KR20230024190A - Method and system for providing advertising - Google Patents

Abstract

The above embodiments relate to an advertisement provision method and system. More specifically, provided is the advertisement provision method which may efficiently transmit advertisements by adjusting a budget ratio by reflecting advertisement performance. The advertisement provision method of the present invention includes the steps of: determining whether there is a learning model for predicting an advertisement budget for each advertisement; predicting an advertisement budget using a learning model for the advertisement; determining whether there is a pre-learned learning model for an advertiser to which the advertisement belongs; predicting an advertising budget using a learning model for the advertiser; collecting performance of advertisements for each advertisement target transmitted according to the predicted advertisement budget; obtaining unit price and price competitiveness for each advertisement target; adjusting the budget ratio; and transmitting an advertisement for each advertisement target.

Description

Advertisement provision method and system {METHOD AND SYSTEM FOR PROVIDING ADVERTISING}

The following embodiments relate to a method and system for providing advertisements, and more specifically, to a method for providing advertisements that efficiently transmit advertisements by adjusting a budget ratio by reflecting advertisement performance.

Advertising is an information delivery activity carried out for a fee through media with the goal of influencing the customer's attitude or behavior and ultimately purchasing his/her product by an advertiser who can be identified.

Advertising aims to have a customer perform an action of value to the advertiser's business, such as making a purchase, and the customer who interacts with the ad (e.g., clicks on a text ad or views a video ad) performs the action targeted by the ad. It's called a conversion. An action that is a target of an advertisement may be defined by an advertiser, and may include, for example, purchase or membership registration.

Recently, with the advent of various online advertising media such as search advertisements, banner advertisements, reward advertisements, and SNS advertisements, it is possible to provide various types of advertisements using various media. Depending on the type of online advertising medium, various characteristics such as advertisement execution cost, advertisement conversion rate, size of advertisement exposure, and characteristics of customers exposed to advertisement may be displayed. Accordingly, there is a demand for development of advertisement publishing technology to maximize return on investment (ROI) within a limited budget in consideration of the characteristics of various media, the size of the budget for advertisement, and the goal of advertisement.

Matters described in this background art section are prepared to enhance understanding of the background of the invention, and may include matters other than prior art already known to those skilled in the art to which this technique belongs.

In order to increase the efficiency of advertising in a limited budget, we will provide technology that automatically adjusts the ratio of the budget to be distributed for each advertising goal by analyzing the performance of the advertisement, and transmits the advertisement for each goal according to the adjusted budget ratio. can

By repeatedly performing the process of analyzing the advertisement performance according to the advertisement transmission, adjusting the budget ratio according to the performance analysis, and sending the advertisement according to the adjusted budget ratio, it is possible to provide a technology that adjusts the budget ratio by reflecting the advertisement analysis result in a timely manner. can

By using a deep learning-based learning model, the optimal advertising budget for each advertising medium is predicted and provided to the advertiser, thereby providing a technology that can increase the cost-effectiveness of advertising.

Problems to be solved by one embodiment are not limited to the problems mentioned above, and other problems not mentioned will be clearly understood by those skilled in the art from the description below.

One embodiment for achieving the above object is an advertisement providing method performed in an advertisement providing server, comprising: determining whether there is a learning model for predicting an advertisement budget for each advertisement; predicting an advertisement budget by using the learning model for the advertisement if there is a learning model for the advertisement; If there is no learning model for the advertisement, determining whether there is a pre-learned learning model for the advertiser to which the advertisement belongs; predicting an advertising budget using the learning model for the advertiser when there is a learning model for the advertiser; collecting performance of advertisements for each advertisement target transmitted according to the predicted advertisement budget; acquiring a unit price and price competitiveness for each advertisement target based on the performance of the advertisement for each advertisement target; adjusting a budget ratio by using price competitiveness for each advertisement target and a unit price for attracting customers; and transmitting an advertisement for each advertisement target based on the adjusted budget ratio.

An embodiment includes determining whether there is a pre-learned learning model for an advertiser group to which the advertiser belongs when there is no learning model for the advertiser; predicting an advertising budget by using the learning model for the advertiser group if there is a learning model for the advertiser group; and predicting an advertising budget based on a record of conversion of a target user when there is no learning model for the advertiser group.

In one embodiment, in the step of predicting the parameter using the learning model for the advertisement, a parameter user is determined using feature data of the target user, and an advertising budget is predicted according to the determined number of parameter users.

An embodiment includes obtaining a bottleneck for each advertisement target based on the performance of the advertisement for each advertisement target, and adjusting the budget ratio is based on price competitiveness for each advertisement target and bottleneck for each advertisement target. Thus, the budget ratio is adjusted.

One embodiment includes at least one processor; and a memory for storing instructions for instructing the at least one processor to perform at least one operation, wherein the at least one operation comprises advertising for each advertisement. determining whether there is a learning model that predicts a budget; predicting an advertisement budget by using the learning model for the advertisement if there is a learning model for the advertisement; If there is no learning model for the advertisement, determining whether there is a pre-learned learning model for the advertiser to which the advertisement belongs; predicting an advertising budget using the learning model for the advertiser when there is a learning model for the advertiser; collecting performance of advertisements for each advertisement target transmitted according to the predicted advertisement budget; acquiring a unit price and price competitiveness for each advertisement target based on the performance of the advertisement for each advertisement target; adjusting a budget ratio by using price competitiveness for each advertisement target and a unit price for attracting customers; and transmitting an advertisement for each advertisement target based on the adjusted budget ratio.

According to one embodiment as described above, by providing a technology for sending out advertisements by automatically distributing an appropriate budget to maximize efficiency within a limited budget according to advertisement performance, analyzing advertisement performance and setting a budget You can reduce the marketing cost required to hire experts who establish marketing strategies such as

It is possible to increase the timeliness of marketing strategy establishment by repeatedly performing advertisement performance analysis and budget ratio setting.

By using a deep learning-based learning model, the optimal advertising budget for each advertising medium is predicted and provided to the advertiser, thereby increasing the cost-effectiveness of advertising.

The effects of one embodiment are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description below.

1 is a configuration diagram of an advertisement providing system according to an exemplary embodiment.
2 is a diagram for explaining an operation of an advertisement providing server performing an advertisement providing method according to an exemplary embodiment.
3 is a diagram for explaining an advertisement transmission step according to an exemplary embodiment.
4 is a diagram for explaining advertisement performance according to an exemplary embodiment.
5 is a configuration diagram of an advertisement providing server according to an exemplary embodiment.
6 is a flowchart of an advertisement providing method according to the first embodiment.
7 is a flowchart of an advertisement providing method according to a second embodiment.
8 and 9 are diagrams for explaining a marketing funnel according to an exemplary embodiment.
10 is a diagram illustrating an advertising budget prediction device and a training device according to an embodiment.
11 is a flowchart of an advertisement providing method according to a third embodiment.
12 is a diagram illustrating a training process of a learning model according to an embodiment.
13 is a configuration diagram of a user terminal according to an embodiment.

Advantages and features of the present invention, and methods of achieving them, will become clear with reference to the detailed description of the following embodiments taken in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but may be implemented in various different forms, only these embodiments are intended to complete the disclosure of the present invention, and are common in the art to which the present invention belongs. It is provided to fully inform the person skilled in the art of the scope of the invention, and the invention is only defined by the scope of the claims.

Terminology used herein is for describing the embodiments and is not intended to limit the present invention. In this specification, singular forms also include plural forms unless specifically stated otherwise in a phrase. As used herein, "comprises" and/or "comprising" does not exclude the presence or addition of one or more other elements other than the recited elements. Like reference numerals throughout the specification refer to like elements, and “and/or” includes each and every combination of one or more of the recited elements. Although "first", "second", etc. are used to describe various components, these components are not limited by these terms, of course. These terms are only used to distinguish one component from another. Accordingly, it goes without saying that the first element mentioned below may also be the second element within the technical spirit of the present invention.

