WO2020231001A2

WO2020231001A2 - Service provision device and method supporting advertisement-related dynamic rewards, and service provision system comprising same

Info

Publication number: WO2020231001A2
Application number: PCT/KR2020/003996
Authority: WO
Inventors: 이관우; 이영호; 이성원
Original assignee: (주)버즈빌
Priority date: 2019-05-10
Filing date: 2020-03-24
Publication date: 2020-11-19
Also published as: KR102245602B1; KR20200129767A; WO2020231001A3

Abstract

The present invention relates to a service provision device and method supporting advertisement-related dynamic rewards, and a service provision system comprising same, and more particularly to: a service provision device and method supporting advertisement-related dynamic rewards for, when an advertisement is provided to a user terminal, increasing a user's activity related to sales while maintaining the user's activity related to engagement with the advertisement, and, at the same time, dynamically changing rewards to induce the user to engage with the advertisement according to the advertisement-related segment characteristic of the user and providing the changed rewards; and a service provision system comprising same.

Description

Service providing device and method supporting dynamic compensation related to advertisement, and service providing system including the same

The present invention relates to a service providing apparatus and method for supporting dynamic compensation related to advertisement, and a service providing system including the same, and more particularly, to a user's sales-related activity while maintaining the user's advertisement participation-related activities when providing advertisements to a user terminal. A service providing device and method that supports dynamic compensation related to advertisement to provide by dynamically changing compensation to induce user participation in advertisement according to the characteristics of user's advertisement-related segment while increasing activity, and service providing system including the same It is about.

Existing advertisements do not have separate benefits for users to watch and participate in advertisements, so users do not want to see annoying advertisements. In most cases, advertisements are blocked through a separate application to block advertisements. There is a problem of extremely low efficiency.

Recently, in order to increase advertising efficiency, an advertisement model has emerged in which a user's reluctance to an advertisement is lowered by paying a reward to the user when a user participates in an advertisement in order to meet the conditions intended for the advertisement.

This advertising model satisfies the conditions of compensation intended in the advertisement, such as watching an advertisement, clicking on an advertisement, or downloading and installing an application that is being advertised in the advertisement, and when the user performs an action related to actual advertisement consumption, By providing the same compensation benefits, users are encouraged to actively participate in advertisement consumption and advertisement efficiency is increased.

However, most of the existing advertisement models use a simple compensation method, such as equally paying a predetermined amount of compensation at a certain time or when satisfying the compensation payment conditions related to advertisements, and this is exploited by users for the purpose of compensation only. There is a problem in that advertising efficiency is degraded because not only there is a lot of potential, but also because the user becomes familiar with the repetitive reward pattern, it becomes difficult to gradually increase the user's activity.

The present invention is an optimal activity compensation model related to activity compensation for providing compensation by calculating the compensation amount according to the user's sales-related activity in relation to the advertisement exposed to the user, but dynamically changing the compensation timing and amount while dividing the compensation amount And provides an optimal induction compensation model related to induction compensation to induce participation in advertisements according to the segment characteristics corresponding to the user, and the activity compensation to maximize sales versus cost according to the correlation between the activity compensation and induction compensation Its purpose is to optimize the model and the induced compensation model.

A service providing apparatus supporting dynamic compensation related to advertisement according to an embodiment of the present invention is an advertisement that receives advertisement information from an advertisement management server that stores a plurality of different advertisement-related advertisement information, and transmits the advertisement information to a user terminal. In response to the provision unit and the advertisement information transmitted to the user terminal, the total compensation amount paid to the user of the user terminal is updated when a preset sales-related event occurs, and the advertisement environment related environment information is provided in correspondence with the updated total compensation amount. Calculated, and calculated for each of the plurality of parameters based on the learning when an event corresponding to a payment condition preset in the advertisement information occurs by learning a plurality of parameters including the total number of rewards and a collapse rate for which the environmental information is improved After determining an activity compensation amount, which is a part of the updated total compensation amount, according to an optimum value, an activity compensation unit may be provided to provide the activity compensation amount corresponding to the user.

As an example related to the present invention, the activity compensation unit receives the event-related event information from the user terminal whenever an event according to the user's activity occurs for each advertisement information transmitted to the user terminal in connection with the advertisement providing unit. You can do it.

As an example related to the present invention, the environment information may include a plurality of preset parameters for each attribute, and the plurality of attributes may include a retention rate, a click rate, sales, and advertisement impressions per user.

As an example related to the present invention, the activity compensation unit calculates a plurality of parameters including the total number of compensation and a collapse rate corresponding to the updated total compensation amount according to a first equation set in advance when the event occurs, and the plurality of The advertisement environment-related environment information is calculated by applying the total number of rewards and the collapse rate calculated in response to the updated total reward amount to the first deep learning algorithm in which the correlation between the parameter of and the advertisement environment is learned, and the environment After calculating the optimum values for each of the plurality of parameters through the first reinforcement learning algorithm for learning the plurality of parameters whose information is improved and the first deep learning algorithm, a preset payment is made to the advertisement information provided through the advertisement provider. When an event corresponding to a condition occurs, an activity compensation amount, which is part of the total compensation amount, may be determined and paid by applying the optimum value to a preset second equation.

As an example related to the present invention, the first formula is

Is,

R is the total compensation amount, T is the total number of compensation, λ is the collapse rate, t is the number of times the advertisement is viewed, and N ₀ may be a constant.

As an example related to the present invention, after the activity compensation unit determines an optimum value for each of the plurality of parameters, the second equation

Depending on the event, the activity compensation amount r _n to be paid to the specific time t _n is determined, where λ is the collapse rate, t is the number of times the advertisement is viewed, and N ₀ is a constant. You can do it.

As an example related to the present invention, a specific group is created by analyzing the advertisement conversion pattern of the user each time the event-related event information is collected in relation to the user, and then grouping the user with one or more other users having similar advertisement conversion patterns. And, by calculating the predicted environmental information corresponding to the specific group, which is the predicted advertising environment, and learning the induction compensation amount for inducing the activity of the user whose predicted environmental information is improved to maximize the expected benefit per cost based on the learning. It may be characterized in that it further comprises an induction compensation unit for determining the amount of induction compensation, setting a payment condition for paying the induction compensation amount in advertisement information provided by the advertisement providing unit, and transmitting the set to the user terminal.

As an example related to the present invention, the induction compensation unit is set for each segment of the specific group in a preset second deep learning algorithm in which a correlation between a plurality of preset segments and the advertisement environment expected corresponding to the segment is learned. A preset second reinforcement learning algorithm that calculates the predicted environment information corresponding to the specific group by applying a value, which is an expected advertisement environment, and learns an induction reward amount for inducing the user's activity in which the predicted environment information is improved, and It may be characterized in that determining an induction compensation amount for maximizing an expected benefit per cost through the second deep learning algorithm.

