phase, the absolute value of the difference between the advertisement bias of the advertisement and the global bias of the advertisement throughout the day is less than a preset threshold during the advertisement maturation phase. The advertisement spans the advertisement initial stage and reaches the advertisement mature stage. The advertisement maturation stage mentioned in the embodiment of the present application refers to the conversion number of advertisement >2 and the consumption of advertisement >2 target_cpa.

(8) Ad bias (bias): also referred to as advertisement conversion bias, is the ratio between the actual conversion number of an advertisement and the estimated conversion number. The whole day deviation of the advertisement is the advertisement deviation corresponding to the data of the whole day.

(9) Same category advertisement: also referred to as similar advertisements, the same category of advertisements in embodiments of the present application refers to advertisements having the same advertiser or the same brand or the same merchandise information.

(10) Parameter optimization modeling: the parameter optimization modeling in the embodiment of the application aims at important parameters related to the pCVR calibration strategy, a parameter optimization model is established, and from data, the parameter optimization model is learned to obtain real-time proper parameters so as to achieve a better calibration effect.

In the cost control factor determining method provided by the embodiment of the application, firstly, advertisement data of advertisements to be controlled are obtained; then, inquiring to obtain statistical data characteristics corresponding to the advertisement data from a pre-constructed data inquiry table; wherein the statistical data features are the data features of advertisement information which are counted in advance for different advertisements; carrying out data coding processing on the statistical data characteristics to obtain aggregation data and smoothing coefficients corresponding to the advertisements to be controlled; wherein the aggregate data is data for characterizing attribute information of the advertisement to be controlled; and then, carrying out smoothing processing on the aggregated data based on the smoothing coefficient to obtain a cost control factor of the advertisement to be controlled. Thus, the query is performed based on the pre-constructed data query table, so that the corresponding statistical data can be queried based on any specific attribute parameter of the advertisement to be controlled, and the accurate judgment of the similar advertisement is realized based on the statistical data characteristics corresponding to the statistical data; and the accurate smooth coefficient can be dynamically fitted through data coding processing, so that the cost control factor can be accurately determined. In the embodiment of the application, the cost control factor of the advertisement to be controlled can be accurately determined, so that the advertisement cost can be reasonably controlled based on the cost control factor.

Aiming at the independent cost control problem of the electronic commerce industry and some existing cost control strategies, the embodiment of the application provides a commodity aggregation attribute optimization strategy and a smooth coefficient automatic calculation strategy by establishing a model, thereby improving the scientificity of decisions. From the data, the effect of cost control is enhanced.

An exemplary application of the cost control factor determining apparatus according to the embodiment of the present application is described below, and the cost control factor determining apparatus provided by the embodiment of the present application may be implemented as a terminal or as a server. In one implementation manner, the cost control factor determining device provided by the embodiment of the application can be implemented as a notebook computer, a tablet computer, a desktop computer, a mobile device (for example, a mobile phone, a portable music player, a personal digital assistant, a special message device, a portable game device), an intelligent robot, an intelligent household appliance, an intelligent vehicle-mounted device and any other terminal with functions of advertisement issuing and popularization, advertisement display, advertisement clicking, consumption and the like; in another implementation manner, the cost control factor determining device provided by the embodiment of the present application may be implemented as a server, where the server may be an independent physical server, or may be a server cluster or a distributed system formed by a plurality of physical servers, or may be a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN, content De livery Network), and basic cloud computing services such as big data and artificial intelligence platforms. The terminal and the server may be directly or indirectly connected through wired or wireless communication, which is not limited in the embodiment of the present application. An exemplary application when the cost control factor determination device is implemented as a server will be described below.

Referring to fig. 2, fig. 2 is an optional architecture schematic diagram of a cost control factor determining system provided by the embodiment of the present application, to realize supporting any one advertisement popularization and application, and to realize advertisement cost control of an advertisement to be controlled by acquiring information such as click operation and conversion data for the advertisement to be controlled on the advertisement popularization and application. The cost control factor determining system 10 at least comprises a terminal 100, a network 200 and a server 300, wherein the server 300 is a server of an advertiser, a server of the platform side or a third party server independent of the advertiser and the platform side, in order to ensure the benefits of the advertiser corresponding to the advertisement to be controlled and the platform side for realizing advertisement promotion application for displaying the advertisement. The server 300 may constitute a cost control factor determination device of an embodiment of the present application. The terminal 100 is connected to the server 300 through the network 200, and the network 200 may be a wide area network or a local area network, or a combination of both. When determining cost control factors of the advertisements to be controlled or when controlling the advertisement cost of the advertisements to be controlled, the terminal 100 acquires advertisement data of the advertisements to be controlled through advertisement popularization and application, and sends the advertisement data to the server 300 through the network 200, and the server 300 inquires from a pre-constructed data inquiry table to obtain statistical data characteristics corresponding to the advertisement data; wherein the statistical data features are the data features of advertisement information which are counted in advance for different advertisements; then, carrying out data coding processing on the statistical data characteristics to obtain aggregation data and smoothing coefficients corresponding to the advertisements to be controlled; and performs smoothing processing on the aggregated data based on the smoothing coefficient to obtain a cost control factor of the advertisement to be controlled, and after obtaining the cost control factor, the server 300 transmits the cost control factor to the terminal.

In some embodiments, after obtaining the cost control factor, the advertisement cost control may be further performed on the advertisement to be controlled based on the cost control factor, so as to obtain an advertisement cost control policy. After obtaining the advertisement cost control policy for the advertisement to be controlled, the server 300 transmits the advertisement cost control policy to the terminal to remind the advertiser and the platform side of controlling the advertisement cost with the advertisement cost control policy.

In some embodiments, the determination of the cost control factor of the advertisement to be controlled may also be implemented by the terminal 100, that is, the cost control factor determination method of the embodiment of the present application is implemented by the terminal as an execution body, to determine the cost control factor of the advertisement to be controlled.

The method for determining the cost control factor according to the embodiment of the present application may be implemented by a cloud technology based on a cloud platform, for example, the server 300 may be a cloud server. And inquiring to obtain statistical data characteristics corresponding to the advertisement data from a pre-constructed data inquiry table through the cloud server, or carrying out data coding processing on the statistical data characteristics through the cloud server to obtain aggregate data and a smoothing coefficient corresponding to the advertisement to be controlled, or determining a cost control factor and the like of the advertisement to be controlled based on the aggregate data and the smoothing coefficient through the cloud server.

In some embodiments, a cloud storage may be further provided, and a pre-constructed data lookup table may be stored in the cloud storage, or information such as advertisement data, cost control factors, advertisement cost control policies, and the like of the advertisement to be controlled may be stored in the cloud storage. In this way, when the cost control factor of the advertisement to be controlled is determined again later, or the advertisement cost of the advertisement to be controlled is controlled, the stored information can be directly obtained from the cloud storage to control the advertisement cost.

Here, cloud technology (Cloud technology) refers to a hosting technology that unifies serial resources such as hardware, software, and networks in a wide area network or a local area network to implement calculation, storage, processing, and sharing of data. The cloud technology is based on the general names of network technology, information technology, integration technology, management platform technology, application technology and the like applied by the cloud computing business mode, can form a resource pool, and is flexible and convenient as required. Cloud computing technology will become an important support. Background services of technical networking systems require a large amount of computing, storage resources, such as video websites, picture-like websites, and more portals. Along with the high development and application of the internet industry, each article possibly has an own identification mark in the future, the identification mark needs to be transmitted to a background system for logic processing, data with different levels can be processed separately, and various industry data needs strong system rear shield support and can be realized only through cloud computing.

Fig. 3 is a schematic structural diagram of a cost control factor determining apparatus provided in an embodiment of the present application, where the cost control factor determining apparatus shown in fig. 3 includes: at least one processor 310, a memory 350, at least one network interface 320, and a user interface 330. The various components in the cost control factor determination device are coupled together by a bus system 340. It is understood that the bus system 340 is used to enable connected communications between these components. The bus system 340 includes a power bus, a control bus, and a status signal bus in addition to the data bus. But for clarity of illustration the various buses are labeled in fig. 3 as bus system 340.

The processor 310 may be an integrated circuit chip with signal processing capabilities such as a general purpose processor, which may be a microprocessor or any conventional processor, or the like, a digital signal processor (DSP, digital Signal Processor), or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or the like.

The user interface 330 includes one or more output devices 331 that enable presentation of media content, and one or more input devices 332.