Unless otherwise defined, all terms (including technical and scientific terms) used in this specification may be used with meanings commonly understood by those skilled in the art to which the present invention belongs. In addition, terms defined in commonly used dictionaries are not interpreted ideally or excessively unless explicitly specifically defined.

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings.

1 is a configuration diagram of an advertisement providing system 10 according to an embodiment.

Referring to FIG. 1 , an advertisement providing system 10 according to an embodiment may include an advertisement providing server 100 , an advertiser terminal 200 and a user terminal 300 .

Communication between various entities included in the advertisement providing system 10 may be performed through a wired/wireless network. Wired/wireless networks may use standard communication technologies and/or protocols.

The advertisement providing server 100, the advertiser terminal 200, and the user terminal 300 in the advertisement providing system 10 include, for example, a computer, an ultra mobile PC (UMPC), a workstation, a net-book, and a PDA. (Personal Digital Assistants), portable computers, web tablets, wireless phones, mobile phones, smart phones, portable multimedia players (PMPs), among other electronic devices. As one, it may include all electronic devices capable of installing and executing applications related to an embodiment. The electronic device may perform overall service operations, such as composition of a service screen, data input, data transmission/reception, and data storage, under the control of an application.

The advertiser terminal 200 may receive advertisement-related information such as advertisement targeting information, advertisement medium, and advertisement material from an advertiser and transmit the received advertisement-related information to the advertisement providing server 100 .

The user terminal 300 may receive an advertisement from the advertisement providing server 100 and output the corresponding advertisement to the user.

The advertisement providing server 100 may predict an advertisement budget, adjust a ratio of the advertisement budget, and publish (transmit to the user terminal 300) an advertisement according to the predicted/adjusted advertisement budget.

2 is a diagram for explaining an operation of an advertisement providing server 100 performing an advertisement providing method according to an exemplary embodiment.

Referring to FIG. 2 , the advertisement providing server 100 according to an exemplary embodiment distributes advertisements for each advertisement target while adjusting a budget ratio between a plurality of advertisement targets according to the advertisement budget of the advertiser terminal 200. may correspond to a server transmitting through The advertisement providing server 100 may include one or more processors for performing an advertisement providing method, and may include a memory for storing data related to the advertisement providing method. Hereinafter, 'advertising target' may be briefly referred to as 'goal'.

An operating method of the advertisement providing server 100 according to an embodiment includes the steps of sending out an advertisement corresponding to the advertiser terminal 200, collecting the results according to the sending of the advertisement, and selecting between a plurality of advertisement targets based on the results. Adjusting the budget ratio and transmitting an advertisement according to the adjusted budget ratio.

According to an embodiment, the operation method of the advertisement providing server 100 may further include receiving information for publishing advertisements from the advertiser terminal 200 . Information for publishing advertisements is information necessary for publishing advertisements corresponding to the advertiser terminal 200, and may include, for example, a total budget for publishing advertisements, advertising materials, and advertisement goals. The advertiser terminal 200 may input information for publishing advertisements into the advertisement providing server 100. For example, through an interface provided by the advertisement providing server 100, the total budget for publishing advertisements, advertising materials, and advertisements You can enter goals.

The advertisement target is an effect to be achieved through advertisement transmission, and may include, for example, increased brand exposure, increased website visits, increased number of members, increased purchases, and the like. According to an embodiment, the advertisement providing server 100 may provide a plurality of predetermined advertisement targets to the advertiser terminal 200 as options through an interface, and the advertiser terminal 200 may provide at least some of the provided advertisement targets. You can set an advertising goal by selecting . Advertisement targets provided to the advertiser terminal 200 from the advertisement providing server 100 may be determined differently for each advertiser by reflecting characteristics of the advertiser, such as type of advertiser and size of budget. According to an embodiment, the advertiser terminal 200 may directly input an advertisement target.

The advertiser terminal 200 may input advertisement material(s) according to a set advertisement goal. Advertising materials are content posted on web pages, apps, or other advertising media that customers can access, and can be composed of various types of content, such as text, images, audio, video, links, and combinations thereof. In addition, advertisement materials may be created in various types suitable for the media in which they are posted. For example, creatives can be generated in various types, such as banner ads displayed on mobile messengers, banner ads displayed on websites, and messages in the form of messages transmitted through mobile messengers, and can be configured to include various contents. there is.

Advertising materials may be provided by the advertiser terminal 200 for each advertising goal, and may be generated to include various types and contents corresponding to a plurality of advertising goals. For example, materials for branding in response to the goal of increasing brand exposure (for example, content including text and images that can draw attention to the brand name), and materials for visits in response to the goal of increasing website visits. (For example, content that includes a link that leads to a website when clicked), materials for membership registration in response to the goal of increasing the number of members (for example, including phrases such as “Let’s sign up now and get 1,000 won”) Event-type content) and material for purchase (for example, event-type content such as providing a short-term discount coupon) may be input in response to a goal of increasing purchase.

The advertising goal set in correspondence with the advertiser terminal 200 may include a goal capable of collecting corresponding results. For example, the number of times and/or the number of customers exposed to an advertisement for a branding creative in response to a goal of increasing brand exposure, the number of clicks on an ad for a creative for visitation in response to a goal of increasing website visits, members The number of member registrations corresponding to the goal of increasing the number, the number of purchases corresponding to the goal of increasing the number of purchases, and/or the increase in sales according to the purchase may be collected as performance for each goal. The performance of the advertisement may be collected by counting the number of times the advertisement is sent and/or the number of user terminals to which the advertisement is sent from the advertisement providing server 100 or by various conversion tracking methods. According to an embodiment, the advertisement providing server 100 may apply a tracking method for each advertisement target set in correspondence with the advertiser terminal 200 to collect the corresponding performance, and display the result of the collected performance to the advertiser terminal 200. ) can be provided.

The advertiser terminal 200 may input a total budget for advertisement transmission, and the advertisement providing server 100 may execute advertisements for each target within the total budget. The cost of publishing an advertisement in the advertisement providing server 100 may be determined according to a predetermined criterion in consideration of the advertisement material, the medium in which the advertisement is displayed, the number of times the advertisement is transmitted, and the number of clicks on the advertisement. The advertisement providing server 100 may publish an advertisement in consideration of a total budget of the advertiser terminal 200 and a cost for publishing an advertisement corresponding to the advertiser terminal 200 .

When a plurality of advertisement targets are set to correspond to the advertiser terminal 200, a budget ratio among the plurality of targets may be determined, and a budget for advertisement execution for each target may be distributed according to the budget ratio. For example, if the total budget of the advertiser terminal 200 is 10 million won and the budget ratio between the first target and the second target is 1:3, the budget for advertising execution of the first target is 1/1 of 10 million won. The budget for 2.5 million won for 4 people and 7.5 million won, which is 3/4 of 10 million won for the second target, can be distributed for each goal. In this case, the advertisement providing server 100 may publish an advertisement corresponding to the first target within a budget of 2.5 million won and publish an advertisement corresponding to the second target within a budget of 7.5 million won. The advertisement providing server 100 may obtain a budget ratio between a plurality of targets in order to publish an advertisement according to a budget for each target.

According to an embodiment, the initial budget ratio before advertisement publishing starts may be determined as a predetermined ratio corresponding to set targets or may be determined as a ratio set by the advertiser terminal 200 . When determined according to a predetermined ratio, the same ratio may be determined corresponding to each target, or may be determined by referring to a budget ratio based on a history of advertisement execution if there is a history of advertisement execution in the past, or a budget ratio corresponding to other advertisers. may be determined by reference. For example, when targets of increased exposure, increased visits, and increased purchases are set in correspondence to the advertiser terminal 200, the advertisement ratio between the targets is set to 1:1:1, or the industry corresponding to the advertiser terminal 200. It is set by following the average, or the average of the budget ratios of some advertisers in the industry corresponding to the advertiser terminal 200 (eg, advertisers with the top 30% of annual advertising volume, advertisers with high relevance of products and / or services) It can be set to follow.