As an example related to the present invention, the set value for each segment excluding segments related to the total activity compensation amount, induction compensation amount, and induction compensation amount obtained for a plurality of different users in connection with the activity compensation unit and the induction compensation unit, and the After applying environmental information to a pre-set regression tree model to group users with similar environmental information into a unique group, the first regression equation for the relationship between the total activity compensation amount and sales and the induction compensation for the unique group Calculate a second regression equation for the relationship between amount and sales, and calculate the optimal activity compensation amount and the optimal derived compensation amount at which the maximum profits obtained from each of the first and second regression equations converge in the regression tree. It may be characterized in that it further comprises an optimization unit calculated through the model and set to the activity compensation unit and the induction compensation unit.

As an example related to the present invention, the optimization unit is calculated in response to the user when at least one of an activity compensation amount and an induction compensation amount according to the satisfaction of the payment condition in connection with the activity compensation unit and the induction compensation unit It may be characterized in that at least one of the optimal activity compensation amount and the induction compensation amount is paid.

A service providing system supporting dynamic compensation related to advertisements according to an embodiment of the present invention includes an advertisement management server that selects and allocates advertisement information to be transmitted to a user terminal, and transmits advertisement information allocated to the user terminal to the user terminal. A service server, the advertisement management server, and the service server for generating rewards to be paid to the user based on event information related to the user's advertisement activity received from the user terminal in response to the advertisement information and accumulating corresponding to the user of the user terminal In conjunction with, the advertisement information to be transmitted to the user terminal is transmitted to the user terminal through the service server, and event information transmitted from the service server in response to the advertisement information is received and transmitted to the user terminal. When a predetermined sales-related event occurs in response to the advertisement information, the total amount of rewards related to the rewards paid to the user of the user terminal is updated, and environmental information related to the advertising environment is calculated in response to the updated total amount of compensation, and the environment When an event corresponding to a payment condition set in advance in the advertisement information occurs by learning a plurality of parameters including the total number of rewards and a collapse rate for which information is improved, the information is calculated according to the optimum value calculated for each of the plurality of parameters. After determining an activity compensation amount, which is a part of the updated total compensation amount, a service providing device for providing the activity compensation amount corresponding to the user may be included.

As an example related to the present invention, the service providing device analyzes the advertisement conversion pattern of the user whenever the event-related event information is collected in relation to the user, and then groups the user with one or more other users having similar advertisement conversion patterns. To create a specific group, calculate the expected environment information corresponding to the specific group, which is an expected advertisement environment, learn the induction reward amount for inducing the user's activity whose expected environment information is improved, and cost based on the learning It may be characterized in that the induction compensation amount for maximizing the expected gain is determined, and a payment condition for paying the induction compensation amount is set in the advertisement information to be transmitted to the user terminal and transmitted to the user terminal.

According to an embodiment of the present invention, a service providing method for supporting dynamic compensation related to advertisement of a service providing device that transmits advertisement information to a user terminal is provided when a sales-related event set in advance occurs in response to advertisement information transmitted to the user terminal. Updating the total compensation amount paid to the user of the terminal, and calculating a plurality of parameters including the total number of compensation and a collapse rate corresponding to the updated total compensation amount according to a predetermined first formula, and the plurality of parameters Calculating the environment information related to the advertisement environment by applying the total number of rewards and the collapse rate calculated in response to the updated total compensation amount to the first deep learning algorithm in which the correlation between the variable and the advertisement environment is learned, and the environment information After calculating the optimal value for each of the plurality of parameters through the first reinforcement learning algorithm for learning the plurality of parameters to be improved and the first deep learning algorithm, a payment condition preset in the advertisement information provided through the advertisement provider And determining and paying an activity compensation amount, which is a part of the total compensation amount, by applying the optimum value to a second equation set in advance when an event corresponding to is generated.

As an example related to the present invention, a specific group is created by analyzing the advertisement conversion pattern of the user each time the event-related event information is collected in relation to the user, and then grouping the user with one or more other users having similar advertisement conversion patterns. And applying a set value for each segment of the specific group to a second preset deep learning algorithm in which a correlation between a plurality of preset segments and an expected advertisement environment corresponding to the segment is learned to correspond to the specific group The cost through the second reinforcement learning algorithm and the second deep learning algorithm set in advance to learn an induction reward amount for inducing the user's activity in which the predicted environment information is improved and calculating the expected environment information, which is an expected advertisement environment. Determining an induction compensation amount for maximizing the expected gain per unit, setting a payment condition for paying the induction compensation amount in advertisement information provided by the advertisement providing unit, and transmitting the payment condition to the user terminal. can do.

As an example related to the present invention, by applying the total activity compensation amount obtained for a plurality of different users, the induction compensation amount, the set value for each segment, and the environment information to a preset regression tree model, users having similar environment information After grouping into a unique group, a first regression equation for the relationship between the total activity compensation amount and sales and a second regression equation for the relationship between the induction compensation amount and sales is calculated for the unique group, and the second The optimal activity compensation amount and the optimal induction compensation amount, in which the maximum profits obtained from each of the 1 regression equation and the second regression equation converge, are calculated through the regression tree model, and when the payment conditions are satisfied, the optimum activity compensation amount and the optimum It may be characterized in that it further comprises the step of paying at least one of the induction compensation amount of.

The present invention provides the best activity compensation for the user's activity related to the advertisement participation, but the total activity compensation amount is divided within the compensation period, and the compensation amount increases according to the user's activity level. By flexibly and dynamically changing the amount of compensation for each compensation period and compensation point through deep learning-based learning, users can maintain maximum ad participation-related activities while maximizing user activities that are linked to sales, thereby increasing revenue. In addition, based on the event generated by the user for the advertisement, the user and other users with similar ad conversion patterns are grouped into a specific group, and the induction compensation amount for inducing the user to participate in the advertisement is included as a segment. The correlation between the characteristics of each segment and the change in the advertising environment according to the change in the induction compensation amount is calculated through deep learning-based learning, and the induction compensation amount that can induce the user's participation in the advertisement for each advertisement for a user in a specific group is calculated. It is possible to increase the activity related to the user's advertisement participation by presenting it, and when the user's advertisement conversion pattern changes, the induction compensation amount is dynamically changed and provided in consideration of the characteristics of the group to which the user is subordinated according to the change in the user's advertisement conversion pattern. It has the effect of greatly increasing advertising efficiency by continuously inducing users to participate in advertising.

In addition, the present invention can provide the optimal activity compensation and induction compensation in which the profits obtained from each of the activity compensation and induction compensation are mutually converged through the regression tree model, and through this, the profit while minimizing the compensation provided through advertisement There is an effect of increasing advertising efficiency by maximizing.

