WO2019083142A1 - Method, server, and computer program for determining whether campaign is exposed - Google Patents

Abstract

The present invention relates to a method, a server, a computer program capable of adjusting a campaign exposure frequency customized for each game user, by using game logs of multiple game users. To this end, a method for determining whether a campaign is exposed according to an embodiment of the present invention may comprise: collecting and learning past game logs of multiple game users for each campaign group; generating a reaction probability prediction model for each campaign group on the basis of a learning result; receiving an exposure or non-exposure inquiry signal from a user terminal, and extracting a reaction probability determination factor for a target campaign from a past game log of a game user by the user terminal; calculating a reaction probability of the game user for the target campaign, by applying the reaction probability determination factor for the game user to a reaction probability prediction model of a campaign group to which the target campaign belongs; and determining, using the reaction probability, whether the target campaign is exposed.

Description

How to determine campaign exposure, server and computer programs

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for determining whether a campaign is exposed, a server and a computer program, and more particularly, A method of controlling exposure frequency, an exposure control server, and a computer program.

Smart devices such as smart TVs and smartphones are becoming popular, and the number of users who enjoy playing games using these smart devices is gradually increasing. In particular, users who use a smartphone capable of Internet communication can enjoy the game through the Internet anytime and anywhere through the mobility and portability of the smartphone.

In order to increase the interest of the game user and increase the sustainability of the game, the mobile game provider may periodically or intermittently perform various events and make various suggestions through the campaign to allow the game user to take desired actions of the mobile game provider. Here, the campaign is a method in which a game user requests a behavior desired by a game provider in a game, and examples thereof are shown in Figs. 1A and 1B.

 Campaigns can occur when a game user connects to a game and when certain conditions are reached during the game. Here, the types of campaigns include the event notification campaigns and purchase proposal campaigns shown in FIGS. 1A and 1B. For example, in addition to the campaigns shown in FIGS. 1A and 1B, the campaign may include an information providing campaign for notifying various game information, a game in-game consumption guide guide campaign, and the like, but the type of campaign is not limited thereto.

These campaigns are common to all game users, and their tendencies and game patterns are different for each game user, making it difficult to satisfy all game users. For example, a novice game guide campaign for a novice may be helpful for a game user who has just been playing the mobile game, but the proposal of the corresponding campaign is unnecessary for a game player who played for a certain period of time. In addition, certain campaigns may be repeatedly provided to game users for a predetermined period of time each time they access the game application. In this case, game users who are not interested in the campaigns should close the campaigns and play the game. In other words, even a game user who is interested in the campaigns has a large number of campaigns, and the frequency of the frequent exposure is so inconvenient that the game is inconvenient.

As shown in FIG. 1A, a specific area may be located at the bottom or top of the once-exposed campaign, and a corresponding campaign may be selected for a preset period (for example, a day) So that it is not exposed during the exposure. However, this method also fails to completely solve the problem of annoying game users who are not interested in the campaign, because the campaign is repeatedly provided to game users at least once every preset period.

Therefore, a new technique for selectively providing a campaign according to the interest of game users is required.

In this respect, Korean Patent Laid-Open Publication No. 2013-0131845 is available.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and it is an object of the present invention to provide a method of determining whether a campaign is exposed by selectively providing a campaign according to the degree of interest of a game user, Server and a computer program.

According to another aspect of the present invention, there is provided a method of determining exposure of a target campaign to be provided to a user terminal equipped with a game application, the method comprising: collecting past game logs of a plurality of game users for each campaign group; ; Learning past game logs; Generating a response probability prediction model for each campaign group based on the learning result; Receiving an exposure / non-contact query signal from a user terminal, and extracting a reaction probability determination factor for a target campaign from a past point game log for a game user of the user terminal; Calculating a response probability of the game user with respect to a target campaign by applying a reaction probability determination factor of the game user to the reaction probability prediction model of the campaign group to which the target campaign belongs; And determining whether the target campaign is exposed using the response probability.

In addition, the past game log includes the campaign exposure time of the past campaigns provided to the game user, and the learning of the past game logs may be performed using the campaign exposure times of the past campaigns provided to a plurality of game users.