According to an embodiment, the advertisement providing server 100 may execute an advertisement within a target-specific budget according to the budget ratio, adjust the budget ratio by reflecting performance collected as a result of advertisement execution, and adjust the budget ratio. Advertising can be executed within the budget for each target adjusted according to Specific steps for adjusting the budget ratio will be described in detail below in FIG. 6 .

3 is a diagram for explaining an advertisement transmission step according to an exemplary embodiment.

The step of sending an advertisement according to an embodiment may correspond to a step of sending an advertisement for each target according to a budget ratio among a plurality of advertisement targets in correspondence to the advertiser terminal 200 .

Referring to FIG. 3 , the advertisement providing server 100 receiving information for advertisement transmission, such as total budget, advertisement materials, and advertisement goals, from the advertiser terminal 200 displays advertisements corresponding to the advertisement goals according to the received information. can be sent The advertisement to be transmitted may include various types of advertising media 211, 212, and 213 corresponding to each of the advertising targets set in correspondence with the advertiser terminal 200, and may include at least one advertising medium in response to a specific advertising target. Composed advertisements can be sent out. For example, the advertising material 211 corresponding to the advertisement goal 1 may include a banner advertisement displayed on a mobile messenger, a banner advertisement displayed on a website, and the like, and may include a plurality of banner advertisements having different content configurations. may be The type and/or content of an advertisement transmitted does not necessarily vary according to a corresponding advertisement target.

The advertiser terminal 200 may select advertisement targets (advertisement targets 1 to n), set advertisement media 211 to 213 corresponding to each advertisement target, and set a total budget for advertisement publication. The advertisement providing server 100 distributes the total budget for each advertisement target according to the budget ratio between the advertisement targets, and advertises for a specific number of times and/or period through an advertisement medium set for each advertisement target within the distributed budget for each target. can be sent.

Referring back to FIG. 2 , the step of collecting the performance of the advertisement sent out according to an embodiment may correspond to the step of collecting the performance of the advertisement sent out according to the budget ratio between the plurality of advertisement goals. As described above, the advertising target set in correspondence with the advertiser terminal 200 may include a target capable of collecting the corresponding performance, and the number of times the advertisement is transmitted from the advertisement providing server 100 or the user's terminal from which the advertisement is transmitted. , etc., or the performance of advertisements can be collected for each advertisement goal according to various conversion tracking methods.

An example of collected advertising performance may refer to FIG. 4 . 4 is a diagram illustrating an example of collected results when goals of increasing brand exposure, increasing website visits, increasing the number of members, and increasing purchases are set in response to advertisers. Hereinafter, the goal of increasing brand exposure may be briefly referred to as exposure, the goal of increasing website visits as visitation, the goal of increasing the number of members as subscription, and the goal of increasing purchase as purchase.

Referring to FIG. 4 , advertisements may be repeatedly published over a plurality of rounds, and the total budget may be set corresponding to each round or all rounds. When the total budget is set corresponding to all rounds, the budget corresponding to each round may be set by uniformly dividing the total budget for each round or by dividing the total budget by assigning a weight to a specific round.

The performance of the advertisement corresponding to the goal of increasing brand exposure 310 is the number of impressions of the advertisement and the number of customers exposed to the advertisement, and the performance of the advertisement corresponding to the goal of increasing website visits 320 is the number of visits to the website of the advertiser The performance 330 of the advertisement corresponding to the target of increasing the number of times and the number of members may include the number of newly registered members and the performance of the advertisement 340 corresponding to the target of increasing purchase may include the number of purchases and the purchase amount. The performance for each advertisement goal collected in correspondence with the advertisement goal may be measured by a conversion tracking method or the like, or may be estimated based on a difference before and after the advertisement. For example, the number of visits to a website may be derived by measuring the number of visits to a website that occurred as a result of a conversion caused by an advertisement according to the tracking method of the advertisement, or by comparing the number of visits before and after the advertisement to visit the website due to the advertisement. It may also be derived by estimating the number of visits.

According to an embodiment, the advertisement providing server 100 may obtain a unit price per target based on the collected advertisement performance. The unit price for attracting customers means an advertising cost required for attracting customers to achieve an advertising goal. For example, the unit price for attracting customers for each goal may be obtained by dividing the budget for each goal by the number of customers included in the performance for each goal. The budget for each target may be obtained by multiplying the total budget by a budget ratio for each target. For example, CPM (Cost Per Mille, Cost Per Thousand Impression), which is the cost of advertising per 1000 impressions, Cost Per Click (CPC), which is the cost of advertising per 1,000 impressions, Cost Per Install (CPI), A cost per registration (CPR), which is an advertising cost per sign-up, and a cost per action (CPA), which is an advertising cost per conversion (eg, purchase), may correspond to the unit price for attracting customers for each target.

For example, in the advertising performance on November 1, 2021 in FIG. 4, assuming that the budget ratio between targets is 1:1:1:1, the CPM is 25000/1000 = 25 won, and the CPC is 25000/1000 = 25 won, CPR can be obtained at 25000/2 = 12500 won, and CPA can be obtained at 25000/1 = 25000 won.

According to an embodiment, the advertisement providing server 100 may obtain a retention rate or a churn rate for each target based on the collected advertisement performance. When a plurality of advertising targets set in response to an advertiser correspond to a series of predetermined sequences, the retention rate corresponding to a specific target is the target compared to the performance of advertising corresponding to the target in the previous step. It means the ratio of the advertisement performance corresponding to, and can be obtained as the performance of the corresponding stage/performance of the previous stage. The exit rate is the opposite concept of the retention rate and can be obtained as (1 - retention rate).

For example, assuming that the goals of exposure, visit, subscription, and purchase correspond to a series of sequences having the order of exposure, visit, subscription, and purchase, visits based on the number of times in the advertising performance on November 1, 2021 in FIG. The retention rate corresponding to 1,000/1000,000 = 0.001, the retention rate corresponding to subscription is 2/1,000 = 0.002, and the retention rate corresponding to purchase is 1/2 = 0.5.

Referring back to FIG. 2 , adjusting a budget ratio between a plurality of advertisement targets according to an embodiment may correspond to adjusting a budget ratio between a plurality of advertisement targets by reflecting collected performance. . The advertisement providing server 100 may analyze the performance of the advertisement for each target and adjust the ratio of the advertisement for each target to be transmitted by adjusting the budget ratio between the targets in order to increase the efficiency of the advertisement in a limited budget. The advertisement providing server 100 may provide an efficient advertisement publishing technique within a limited budget by performing an advertisement providing method of repeatedly publishing advertisements by adjusting a budget ratio according to advertisement performance. The advertiser terminal 200 can efficiently publish advertisements for each target within a limited budget without analyzing advertisement performance or resetting a budget ratio by using the advertisement providing server 100 that automatically publishes advertisements. The step of adjusting the budget ratio by reflecting the performance of the advertisement will be described in detail below in FIG. 6 .

5 is a configuration diagram of an advertisement providing server 100 according to an embodiment.

Referring to FIG. 5 , the advertisement providing server 100 according to an embodiment includes a communication unit 110, an input unit 120, an output unit 130, a memory 140, a power supply unit 150, and a control unit 160. can include

The configurations shown in FIG. 5 are exemplary diagrams for implementing the embodiments of the present invention, and appropriate hardware/software configurations that are apparent to those skilled in the art may be additionally included in the advertisement providing server 100.

The communication unit 110 may communicate with an external device through various communication methods. For example, the communication unit 110 may communicate with the advertiser server 200 and the user terminal 300 to transmit and receive various data.

The input unit 120 may receive various inputs from an operator who is a user of the advertisement providing server 100 and transmit them to the control unit 160 . In particular, the input unit 120 may include a touch sensor, a (digital) pen sensor, a pressure sensor, a key, or a microphone. The touch sensor may use, for example, at least one of a capacitive type, a pressure-sensitive type, an infrared type, or an ultrasonic type. The (digital) pen sensor may be, for example, a part of the touch panel or may include a separate recognition sheet. Keys may include, for example, physical buttons, optical keys, or keypads. The microphone is a configuration for receiving operator voice and may be provided inside the advertisement providing server 100, but this is only one embodiment, and is provided outside the advertisement providing server 100 to communicate with the advertisement providing server 100. can be electrically connected.