Memory 350 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid state memory, hard drives, optical drives, and the like. Memory 350 optionally includes one or more storage devices physically located remote from processor 310. Memory 350 includes volatile memory or nonvolatile memory, and may also include both volatile and nonvolatile memory. The nonvolatile Memory may be a Read Only Memory (ROM), and the volatile Memory may be a random access Memory (RAM, random Access Memory). The memory 350 described in embodiments of the present application is intended to comprise any suitable type of memory. In some embodiments, memory 350 is capable of storing data to support various operations, examples of which include programs, modules and data structures, or subsets or supersets thereof, as exemplified below.

The operating system 351 including system programs for handling various basic system services and performing hardware-related tasks, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and handling hardware-based tasks;

network communication module 352 for reaching other computing devices via one or more (wired or wireless) network interfaces 320, exemplary network interfaces 320 include: bluetooth, wireless compatibility authentication (WiFi), and universal serial bus (USB, universal Serial Bus), etc.;

An input processing module 353 for detecting one or more user inputs or interactions from one of the one or more input devices 332 and translating the detected inputs or interactions.

In some embodiments, the apparatus provided by the embodiments of the present application may be implemented in software, and fig. 3 shows a cost control factor determining apparatus 354 stored in a memory 350, where the cost control factor determining apparatus 354 may be a cost control factor determining apparatus in a cost control factor determining device, and may be software in the form of a program and a plug-in, and includes the following software modules: the acquisition module 3541, the query module 3542, the data encoding module 3543, and the smoothing module 3544 are logical, and thus may be arbitrarily combined or further split according to the implemented functions. The functions of the respective modules will be described hereinafter.

In other embodiments, the apparatus provided by the embodiments of the present application may be implemented in hardware, and by way of example, the apparatus provided by the embodiments of the present application may be a processor in the form of a hardware decoding processor that is programmed to perform the cost control factor determination method provided by the embodiments of the present application, e.g., the processor in the form of a hardware decoding processor may employ one or more application specific integrated circuits (ASIC, application Specific Integrated Circuit), DSPs, programmable logic devices (PLD, programmable Logic Device), complex programmable logic devices (CPLD, complex Programma ble Logic Device), field programmable gate arrays (FPGA, field-Programmable Gate Array), or other electronic components.

The cost control factor determining method provided by the embodiments of the present application may be performed by a cost control factor determining device, where the cost control factor determining device may be any kind of terminal having functions of advertisement delivery and promotion, advertisement display, advertisement click, consumption, etc., or may also be a server, that is, the cost control factor determining method of the embodiments of the present application may be performed by the terminal, or may be performed by the server, or may also be performed by the terminal interacting with the server.

Referring to fig. 4, fig. 4 is a schematic flow chart of an alternative method for determining a cost control factor according to an embodiment of the present application, and the steps shown in fig. 4 will be described below, where the method for determining a cost control factor in fig. 4 is described by taking a server as an execution body as an example.

Step S401, advertisement data of advertisements to be controlled are acquired.

Here, the advertisement to be controlled may be any type of advertisement, and the advertisement data includes: advertisement attribution (e.g., advertiser group, advertising industry, advertising category, advertising region, etc.), advertisement content (e.g., price, creative, title, picture, video, template, specification, brand, label, etc.), merchandise information, advertisement audience identification (e.g., gender, age, region, history, browsing information, click information, installation information, etc.) of the advertisement to be controlled. The advertisement data constitutes attribute information of the advertisement to be controlled.

Step S402, inquiring to obtain statistical data characteristics corresponding to advertisement data from a pre-constructed data inquiry table; wherein the statistical data characteristic is the data characteristic of advertisement information which is counted in advance for different advertisements.

Here, the data lookup table is a pre-constructed query dictionary, and the data lookup table includes statistics of advertisements with different attribute information, where the statistics of the advertisements include, but are not limited to: advertisement attribution, advertisement content, commodity information, advertisement audience identification, advertisement exposure, click-through amount, conversion amount and other information.

In the embodiment of the application, when inquiring from the data inquiry table, advertisement data can be used as a keyword, and statistical data corresponding to the advertisement data can be inquired from the data inquiry table. The statistical data may be data of similar advertisements of the same type as the advertisement to be controlled, or data of the advertisement to be controlled which is statistically stored in a history period.

In some embodiments, after the statistical data is obtained by querying, feature extraction is performed on the statistical data to obtain statistical data features, where the statistical data features are data features of advertisement information pre-counted for different advertisements, that is, the statistical data features are obtained by feature extraction of data of advertisement information pre-counted for different advertisements.

In the embodiment of the application, based on the pre-constructed data lookup table, the statistical data corresponding to any attribute information of the advertisement to be controlled can be queried, and further the statistical data characteristics can be extracted. Therefore, the statistical data query can be carried out aiming at different attribute information, the data which can accurately reflect the attribute of the advertisement to be controlled can be obtained, the data which are matched with different commodities and the commodities of the advertisement to be controlled can be obtained through screening, namely, the useful attribute data of the cost control factor can be effectively calculated, and the cost control factor of the advertisement to be controlled can be accurately calculated based on the data obtained through screening.

And S403, carrying out data coding processing on the statistical data characteristics to obtain aggregation data and smoothing coefficients corresponding to the advertisements to be controlled.

The data coding process can be realized by adopting any multi-layer neural network, and the purpose of the data coding process is to further screen the statistical data characteristics so as to screen the data characteristics most suitable for calculating the cost control factor from the statistical data characteristics obtained by inquiry; another aspect is to get a more accurate smoothing coefficient for the fit.

In the embodiment of the application, the aggregated data is data which is obtained after data screening is performed on the statistical data characteristics and is used for representing attribute information of the advertisement to be controlled. That is, based on the statistical data characteristics obtained by the query, useful attribute information for calculating the cost control factor is determined by the data encoding process, and useless characteristics in the statistical data characteristics are removed, that is, useful attribute information of the advertisement to be controlled is automatically identified by the data encoding process.

In the embodiment of the application, because the statistical data characteristics are inquired from the pre-constructed data inquiry table, and the statistical data characteristics are the data characteristics of advertisement information which are pre-counted for different advertisements. Therefore, in the embodiment of the application, the data characteristics of other queried advertisements are screened out through data coding processing and used for representing the attribute information of the advertisements to be controlled, that is, the data coding processing realizes the function of representing the attribute information of the advertisements to be controlled by adopting the data characteristics of similar advertisements similar to the advertisements to be controlled, thereby representing the attribute information of the advertisements to be controlled by using richer data in a pre-constructed data query table and improving the accuracy of the data.

The smoothing coefficient is used for smoothing the aggregate data, that is, when the cost control factor is calculated, the smoothing coefficient is used for predicting the development trend of the cost control factor in a certain period, and the smoothing coefficient can be used as the calculation coefficient when the cost control factor is predicted and multiplied with the determined aggregate data. In the embodiment of the present application, the smoothing coefficient may be any value between 0 and 1.

And step S404, carrying out smoothing processing on the aggregated data based on the smoothing coefficient to obtain a cost control factor of the advertisement to be controlled.

Here, the cost control factor of the advertisement to be controlled is determined based on the aggregate data and the smoothing coefficient, which may be a smoothing process implemented based on a predefined calculation formula, the aggregate data obtained after the data encoding process and the smoothing coefficient are input into the calculation formula, and the aggregate data is smoothed by the smoothing coefficient, so as to calculate the cost control factor of the advertisement to be controlled.

In the embodiment of the application, the smoothing processing is a time sequence analysis prediction method, and when the smoothing processing is carried out, the weighted average calculation of the time dimension is carried out on the aggregated data through the smoothing coefficient, so that the correlation degree between the calculated cost control factor and different attribute information of the advertisement to be controlled is larger than a correlation degree threshold value, and the more accurate advertisement cost control on the advertisement to be controlled can be realized based on the obtained cost control factor.

The cost control factor determining method of the embodiment of the application can be at least applied to advertisement cost control scenes, namely, after the cost control factor is calculated, advertisement cost control can be carried out on advertisements to be controlled based on the cost control factor. In the embodiment of the application, after the cost control factor is obtained, an advertisement cost control strategy can be formulated based on the cost control factor, and then the advertisement cost control process is realized based on the advertisement cost control strategy. For example, if it is determined that the advertisement cost control policy is to increase the advertisement fee based on the cost control factor, the advertiser may be reminded to increase the advertisement fee, and the reminder information is sent to the advertiser's terminal as the advertisement cost control policy. Or determining that the advertisement cost control strategy is to improve the advertisement exposure based on the cost control factor, reminding the platform side to improve the advertisement exposure, and sending the reminding information to the terminal of the platform side as the advertisement cost control strategy.