The step of learning past game logs may include arranging campaign exposure times of past campaigns according to the size of time; Determining a first time and a second time by analyzing the alignment result; And filtering campaign exposures of past campaigns that are less than or greater than the first time.

In the response probability prediction model for each campaign group, the response probability may be modeled such that the response probability decreases as the campaign exposure time approaches the first time, and increases as the campaign exposure time approaches the second time.

In addition, the step of generating the reaction probability prediction model for each campaign group may model the response probability for the campaign exposure times that is less than the first time or exceeds the second time to be zero.

In addition, the response probability determination factor may be derived based on campaign exposure times of past campaigns provided to the game user of the user terminal.

Also, the past game log may include information on whether or not to accept the proposal according to past campaigns provided to each game user, and the step of learning past game logs may be performed by using a plurality of game users' proposal acceptance information.

The step of learning past game logs may include extracting feature values and proposal acceptance information of each game user in past game logs, And learning the feature values and proposal acceptability information of each game user.

In addition, the feature values represent values for predefined features of game users, and the features include game user level, in-game good, in-game good, purchase probability, user status information, campaign exposure time, Or the like.

In addition, the step of determining whether the target campaign is exposed may include: determining an exposure determination value; And determining that the target campaign is to be exposed when the response probability of the game user exceeds the exposure determination value.

In addition, the step of determining the exposure determination value may be performed by determining a reaction probability of a game user corresponding to a predetermined upper percentage of all game users as the exposure determination value.

The method further includes determining that the target campaign is to be exposed when the probability of the game user's reaction exceeds the exposure determination value and the difference between the current time and the latest exposure time exceeds the minimum exposure time, It may be the time of presentation of the last campaign provided to the user terminal of the campaign group including the target campaign.

The method may further include confirming that the target campaign is to be exposed when a reaction probability of a game user is less than or equal to the exposure determination value and a difference between the current time and the latest exposure time exceeds a maximum exposure frequency time.

In addition, the step of determining whether to expose the target campaign may include deriving a random value between 0 and 1 through a random function; And determining that the target campaign is to be exposed if the response probability of the game user exceeds a random value.

According to another aspect of the present invention, there is provided an exposure control server comprising: a game log collecting unit collecting past game logs of a plurality of game users for each campaign group; A learning unit for learning past game logs; A model generation unit that generates a response probability prediction model for each campaign group based on the learning result; A request receiving unit for receiving an exposure / non-contact query signal from a user terminal; The response probability determination factor for the target campaign is extracted from the past game log of the game terminal for the game user of the user terminal and the response probability determination factor of the game user is applied to the reaction probability prediction model of the campaign group to which the target campaign belongs, A reaction probability calculation unit for calculating a reaction probability of a game user; And an exposure determination unit for determining whether to expose the target campaign using the response probability, and each step according to the above-described method is executed.

In addition, a computer program stored on a computer-readable recording medium for executing the above-described method on a computer is provided.

Further, a computer-readable recording medium on which a computer program for performing the above-described method is recorded is provided.

The method for determining whether a campaign is exposed or not, the exposure control server and the computer program of the present invention can be applied to not only the past game logs of a plurality of game users but also the campaigns optimized for the game users There is an effect that the exposure frequency can be adjusted.

In addition, the method of determining whether or not a campaign is exposed according to the present invention, the exposure control server and the computer program can be performed through a separate exposure control server rather than a user terminal, thereby minimizing the load on the user terminal, Game play becomes smooth.

In addition, the method of determining whether or not the campaign is exposed according to the present invention, the exposure control server and the computer program utilize the data according to the provision of the campaign group to which the target campaign belongs to the target campaign to be provided to the game user. You can get a more accurate picture of your interest. Of course, this interest can be extended to campaign units rather than to grouping campaigns.

In addition, the method for determining whether or not a campaign is exposed according to the present invention, the exposure control server and the computer program use a game log of a past time point, and only use a game log within a predetermined time period based on the current time point, And the frequency of exposure of the campaign is also determined based on the recent behavior of the game user, so that the effect can be adaptively changed to the user behavior change.