The output unit 130 may provide various screens.

The memory 140 may store commands or data related to at least one other component of the advertisement providing server 100 . In particular, the memory 140 may be implemented as a non-volatile memory, a volatile memory, a flash-memory, a hard disk drive (HDD), or a solid state drive (SSD). The memory 140 is accessed by the control unit 160, and reading/writing/modification/deleting/updating of data by the control unit 160 can be performed. In the present invention, the term memory refers to a memory 140, a ROM (not shown), a RAM (not shown) in the controller 160, or a memory card (not shown) mounted on the advertisement providing server 100 (eg, micro SD card, memory stick). In addition, the memory 140 may store programs and data for configuring various screens to be displayed on the display area of the output unit 130 .

The power supply unit 150 receives external power and internal power under the control of the control unit 160 and supplies power necessary for the operation of each component.

The control unit 160 is electrically connected to the communication unit 110, the input unit 120, the output unit 130, the memory 140, and the power supply unit 150, and controls overall operations and functions of the advertisement providing server 100. can do. In particular, the controller 160 may provide functions described later using various modules stored in the memory 140 .

Various operations of the advertisement providing server 100 described below may be performed under the control of the control unit 160 .

6 is a flowchart of an advertisement providing method according to the first embodiment, and is a diagram for explaining an advertisement providing method in which an advertisement is transmitted by adjusting a budget ratio.

Referring to FIG. 6 , the advertisement providing method according to the first embodiment includes obtaining a budget ratio between a plurality of advertisement targets in response to a target advertiser (S100), and performance of advertisements for each target transmitted according to the budget ratio. Collecting (S110), acquiring a unit price for each target based on the advertisement performance for each target (S120), acquiring price competitiveness for each target based on the unit price for each target (S130), Adjusting a budget ratio based on price competitiveness (S140) and transmitting an advertisement for each target based on the adjusted budget ratio (S150) may be included.

According to an embodiment, the step of acquiring price competitiveness for each target ( S130 ) may include acquiring price competitiveness for each target based on the reference value of the unit price for each target and the unit price for each target. The reference value of the unit price for attracting customers for each target may correspond to a value determined by referring to the unit price for attracting customers of at least one other advertiser for each advertising target. It may include an average of unit prices for attracting customers corresponding to the target of the advertiser. Price competitiveness may be determined as a relative value of the unit price of customer acquisition corresponding to the target advertiser compared to the unit price of customer acquisition of other advertisers, and may be determined as a larger value when the unit price of customer acquisition for each target of the target advertiser is smaller than the reference value of the unit price of customer attraction.

According to an embodiment, the step of adjusting the budget ratio based on the price competitiveness per target (S140) is to adjust the budget ratio by setting the ratio of price competitiveness per target among the plurality of advertisement targets as the budget ratio. steps may be included. The ratio of price competitiveness for each target among the plurality of advertising targets is the ratio of the price competitiveness of the target to the total price competitiveness of the plurality of targets. For example, the price competitiveness corresponding to the target corresponds to each of the plurality of targets. It can be obtained by dividing by the sum of the price competitiveness.

For example, when a plurality of targets corresponding to a target advertiser are impressions and visits, a method of adjusting a budget ratio based on price competitiveness will be described.

As a result of running ads for each advertising goal of impressions and visits at a budget ratio of x:y, the number of impressions, which is the advertising performance corresponding to the advertising goal of impressions, is a, and the number of visits, which is the advertising performance corresponding to the advertising goal of visits, is b. In this case, the budget ratio may be adjusted based on price competitiveness.

In the case of an advertisement target for impressions, the unit price for attracting customers (CPM) may be obtained by dividing the advertisement budget for each target (total budget * x) by multiplying the total budget by the advertisement rate (x) by the number of impressions (a). In the case of an advertising goal of a visit, the customer acquisition unit price (CPC) may be obtained by dividing the advertising budget for each target (total budget * y) by multiplying the total budget by the advertising rate (y) by the number of impressions (b).

Price competitiveness by advertising goal is to determine whether the target advertiser's unit price for each advertising goal is competitive in terms of price compared to at least one other advertiser's unit price for each advertising goal. The lower the value, the higher the value may be determined. When the average of the CPM values of other advertisers is CPM' and the average of the CPC values of other advertisers is CPC', the price competitiveness (pcm) of the impression target is CPM'/CPM, and the price competitiveness (pcc) of the visit target is CPC' /CPC can be determined. Here, the other advertisers may mean all advertisers, advertisers in the same industry as the target advertiser, or advertisers with a higher portion of sales. Although an example of deriving price competitiveness by dividing the average customer unit price of other advertisers by the customer unit price of the target advertiser is shown, price competitiveness is derived by multiplying the customer unit price of other advertisers and/or the customer unit price of the target advertiser by a parameter. Various methods may be applied to derive the price competitiveness of the target advertiser's unit price compared to that of other advertisers.

A ratio of price competitiveness obtained for each target may be determined as an adjusted budget ratio. When the adjusted budget ratio is x':y', x', the budget ratio of the exposure target, can be determined as pcm/(pcm+pcc), which is the ratio of pcm, the price competitiveness of the exposure target, and the budget ratio of the visit target. The value y' may be determined as pcc/(pcm+pcc), which is a ratio of pcc, which is the price competitiveness of the visit target. An example of calculating the ratio of price competitiveness by dividing the unit price for each target by the sum of the unit prices for each target is explained, but in order to derive the ratio of price competitiveness for each target Various methods may be applied.

The method of adjusting the budget ratio based on price competitiveness may correspond to a method of adjusting the budget ratio by increasing the proportion of advertisement targets whose unit price is lower than the average value of other advertisers and reducing the proportion of the value exceeding the average value of other advertisers. For example, as a result of collecting the performance of the target advertiser running ads with a budget ratio of 50% impressions and 50% visits, the CPM is 100 won, the CPC is 200 won, and the average CPM of other advertisers is 150 won and the average CPC is 100 won. is 190 won, the CPM is executed at a price 33% (1 - 100/150) lower than the market and the CPC is executed at a higher price by 5.2% (1 - 200/190). The budget ratio may be adjusted in such a way as to decrease the budget ratio of the target and increase the budget ratio of the advertising target of low-cost exposure. Through budget adjustment, the ratio of advertisements corresponding to a plurality of targets may be managed so as to increase efficiency within a limited budget by increasing the proportion of advertisements that can achieve advertisement efficiency at a low price.

Based on the adjusted budget for each advertising goal according to the adjusted budget ratio, advertisements can be re-sent for each goal (S150), and the performance of the advertisement transmission is collected (S110) to acquire unit price and price competitiveness (S120, S130). ), the process of adjusting the budget ratio (S140) and sending out the advertisement (S150) may be repeated.

According to an embodiment, after the advertisement is sent (S150), it may be determined whether the termination condition is satisfied (S160), and if the termination condition is satisfied, the process of the advertisement providing method may be terminated, and In this case, the process of steps S110 to S150 may be repeated. An end condition according to an embodiment may be set to the number of iterations, total budget, achievement or not. For example, if the number of advertisements is set to 10, it may be terminated when the condition for 10 advertisements is satisfied, and if the total budget for advertisement execution is set to a specific amount, it may be terminated when the set amount is exhausted, , If a goal to be achieved through advertisement publication is set, it may end when the goal is achieved.

An advertisement providing method according to another embodiment may further include obtaining a residual rate for each target based on the performance of an advertisement for each target, and acquiring a bottleneck for each target based on the residual rate for each target. In this case, adjusting the budget ratio according to an embodiment ( S140 ) may include adjusting the budget ratio based on price competitiveness for each target and bottleneck for each target.

According to another embodiment, the plurality of goals corresponding to the target advertiser may correspond to a series of sequences having a predetermined order, and the step of obtaining the retention rate for each goal corresponds to each of the plurality of goals, The method may further include obtaining a retention rate of the corresponding target by dividing the number of customers of the target by the number of customers of the previous target.