According to the cost control factor determining method provided by the embodiment of the application, the statistical data characteristics corresponding to the advertisement data of the advertisement to be controlled are inquired from the pre-constructed data inquiry table, the statistical data characteristics are subjected to data coding processing, and the aggregation data and the smoothing coefficients corresponding to the advertisement to be controlled are obtained, so that the cost control factor of the advertisement to be controlled is determined based on the aggregation data and the smoothing coefficients. Thus, the query is performed based on the pre-constructed data query table, so that the corresponding statistical data can be queried based on any specific attribute parameter of the advertisement to be controlled, and the accurate judgment of the similar advertisement is realized based on the statistical data characteristics corresponding to the statistical data; and the accurate smooth coefficient can be dynamically fitted through data coding processing, so that the cost control factor can be accurately determined.

In some embodiments, the cost control factor determining system at least includes a terminal and a server, and the terminal is provided with an advertisement popularization application, so as to support any advertisement popularization application, and by acquiring information such as clicking operation and conversion data for an advertisement to be controlled on the advertisement popularization application, the advertisement popularization application in the embodiment of the application can be any application capable of receiving the advertisement and displaying the advertisement, for example, a shopping application, an instant messaging application, an information browsing application, an audio-video application and the like. The advertisement popularization and application form a platform side for advertisement popularization and exposure, and the platform side can expose the advertisement to be controlled according to a certain exposure amount in the process of running the advertisement popularization and application by the terminal, and collect a certain advertisement fee of the advertisement side based on the exposure amount, click amount, conversion amount and other data. A feasible implementation mode of the cost control factor determining method of the embodiment of the application is to control reasonable charging of the advertising expense based on the cost control factor after calculating the cost control factor, thereby realizing an advertising cost control strategy which is mutually beneficial to advertisers and platform sides.

In the embodiment of the application, in order to ensure the benefits of the advertiser corresponding to the advertisement to be controlled and the platform side for realizing the advertisement promotion application for displaying the advertisement, the server can be a server of the advertiser (for example, the server can be a server of a shopping application aiming at any commodity in the shopping application), a server of the platform side (namely, a server of the advertisement promotion application), or a third party server independent of the advertiser and the platform side. When the advertisement cost control of the advertisement to be controlled is realized, the terminal acquires advertisement data of the advertisement to be controlled through advertisement popularization and application, and sends the advertisement data to the server, so that the advertisement cost control of the advertisement to be controlled is realized.

In the embodiment of the application, the conversion process of the advertisement to be controlled comprises an advertisement initial stage and an advertisement maturation stage; in the initial stage of the advertisement, the absolute value of the difference between the advertisement deviation of the advertisement to be controlled and the whole-day deviation of the advertisement to be controlled is larger than or equal to a preset threshold value; in the advertisement maturation stage, the absolute value of the difference between the advertisement deviation of the advertisement to be controlled and the whole day deviation of the advertisement to be controlled is smaller than a preset threshold value.

Based on the cost control factor determining method of the embodiment of the application, an advertisement cost control method is provided. Fig. 5 is a schematic flow chart of an alternative advertisement cost control method provided in an embodiment of the present application in an advertisement initial stage, as shown in fig. 5, the method includes the following steps:

in step S501, the terminal collects advertisement data of advertisements to be controlled.

Here, the advertisement data includes: advertisement attribution, advertisement content, commodity information, advertisement audience identification, advertisement click quantity, browsing quantity, exposure quantity, conversion quantity and other information of the advertisement to be controlled. The advertisement data may be advertisement data over a whole day or advertisement data over a specific period of time. In the embodiment of the application, each advertisement corresponds to a data storage unit, when the advertisement data is collected, the advertisement data can be collected in real time or periodically, and after the advertisement data is collected, the collected advertisement data is stored into the data storage unit corresponding to the advertisement to be controlled, and when any type of advertisement data is updated, the type of advertisement data in the data storage unit is updated in an overwriting manner.

Step S502, the terminal encapsulates the advertisement data to form an advertisement cost control request.

In step S503, the terminal transmits an advertisement cost control request to the server.

In step S504, the server acquires advertisement attribute parameters of the advertisement to be controlled from the advertisement data in response to the advertisement cost control request.

Here, the advertisement cost control request may be parsed to obtain advertisement attribute parameters, where the advertisement attribute parameters are predefined attribute parameters of the advertisement to be controlled, which needs to be queried. That is, for each advertisement, there may be a plurality of attribute parameters, which attribute parameter needs to be queried based on when generating the advertisement cost control request may be predefined, that is, a target attribute parameter is predefined, and when analyzing the advertisement cost control request, the target attribute parameter is analyzed, and the target attribute parameter is determined as the advertisement attribute parameter of the advertisement to be controlled.

In step S505, the server queries global data and local data corresponding to the advertisement attribute parameter from the data lookup table.

Here, the global data may include global data throughout the day, and the local data may include recent data within a nearest neighbor preset period of time before the current time, for example, data within two hours before the current time; each data in the data lookup table has a data timestamp.

In some embodiments, step S505 may be implemented by: and inquiring to obtain the whole data and the recent data in the whole day from the data inquiry table by taking the advertisement attribute parameter as a keyword based on the current moment and the data time stamp of each data.

In step S506, the server performs feature extraction on the global data and the local data, so as to obtain a global data feature and a local data feature.

Here, the global data feature is a data feature obtained by extracting features of the whole data in the whole day; the local data feature is obtained by extracting the features of the recent data.

In step S507, the server determines the global data feature and the local data feature as statistical data features corresponding to the advertisement attribute parameters in the advertisement data.

Here, the statistical data feature is a data feature of advertisement information that is counted in advance for different advertisements.

Step S508, performing data encoding processing on the global data features and the local data features by using a first encoder to correspondingly obtain global aggregation data and local aggregation data; the global aggregation data and the local aggregation data form aggregation data corresponding to advertisements to be controlled.

In an embodiment of the present application, the first encoder is a first neural network having N layer data processing layers, where N is an integer greater than 1. The data encoding process of the global data feature using the first encoder in step S508 may be implemented by the following steps S5081 to S5083 (not shown in the figure):

in step S5081, the data characteristics of the whole data throughout the day are input to the first neural network.

In step S5082, each data processing layer in the first neural network is used to perform data processing on the data features of the whole data in all days, so as to obtain all-day conversion data and all-day advertisement aggregation data.

Here, the conversion data of the whole day is the estimated conversion number of the advertisement of the whole day; the total day advertisement aggregate data is the actual conversion number of the total day.

When the advertisement to be controlled is CPC advertisement, the whole-day advertisement aggregation data is the product of the whole-day advertisement conversion rate estimated value and the industry factor of the advertisement to be controlled; when the advertisement to be controlled is a CPM advertisement, the all-day advertisement aggregate data is the product of an all-day advertisement conversion rate forecast value (pCVR), an all-day page access click rate (pCTR), and an industry factor of the advertisement to be controlled.

Step S5083, the all-day conversion data and the all-day advertisement aggregation data are determined as global aggregation data.

In some embodiments, the data encoding process of the local data feature in step S508 using the first encoder may be implemented by the following steps S5084 to S5086 (not shown in the figure):

in step S5084, the data characteristics of the recent data are input to the first neural network.

Step S5085, performing data processing on the data features of the recent data by using each data processing layer in the first neural network to obtain the recent conversion data and the recent advertisement aggregation data in the nearest neighbor preset time period.

Here, the recent conversion data is the recent advertisement estimated conversion number; recent advertisement aggregate data is the number of recent actual conversions.

When the advertisement to be controlled is a CPC advertisement, the recent advertisement aggregation data is the product of a recent advertisement conversion rate estimated value in the nearest neighbor preset time period and an industry factor of the advertisement to be controlled; when the advertisement to be controlled is a CPM advertisement, the recent advertisement aggregate data is the product of the recent advertisement conversion rate predicted value, the recent page access click rate and the industry factor of the advertisement to be controlled.

In step S5086, the near-term conversion data and the near-term advertisement aggregation data are determined as local aggregation data.

Step S509, the server performs stitching processing on the global data feature and the local data feature to obtain a first stitching feature.

In the embodiment of the application, the global data characteristic and the local data characteristic are both expressed in the form of vectors. Here, the stitching process refers to connecting a global data feature vector corresponding to a global data feature and a local data feature vector corresponding to a local data feature, so as to form a first stitching feature with a higher dimension. The first stitching feature is also represented in the form of a vector, the dimensions of the first stitching feature being equal to the sum of the dimensions of the global data feature vector and the local data feature vector.

And S510, performing data encoding processing on the first splicing characteristic by adopting a second encoder to obtain a smooth coefficient.