1 is a block diagram of a service providing system supporting dynamic compensation related to advertisements according to an embodiment of the present invention.

2 is a block diagram of a service providing apparatus supporting dynamic compensation related to advertisements according to an embodiment of the present invention.

3 is a detailed configuration and operation related example diagram of an activity compensation unit configured in a service providing apparatus supporting dynamic compensation related to advertisement according to an embodiment of the present invention.

4 is a diagram illustrating a detailed configuration and operation of an induction compensation unit configured in a service providing apparatus supporting dynamic compensation related to advertisement according to an embodiment of the present invention.

5 is a flowchart illustrating an operation of an optimization unit configured in a service providing apparatus supporting dynamic compensation related to advertisements according to an embodiment of the present invention.

Hereinafter, detailed embodiments of the present invention will be described with reference to the drawings.

As shown, an advertisement management server 10 for storing advertisement-related advertisement information received from the advertiser terminal in an advertisement DB by communicating with a plurality of different advertiser terminals through a communication network, and the advertisement management server 10 through a communication network. ), and transmitting the advertisement information to a user terminal through a plurality of different media servers, or directly transmitting the advertisement information.

In this case, the service providing device 100 may communicate with the plurality of media servers through a communication network, and the media server is an application installed and executed in the user terminal and the advertisement received from the service providing device 100 By transmitting information, when the application is executed in the user terminal, the advertisement information may be displayed and provided in the user terminal through the application.

In addition, the communication network may employ a variety of well-known wired and wireless communication methods such as wired local area network (LAN), wired wide area network (WAN), Ethernet, 4G mobile communication service, and 5G mobile communication service.

In addition, the service providing device 100 may directly transmit the advertisement information through a communication network to a specific application installed and executed in the user terminal, and the user terminal in a state in which the specific application is executed transmits the advertisement information to the user. It can be displayed through the display of the terminal.

As described above, the service providing apparatus 100 may transmit and display advertisement information to a user terminal, thereby exposing an advertisement according to the advertisement information to a user of the user terminal and providing the user to view the advertisement.

However, since it is difficult to determine whether the user has been properly recognized for the advertisement because the user is exposed to the advertisement only by simple viewing, the service providing device 100 clicks an advertisement image or link included in the advertisement information. Reward in response to a user's activity-related event (input event), such as a click event, an application installation event corresponding to advertisement information, or a conversion event such as a purchase of a product corresponding to advertisement information. ) Can be provided.

Through this, the service providing device 100 may provide a reward as a reward according to the user's activity related to advertisement, and provide the reward to the user so that the reward can be used for product purchase discounts or product exchange. I can.

In such a reward advertisement, the user's activity is very different depending on the time when the reward is given and the amount of the reward. If a simple compensation payment method (for example, a predetermined compensation is paid equally at a certain time) is used, it is likely to be abused by users who only want compensation. In addition, as users become accustomed to repetitive reward patterns, it becomes difficult to gradually increase the user's activity.

Accordingly, the service providing device 100 according to the present invention efficiently calculates the compensation amount and payment timing according to the user's activity status for the total compensation amount determined based on the sales generated according to the user's advertisement-related activity. In addition, by dividing the total amount of compensation, the user's participation in the advertisement is maintained, and when the advertisement is provided, the induction reward that induces the user's activity related to the advertisement is dynamically provided to actively promote the user whose activity related to the advertisement is degraded. By inducing participation, it is possible to support to increase advertising efficiency and increase sales according to advertising.

A detailed operation example of the service providing apparatus 100 supporting dynamic compensation related to advertisement according to an embodiment of the present invention based on the above-described configuration will be described below with reference to the drawings.

First, FIG. 2 is a detailed configuration diagram of a service providing apparatus 100 according to an embodiment of the present invention.

As shown, the service providing apparatus 100 may include an advertisement providing unit 110, an activity compensation unit 120, an induction compensation unit 130, and an optimization unit 140.

In this case, any one of the constituent units constituting the service providing device 100 may be configured as a control unit that controls the service providing device 100.

The control unit may execute an overall control function of the service providing device 100 using programs and data previously stored in the service providing device 100, and the control unit may include a RAM, ROM, CPU, GPU, and bus. In addition, RAM, ROM, CPU, GPU, etc. can be connected to each other through a bus.

In addition, at least one of the constituent units constituting the service providing apparatus 100 may be included in another constituent unit.

First, the advertisement providing unit 110 may transmit advertisement information received from the advertisement management server 10 to the user terminal.

In this case, the advertisement providing unit 110 may be configured to include a media linking unit 111 that communicates with the plurality of media servers, and the media linking unit 111 is configured as described above through the media server. Advertisement information can be transmitted to the user terminal.

Accordingly, the media server, which has received the advertisement information from the service providing device 100, executes the advertisement information in the application of the user terminal corresponding to the media server and displays the advertisement information on the user terminal. Can be transferred to.

In addition, the activity compensation unit 120 may provide a portion of the sales as a reward to the user in response to the user's activity related to sales based on the user's activity participating in the advertisement.

In addition, the activity compensation unit 120 not only can continuously modify and pay the total compensation amount determined according to the user's activity, but also divide the total compensation amount for a predetermined period, and divide the total compensation amount. When payment is made, the amount of compensation paid at a specific point in time and the amount of compensation paid at a different point in time different from the specific point in time are paid differently according to the user's activity, thereby maintaining the activities related to the user's participation in advertisements, A higher compensation amount may be provided to induce a user's active advertisement participation activity, which will be described in detail with reference to FIG. 3 for a detailed configuration of the activity compensation unit 120.

First, the activity compensation unit 120 may calculate a total compensation amount, which is an action reward to be paid during a unit period.

As an example, the activity compensation unit 120 may provide a basic compensation necessary for maintaining the user's advertisement participation activity, and after dividing the user's group according to the service usage index set in advance related to the advertisement, the basic compensation amount for each user group Can be calculated.

For example, a basic compensation amount can be set separately by dividing a group of users who generate a lot of content in a service and a user group who only consumes.

In addition, the activity compensation unit 120 may pay a certain percentage of sales generated by the user's participation in advertisement as a compensation amount. For example, when 1,000 won of sales are generated through the user's participation in advertisements, 10% of this, or 100 won, is added to the user's total compensation amount. In addition, after assigning a sales value to activities such as inviting friends or creating content, the sales compensation amount according to the corresponding activity may be calculated and added to the total compensation amount.

In addition, the activity compensation unit 120 may store the total compensation amount calculated in response to the user in the DB as described above.

In this case, if the entire compensation amount is paid all at once immediately after the activity, the usual compensation amount felt by the user is too low. For this reason, the retention rate of the user may decrease. To solve this problem, the entire compensation amount is divided over a period of time and paid.