According to another embodiment, the acquiring of the bottleneck for each target may include obtaining the bottleneck for each target based on the reference value of the residual rate for each target and the residual rate for each target. The reference value of the retention rate for each target may correspond to a value determined by referring to the retention rate of at least one other advertiser for each advertising target. For example, the average retention rate corresponding to the corresponding target of all advertisers, The average retention rate corresponding to the target of the advertiser may be included. The bottleneck degree may be determined as a relative value of the retention rate corresponding to the target advertiser compared to that of other advertisers, and may be determined as a larger value when the target advertiser's retention rate for each target is smaller than the reference value.

According to another embodiment, the step of adjusting the budget ratio based on the price competitiveness and bottleneck for each target sets an average of the price competitiveness ratio and bottleneck ratio for each target among a plurality of advertising targets as the budget ratio. By doing so, the step of adjusting the budget ratio may be included. The bottleneck ratio for each target among the plurality of advertising targets is the ratio of the bottleneck of the corresponding target to the overall bottleneck of the plurality of targets. For example, the bottleneck corresponding to the target corresponds to each of the plurality of targets. It can be obtained by dividing by the sum of the bottlenecks. The average of the price competitiveness ratio and the bottleneck ratio for each target among the plurality of advertising targets may include an arithmetic mean, geometric mean, harmonic average, etc., and different weights are applied to the price competitiveness ratio and the bottleneck ratio. You can also calculate an average of two. Hereinafter, a case in which a budget ratio is adjusted based on a harmonized average of a price competitiveness ratio and a bottleneck ratio for each target among a plurality of advertisement targets will be described as an example.

For example, when a plurality of targets corresponding to a target advertiser are exposure, visit, and conversion, a method of adjusting a budget ratio based on price competitiveness and bottleneck will be described.

As a result of running ads for each advertising goal of impressions, visits, and conversions at a budget ratio of x:y:z, the number of impressions, which is the advertising performance corresponding to the advertising goal of impressions, is a, and the advertising performance corresponding to the advertising goal of visits, visits. If the number is b and the number of conversions, which is the advertising performance corresponding to the advertising goal of conversion, is c, the budget ratio may be adjusted based on price competitiveness and bottleneck.

The step of obtaining the unit price, price competitiveness, and rate of price competitiveness based on the advertising performance is the same as the above step.

When the advertising goals corresponding to the target advertiser form a sequence of exposure, visit, and conversion, in the case of exposure, which is the first advertising goal, the retention rate cannot be obtained because the previous goal does not exist. In the case of visits, which are the second advertising goal, a value (b/a) obtained by dividing the number of visitors (b) by the number of visitors (a) of exposures, which is the previous goal, may be obtained as a retention rate (RC) of the advertising goal. In the case of conversion, which is the third advertising target, a value (c/b) obtained by dividing the number of conversion customers (c) by the previous target, the number of customers receiving exposure (b), may be obtained as the retention rate (RA) of the advertising target. In FIG. 6, the case where the number of customers of the previous target is the number of customers of the immediately previous stage is described as an example. It can be calculated as the sum of the number of customers of all previous targets. For example, the retention rate of a conversion can be calculated as c/(a+b) rather than c/b.

The bottleneck by advertising goal is to determine whether the retention rate by advertising goal of a target advertiser is lower than that of at least one other advertiser by advertising goal. can be determined When the average retention rate of visits from other advertisers is RC' and the average retention rate of conversions from other advertisers is RA', the bottleneck of the visit goal (brc) is RC'/RC, and the bottleneck of the conversion goal (bra ) may be determined as RA'/RA. Here, the other advertisers may mean all advertisers, advertisers in the same industry as the target advertiser, or advertisers with a higher portion of sales. Although an example of deriving the bottleneck by dividing the average retention rate of other advertisers by the retention rate of the target advertiser is shown, the bottleneck is derived by multiplying the retention rate of other advertisers and / or the retention rate of the target advertiser by a parameter Various methods may be applied to derive the bottleneck of the target advertiser's retention rate compared to that of other advertisers.

An adjustment coefficient for each target may be obtained as a harmonized average of the ratio of price competitiveness and the ratio of bottleneck obtained for each target, and the ratio of the adjustment coefficient may be determined as the adjusted budget ratio. In the case of exposure, which is the first step, since the residual rate and bottleneck cannot be obtained, the adjustment coefficient (rm) of exposure can be obtained as the ratio (pm) of price competitiveness. The adjustment coefficient of visits (rc) can be determined as the harmonic mean (2pcqc/(pc + qc)) of the ratio of price competitiveness (pc) and the ratio of bottlenecks (qc), and the adjustment coefficient of conversion (ra) is price competitiveness. It can be determined as the harmonic average (2paqa/(pa + qa)) of the ratio of (pa) and the ratio of bottlenecks (qa).

When the adjusted budget ratio is x':y':z', the budget ratio x' of the exposure target can be determined as rm/(rm+rc+ra), which is the ratio of the adjustment coefficient (rm) of the exposure target , The budget ratio of the visit target, y', can be determined as the ratio of the adjustment coefficient (rc) of the visit target, rc/(rm+rc+ra), and the budget ratio of the conversion target, z', is the adjustment coefficient of the conversion target ( It can be determined as ra / (rm + rc + ra), which is the ratio of ra). Although an example of calculating the ratio of the adjustment factor by dividing the adjustment factor of the corresponding target by the sum of the adjustment coefficients for each target is explained, in order to derive the ratio of the adjustment factor for each target, such as putting a weight on the adjustment coefficient for each target, Various methods may be applied.

FIG. 7 is a flow chart of an advertisement providing method according to a second embodiment, and is a diagram for explaining an advertisement providing method in which advertisements are repeatedly transmitted by adjusting a budget ratio to solve a bottleneck stage in a marketing funnel. According to another embodiment, the bottleneck stage may refer to any one advertising goal that limits the overall performance of an advertisement among a plurality of advertising goals included in a marketing funnel.

Referring to FIG. 7 , the method for providing advertisements according to the second embodiment includes the steps of collecting performance of advertisements for each target sent out according to a budget ratio between a plurality of advertisement targets in response to a target advertiser (S200); Acquiring unit price for each target and retention rate for each target based on the performance of the advertisement (S210), and based on the unit price for each target, price competitiveness for each target is acquired, and based on the retention rate for each target, price competitiveness for each target is acquired Obtaining a bottleneck diagram (S220), identifying a bottleneck stage in the marketing funnel corresponding to the performance of advertisements for each goal based on at least one of price competitiveness for each goal and bottleneck for each goal (S230), It may include adjusting a budget ratio to resolve (S240), and transmitting an advertisement for each target based on the adjusted budget ratio (S250).

'Marketing Funnel' refers to a series of processes from potential customers to become aware of a product or service, visit a website related to the product or service, and convert, such as membership registration and purchase, by stages. It shows the number of customers who have reached the stage in a funnel shape.

Referring to FIG. 8 , the number of customers who have reached each stage 810 to 850 can be visualized and displayed for each stage. The marketing funnel can be used to establish a marketing strategy to increase the conversion rate by identifying at which stage a customer is leaving and improving the cause of the customer's departure.

Referring to FIG. 9 , the marketing funnel may consist of an ATTENTION step, an INTEREST step, a SEARCH step, an ACTION step, and a SHARE step. Each step is described. The ATTENTION step is a step in which the user needs to recognize the advertisement's brand or company, and a marketing strategy to attract the user to the advertisement-related site. In the INTEREST and SEARCH stages, when a user enters an advertisement-related site, a marketing strategy is required to induce the user to have interest and interest in advertisement-related products or services and to actively search for advertisement-related products or services. am. ACTION phase, SHARE phase Based on the judgment that the user is interested in advertising-related products or services, a marketing strategy is required to encourage repurchase by encouraging users to directly purchase products or services and share their purchase experience.