In the embodiment of the present application, the second encoder is a second neural network with M layer data processing layers, where M is an integer greater than 1, and M may be the same as N or different from N. Step S510 may be implemented by the following steps S5101 to S5103 (not shown in the figure):

in step S5101, the first stitching feature is input into the second neural network.

In step S5102, each data processing layer in the second neural network is used to perform data processing on the first stitching feature, so as to obtain a global smoothing coefficient and a first smoothing coefficient.

Here, the global smoothing coefficient is the smoothing coefficient of the whole day, and the first smoothing coefficient is the unified smoothing coefficient.

In step S5103, the global smoothing coefficient and the first smoothing coefficient are determined as smoothing coefficients.

In step S511, the server determines a cost control factor for the advertisement to be controlled based on the aggregated data and the smoothing coefficient.

Here, a predefined calculation formula may be adopted, and the aggregate data and the smoothing coefficient may be input as variable values into the calculation formula, so as to calculate a cost control factor of the advertisement to be controlled in the advertisement initial stage.

In step S512, the server determines an advertisement cost control policy based on the cost control factor.

In step S513, the server transmits the advertisement cost control policy to the terminal.

Step S514, the terminal displays the advertisement cost control strategy to remind the user to control the advertisement cost according to the advertisement cost control strategy.

According to the advertisement cost control method provided by the embodiment of the application, in the initial stage of advertisement, the encoding processing process of the statistical data characteristics is realized through the first encoder and the second encoder, so that accurate aggregate data and smooth coefficients are obtained through prediction, and accurate cost control factors are obtained through calculation based on the aggregate data and the smooth coefficients, so that reasonable advertisement cost control is carried out on advertisements to be controlled based on the cost control factors, and the use experience of advertisers and platform sides is improved.

In some embodiments, determining a cost control factor for an advertisement to be controlled in an advertisement initial stage may be implemented by an advertisement initial stage model. The advertisement initial stage model at least comprises a data query network, a data coding network and a data calculation layer, wherein the data query network is used for realizing data query on sample advertisement data of sample advertisements to obtain sample statistical data, and extracting characteristics of the sample statistical data to obtain sample statistical data characteristics; the data coding network is used for carrying out data coding processing on the sample statistical data characteristics so as to screen and extract sample aggregate data and sample smoothing coefficients; the data calculation layer is used for calculating and obtaining a sample cost control factor.

The embodiment of the application provides a training method of an advertisement initial stage model, which is used for training the advertisement initial stage model through the training method of the advertisement initial stage model to obtain a trained advertisement initial stage model, so that a cost control factor is obtained based on the prediction and calculation of the trained advertisement initial stage model, and further, the advertisement cost control to be controlled based on the determined cost control factor is realized.

Fig. 6 is a schematic flow chart of an implementation of an advertisement initial stage model training method according to an embodiment of the present application, as shown in fig. 6, the method includes the following steps:

step S601, inputting the first sample advertisement data into the advertisement initial stage model.

Here, the first sample advertisement data includes a positive sample, which is sample data having a response operation and having conversion data for the first sample advertisement, and a negative sample, which is sample data having a response operation and not having conversion data for the first sample advertisement.

Step S602, inquiring to obtain first sample statistical data characteristics corresponding to first sample advertisement data from a pre-constructed data inquiry table through a data inquiry network of an advertisement initial stage model.

Step S603, performing data encoding processing on the first sample statistical data feature through the data encoding network of the advertisement initial stage model to obtain first sample aggregate data and a first sample smoothing coefficient corresponding to the first sample advertisement.

Step S604, smoothing the first sample aggregate data based on the first sample smoothing coefficient by a data calculation layer of the advertisement initial stage model to obtain a first sample cost control factor of the first sample advertisement.

Step S605, the first sample cost control factor is input into a preset loss model, and the loss calculation is performed on the first sample cost control factor through the preset loss model, so as to obtain a first loss result.

Step S606, correcting parameters in the advertisement initial stage model according to the first loss result to obtain a trained advertisement initial stage model.

In the embodiment of the application, the loss function is arranged in the preset loss model, the loss calculation is carried out on the first sample cost control factor through the loss function in the preset loss model, the difference value between the first sample cost control factor and the real cost control factor is determined, and when the difference value is larger than the distance threshold value, the current advertisement initial stage model can not accurately predict the first sample cost control factor, so that the advertisement initial stage model needs to be continuously trained.

In the embodiment of the application, the loss calculation is performed on the first sample cost control factor through the preset loss model, so that the parameter in the advertisement initial stage model is reversely transmitted, the optimization of the advertisement initial stage model is realized, the cost control factor generated by the advertisement initial stage model obtained through training can be ensured to be more accurate, and a more proper advertisement cost control strategy can be determined based on the predicted cost control factor.

Based on the cost control factor determining method of the embodiment of the application, an advertisement cost control method is provided. FIG. 7 is a schematic flow chart of an alternative advertisement cost control method in the advertisement maturation stage according to an embodiment of the present application, as shown in FIG. 7, the method includes the following steps:

in step S701, the terminal collects advertisement data of an advertisement to be controlled.

In the embodiment of the application, each advertisement corresponds to a data storage unit, when the advertisement data is collected, the advertisement data can be collected in real time or periodically, and after the advertisement data is collected, the collected advertisement data is stored into the data storage unit corresponding to the advertisement to be controlled, and when any type of advertisement data is updated, the type of advertisement data in the data storage unit is updated in an overwriting manner.

In step S702, the terminal encapsulates the advertisement data to form an advertisement cost control request.

In step S703, the terminal transmits an advertisement cost control request to the server.

In step S704, the server acquires advertisement attribute parameters of the advertisement to be controlled from the advertisement data in response to the advertisement cost control request.

In step S705, the server queries global data corresponding to the advertisement attribute parameter from the data lookup table.

In step S706, the server acquires advertisement conversion data and advertisement aggregation data of the advertisement to be controlled.

Here, the advertisement conversion data is the estimated conversion number of the advertisement itself; the advertisement aggregate data is the actual conversion number of the advertisement itself.

In step S707, the server queries similar advertisement data of the advertisement to be controlled from the data lookup table based on the advertisement conversion data and the advertisement aggregation data.

Here, the similar advertisement data is data of a similar advertisement belonging to the same category as the advertisement to be controlled, that is, the similar advertisement belonging to the same category as the advertisement to be controlled is determined from the data lookup table based on the advertisement conversion data and the advertisement aggregation data, and then the advertisement data of the similar advertisement is extracted.

In the embodiment of the application, the similar advertisements of the advertisements to be controlled can be queried from the data query table through the pre-constructed neural network.

In step S708, the server performs feature extraction on the global data and the similar advertisement data, respectively, to obtain global data features and similar advertisement data features.

Here, the global data feature is a data feature obtained by extracting features of the whole data in the whole day; the similar advertisement data feature is obtained by extracting the feature of the similar advertisement data.

In step S709, the server determines the global data feature and the similar advertisement data feature as statistical data features corresponding to the advertisement attribute parameters in the advertisement data.

In step S710, the server performs data encoding processing on the global data feature by using a third encoder, so as to obtain global aggregate data correspondingly. In the embodiment of the application, the global aggregated data forms the aggregated data corresponding to the advertisement to be controlled.

Step S711, the server performs a stitching process on the global data feature and the similar advertisement data feature, to obtain a second stitching feature.

In the embodiment of the application, the global data characteristics and the similar advertisement data characteristics are expressed in the form of vectors. Here, the splicing process refers to connecting the global data feature vector corresponding to the global data feature and the similar advertisement data feature vector corresponding to the similar advertisement data feature to form a second splicing feature with higher dimension. The second stitching feature is also represented in the form of a vector, the dimensions of the second stitching feature being equal to the sum of the dimensions of the global data feature vector and the similar advertisement data feature vector.

In step S712, the server performs data encoding processing on the second splicing feature by using a fourth encoder, to obtain a smoothing coefficient.

In the embodiment of the present application, the fourth encoder is a third neural network having an L-layer data processing layer, where L is an integer greater than 1, and the three layers L, M and N may be the same or different. Step S712 may be implemented by the following steps S7121 to S7123 (not shown in the drawings):

step S7121, inputting a second stitching feature into the third neural network.

And step S7122, adopting each data processing layer in the third neural network to perform data processing on the second splicing characteristic to obtain a basic smoothing coefficient and a smoothing index.

In step S7123, the base smoothing coefficient and the smoothing index are determined as smoothing coefficients.

In step S713, the server determines a cost control factor for the advertisement to be controlled based on the aggregated data and the smoothing coefficients.