On the other hand, if the same amount is simply divided and paid for a certain period of time, it becomes difficult to induce users to participate in advertisements. It is necessary to recognize that the reward increases when the user does an activity.

To this end, the activity compensation unit 120 pays a relatively large amount of compensation when a sales-related activity (for example, conversion) occurs, and executes a strategy of gradually reducing the compensation during a period of inactivity. Through this, it is possible to induce additional activities of the user.

To this end, the activity compensation unit 120 may calculate a compensation amount paid at a time when one advertisement is clicked according to Equation 1 (Exponential decay) below.

In this case, t is the number of times the advertisement has been viewed, and when the total compensation amount is increased by the user's activity, t may be initialized to 0.

In other words, when the user sees the first ad, t is 1, and when the user sees the next ad, t becomes 2, and the total amount of compensation (R) when a sales-related event (for example, ad conversion) occurs. As this is newly calculated, t becomes zero again.

In addition, the activity compensation unit 120 may calculate N ₀ through Equation 2 below.

In this case, R is the total compensation amount, T is the total number of compensation, λ is the collapse rate, and t may be the number of times the advertisement has been viewed.

In addition, N ₀ may be a constant calculated through Equation 2.

Based on the above-described configuration, an embodiment of the activity compensation unit 120 that dynamically divides and pays the total compensation amount according to the user's activity will be described.

As shown, the activity compensation unit 120 corresponds to the advertisement information for each advertisement information transmitted to the user terminal in connection with the advertisement providing unit 110, so that each time an event according to the user's activity occurs, the Event-related event information may be received, and the event information may be stored in a DB in response to the user.

In addition, the activity compensation unit 120 updates the total compensation amount paid to the user of the user terminal when a preset sales-related event occurs in response to advertisement information transmitted to the user terminal based on the event information, and According to the first equation, the total number of compensation and the collapse rate corresponding to the updated total compensation amount may be calculated.

In this case, the first equation may include Equations 1 and 2, and N _O may be a constant obtained according to the above-described first equation.

In addition, the total number of compensation (T) is the number of times to be paid by dividing the total compensation amount, and determines an interval between compensation. If the total number of compensation (T) is small, a relatively large amount is paid in a short period, and if the total number of compensation (T) is large, a relatively small amount is paid for a long period of time.

In addition, the exponential decay constant (λ) determines the degree to which the compensation decreases over time. When the collapse rate is large, the activity compensation unit 120 initially pays a large amount of compensation and then quickly reduces the compensation amount. When the collapse rate is small, the activity compensation unit 120 pays less initial compensation, but slowly reduces the compensation amount.

In addition, the activity compensation unit 120 is a first preset deep learning algorithm in which a correlation between an advertisement environment and a plurality of parameters including the total number of compensation and a collapse rate used in the first equation is learned. The advertisement environment-related environmental information may be calculated by applying the total number of compensations and the collapse rate calculated in correspondence to the updated total compensation amount.

In this case, the environmental information may include a plurality of parameters for each attribute including a retention rate, a click through rate, a revenue, an impression count per user, and the like.

In addition, the retention rate may mean the user retention rate of all advertisement-related services, and the click rate may mean the advertisement click rate of all users, and the sales may mean the sales of the entire service, and the advertisement per user The number of impressions may mean the number of advertisement impressions for each user.

In this case, an increase in the number of advertisement impressions per user may be interpreted as an increase in user activity.

In addition, the activity compensation unit 120 may be configured to include a first reinforcement learning unit 121, and the first reinforcement learning unit 121 has a correlation between the environment information and the plurality of parameters. By applying the environmental information calculated through the first deep learning algorithm in response to the user's sales-related event to the first pre-set reinforcement learning algorithm in the learned state, each parameter value of the total number of compensation and the collapse rate is calculated. can do.

In this case, the first reinforcement learning algorithm may refer to a neural network related to reinforcement learning in which a result is viewed and a reward is given if the result is good and a penalty is given if the result is not good. That is, if the result is good, a larger weight is given to the current parameter, and if not, the weight is decreased.

Due to the nature of the advertising system, it is difficult for any indicator to have an optimal or answer value. That is, it is difficult to know how wrong the index is, but since reinforcement learning continuously searches for a better index value, it is possible to find the optimal parameter values for each of the total number of compensation and the collapse rate.

In addition, the activity compensation unit 120 is calculated by applying the total number of compensation and the collapse rate calculated through the first reinforcement learning unit 121 in response to the user's sales-related event to the first deep learning algorithm. The process of recalculating the total number of compensation and the collapse rate by repeatedly applying environmental information to the first reinforcement learning algorithm of the first reinforcement learning unit 121 and applying the same to the first deep learning algorithm may be repeated. The first reinforcement learning unit 121 may find and continuously calculate parameter values for each of the total number of rewards and the collapse rate for which the environmental information is improved through reinforcement learning using the first reinforcement learning algorithm.

In addition, the activity compensation unit 120 applies to the first deep learning algorithm whenever a plurality of parameter values are calculated by the first reinforcement learning unit 121 to calculate environmental information, and then the first reinforcement learning unit 121 The plurality of parameter values may be calculated by applying to the reinforcement learning unit 121, and when the plurality of parameter values calculated through the first reinforcement learning unit 121 satisfy a first condition set in advance, the first The plurality of parameter values calculated when the 1 condition is satisfied may be determined as optimal parameter values (optimal values) for the total number of compensations and the collapse rate, respectively.

As an example, when the first reinforcement learning unit 121 repeatedly calculates a plurality of parameter values identical to those of the existing (just before), the activity compensation unit 120 determines the total number of compensation and the collapse rate. It can be determined as the optimal parameter value (optimum value) for each.

That is, the activity compensation unit 120 evaluates the advertisement environment created by the number of compensations determined through the first equation and the collapse rate in relation to the total compensation amount increased by the sales-related event generated by providing advertisements to the user. By doing so, it is possible to optimize the total number of compensation and the collapse rate in a direction in which the corresponding advertisement environment is improved, and through this, the total compensation amount generated for the user may be dynamically divided and paid according to the total number of compensation and the collapse rate.

As an example, as described above, the activity compensation unit 120 may use the first reinforcement learning algorithm and the first deep learning algorithm to learn the plurality of parameters for improving the advertising environment according to the environment information. After calculating the optimum value for each parameter, when an event corresponding to a payment condition set in advance for the payment of the compensation amount to the advertisement information provided through the advertisement providing unit 110 occurs, the optimum value is applied to the second equation After determining (calculating) an activity compensation amount, which is a part of the total compensation amount, the activity compensation amount may be paid in response to the user.

At this time, the activity compensation unit 120 determines the optimal value for each of the plurality of parameters, and then according to the second equation, the following equation (3), the activity compensation amount r, which is to be paid to the specific time point t _n _n can be determined.