In an advertisement providing method according to another exemplary embodiment, an advertisement target to increase a budget ratio may be identified as a bottleneck step in a marketing funnel based on at least one of price competitiveness for each target and bottleneck for each target. For example, when the price competitiveness of a specific target is higher than that of other targets, the corresponding target may be identified as a bottleneck stage, and a budget ratio for advertisement transmission of the corresponding target may be increased. Alternatively, when the bottleneck of a specific target is higher than that of other targets, the corresponding target may be identified as a bottleneck stage, and a budget ratio for advertisement transmission of the target may be increased.

The budget ratio according to an embodiment may be adjusted according to the method described above. According to the budget rate adjustment method, advertisements for each goal may be executed in a direction in which the bottleneck stage is resolved by increasing the budget rate of the target identified as the bottleneck stage and increasing the proportion of advertisements for the target identified as the bottleneck stage.

10 is a diagram illustrating an advertisement budget prediction device 102 and a training device 101 according to an embodiment.

An embodiment relates to a method and apparatus for predicting an advertising budget using a learning model 103 and a method and apparatus for training the learning model 103, and trains the learning model 103 for predicting an advertising budget using the characteristics of a parameter user. and a method and apparatus for predicting an advertising budget to be used for an actual advertisement using a trained learning model (103).

The parameter user refers to a user with a high probability of converting to a specific advertisement. When a conversion occurs, the target user accesses the specific advertisement, and the event related to the advertiser's sales on the advertiser's homepage of the specific advertisement (e.g., membership registration, purchase, app installation, app purchase, shopping cart, talk participation, talk signup) etc.)

Therefore, even if the target user accesses a specific advertisement, it cannot be regarded as a user whose conversion has occurred again without performing an event related to sales. That is, the conversion event refers to an event related to the advertiser's sales.

Referring to FIG. 10 , a training device 101 trains a learning model 103 using learning data 104, and an advertising budget prediction device 102 uses the trained learning model 103 to train a parameter user's An advertising budget is predicted for each advertisement according to feature data (feature) 105 .

For example, as shown in FIG. 10, when there are three advertisements A, B, and C, a learning model 103 is created for each advertisement, and the advertisement providing device 100 sets each learning model 103 for all advertisements. Ad budgets 106, 107, and 108 for each advertisement are predicted by using.

The training data 104 is composed of seed user and non-seed user feature data. Seed users refer to users who convert to advertisements during a certain period of time, and non-seed users mean users to whom conversion to advertisements does not occur during a certain period of time. Accordingly, the seed user and the non-seed user may be different for each advertisement, and even the same user may be a seed user or a non-seed user depending on the advertisement.

Specifically, the characteristic data of seed users whose conversions actually occurred in the learning data 104 is used as the correct answer label, and the training device 101 determines that the learning model 103 is composed of seed users and non-seed users. Based on the feature data, the learning model 103 is trained to predict a seed user among training users as a parameter user and to predict an advertising budget according to the number of predicted parameter users. In this case, since non-seed users are more numerous than seed users, they may be arbitrarily identified through negative sampling.

The feature data 105 may refer to log records of actions (clicks, inputs, etc.) performed in relation to advertisements online by seed users, non-seed users, and target users during a certain period of time. For example, feature data may be obtained through SDK & Pixel, Mobile App Tracking (MAT), Mobile Measurement Partner (MMP), and the like. Accordingly, seed users or non-seed users may be previously determined for each advertisement based on previously collected characteristic data of users.

Characteristic data may be different for each advertisement, and include an event for conversion. For example, a click on an advertisement may not be a conversion event, but may be a type of feature data. In addition, actions such as clicking a specific banner after accessing the advertiser's homepage may also be a type of feature data. In addition, conversion events such as subscription or shopping cart may also be a type of feature data.

At this time, the characteristic data may be determined based on the actions performed by the user on a specific homepage or app, or based on a specific browser (eg Explorer) or hardware (eg PC, smartphone). action can be determined.

For example, in the learning process, the training device 101 may identify seed users and non-seed users at a ratio of 1:2, divide them into training and validation sets and test sets, and use them for training of the learning model 103. there is.

However, in order to train the learning model 103 for a specific advertisement, more than a certain number of seed users are required. For example, the training process of the learning model 103 may be processed for advertisements with 500 or more seeds. This is because it is difficult for the learning model 103 to accurately predict the parameter user when the number of seed users is less than a certain number.

Also, when the number of seed users for a specific advertisement is less than a predetermined number, seed users of other advertisements included in the advertiser to which the specific advertisement belongs may be used. Specifically, the training device 101 may train the learning model 103 for a specific advertiser using learning data 104 of seed users for advertisements belonging to the specific advertiser.

Also, when the number of seed users for advertisements belonging to a specific advertiser is less than a predetermined number, seed users of other advertisements included in the advertiser group, which is a group of advertisers having the same characteristics as the specific advertiser, may be used. Specifically, the training device 101 may train the learning model 103 for a specific advertiser using the learning data 104 of seed users for advertisements belonging to a specific advertiser group.

That is, in the advertising budget prediction process, when there is no learning model 103 trained for a specific advertisement, but there is a learning model 103 for an advertiser or advertiser group to which a specific advertisement belongs, the advertising budget prediction device 102 The learning model 103 for an advertiser or group of advertisers is used to predict the parametric user for that particular advertisement.

As another example, the advertising budget prediction device 102 may predict an advertising budget for a specific advertisement by inputting the target user's characteristic data 105 to the learning model 103 for an advertiser or advertiser group.

Both the advertisement budget prediction device 102 and the training device 101 correspond to the processor of the advertisement providing server 100, and may correspond to the same processor or to different processors. And, the advertising budget prediction device 102 performs an advertising budget prediction method according to an embodiment, and the training device 101 performs a training method according to an embodiment.

FIG. 11 is a flow chart of a method for providing advertisements according to a third embodiment, and is a flowchart illustrating a method for predicting an advertisement budget according to a third embodiment.

In step S300, the advertising budget prediction device 102 determines whether there is a learning model for predicting the advertising budget for each advertisement. As explained earlier, there is no trained learning model if there are not enough seed users for a particular advertisement. Accordingly, the advertising budget predicting device 102 uses a different advertising budget prediction method for each advertisement.

In step S310, the advertising budget prediction device 102 predicts the advertising budget by using the learning model for advertising when there is a learning model for advertising. At this time, the learning model is a learning model pre-trained for the advertisement.

The advertisement budget prediction device 102 collects characteristic data of target users and extracts characteristic data associated with a conversion for the advertisement from the characteristic data of the target user. Then, the advertising budget prediction unit 102 determines a parameter user among target users by inputting the extracted feature data to the trained learning model, and determines an advertising budget according to the determined parameter user.

In step S320, if there is no learning model for the advertisement, the advertisement budget prediction unit 102 determines whether there is a learning model pre-learned for the advertiser to which the advertisement belongs. For example, in the case of an advertisement related to manufacturer A's shoes, even if there is a shortage of seed users for maker A, there may be seed users for various advertisements such as clothes and sunglasses of maker A as an advertiser.

In step S330, if there is a learning model for the advertiser, the advertising budget prediction unit 102 predicts the advertising budget by using the learning model for the advertiser.

The advertisement budget prediction device 102 collects characteristic data of target users, and extracts characteristic data associated with conversion of advertisements belonging to the category from the characteristic data of target users. Also, the advertising budget prediction device 102 may determine a parameter user among target users by inputting the extracted feature data to the trained learning model, and may determine an advertising budget according to the determined parameter user.

The advertisement budget prediction unit 102 may determine an advertisement budget according to the determined number of user parameters. For example, the advertisement budget prediction unit 102 may increase the budget or budget ratio of the advertisement as the number of user parameters increases, and may decrease the budget or budget ratio of the advertisement as the number of users decreases.

The advertising budget predicting device 102 may determine an advertising budget according to the number of parameter users having a specific characteristic among the parameter users, without using all of the determined parameter users. For example, the advertising budget prediction device 102 may predict an advertising budget using the number of male users among the determined user parameters, and may predict an advertising budget using the number of university students among the determined user parameters.