Here, a predefined calculation formula may be adopted, and the aggregate data and the smoothing coefficient may be input as variable values into the calculation formula, so as to calculate a cost control factor of the advertisement to be controlled in the advertisement maturation stage.

In step S714, the server determines an advertisement cost control policy based on the cost control factor.

In step S715, the server transmits the advertisement cost control policy to the terminal.

In step S716, the terminal displays the advertisement cost control policy to remind the user to control the advertisement cost according to the advertisement cost control policy.

According to the advertisement cost control method provided by the embodiment of the application, in the advertisement maturation stage, the encoding processing process of the statistical data characteristics is realized through the third encoder and the fourth encoder together, so that accurate aggregate data and smooth coefficients are obtained through prediction, and accurate cost control factors are obtained through calculation based on the aggregate data and the smooth coefficients, so that reasonable advertisement cost control is carried out on advertisements to be controlled based on the cost control factors, and the use experience of advertisers and platform sides is improved.

In some embodiments, determining a cost control factor for an advertisement to be controlled at an advertisement maturity stage may be accomplished through an advertisement maturity stage model. The advertisement maturity stage model at least comprises a data query network, a data coding network and a data calculation layer, wherein the data query network is used for realizing data query on sample advertisement data of sample advertisements to obtain sample statistical data, and extracting characteristics of the sample statistical data to obtain sample statistical data characteristics; the data coding network is used for carrying out data coding processing on the sample statistical data characteristics so as to screen and extract sample aggregate data and sample smoothing coefficients; the data calculation layer is used for calculating and obtaining a sample cost control factor.

The embodiment of the application provides a training method of an advertisement maturity stage model, which is used for training the advertisement maturity stage model through the training method of the advertisement maturity stage model to obtain a trained advertisement maturity stage model, so that a cost control factor of the advertisement maturity stage is obtained based on the prediction and calculation of the trained advertisement maturity stage model, and further, the advertisement cost control to be controlled based on the determined cost control factor is realized.

Fig. 8 is a schematic flow chart of an implementation of the advertisement maturity stage model training method according to an embodiment of the present application, as shown in fig. 8, the method includes the following steps:

step S801, second sample advertisement data is input into the advertisement maturity stage model.

Here, the second sample advertisement data includes a positive sample and a negative sample, wherein the positive sample is sample data having a response operation and having conversion data for the second sample advertisement, and the negative sample is sample data having a response operation and not having conversion data for the second sample advertisement.

Step S802, inquiring to obtain second sample statistical data characteristics corresponding to second sample advertisement data from a pre-constructed data inquiry table through a data inquiry network of the advertisement maturity stage model.

Step 803, performing data encoding processing on the second sample statistical data feature through the data encoding network of the advertisement maturity stage model to obtain second sample aggregate data and a second sample smoothing coefficient corresponding to the second sample advertisement.

Step S804, smoothing the second sample aggregate data based on the second sample smoothing coefficient by the data calculation layer of the advertisement maturity stage model to obtain a second sample cost control factor of the second sample advertisement.

Step S805, inputting the second sample cost control factor into a preset loss model, and calculating the loss of the second sample cost control factor through the preset loss model to obtain a second loss result.

And step S806, correcting parameters in the advertisement maturity stage model according to the second loss result to obtain a trained advertisement maturity stage model.

In the embodiment of the application, the loss function is arranged in the preset loss model, the loss calculation is carried out on the second sample cost control factor through the loss function in the preset loss model, the difference value between the second sample cost control factor and the real cost control factor is determined, and when the difference value is larger than the distance threshold value, the current advertisement maturation stage model can not accurately predict the second sample cost control factor, so that the advertisement maturation stage model needs to be continuously trained.

In the embodiment of the application, the loss calculation is performed on the second sample cost control factor through the preset loss model, so that the parameter in the advertisement maturity stage model is reversely transmitted, the optimization of the advertisement maturity stage model is realized, the cost control factor generated by the advertisement maturity stage model obtained through training can be ensured to be more accurate, and a more proper advertisement cost control strategy can be determined based on the predicted cost control factor.

In the following, an exemplary application of the embodiment of the present application in a practical application scenario will be described.

The E-commerce industry calculates the industry independent cost control factors, and the calculation is divided into an advertisement initial stage and an advertisement maturation stage. Fig. 9 and 10 are advertisement deviation variation curves of two advertisements of commodity a and commodity B of a brand e-commerce respectively in one day, fig. 11 and 12 are advertisement deviation variation curves of two advertisements of commodity C and commodity D of a direct e-commerce respectively in one day, and as shown in fig. 9 to 12, from the data analysis, the deviation of the advertisement in the insufficient period of conversion (i.e. the initial period of advertisement) is greatly different from the deviation of the whole day, and after a certain amount of conversion is accumulated in the advertisement (i.e. the mature period of advertisement), the advertisement deviation is similar to the whole deviation.

Based on such variability, different strategies are required to calculate independent cost control factors at different stages.

In the initial stage of the advertisement, the calculation of the control factors is completed mainly by combining recent data of similar advertisements with the whole data. In the initial stage of the advertisement, the cost control factor is calculated, and the optimization is approximately performed in four steps, as shown in a schematic diagram of the process of calculating the cost control factor in the initial stage of the advertisement shown in fig. 13:

step S131, the advertisements are layered and aggregated according to commodity attributes and flow attributes, which aggregation class the advertisements to be controlled belong to is determined, and the statistical data of the aggregation class is used for calculating the cost control factor.

As shown in fig. 13, merchandise attributes include, but are not limited to: advertiser groups, price intervals, brands; traffic attributes include, but are not limited to: traffic such as order, payment, and activation generated by instant messaging applications, shopping applications, information recommendation applications, and the like.

Step S132, adding new and old advertisement attributes based on the previous step, and distinguishing external different competition environments according to the categories of the new and old advertisements.

Here, the categories of the new and old advertisements include: the advertisement has no consumption yesterday, the advertisement has consumption yesterday, and the like.

In step S133, the statistical method weighted by time is added to fade the effect of the data before a long time.

Step S134, considering both the influence of recent data and all day data, emphasizes the recent data, and facilitates capturing the recent cost mutation.

The calculation formula of the finally obtained calculation independent cost control factor C is the following formula (1):

wherein smoothened_total_conveyances are represented by the following formula (2):

where, event_represents recent (e.g., two hours) data; the transitions represent the estimated conversion number of advertisements; total_represents the whole day data. coef is advertisement aggregate data (i.e., actual conversion number of advertisement), for CPC advertisement, coef is pCVR, which is the product of pCVR and industry factor (industry factor); for CPM advertising, coef is the product pCTR, pCVR and industry factor pCVR industriyfactor.

Then in equations (1) and (2) above, the current_transitions represent the number of recent ad campaign conversions; the current_coef indicates the actual conversion number in the near term; total_coef represents the actual conversion number throughout the day; total_transitions represents the estimated conversion number of advertisements throughout the day. Both total_smooth and smooth are smoothing coefficients. Wherein total_smooth represents the smoothing coefficient of the whole day; smooths represent unified smoothing coefficients.

In the advertisement maturation stage, the data of the advertisement itself can be utilized to combine with the data of similar advertisements to complete the cost control. Here, the advertisement maturation stage means that the advertisement has a lot of data such as exposure conversion, and the advertisement cost can be controlled mainly by using the data of the advertisement itself.

The control factor calculation at this stage is approximately 3 steps, and the advertisement maturation stage cost control factor calculation process is schematically shown in fig. 14:

in step S141, the calculation of the cost control factor is completed using the data of the advertisement itself. An important assumption in this way is that future deviations of the advertisement remain consistent with previous deviations, but this assumption is difficult to guarantee as the competing environment changes, because even if the conversion is sufficient, the data is not completely sufficient.

At this time, the calculation formula of the cost control factor C is the following formula (3):

C＝(ad_conversions+smooth_base)/(ad_coef+smooth_base) (3)；

wherein ad_represents statistical data of the advertisement itself, advertisement bias ad_bias=ad_coef/ad_conveyances;

in the above formula (3), ad_convertions represent the estimated conversion number of the advertisement itself; ad_coef represents the actual conversion number of the advertisement itself.

In step S142, data for similar advertisements of the same type is imported. The cost control factor is calculated by utilizing the advertisement self data and the aggregation data of similar advertisements.

At this time, the calculation formula of the cost control factor C is the following formula (4):

C＝(ad_conversions*history_bias_factor+smooth_base)/(ad_coef+smooth_base) (4)；

wherein, the parameter history_bias_factor is represented by the following formula (5):

wherein category_represents aggregated statistics of similar advertisements of the same kind.

Then in equations (4) and (5) above, category_conversio represents the estimated conversion number of the similar advertisement; category_coef represents the actual conversion number of similar advertisements.