Here, λ is the collapse rate, t is the number of times the advertisement is viewed, and N ₀ may be a constant.

In the above configuration, the activity compensation unit 120 may perform reinforcement learning on all data stored in the DB to find an initial decay rate λ, and then distribute a model having this value as an initial setting to each user.

In addition, the activity compensation unit 120 may manage the corresponding model to design an optimal compensation distribution strategy by identifying a user's pattern.

For example, the activity compensation unit 120 applies the initial λ ₁ to the user. After waiting for a certain period of time T (for example, 24 hours, 48 hours, etc.), the value proportional to the increase or decrease value of the user's sales (R ₁ -R ₀ , sales changed from R ₀ to R ₁ ) Apply to

That is, the activity compensation unit 120 applies λ ₂ = λ ₁ + k·(R ₁ -R ₀ ) to the user as the next training step of reinforcement learning. Where k is a preset hyperparameter.

Meanwhile, the induction compensation unit 130 may calculate an Attraction Reward amount based on an expected value of sales that may occur when a specific advertisement is displayed to the user.

In this case, the induction compensation amount is a compensation paid to induce advertisement participation based on the possibility of user participation in advertisement, and the compensation amount is dynamically calculated at the time the advertisement is delivered.

As an example, the induction compensation unit 130 may calculate the expected sales value by estimating the possibility of participation of each advertisement for the user based on this while learning the user's interest and the advertisement participation activity, and the service provider ( Alternatively, the induction compensation amount can be calculated by multiplying the user sharing ratio set by the advertiser).

For example, if the probability of the user participating in the advertisement is 5%, and the sales when participating in the advertisement is 1,000 won, the expected sales value is 50 won. If the user sharing ratio is 10%, 5 won can be calculated as the induction compensation amount.

At this time, when the user purchases a specific product, the induction compensation unit 130 may grant a high probability of participation in advertisements for other products similar to the purchased product, and the current user May give a higher probability of participating in advertisements for items not purchased.

Based on the above-described configuration, the induction compensation unit 130 dynamically provides an induction compensation for inducing the user's activity when providing an advertisement in connection with the advertisement providing unit 110 to prevent a user whose activity related to advertisement participation is degraded. By inducing active participation in advertisement, it is possible to support to increase advertisement efficiency and increase sales according to advertisement, which will be described in detail with reference to FIG. 4.

The induction compensation unit 130 is associated with at least one of the advertisement providing unit 110 and the activity compensation unit 120 to correspond to the user, and each time the event-related event information is collected, a plurality of After analyzing the advertisement conversion pattern of the user based on event information, a specific group may be generated by grouping the user with one or more other users having similar advertisement conversion patterns.

In this case, the advertisement conversion pattern may mean an advertisement conversion related pattern including advertisement selection, membership registration, purchase of an advertisement target product, installation of an advertisement target application, and execution of an advertisement target application.

In addition, the induction compensation unit 130 generates the specific group in a preset second deep learning algorithm in a state in which a correlation between a plurality of preset segments and an advertisement environment expected corresponding to the segment is learned. By applying the set value for each segment of the specific group, which is determined at the time, predicted environment information related to the predicted advertisement environment expected in correspondence with the specific group may be calculated.

In this case, the segment may include a user's propensity set in the group, a product activity characteristic, an interest characteristic, an advertisement participation characteristic, an induction compensation amount, etc. as a criterion for classification and setting of the group.

In addition, the propensity of the user can be calculated as a score for each item purchased by the user compared with the purchase of items by other users similar to the user, and the product activity characteristic is an advertisement for a product similar to a specific product purchased by the user. The score is calculated based on the record of the selection or purchase, the interest characteristic is the relevance score calculated based on the interest analyzed by the activity such as the news article clicked by the user, and the advertisement participation characteristic is the record of participation in the advertisement. The score can be calculated based on.

In addition, the expected environment information includes a plurality of the number of ad impressions per user, such as an expected click rate (Expected CTR), an expected conversion rate, an expected revenue, an expected retention rate (or retention rate), etc. It may include parameters for each attribute of.

At this time, the expected click-through rate is the probability that the user will see and click on the advertisement, the expected conversion rate is the probability that the user will see the advertisement and participate in the advertisement to generate sales, and the expected sales are the product of the advertisement unit price and the expected conversion rate.

In addition, the induction compensation unit 130 may be configured to include a second reinforcement learning unit 131, and the second reinforcement learning unit 131 includes the expected environment information and an induction compensation that is one of the plurality of segments. The induction compensation amount for the user is calculated by applying the predicted environment information calculated through the second deep learning algorithm in response to the user to a preset second reinforcement learning algorithm in a state in which the correlation between amounts is learned. I can.

In this case, the second reinforcement learning algorithm refers to a neural network related to reinforcement learning that sees the result and learns by giving a reward if the result is good, and giving a penalty if the result is not good, like the first reinforcement learning algorithm. I can. That is, if the result is good, a larger weight is given to the current induction compensation amount, and if not, the weight is decreased.

In addition, the induction compensation unit 130 is a result of the induction compensation amount calculated through the second reinforcement learning unit 131 corresponding to the user, and is an induction compensation amount among the set values for each segment (a plurality of segments). After substituting the relevant segment (segment value), it is applied to the second deep learning algorithm, and the predicted environment information calculated by applying the set value for each segment replaced as a result of the induction compensation amount to the second deep learning algorithm As a new result for the induction compensation amount calculated by repeatedly applying the second reinforcement learning algorithm of the second reinforcement learning unit 131, the existing result for the induction compensation amount, which is one of the set values for each segment, is replaced again. After that, the process of applying the set value for each segment reflecting the new result (replaced with the new result) to the second deep learning algorithm may be repeated, and the second reinforcement learning unit 131 may perform the second reinforcement learning. Through reinforcement learning using an algorithm, a value for an induction compensation amount for which the predicted environment information is improved may be continuously calculated.

In addition, the induction compensation unit 130 calculates predicted environment information by applying it to the second deep learning algorithm together with other segments whenever the induction compensation amount is calculated by the second reinforcement learning unit 131, The induction compensation amount may be calculated by applying it to the second reinforcement learning unit 131 again, and when the induction compensation amount calculated through the second reinforcement learning unit 131 satisfies a preset second condition, the second 2 The induction compensation amount calculated when the conditions are satisfied can be determined as the final induction compensation amount.

For example, when the second reinforcement learning unit 131 repeatedly calculates the same induction compensation amount as before, the induction compensation unit 130 may determine the corresponding induction compensation amount as the final induction compensation amount.