In step S340, if there is no learning model for the advertiser, the advertising budget prediction device 102 determines whether there is a pre-learned learning model for the advertiser group. For example, in the case of an advertisement related to advertiser A, there may be seed users for various advertisements of the advertiser group, which is a group of advertisers having the same characteristics as advertiser A.

In step S350, if there is a learning model for the advertiser group, the advertising budget prediction device 102 predicts the advertising budget by using the learning model for the advertiser group.

Then, in step S360, if there is no learning model for the advertiser group, the advertising budget prediction device 102 predicts users with high purchasing power as parameter users, and predicts the advertising budget using the predicted parameter users. . Specifically, when there is no learning model for an advertiser group, the advertising budget predicting device 102 predicts a parameter user based on a record of conversion of target users, and predicts an advertising budget using the predicted parameter user.

For example, a user with high purchasing power refers to a user who has converted more than other users with respect to advertisement exposure regardless of the type or category of advertisement, and random advertisement for a certain period of time based on records of conversion of target users. Users with the highest number of conversions for can be predicted as a parameter.

According to the above-described third embodiment, an embodiment may predict an advertisement budget and transmit an advertisement based on the predicted advertisement budget. In addition, an embodiment may obtain performance for each transmitted advertisement, and adjust a budget ratio for each advertisement based on the performance for each advertisement obtained according to the first or second embodiment.

12 is a diagram illustrating a training process of a learning model according to an embodiment.

In the process of training the learning model, the feature data 411 to 412 of the training user 400 composed of seed users 402 and 403 and non-seed users 401 are input to the learning model of the training device 101 . 12, the seed user 402 for Ad 1 is a non-seed user for Ad 2. Similarly, the seed user 403 for Ad 2 is a non-seed user for Ad 1. Also, the non-seed user 401 is not a seed user for any advertisement.

The training device 101 determines characteristic data highly correlated with occurrence of a conversion in the characteristic data 411 and 412 of seed users for each advertisement for each advertisement. Feature data that is highly correlated with conversion occurrence is determined by the importance of the feature data.

In order to determine the importance of specific feature data, the training apparatus 101 determines the importance of specific feature data by comparing seed users with specific feature data with non-seed users with specific feature data, and determining the importance of specific feature data with high importance. Extract feature data in order.

Specifically, the training device 101 determines the importance of specific feature data by using the difference between the ratio of seed users having specific feature data among seed users and the ratio of non-seed users having specific feature data among non-seed users. can decide

In other words, as the proportion of seed users with specific feature data among seed users increases, the importance of specific feature data is determined to be high, and as the proportion of non-seed users with specific feature data among non-seed users increases, The importance of feature data is determined to be low.

The training device 101 performs training of a learning model for a specific advertisement using feature data having the highest importance for the advertisement. As an example, referring to FIG. 12 , when the characteristic data 411 of the seed user 402 for advertisement 1 includes A, B, C, D, etc., the training device 101 provides data of each characteristic data 411. The importance of advertisement 1 is determined, and feature data B and D 421 having the highest importance are used for training of the learning model.

As another example, when the characteristic data 412 of the seed user 403 for advertisement 2 includes E, B, C, H, etc., the training device 101 calculates the importance of each characteristic data 412 for advertisement 2. is determined, and feature data E and H (422) having the highest importance are used for training of the learning model. At this time, feature data B is feature data commonly included in the seed users 402 and 403, but is not determined as feature data of high importance in advertisement 2 because the importance of feature data varies depending on the advertisement.

Further, the training device 101 may delete the record of the extracted feature data when the number of feature data extracted during a certain period exceeds a reference range. This is to prevent memory waste and improve performance.

The training device 101 again identifies characteristic data of seed users for each advertisement. Then, the training device 101 extracts again feature data associated with the conversion to the advertisement among the identified feature data.

The training device 101 selects features 421-422 of high importance among feature data 411-412 of the training user 400 including seed users 402 and 403 and non-seed users 401 as a learning model ( 431, 432). At this time, not all non-seed 401 users may be input as training users 400, and may be randomly sampled in consideration of a ratio with seed users 402 and 403. In this case, negative sampling may be used.

The training device 101 trains the learning models 431 and 432 to predict the seed users 402 and 403 among the training users 400 as parameters by using the feature data of the seed users 402 and 403 as correct labels. do. That is, the training device 101 trains the learning models 431 and 432 such that a difference between the parameter user predicted by the learning models 431 and 432 and the seed user is minimized.

Specifically, among the feature data of training users, feature data 421 and 422 of high importance are input to a learning model, and the learning model is trained to predict a seed user based on the feature data 421 and 422 of training users. .

Then, the training device 101 uses loss functions 441 and 442 for calculating the difference between the parameter user predicted by the learning model and the seed user. Specifically, the training apparatus 101 may train the learning model by updating weights included in the learning model so that the value of the loss function is minimized.

For example, the training device 101 may determine a probability of being a seed of training users as a learning model for a specific advertisement. Then, the training device 101 may determine the parameter user among the training users in the order of highest probability. At this time, a certain number of top training users with the highest determined probability may be stored.

And, if most of the upper parameters with the highest determined probability are determined as parameters, overfitting is considered to have occurred, and resets in the direction of reducing epochs and iterations. In order to prevent overfitting from occurring, the training device 101 presets a minimum criterion, and stops training for the advertisement when the value of the loss function is smaller than the minimum criterion.

The minimum criterion is determined based on the number of seed users. As the number of seed users increases, the minimum criterion is determined with a small value, and as the number of seed users decreases, the minimum criterion is determined with a high value. Then, the training device 101 can correct the determined score by adjusting the negative sampling rate of non-seed users constituting the training users.

For example, the learning model of the present invention may be a deep learning-based network model. Specifically, the learning model includes i) a learning model that outputs a regression analysis result for input data (e.g., logistic regression) and ii) a learning model that has a plurality of layers and outputs a generalized result for the input data (eg, logistic regression) e.g., deep neural networks) can be combined to provide complementary learning models.

As an example, referring to FIG. 12 , when the learning model 431 is used for advertisement 1 and the learning model 432 is used for advertisement 2, the learning model 431 includes the training user 400's characteristic data ( 421) is input, and the characteristic data of the seed user 402 is input as the correct answer label. Characteristic data 422 of the training user 400 is input to the learning model 432, and characteristic data of the seed user 403 is input as a correct answer label.

13 is a configuration diagram of a user terminal 300 according to an embodiment. Hereinafter, components constituting the user terminal 300 shown in FIG. 13 will be sequentially described.

The wireless communication unit 310 may include one or more components that perform wireless communication between the user terminal 300 and a wireless communication system or between the user terminal 300 and a network in which the user terminal 300 is located. . For example, the wireless communication unit 310 may include a broadcast reception module 311, a mobile communication module 312, a wireless Internet module 313, a short-distance communication module 314, a location information module 315, and the like. .

The broadcast reception module 311 receives a broadcast signal and/or broadcast-related information from an external broadcast management server through a broadcast channel. Here, the broadcast channel may include a satellite channel and a terrestrial channel. Meanwhile, broadcast related information may be provided through a mobile communication network, and in this case, it may be received by the mobile communication module 312 .

In addition, the mobile communication module 312 transmits and receives a radio signal with at least one of a base station, an external terminal, and a server on a mobile communication network. Here, the wireless signal may include a voice call signal, a video call signal, or various types of data according to text/multimedia message transmission/reception.

The wireless Internet module 313 refers to a module for wireless Internet access, and may be built into or external to the user terminal 300 .

The short-distance communication module 314 refers to a module for short-distance communication. As a short-range communication technology, Bluetooth, Radio Frequency Identification (RFID), Infrared Data Association (IrDA), Ultra Wideband (UWB), ZigBee, or the like may be used.

Also, the location information module 115 is a module for confirming or obtaining the location of the user terminal 300 . One example is a Global Positioning System (GPS) module. The GPS module receives location information from a plurality of satellites. Here, the location information may include coordinate information indicated by latitude and longitude.