In step S143, the smoothing coefficients used in the previous two steps are intelligently adjusted.

At this time, the calculation formula of the cost control factor C is the following formula (6):

wherein the parameters smooth_bias and smooth_base are represented by the following formulas (7) and (8):

smooth_biae＝category_conversions/category_coef (7)；

/>

wherein threshold is the conversion number threshold; basic_smooth_base represents a basic smoothing coefficient; smooth_ exact represents the smoothness index.

Fig. 15 is an aggregation statistics manner of similar advertisement data in the advertisement maturation stage provided by the embodiment of the present application, as shown in fig. 15, similar advertisement data may be aggregated and counted based on commodity attributes (for example, advertiser sets, price intervals, brands), in combination with flow attributes (for example, flows of ordering, payment, activation, etc. generated by instant messaging applications, shopping applications, information recommendation applications, etc.), in combination with new and old advertisement categories (for example, advertisement no consumption yesterday, advertisement consumption before yesterday, etc.), and the like.

In the embodiment of the application, to obtain the aggregated data of similar advertisements of the same kind, how to identify similar advertisements of the same kind and how to determine the smoothing coefficients playing an important role are all needed in the calculation of the independent cost control factors of different stages, and the problem to be solved in the following cost control factor determination method is solved.

Based on the above, the aggregation manner of similar advertisements of the same kind can be realized by adopting the manner of fig. 15; independent cost control at the initial stage of advertisement can be realized based on the formulas (1) and (2) when cost control factors are calculated; independent cost control of the ad maturation stage, when calculating cost control factors, may be implemented based on equations (6), (7) and (8) above.

The method for determining the cost control factor provided by the embodiment of the application provides a mode for automatically identifying similar advertisements and a strategy for automatically calculating the smoothing coefficient aiming at the two problems. The smoothing coefficients in equations (1) and (2), and equations (6), (7) and (8) are all smoothing coefficients that need to be calculated.

The judgment of the similar advertisements mainly determines what commodity attributes are used as clustering indexes of the similar advertisements. FIG. 16 is a probability distribution function (PDF, probability Distribution Fun ctions) of the ratio sum_convs/sum_coefs of the sum of estimated conversion numbers sum_convs to the sum of actual conversion numbers sum_coefs for different products according to an embodiment of the present application. It can be seen that different merchandise attributes represent the differences between advertisements: the good quality commodity attributes can distinguish the similar points and the different points among advertisements.

Meanwhile, the quality commodity attributes vary with industry and competing environments. Fig. 17 is a diagram of classifying properties of goods based on different categories, wherein under the category of advertisement attribution 171, the method comprises: commodity attributes such as advertisers, advertiser groups, advertising industry, advertising categories, advertising regions, and the like. Under the advertisement audience 172 category, include: gender, age, region, academic, personnel, browsing, clicking and installing, etc. Under the advertising content 173 category, include: price, creative, title, picture, template, specification, brand, label, etc. By selecting commodity attributes in real time in different industries, it is determined which commodity attributes can be adopted, and which commodity attributes can be adopted specifically can be dynamically determined.

In the embodiment of the application, similar advertisement characteristics are screened by establishing a model, and super parameters are fitted. The advertisement initial stage and the advertisement maturation stage need to be modeled separately because of the different methods of calculating the cost control factors.

Because the advertisement initial stage uses the data of all the time slots of the day and the data of the last two hours; the advertisement maturation stage temporarily does not consider the influence of the latest data, so that the smoothing coefficients of the two differ.

In the embodiment of the application, the model comprises an advertisement initial stage model and an advertisement mature stage model, wherein a positive sample of the model is data with effective click conversion, and a negative sample of the model is data without effective click conversion. The loss function is a cross entropy loss function. All parameters required for cost control factor calculation, including smoothing coefficients, are calculated from a plurality of commodity attribute aggregation data through the construction of a deep learning network. The whole process is to screen aggregate data combining different commodity attributes through a neural network and dynamically fit smooth coefficients.

Fig. 18 is a schematic structural diagram of an advertisement initial stage model according to an embodiment of the present application, and as shown in fig. 18, the advertisement initial stage model is mainly divided into three parts. From the left, the first portion 181 is the calculated whole day data, the second portion 182 is the calculated last two hours of advertising data, and the third portion 183 is the calculated smoothing factor. These calculated data are then used to carry over into cost control calculations (1) and (2) to obtain cost control factor p (x) (i.e., C) above, and then carried over into loss function loss to calculate the loss.

The look-up table (look-up table) is a data table obtained by aggregating commodity attributes such as site set, optimization target, time slicing, advertiser, brand, etc.

FIG. 19 is a schematic diagram of an advertisement maturation stage model according to an embodiment of the present application, where, as shown in FIG. 19, the advertisement maturation stage model is mainly divided into two parts, and from the left, the first part 191 is calculated data of all days, and is mainly the best aggregate statistics obtained by screening from various attribute aggregate data of advertisements; the second portion 192 is to calculate smoothing coefficients. After these smoothing coefficients are obtained, the result is taken into cost control factor calculation formulas (6), (7), (8) and (5), and after the cost control factor p (x) (i.e., C) is obtained, the result is taken into a loss function loss to calculate a loss.

Table 1 shows the results of the method of the present application applied to the direct-nutrient electronic commerce and the brand electronic commerce respectively, wherein the ratio of the total commodity transaction amount (GMV, gross Merchandise Volume) to the cost (cost) is increased by 1.05% and the non-excessive cost rate (i.e. the consumption dimension) is increased by 2.5% in the direct-nutrient electronic commerce as shown in Table 1; in the brand e-commerce, the GMV/cost is increased by 2.2%, and the non-superstration cost (namely the consumption dimension) is increased by 3.0%.

TABLE 1

Industry (e.g.)	GMV/COST	Without exceeding cost rate (consumption dimension)
			Direct nutrient electricity merchant	+1.05％	+2.5％
Brand E-commerce	+2.2％	+3.0％

It should be noted that, by establishing a model, the strategy of learning the aggregation property and the smoothing coefficient from the data is within the protection scope of the embodiment of the present application, no matter what type of network structure is established.

It will be appreciated that in embodiments of the present application, where information related to user information, such as advertisement data, cost control factors, advertisement cost control policies, clicking actions of users, consumption actions, etc., is involved, if data related to user information or business information is involved, user permissions or consents need to be obtained when embodiments of the present application are applied to specific products or technologies, and the collection, use and processing of the relevant data is subject to relevant legal regulations and standards of the relevant country and region.

Continuing with the description below, the cost control factor determining device 354 provided by embodiments of the present application is implemented as an exemplary structure of a software module, and in some embodiments, as shown in fig. 3, the cost control factor determining device 354 includes:

In some embodiments, the conversion process of the advertisement to be controlled comprises an advertisement initial stage; in the initial stage of the advertisement, the absolute value of the difference value between the advertisement deviation of the advertisement to be controlled and the whole-day deviation of the advertisement to be controlled is larger than or equal to a preset threshold value; the query module is further configured to, when in the advertisement initiation phase: acquiring advertisement attribute parameters of the advertisement to be controlled from the advertisement data; querying global data and local data corresponding to the advertisement attribute parameters from the data query table; respectively carrying out feature extraction on the global data and the local data to obtain global data features and local data features; and determining the global data characteristic and the local data characteristic as statistical data characteristics corresponding to advertisement attribute parameters in the advertisement data.

In some embodiments, the global data comprises global data throughout the day, and the local data comprises recent data within a nearest neighbor preset time period prior to the current time; each data in the data lookup table has a data timestamp; the query module is further configured to: and inquiring the whole data and the recent data in the whole day based on the current moment and the data time stamp of each data in the data inquiry table by taking the advertisement attribute parameter as a keyword.

In some embodiments, the data encoding module is further to: respectively carrying out data coding processing on the global data features and the local data features by adopting a first coder to correspondingly obtain global aggregation data and local aggregation data; the global aggregation data and the local aggregation data form aggregation data corresponding to the advertisement to be controlled; performing splicing processing on the global data features and the local data features to obtain first splicing features; and adopting a second encoder to carry out data encoding processing on the first splicing characteristic to obtain the smoothing coefficient.