Accordingly, the induction compensation unit 130 determines the user's belonging through a preset second reinforcement learning algorithm and a second deep learning algorithm for learning the induction compensation amount for inducing the user's activity in which the expected environment information is improved. The induction compensation amount that maximizes the expected gain per cost (= expected revenue/cost) while inducing user activity according to the segment of the group can be determined as the final induction compensation amount.

In addition, the induction compensation unit 130 interlocks with the advertisement providing unit 110 to set a payment condition for paying the final induction compensation amount in the advertisement information provided by the advertisement providing unit 110, and the user terminal Can be transferred to.

In addition, when the induction compensation unit 130 receives event information satisfying a payment condition corresponding to the final induction compensation amount from the user terminal in connection with the advertisement providing unit 110, the final induction compensation amount Can be paid.

On the other hand, activity reward and induction reward change fluidly under the influence of each other. For example, when the activity reward is very large, even if a lot of induction rewards are given, the user does not feel any difference.

Accordingly, the service providing device 100 may be configured to include an optimization unit 140, and the optimization unit 140 may be configured to perform the activity compensation and induction compensation unit 130 determined by the activity compensation unit 120. An optimal reward, which is an optimal value for each of the activity reward and the induction reward, may be determined so that the amount of the induction reward determined by the above can maximize the influence of each other.

In this case, the optimal compensation of the optimization unit 140 is started after sufficient data for activity compensation and induction compensation are created. Once enough data is created, it is possible to determine whether induction rewards are meaningful in current activity rewards as well. In other words, if a certain reward is reduced but the same profit comes out, you can choose to reduce the reward, or increase the reward if the expected profit becomes larger even though the reward is slightly increased.

The detailed configuration of the optimization unit 140 will be described in detail with reference to FIG. 5.

As illustrated, the optimization unit 140 interlocks with the activity compensation unit 120 and the induction compensation unit 130 to obtain a total activity compensation amount and an induction compensation amount (final induction compensation amount) for a plurality of different users. ), by applying the set value for each segment of a specific group excluding the segment related to the induction compensation amount and the environment information to a preset regression tree model, users with similar environment information are grouped into a unique group, and the unique For a group, a first regression equation for the relationship between the total activity compensation amount and sales and a second regression equation for the relationship between the induced compensation amount and sales are calculated, and the first regression equation and the second regression equation An optimal activity compensation amount and an optimal induction compensation amount in which the maximum profits obtained from each converge are calculated through the regression tree model and set in the activity compensation unit 120 and the induction compensation unit 130.

In this case, the regression tree model is an analysis technique for inferring a regression equation from data of a finally grouped model after grouping data of similar types.

Also, the total activity compensation amount may mean a total activity compensation amount to be paid for a unit period based on the user's activity.

In addition, the user's propensity, which is one of the plurality of segments, may mean a score assigned to each advertisement compared with advertisement conversion of other users similar to the user.

That is, the optimization unit 140 may select whether the compensation provided by the activity compensation unit 120 and the induction compensation unit 130 is appropriate or whether additional benefits can be obtained, and the activity compensation unit 120 and induction The value calculated by the compensation unit 130 may calculate an optimal activity compensation amount and an induction compensation amount for obtaining additional gains.

For example, the optimizer 140 may group users of types having similar environment information into a unique group through a regression tree model based on environment information for each user obtained through the activity analysis unit (S1).

In addition, the optimization unit 140 fixes the activity compensation amount (or the total activity compensation amount) within the unique group, and interlocks with the induction compensation unit 130 to determine the relationship between the induction compensation amount and the sales as a first regression equation. After the calculation (S2), a first process of calculating the maximum profit A that can be obtained when the amount of induction compensation is reduced by the first regression equation may be performed (S3).

In addition, the optimization unit 140 fixes the induction compensation amount and calculates a second regression equation for the relationship between the activity compensation amount (or the total activity compensation amount) and sales in connection with the activity compensation unit 120 (S4) A second process of calculating the maximum profit B that can be obtained when the activity compensation is reduced by the second regression equation may be performed (S5).

In addition, the optimization unit 140 repeatedly performs the first and second processes, so that the maximum profit A and the maximum profit B converge (S6) according to the repeated execution of the first and second processes. It is possible to calculate the compensation amount and the optimal induction compensation amount (S7).

In addition, the optimization unit 140 interlocks with the activity compensation unit 120 and the induction compensation unit 130 to provide the user with at least one of an activity compensation amount and an induction compensation amount according to the satisfaction of the payment condition. At least one of the optimal activity compensation amount and the optimal induction compensation amount calculated correspondingly may be paid in correspondence to the user.

Meanwhile, according to the above-described configuration, the optimizer 140 may group illegal users who indiscriminately click on advertisements for compensation only without participating in advertisements into a specific group through the regression tree model, and the regression tree In the case of fraudulent users grouped into the specific group through the model, sales are calculated very low compared to the compensation amount, so it is easy to optimize the activity compensation amount and the induction compensation amount to a low value or not to pay compensation for such groups of fraudulent users. Can block fraudulent users.

In addition, the first and second deep learning algorithms described in the present invention may be composed of one or more neural network models, and the neural network model (or neural network) includes an input layer, one or more hidden layers, and an output layer ( Output Layer).

In addition, as an example of the neural network model, various types of neural networks such as a deep neural network (DNN), a recurrent neural network (RNN), a convolutional neural network (CNN), and a support vector machine (SVM) may be applied.

In addition, the activity compensation amount and the user compensation amount described in the present invention may be rewards, and in this case, the rewards paid to users may be self-rewards that can be used only in a specific service or an integrated reward that is universally used in various services.

In addition, the reward may be a cache in the case of web, and is mapped to values that can distinguish users such as IFA and GAID in the case of mobile, and is mapped to the user through login and phone number authentication, and can be integrated and managed for each user. I can.

In addition, in the case of an integrated reward, the service providing device 100 supports the user to select and selectively receive a reward that the user wants to acquire, and can support converting the reward into another reward.

In addition, the service providing device 100 may or may not provide dynamic rewards as described above for contents aimed at promoting user participation, such as various contents, events, leaflets, as well as advertisements, and dynamically increase the rewards. Can be reduced or reduced.

In addition, the service providing apparatus 100 may be configured in various platforms that attract participation of content through the above-described configuration.