Meanwhile, an audio/video (A/V) input unit 320 is for inputting an audio signal or a video signal, and may include a camera 321 and a microphone 322. The camera 321 processes an image frame such as a still image or a moving image obtained by an image sensor in a video call mode or a photographing mode. And, the processed image frame may be displayed on the display unit 351 .

An image frame processed by the camera 321 may be stored in the memory 360 or transmitted to the outside through the wireless communication unit 310 . Two or more cameras 321 may be provided according to configuration aspects of the user terminal 300 .

The microphone 322 receives an external sound signal through a microphone in a call mode, a recording mode, or a voice recognition mode, and processes it into electrical voice data. In addition, the processed voice data may be converted into a form transmittable to a mobile communication base station through the mobile communication module 312 and output in the case of a call mode. The microphone 322 may implement various noise cancellation algorithms for removing noise generated in the process of receiving an external sound signal.

The user input unit 330 receives an input operation from the user and generates input data for controlling the operation of the user terminal 300 .

The sensing unit 340 detects the current state of the user terminal 300, such as the location of the user terminal 300, presence or absence of contact with the user, direction of the user terminal 300, acceleration/deceleration of the user terminal 300, etc. A sensing signal for controlling the operation of the terminal 300 is generated.

The interface unit 370 serves as an interface with all external devices connected to the user terminal 300 . For example, a wired/wireless headset port, an external charger port, a wired/wireless data port, a memory card port, a port for connecting a device having an identification module, an audio I/O (Input/Output) port, A video I/O (Input/Output) port, an earphone port, and the like may be included.

The output unit 350 is for outputting an audio signal, video signal, or alarm signal, and may include a display unit 351, a sound output module 352, an alarm unit 353, and the like.

The display unit 351 displays and outputs information processed by the user terminal 300 . For example, when the terminal is in a call mode, a UI (User Interface) or GUI (Graphic User Interface) related to a call is displayed. In addition, when the user terminal 300 is in a video call mode or a shooting mode, a photographed or/and received image, UI, or GUI is displayed.

Meanwhile, as described above, when the display unit 351 and the touch pad form a mutual layer structure to form a touch screen, the display unit 351 may be used as an input device as well as an output device. The display unit 351 includes a liquid crystal display, a thin film transistor-liquid crystal display, an organic light-emitting diode, a flexible display, a 3D display ( 3D display) may include at least one. Also, depending on the implementation form of the user terminal 300, two or more display units 351 may exist. For example, the user terminal 300 may include an external display unit (not shown) and an internal display unit (not shown) at the same time.

The audio output module 352 outputs audio data received from the wireless communication unit 310 or stored in the memory 360 in a call signal reception mode, a call mode or a recording mode, a voice recognition mode, a broadcast reception mode, and the like. In addition, the sound output module 352 outputs sound signals related to functions performed by the user terminal 300 (eg, call signal reception sound, message reception sound, etc.). The sound output module 352 may include a speaker, a buzzer, and the like.

The alarm unit 353 outputs a signal for notifying the occurrence of an event in the user terminal 300 . Examples of events generated in the terminal include call signal reception, message reception, key signal input, and the like.

The memory 360 may store programs for processing and control of the controller 380, and provides a function for temporarily storing input/output data (eg, phonebook, message, still image, video, etc.). can also be done

The memory 360 is a flash memory type, a hard disk type, a multimedia card micro type, a card type memory (eg SD or XD memory, etc.), RAM (RAM, Random Access Memory) SRAM (Static Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), PROM (Programmable Read-Only Memory) magnetic memory, magnetic disk, It may include at least one type of storage medium among optical disks.

The control unit 380 typically controls overall operations of the terminal. For example, it performs related control and processing for voice calls, data communications, video calls, and the like. Also, the controller 380 may include a multimedia module 181 for playing multimedia. The multimedia module 381 may be implemented within the control unit 380 or may be implemented separately from the control unit 380.

The controller 380 controls various operations of the terminal to implement the aforementioned advertisement providing method.

The power supply unit 290 receives external power and internal power under the control of the controller 280 and supplies power required for operation of each component.

Meanwhile, at least some or all of the operations of the advertisement providing server 100 described above may be implemented in the user terminal 300 . At this time, the user terminal 300 may be pre-installed with an application for communicating with the advertisement providing server 100 to perform the operation of the advertisement providing server 100 described above.

Claims (8)

As an advertisement providing method performed by an advertisement providing server,
Determining whether there is a learning model for predicting an advertisement budget for each advertisement;
predicting an advertisement budget by using the learning model for the advertisement if there is a learning model for the advertisement;
If there is no learning model for the advertisement, determining whether there is a pre-learned learning model for the advertiser to which the advertisement belongs;
predicting an advertising budget using the learning model for the advertiser when there is a learning model for the advertiser;
collecting performance of advertisements for each advertisement target transmitted according to the predicted advertisement budget;
acquiring a unit price and price competitiveness for each advertisement target based on the performance of the advertisement for each advertisement target;
adjusting a budget ratio by using price competitiveness for each advertisement target and a unit price for attracting customers; and
and transmitting an advertisement for each advertisement target based on the adjusted budget ratio.
According to claim 1,
determining whether there is a pre-learned learning model for an advertiser group to which the advertiser belongs, when there is no learning model for the advertiser;
predicting an advertising budget by using the learning model for the advertiser group if there is a learning model for the advertiser group; and
And if there is no learning model for the advertiser group, predicting an advertising budget based on a record of conversion of a target user.
According to claim 1,
The step of predicting the parameters using the learning model for the advertisement,
A method for providing advertisements, wherein a parameter user is determined using feature data of a target user, and an advertising budget is predicted according to the determined number of parameter users.
According to claim 1,
Obtaining a bottleneck for each advertisement target based on the performance of the advertisement for each advertisement target;
Adjusting the budget ratio,
The advertisement providing method of adjusting the budget ratio based on the price competitiveness for each advertisement target and the bottleneck for each advertisement target.
A non-transitory recording medium on which a program for executing the method according to claim 1 is recorded and can be read by a computer.
A computer program recorded on a non-transitory recording medium to execute the method according to claim 1 in an advertisement providing server.
at least one processor; and
An advertisement providing server comprising a memory for storing instructions instructing the at least one processor to perform at least one operation,
At least one operation,
Determining whether there is a learning model for predicting an advertisement budget for each advertisement;
predicting an advertisement budget by using the learning model for the advertisement if there is a learning model for the advertisement;
If there is no learning model for the advertisement, determining whether there is a pre-learned learning model for the advertiser to which the advertisement belongs;
predicting an advertising budget using the learning model for the advertiser when there is a learning model for the advertiser;
collecting performance of advertisements for each advertisement target transmitted according to the predicted advertisement budget;
acquiring a unit price and price competitiveness for each advertisement target based on the performance of the advertisement for each advertisement target;
adjusting a budget ratio by using price competitiveness for each advertisement target and a unit price for attracting customers; and
And transmitting an advertisement for each advertisement target based on the adjusted budget ratio.
As an advertisement providing system,
an advertiser terminal receiving advertisement-related information from an advertiser and providing the received advertisement-related information to an advertisement providing server;
an advertisement providing server that adjusts a budget ratio between a plurality of advertisement targets according to a budget and transmits an advertisement for each of the advertisement targets using a plurality of advertisement media; and
Including a user terminal receiving advertisements from the advertisement providing server,
The advertisement providing server uses at least one processor,
Determining whether there is a learning model for predicting an advertisement budget for each advertisement;
predicting an advertisement budget by using the learning model for the advertisement if there is a learning model for the advertisement;
If there is no learning model for the advertisement, determining whether there is a pre-learned learning model for the advertiser to which the advertisement belongs;
predicting an advertising budget using the learning model for the advertiser when there is a learning model for the advertiser;
collecting performance of advertisements for each advertisement target transmitted according to the predicted advertisement budget;
acquiring a unit price and price competitiveness for each advertisement target based on the performance of the advertisement for each advertisement target;
adjusting a budget ratio by using price competitiveness for each advertisement target and a unit price for attracting customers; and
and transmitting an advertisement for each advertisement target based on the adjusted budget ratio.

Images