In some embodiments, the first encoder is a first neural network having N-layer data processing layers; the data encoding module is further configured to: inputting the data characteristics of the whole data in the whole day into the first neural network; adopting each data processing layer in the first neural network to perform data processing on the data characteristics of the whole data in all days to obtain all-day conversion data and all-day advertisement aggregation data; when the advertisement to be controlled is a CPC advertisement, the all-day advertisement aggregation data is the product of an all-day advertisement conversion rate predicted value and an industry factor of the advertisement to be controlled; when the advertisement to be controlled is a CPM advertisement, the all-day advertisement aggregation data is the product of an all-day advertisement conversion rate estimated value, an all-day page access click rate and an industry factor of the advertisement to be controlled; determining the all-day conversion data and the all-day advertisement aggregation data as the global aggregation data; wherein N is an integer greater than 1.

In some embodiments, the first encoder is a first neural network having N-layer data processing layers; the data encoding module is further configured to: inputting data characteristics of the recent data into the first neural network; adopting each data processing layer in the first neural network to perform data processing on the data characteristics of the recent data to obtain the recent conversion data and the recent advertisement aggregation data in the nearest neighbor preset time period; when the advertisement to be controlled is a CPC advertisement, the recent advertisement aggregation data is the product of a recent advertisement conversion rate estimated value in the nearest neighbor preset time period and an industry factor of the advertisement to be controlled; when the advertisement to be controlled is a CPM advertisement, the recent advertisement aggregation data is the product of a recent advertisement conversion rate estimated value, a recent page access click rate and an industry factor of the advertisement to be controlled; determining the recent conversion data and the recent advertisement aggregation data as the local aggregation data; wherein N is an integer greater than 1.

In some embodiments, the second encoder is a second neural network having M layer data processing layers; the data encoding module is further configured to: inputting the first stitching feature into the second neural network; adopting each data processing layer in the second neural network to perform data processing on the first splicing characteristics to obtain a global smoothing coefficient and a first smoothing coefficient; determining the global smoothing coefficient and the first smoothing coefficient as the smoothing coefficient; wherein M is an integer greater than 1.

In some embodiments, determining the cost control factor for the advertisement to be controlled is accomplished by an advertisement initial stage model; the advertisement initial stage model is trained by the following steps: inputting first sample advertisement data into the advertisement initial stage model; inquiring to obtain first sample statistical data characteristics corresponding to the first sample advertisement data from a pre-constructed data inquiry table through a data inquiry network of the advertisement initial stage model; performing data coding processing on the first sample statistical data characteristic through a data coding network of the advertisement initial stage model to obtain first sample aggregate data and a first sample smoothing coefficient corresponding to a first sample advertisement; performing smoothing processing on the first sample aggregate data based on the first sample smoothing coefficient through a data calculation layer of the advertisement initial stage model to obtain a first sample cost control factor of the first sample advertisement; inputting the first sample cost control factor into a preset loss model, and carrying out loss calculation on the first sample cost control factor through the preset loss model to obtain a first loss result; and correcting parameters in the advertisement initial stage model according to the first loss result to obtain a trained advertisement initial stage model.

In some embodiments, the conversion process of the advertisement to be controlled includes an advertisement maturation stage; wherein, in the advertisement maturation stage, the absolute value of the difference between the advertisement deviation of the advertisement to be controlled and the whole day deviation of the advertisement to be controlled is smaller than a preset threshold; the query module is further configured to, when in the advertisement maturation stage: acquiring advertisement attribute parameters of the advertisement to be controlled from the advertisement data; querying global data corresponding to the advertisement attribute parameters from the data query table; acquiring advertisement conversion data and advertisement aggregation data of the advertisement to be controlled; inquiring similar advertisement data of the advertisement to be controlled from the data inquiry table based on the advertisement conversion data and the advertisement aggregation data; respectively extracting features of the global data and the similar advertisement data to obtain global data features and similar advertisement data features; and determining the global data characteristic and the similar advertisement data characteristic as statistical data characteristics corresponding to advertisement attribute parameters in the advertisement data.

In some embodiments, the data encoding module is further to: performing data encoding processing on the global data features by adopting a third encoder to correspondingly obtain global aggregated data; the global aggregation data form aggregation data corresponding to the advertisement to be controlled; performing splicing processing on the global data features and the similar advertisement data features to obtain second splicing features; and adopting a fourth encoder to carry out data encoding processing on the second splicing characteristic to obtain the smoothing coefficient.

In some embodiments, the fourth encoder is a third neural network having L layer data processing layers; the data encoding module is further configured to: inputting the second stitching feature into the third neural network; adopting each data processing layer in the third neural network to perform data processing on the second splicing characteristics to obtain a basic smoothing coefficient and a smoothing index; determining the base smoothing coefficient and the smoothing index as the smoothing coefficient; wherein L is an integer greater than 1.

In some embodiments, determining the cost control factor for the advertisement to be controlled is accomplished by an advertisement maturity stage model; the advertisement maturity stage model is trained by the steps of: inputting second sample advertisement data into the advertisement maturity stage model; inquiring to obtain second sample statistical data characteristics corresponding to the second sample advertisement data from a pre-constructed data inquiry table through a data inquiry network of the advertisement maturity stage model; performing data coding processing on the second sample statistical data characteristics through a data coding network of the advertisement maturity stage model to obtain second sample aggregate data and second sample smoothing coefficients corresponding to second sample advertisements; performing smoothing processing on the second sample aggregate data based on the second sample smoothing coefficient through a data calculation layer of the advertisement maturity stage model to obtain a second sample cost control factor of the second sample advertisement; inputting the second sample cost control factor into a preset loss model, and carrying out loss calculation on the second sample cost control factor through the preset loss model to obtain a second loss result; and correcting parameters in the advertisement maturity stage model according to the second loss result to obtain a trained advertisement maturity stage model.

It should be noted that, the description of the apparatus according to the embodiment of the present application is similar to the description of the embodiment of the method described above, and has similar beneficial effects as the embodiment of the method, so that a detailed description is omitted. For technical details not disclosed in the present apparatus embodiment, please refer to the description of the method embodiment of the present application for understanding.

Embodiments of the present application provide a computer program product or computer program comprising executable instructions, the executable instructions being a computer instruction; the executable instructions are stored in a computer readable storage medium. The cost control factor determination device is caused to perform the method of the embodiments of the present application described above when the processor of the cost control factor determination device reads the executable instructions from the computer-readable storage medium and the processor executes the executable instructions.

Embodiments of the present application provide a storage medium having stored therein executable instructions which, when executed by a processor, cause the processor to perform a method provided by embodiments of the present application, for example, as shown in fig. 4.

In some embodiments, the storage medium may be a computer readable storage medium, such as a ferroelectric Memory (FRAM, ferromagnetic Random Access Memory), read Only Memory (ROM), programmable Read Only Memory (PROM, programmable Read Only Memory), erasable programmable Read Only Memory (EPROM, erasable Programmable Read Only Memory), electrically erasable programmable Read Only Memory (EEPROM, electrically Erasable Programmable Read Only Memory), flash Memory, magnetic surface Memory, optical Disk, or Compact Disk-Read Only Memory (CD-ROM), or the like; but may be a variety of devices including one or any combination of the above memories.

In some embodiments, the executable instructions may be in the form of programs, software modules, scripts, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and they may be deployed in any form, including as stand-alone programs or as modules, components, subroutines, or other units suitable for use in a computing environment.

As an example, the executable instructions may, but need not, correspond to files in a file system, may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a hypertext markup language (HTML, hyper Text Markup Language) document, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files that store one or more modules, sub-programs, or portions of code). As an example, the executable instructions may be deployed to execute on one computing device (which may be a job run-length determining device) or on multiple computing devices located at one site, or on multiple computing devices distributed across multiple sites and interconnected by a communication network.

The foregoing is merely exemplary embodiments of the present application and is not intended to limit the scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and scope of the present application are included in the protection scope of the present application.

Claims

1. A method of cost control factor determination, the method comprising:

acquiring advertisement data of advertisements to be controlled;

2. The method according to claim 1, wherein the conversion process of the advertisement to be controlled comprises an advertisement initiation phase; in the initial stage of the advertisement, the absolute value of the difference value between the advertisement deviation of the advertisement to be controlled and the whole-day deviation of the advertisement to be controlled is larger than or equal to a preset threshold value;

And when the advertisement is in the initial stage of advertisement, inquiring to obtain statistical data characteristics corresponding to the advertisement data from a pre-constructed data inquiry table, wherein the statistical data characteristics comprise:

acquiring advertisement attribute parameters of the advertisement to be controlled from the advertisement data;

querying global data and local data corresponding to the advertisement attribute parameters from the data query table;

respectively carrying out feature extraction on the global data and the local data to obtain global data features and local data features;

and determining the global data characteristic and the local data characteristic as statistical data characteristics corresponding to advertisement attribute parameters in the advertisement data.