As described above, the present invention provides the user with an activity compensation according to the advertisement participation for the user's activity related to the advertisement participation, but divides the total activity compensation amount within the compensation period to maintain the user's advertisement participation-related activity. Through deep learning-based learning, the best reward period and the reward amount for each reward point that can be recognized by the user that the reward amount increases according to the level of activity are dynamically changed through deep learning-based learning, so that the user can participate in advertising activities. Not only can you increase revenue by maximizing the user's activity that is linked to sales while maintaining it as much as possible, but also other users and users with similar ad conversion patterns are grouped into a specific group based on the events the user generates for advertisements. After calculating the correlation between the segment-specific characteristics of the specific group including the induction compensation amount for inducing the user to participate in the advertisement as a segment and the change in the advertisement environment according to the change in the induction compensation amount through deep learning-based learning, a specific group Whenever an advertisement for a user belonging to a user belongs to an advertisement, it is possible to increase the user's advertisement participation-related activities by presenting an induction reward amount that can induce the user's participation in the advertisement, and according to the change of the user's advertisement conversion pattern when the user's advertisement conversion pattern changes. By dynamically changing and providing the induction compensation amount in consideration of the characteristics of the subordinate group, the advertisement efficiency can be greatly increased by continuously inducing the user to participate in advertisement.

In addition, the present invention can provide the optimal activity compensation and induction compensation in which the profits obtained from each of the activity compensation and induction compensation are mutually converged through the regression tree model, and through this, the profit while minimizing the compensation provided through advertisement You can maximize advertising efficiency.

On the other hand, based on the above-described configuration, it is applied to various advertisement models based on various advertisement platforms that are already provided to provide a fixed compensation for each advertisement participation. By applying a dynamic compensation model composed of a service providing device supporting compensation to the corresponding advertisement model, it is possible to replace the existing static compensation and support dynamic compensation and induction compensation to be performed in the corresponding advertisement model.

To this end, the service providing system supporting dynamic compensation related to advertisements according to an embodiment of the present invention includes an advertisement management server that selects and allocates advertisement information to be transmitted to a user terminal, and advertisement information allocated to the user terminal to the user terminal. A service server that generates rewards for payment to a user based on event information related to the user's advertisement activity received from the user terminal in response to the advertisement information and accumulates corresponding to the user of the user terminal, the advertisement management server, and By interlocking with the service server, the advertisement information to be transmitted to the user terminal is transmitted to the user terminal through the service server, and event information transmitted from the service server in response to the advertisement information is received, and the user In response to the advertisement information transmitted to the terminal, when a preset sales-related event occurs, the total reward amount related to the rewards paid to the user of the user terminal is updated, and in response to the updated total compensation amount, the service server is activated dynamically After determining the compensation amount, it may be configured to include the service providing device corresponding to the user.

In addition, the service providing device is associated with the user, analyzes the advertisement conversion pattern of the user each time the event-related event information is collected, and then creates a specific group by grouping the users with one or more other users having similar advertisement conversion patterns. , Maximizing the expected benefit per cost based on the learning by calculating the predicted environmental information corresponding to the specific group, which is the predicted advertising environment, and learning an induction reward amount for inducing the user's activity in which the predicted environmental information is improved. An induction compensation amount may be determined, a payment condition for paying the induction compensation amount may be set in the advertisement information to be transmitted to the user terminal, and transmitted to the user terminal through the service server.

Through the above-described configuration, the present invention also applies to a service server that operates and executes an advertisement model configured to perform advertisement by transmitting advertisement information through a specific application executed in a user terminal such as SNS (Social Networking Service) or a game. The dynamic compensation model according to the embodiment can be applied, and through this, advertising efficiency can be greatly improved as described above.

At this time, the service providing device supporting dynamic compensation related to advertisement according to an embodiment of the present invention is configured to communicate with a service server operating the advertisement model through a communication network or to be included as a module in the service server, and Dynamic compensation and induction compensation may be applied to the advertisement information provided through the advertisement to perform advertisement, as well as dynamic compensation and induction compensation by being installed or inserted in the service server or advertisement model in the form of software.

The various devices and components described herein may be implemented by hardware circuitry (eg, CMOS-based logic circuitry), firmware, software, or a combination thereof. For example, it may be implemented using transistors, logic gates, and electronic circuits in the form of various electrical structures.

The above contents may be modified and modified without departing from the essential characteristics of the present invention by those of ordinary skill in the technical field to which the present invention belongs. Accordingly, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention, but to explain the technical idea, and the scope of the technical idea of the present invention is not limited by these embodiments. The scope of protection of the present invention should be interpreted by the following claims, and all technical ideas within the scope equivalent thereto should be interpreted as being included in the scope of the present invention.

Claims

An advertisement providing unit for receiving advertisement information from an advertisement management server that stores a plurality of different advertisement-related advertisement information, and transmitting the advertisement information to a user terminal; And

In response to the advertisement information transmitted to the user terminal, when a preset sales-related event occurs, the total compensation amount paid to the user of the user terminal is updated, and advertisement environment-related environment information is calculated in correspondence with the updated total compensation amount, When an event corresponding to a payment condition set in advance in the advertisement information occurs by learning a plurality of parameters including the total number of rewards and a collapse rate for which the environmental information is improved, the optimal value calculated for each of the plurality of parameters is determined based on the learning. Accordingly, an activity compensation unit that determines an activity compensation amount, which is part of the updated total compensation amount, and then provides the activity compensation amount corresponding to the user.

A service providing device supporting dynamic compensation related to advertisements comprising a.
The method according to claim 1,

The activity compensation unit supports dynamic advertisement-related compensation, characterized in that it receives the event-related event information from the user terminal whenever an event according to the user's activity occurs for each advertisement information transmitted to the user terminal in connection with the advertisement providing unit. Service providing device.
The method according to claim 1,

The environment information includes a plurality of preset parameters for each attribute, and the plurality of attributes includes a retention rate, click rate, sales, and number of advertisement impressions per user.
The method according to claim 1,

The activity compensation unit

When the event occurs, a plurality of parameters including the total number of compensations and a collapse rate corresponding to the updated total compensation amount are calculated according to a preset first equation, and the correlation between the plurality of parameters and the advertisement environment is learned. Computing the advertising environment-related environment information by applying the total number of rewards and the collapse rate calculated in correspondence with the updated total reward amount to the first deep learning algorithm, and learning the plurality of parameters for which the environmental information is improved. After calculating the optimum values for each of the plurality of parameters through the first reinforcement learning algorithm and the first deep learning algorithm, the second formula is set in advance when an event corresponding to a payment condition set in advance in the advertisement information provided through the advertisement providing unit occurs. A service providing apparatus for supporting dynamic compensation related to advertisement, characterized in that, by applying the optimum value, an activity compensation amount, which is a part of the total compensation amount, is determined and paid.
The method of claim 4,

The first formula is

Is,

R is the total compensation amount, T is the total number of compensation, λ is the collapse rate, t is the number of times the advertisement is viewed, and N 0 is a service providing apparatus supporting dynamic compensation related to advertisement, characterized in that a constant.
The method of claim 4,

After determining the optimum value for each of the plurality of parameters, the activity compensation unit