Figures 1a and 1b are illustrations of examples for illustrating the concept of a campaign.

2 is a conceptual diagram of a system for determining whether or not a campaign is exposed according to an embodiment of the present invention.

3 is a conceptual diagram for explaining the concept of a campaign group.

4 is a block diagram of a user terminal according to an embodiment of the present invention.

5 is a block diagram of an exposure control server according to an embodiment of the present invention.

FIG. 6 and FIG. 7 illustrate a method of generating a response probability prediction model for each campaign group through the exposure control server according to an embodiment of the present invention.

8A and 8B are views for explaining a method of generating a response probability prediction model for each campaign group through the exposure control server according to an embodiment of the present invention.

9 is a flowchart illustrating a method of determining whether a campaign is exposed through the exposure control server according to an exemplary embodiment of the present invention.

10 is a flowchart of a learning step according to an embodiment of the present invention.

11 is a flowchart of a learning step according to another embodiment of the present invention.

12 is a flowchart illustrating a step of determining whether a target campaign is exposed according to an exemplary embodiment of the present invention.

13 is a flowchart illustrating a step of determining whether a target campaign is exposed according to another exemplary embodiment of the present invention.

The present invention will now be described in detail with reference to the accompanying drawings. Hereinafter, a repeated description, a known function that may obscure the gist of the present invention, and a detailed description of the configuration will be omitted. Embodiments of the present invention are provided to more fully describe the present invention to those skilled in the art. Accordingly, the shapes and sizes of the elements in the drawings and the like can be exaggerated for clarity.

Hereinafter, a campaign exposure determination system according to an embodiment of the present invention will be described.

2 is a conceptual diagram of a system 1000 for determining whether or not a campaign is exposed according to an exemplary embodiment of the present invention. 2, the campaign exposure determination system 1000 may include user terminals 100 and 200 and an exposure control server 300 according to an exemplary embodiment of the present invention. In FIG. 2, the number of user terminals and the number of game users are shown as two, but it should be understood that these are abbreviated to facilitate understanding of the present invention. In addition, since the flow of operation through each of the user terminals 100 and 200 is substantially the same, the operation of the user terminal 100 will be described below.

The user terminal 100 represents a terminal that the game user has and in which the game application is installed. For example, the user terminal 100 can be any type of device as long as it is a device capable of installing and executing a mobile device or a game application such as a smart phone or a tablet PC. The game application may include a game running on a device such as a computer, a laptop computer, a console game machine, etc., and a mobile game running on a portable terminal (e.g., a smart phone or a tablet PC) When the execution input of the application is received, the user terminal 100 executes the corresponding game application and accesses the game server 10 through the network 20. [

Here, the game application may include a plurality of campaigns referred to in the background description item together with the game. The campaigns may be stored in advance in the user terminal at the time of installation of the game application. In addition, the mobile game provider may periodically or intermittently provide updates for the game application, which may be added to the update and stored in the user terminal. The campaign may be directly transmitted from the game server 10 and stored in the user terminal 100 when the connection between the user terminal 100 and the game server 10 is performed.

When the connection between the user terminal 100 and the game server 10 is completed, the user terminal 100 determines whether the campaign exposure time of the campaign stored in the user terminal 100 or transmitted from the game server 10 has arrived. As a result of the determination, when it is determined that the campaign exposure time has arrived, the user terminal 100 selects at least one target campaign among the stored or received campaigns, and inquires whether the selected target campaign is exposed.

As described above, the campaign exposure determination system 1000 according to an embodiment of the present invention provides a customized campaign to each game user rather than providing the same campaign to each game user. For this customization process, the campaign exposure determination system 1000 according to an embodiment of the present invention includes an exposure control server 300, and the exposure control server 300 determines whether or not the user queries the user terminals 100 and 200 It determines the exposure of the target campaign.