3. The method of claim 2, wherein the global data comprises global data throughout the day and the local data comprises recent data within a nearest neighbor preset time period prior to the current time instant; each data in the data lookup table has a data timestamp;

the querying global data and local data corresponding to the advertisement attribute parameters from the data query table comprises:

and inquiring the whole data and the recent data in the whole day based on the current moment and the data time stamp of each data in the data inquiry table by taking the advertisement attribute parameter as a keyword.

4. A method according to claim 3, wherein the data encoding the statistical data feature to obtain aggregate data and a smoothing coefficient corresponding to the advertisement to be controlled comprises:

respectively carrying out data coding processing on the global data features and the local data features by adopting a first coder to correspondingly obtain global aggregation data and local aggregation data; the global aggregation data and the local aggregation data form aggregation data corresponding to the advertisement to be controlled;

performing splicing processing on the global data features and the local data features to obtain first splicing features;

and adopting a second encoder to carry out data encoding processing on the first splicing characteristic to obtain the smoothing coefficient.

5. The method of claim 4, wherein the first encoder is a first neural network having N layers of data processing layers; and adopting a first encoder to perform data encoding processing on the global data characteristics to correspondingly obtain global aggregate data, wherein the method comprises the following steps:

inputting the data characteristics of the whole data in the whole day into the first neural network;

adopting each data processing layer in the first neural network to perform data processing on the data characteristics of the whole data in all days to obtain all-day conversion data and all-day advertisement aggregation data;

When the advertisement to be controlled is a CPC advertisement, the all-day advertisement aggregation data is the product of an all-day advertisement conversion rate predicted value and an industry factor of the advertisement to be controlled; when the advertisement to be controlled is a CPM advertisement, the all-day advertisement aggregation data is the product of an all-day advertisement conversion rate estimated value, an all-day page access click rate and an industry factor of the advertisement to be controlled;

determining the all-day conversion data and the all-day advertisement aggregation data as the global aggregation data;

wherein N is an integer greater than 1.

6. The method of claim 4, wherein the first encoder is a first neural network having N layers of data processing layers; and adopting a first encoder to perform data encoding processing on the local data characteristics to correspondingly obtain local aggregate data, wherein the method comprises the following steps:

inputting data characteristics of the recent data into the first neural network;

adopting each data processing layer in the first neural network to perform data processing on the data characteristics of the recent data to obtain the recent conversion data and the recent advertisement aggregation data in the nearest neighbor preset time period;

when the advertisement to be controlled is a CPC advertisement, the recent advertisement aggregation data is the product of a recent advertisement conversion rate estimated value in the nearest neighbor preset time period and an industry factor of the advertisement to be controlled; when the advertisement to be controlled is a CPM advertisement, the recent advertisement aggregation data is the product of a recent advertisement conversion rate estimated value, a recent page access click rate and an industry factor of the advertisement to be controlled;

Determining the recent conversion data and the recent advertisement aggregation data as the local aggregation data;

wherein N is an integer greater than 1.

7. The method of claim 4, wherein the second encoder is a second neural network having M layer data processing layers;

the step of performing data encoding processing on the first splicing characteristic by using a second encoder to obtain the smoothing coefficient includes:

inputting the first stitching feature into the second neural network;

adopting each data processing layer in the second neural network to perform data processing on the first splicing characteristics to obtain a global smoothing coefficient and a first smoothing coefficient;

determining the global smoothing coefficient and the first smoothing coefficient as the smoothing coefficient;

wherein M is an integer greater than 1.

8. The method according to any one of claims 2 to 7, wherein determining a cost control factor for the advertisement to be controlled is achieved by an advertisement initial stage model;

the advertisement initial stage model is trained through the following steps:

inputting first sample advertisement data into the advertisement initial stage model;

Inquiring to obtain first sample statistical data characteristics corresponding to the first sample advertisement data from a pre-constructed data inquiry table through a data inquiry network of the advertisement initial stage model;

performing data coding processing on the first sample statistical data characteristic through a data coding network of the advertisement initial stage model to obtain first sample aggregate data and a first sample smoothing coefficient corresponding to a first sample advertisement;

performing smoothing processing on the first sample aggregate data based on the first sample smoothing coefficient through a data calculation layer of the advertisement initial stage model to obtain a first sample cost control factor of the first sample advertisement;

inputting the first sample cost control factor into a preset loss model, and carrying out loss calculation on the first sample cost control factor through the preset loss model to obtain a first loss result;

and correcting parameters in the advertisement initial stage model according to the first loss result to obtain a trained advertisement initial stage model.

9. The method of claim 1, wherein the conversion process of the advertisement to be controlled includes an advertisement maturation stage; wherein, in the advertisement maturation stage, the absolute value of the difference between the advertisement deviation of the advertisement to be controlled and the whole day deviation of the advertisement to be controlled is smaller than a preset threshold;

And when the advertisement is in the advertisement maturation stage, inquiring from a pre-constructed data inquiry table to obtain statistical data characteristics corresponding to the advertisement data, wherein the statistical data characteristics comprise:

querying global data corresponding to the advertisement attribute parameters from the data query table;

acquiring advertisement conversion data and advertisement aggregation data of the advertisement to be controlled;

inquiring similar advertisement data of the advertisement to be controlled from the data inquiry table based on the advertisement conversion data and the advertisement aggregation data;

respectively extracting features of the global data and the similar advertisement data to obtain global data features and similar advertisement data features;

and determining the global data characteristic and the similar advertisement data characteristic as statistical data characteristics corresponding to advertisement attribute parameters in the advertisement data.

10. The method of claim 9, wherein the data encoding the statistical data feature to obtain aggregate data and a smoothing coefficient corresponding to the advertisement to be controlled comprises:

performing data encoding processing on the global data features by adopting a third encoder to correspondingly obtain global aggregated data; the global aggregation data form aggregation data corresponding to the advertisement to be controlled;

Performing splicing processing on the global data features and the similar advertisement data features to obtain second splicing features;

and adopting a fourth encoder to carry out data encoding processing on the second splicing characteristic to obtain the smoothing coefficient.

11. The method of claim 10, wherein the fourth encoder is a third neural network having L layer data processing layers;

and performing data encoding processing on the second splicing characteristic by adopting a fourth encoder to obtain the smoothing coefficient, wherein the method comprises the following steps:

inputting the second stitching feature into the third neural network;

adopting each data processing layer in the third neural network to perform data processing on the second splicing characteristics to obtain a basic smoothing coefficient and a smoothing index;

determining the base smoothing coefficient and the smoothing index as the smoothing coefficient;

wherein L is an integer greater than 1.

12. The method according to any of claims 9 to 11, wherein determining a cost control factor for the advertisement to be controlled is achieved by an advertisement maturity stage model;

wherein the advertisement maturity stage model is trained by:

Inputting second sample advertisement data into the advertisement maturity stage model;

inquiring to obtain second sample statistical data characteristics corresponding to the second sample advertisement data from a pre-constructed data inquiry table through a data inquiry network of the advertisement maturity stage model;

performing data coding processing on the second sample statistical data characteristics through a data coding network of the advertisement maturity stage model to obtain second sample aggregate data and second sample smoothing coefficients corresponding to second sample advertisements;

performing smoothing processing on the second sample aggregate data based on the second sample smoothing coefficient through a data calculation layer of the advertisement maturity stage model to obtain a second sample cost control factor of the second sample advertisement;

inputting the second sample cost control factor into a preset loss model, and carrying out loss calculation on the second sample cost control factor through the preset loss model to obtain a second loss result;

and correcting parameters in the advertisement maturity stage model according to the second loss result to obtain a trained advertisement maturity stage model.

13. A cost control factor determination apparatus, the apparatus comprising:

The acquisition module is used for acquiring advertisement data of the advertisement to be controlled;

the query module is used for querying and obtaining statistical data characteristics corresponding to the advertisement data from a pre-constructed data query table; wherein the statistical data features are data features of advertisement information which are counted in advance for different advertisements;

the data coding module is used for carrying out data coding processing on the statistical data characteristics to obtain aggregate data and a smooth coefficient corresponding to the advertisement to be controlled; wherein the aggregate data is data for characterizing attribute information of the advertisement to be controlled;

and the smoothing processing module is used for carrying out smoothing processing on the aggregated data based on the smoothing coefficient to obtain the cost control factor of the advertisement to be controlled.

14. A cost control factor determination apparatus, comprising:

a memory for storing executable instructions; a processor for implementing the cost control factor determination method of any of claims 1 to 12 when executing executable instructions stored in the memory.

15. A computer readable storage medium, characterized in that executable instructions are stored for causing a processor to execute the executable instructions for implementing the cost control factor determination method of any of claims 1 to 12.