Depending on the event, the activity compensation amount r n to be paid to the specific time t n is determined, where λ is the collapse rate, t is the number of times the advertisement is viewed, and N 0 is a constant. A service providing device that supports dynamic compensation related to advertising.
The method according to claim 1,

Whenever the event-related event information related to the user is collected, an advertisement conversion pattern of the user is analyzed, and then at least one other user having a similar advertisement conversion pattern and the user are grouped to create a specific group, and corresponding to the specific group Calculate expected environmental information, which is an expected advertising environment, and determine an induction compensation amount for maximizing an expected benefit per cost based on the learning by learning an induction compensation amount for inducing a user's activity in which the expected environmental information is improved, And an induction compensation unit for setting a payment condition for paying the induction compensation amount in advertisement information provided by the advertisement providing unit and transmitting it to the user terminal.
The method of claim 7,

The induction compensation unit

By applying the set value for each segment of the specific group to a preset second deep learning algorithm in which the correlation between a plurality of preset segments and the expected advertisement environment corresponding to the segment is learned, it is expected to correspond to the specific group. Estimated gain per cost through a preset second reinforcement learning algorithm and the second deep learning algorithm that calculates the predicted environment information, which is an advertisement environment, and learns an induction reward amount for inducing a user's activity in which the predicted environment information is improved A service providing device supporting dynamic compensation related to advertisement, characterized in that determining an induction compensation amount that maximizes.
The method of claim 7,

A regression tree model pre-set with the set value for each segment and the environment information excluding segments related to the total activity compensation amount, the induction compensation amount, and the induction compensation amount obtained for a plurality of different users in connection with the activity compensation unit and the induction compensation unit After grouping users with similar environmental information into a unique group, a first regression equation for the relationship between the total activity compensation amount and sales for the unique group, and the relationship between the induction compensation amount and sales The activity compensation is calculated by calculating a second regression equation, and calculating an optimal activity compensation amount and an optimal induction compensation amount in which the maximum profits obtained from each of the first and second regression equations converge through the regression tree model. A service providing apparatus for supporting dynamic compensation related to advertisements, further comprising an optimization unit configured to set the sub and induction compensation unit.
The method of claim 9,

The optimization unit interlocks with the activity compensation unit and the induction compensation unit, and when at least one of an activity compensation amount and an induction compensation amount according to the satisfaction of the payment condition is required, the optimal activity compensation amount calculated in response to the user and induction A service providing device supporting dynamic compensation related to advertisement, characterized in that at least one of the compensation amounts is paid.
An advertisement management server for selecting and assigning advertisement information to be transmitted to the user terminal;

Respond to the user of the user terminal by transmitting the advertisement information allocated to the user terminal to the user terminal and generating a reward for payment to the user based on the event information related to the user's advertisement activity received from the user terminal in response to the advertisement information A service server to be accumulated; And

Event information transmitted to the user terminal through the service server by interworking with the advertisement management server and the service server, and transmitted from the service server to the user terminal in response to the advertisement information In response to the advertisement information transmitted to the user terminal by receiving and updating the total compensation amount related to the rewards paid to the user of the user terminal when a preset sales-related event occurs, and corresponding to the updated total compensation amount, advertising environment related The plurality of parameters based on the learning when an event corresponding to a payment condition preset in the advertisement information occurs by learning a plurality of parameters, including the total number of rewards and a collapse rate for which the environmental information is improved and the environmental information is improved A service providing device that determines an activity compensation amount, which is a part of the updated total compensation amount according to the optimum value calculated for each, and then pays the activity compensation amount in correspondence to the user

A service providing system that supports dynamic compensation related to advertisements including a.
The method of claim 11,

The service providing device analyzes the advertisement conversion pattern of the user each time the event-related event information is collected in relation to the user, and then creates a specific group by grouping the user with one or more other users having similar advertisement conversion patterns, and the Induction compensation for maximizing the expected benefit per cost based on the learning by calculating the expected environment information, which is an expected advertising environment corresponding to a specific group, and learning the amount of induction compensation for inducing the activity of the user whose expected environment information is improved A service providing system supporting dynamic compensation related to advertisement, characterized in that the payment condition for determining an amount and paying the induction compensation amount is set in the advertisement information to be transmitted to the user terminal and transmitted to the user terminal.
In a service providing method for supporting dynamic compensation related to advertisement of a service providing device that transmits advertisement information to a user terminal,

When a predetermined sales-related event occurs in response to the advertisement information transmitted to the user terminal, the total compensation amount paid to the user of the user terminal is updated, and the total number of compensation corresponding to the updated total compensation amount according to the first predetermined formula And calculating a plurality of parameters including a decay rate.

Calculating environment information related to the advertisement environment by applying the total number of compensations and a collapse rate calculated in correspondence with the updated total compensation amount to a first deep learning algorithm in which the correlation between the plurality of parameters and the advertisement environment is learned ; And

After calculating the optimum values for each of the plurality of parameters through the first reinforcement learning algorithm for learning the plurality of parameters for which the environmental information is improved and the first deep learning algorithm, the advertisement information provided through the advertisement providing unit When an event corresponding to a set payment condition occurs, applying the optimum value to a preset second formula to determine and pay an activity compensation amount, which is a part of the total compensation amount

A service providing method that supports dynamic compensation related to advertisements including a.
The method of claim 13,

Analyzing the advertisement conversion pattern of the user each time the event-related event information is collected related to the user, and then grouping the user with one or more other users having similar advertisement conversion patterns to create a specific group;

Ads expected in response to the specific group by applying a set value for each segment of the specific group to a second preset deep learning algorithm in which a correlation between a plurality of preset segments and an expected advertisement environment corresponding to the segment is learned Calculating predicted environment information as an environment; And

Determine an induction compensation amount for maximizing an expected benefit per cost through a preset second reinforcement learning algorithm and the second deep learning algorithm for learning an induction compensation amount for inducing an activity of a user whose expected environment information is improved, and the Setting a payment condition for paying an induction compensation amount in advertisement information provided by the advertisement providing unit and transmitting it to the user terminal

A service providing method for supporting dynamic compensation related to advertisement, characterized in that it further comprises.
The method of claim 14,

After applying the total activity compensation amount obtained for a plurality of different users, the induction compensation amount, the segment-specific setting value, and the environment information to a preset regression tree model, users with similar environment information are grouped into a unique group, and the unique For a group, a first regression equation for the relationship between the total activity compensation amount and sales and a second regression equation for the relationship between the induced compensation amount and sales are calculated, and the first regression equation and the second regression equation By calculating the optimal activity compensation amount and the optimal induction compensation amount, in which the maximum profits obtained from each mutually converge, through the regression tree model, at least one of the optimal activity compensation amount and the optimal induction compensation amount when the payment condition is satisfied. Steps to pay

A service providing method for supporting dynamic compensation related to advertisement, characterized in that it further comprises.