The exposure control server 300 is different from the game server 10 in that the game server 10 processes and returns data so that the game can be played during the progress of the game while the exposure control server 300 Uses the game log indicating the game play situation from the game application to determine whether or not the target campaign that is queried by each of the user terminals 100 and 200 is exposed. Here, the game log refers to all the raw data generated in the process of installing, executing, or removing a game application on a user terminal of a game user. In one embodiment, the reaction information of the game user with respect to a campaign to be described later may also be included in the game log.

3 is a conceptual diagram for explaining the concept of a campaign group.

The exposure control server 300 determines whether to expose the target campaign using the past game logs of a plurality of game users. Specifically, the exposure control server 300 may determine whether to expose the target campaign using past game logs of a plurality of game users according to provision of past campaigns belonging to the same campaign group as the target campaign. Here, the past game logs (i.e., past game logs based on the provision of past campaigns belonging to the same campaign group as the target campaign) are used because the game users have different tendencies and game patterns for each game user. For example, some game users may be interested in event notification campaigns but not others, and some other game users may be interested in purchase offer campaigns but not others. Therefore, when determining whether to expose a target campaign, it is desirable to utilize reference data having the same or similar purpose as the target campaign.

Referring to FIG. 3, the exposure control server 300 according to an exemplary embodiment of the present invention can group campaigns into a plurality of campaign groups (cg ₁ , cg ₂ , cg _m ) according to its purpose. As shown in Figure 3, includes a plurality of campaigns, each campaign is similar purpose group _{(cg 1, cg 2, cg} m). For example, the first campaign group cg ₁ may be a group of campaigns whose purpose is an event notification, the second campaign group cg ₂ may be a group of campaigns whose purpose is a purchase proposal, (cg _m ) may be a group of campaigns whose objectives are game guides.

Thereafter, the exposure control server 300 generates the reaction probability prediction models for each campaign group by learning past game logs according to the provision of the past campaigns included in the respective campaign groups cg ₁ , cg ₂ , cg _m . For example, if the number of campaign groups is three, the exposure control server 300 may generate at least one response probability prediction model for each campaign group. In this example, A reaction probability prediction model of the second campaign group, and a reaction probability prediction model of the third campaign group. Of course, if there are a plurality of campaign groups, the number of reaction probability prediction models may be changed by the number of campaign groups.

The exposure control server 300 may use at least one of the campaign exposure time and the proposal acceptance information in the information included in the past game logs at the time of learning the past game logs. Specifically, the exposure control server 300 collects past game logs of a plurality of game users according to the provision of past campaigns belonging to each campaign group for each campaign group, calculates campaign exposure times and suggestions The reaction probability prediction model may be generated or updated using at least one of the acceptability information. The exposure control server 300 can determine whether to expose the target campaign using the reaction probability prediction model generated or updated. Here, the reaction probability prediction model is a model generated through a preset learning technique, and predicts a game user's interest (i.e., reaction probability) with respect to the target campaign.

Thereafter, the exposure control server 300 extracts a reaction probability determination factor for the target campaign in the game log for the game user of the user terminal 100 that has transmitted the exposure / non-query signal. Then, the exposure control server 300 calculates the response probability for the target user's campaign by applying the extracted response probability determination factor to the reaction probability prediction model of the campaign group to which the target campaign belongs. For example, if the target campaign is a campaign belonging to the second campaign group, the exposure control server 300 applies the response probability determination factor to the response probability prediction model of the second campaign group among the plurality of reaction probability prediction models, Probability can be calculated. When the calculation of the reaction probability is completed, the exposure control server 300 can determine whether to expose the target campaign using the calculated reaction probability.

The target campaign in which the exposure is inquired may be exposed through the operations described above with respect to the exposure control server 300, and may be selectively displayed on the user terminal 100 according to the exposure. In addition, although a description has been made for one target campaign in the above, the operations can be respectively performed for a plurality of target campaigns, so that the user terminal 100 and the user terminal 200 are exposed Target campaigns can be different.

For example, assume that two user terminals exist in the user terminal 100 and the user terminal 200. As a result of the determination by the exposure control server 300, the exposure control server 300 determines that the first target campaign is exposed but the second target campaign is not exposed to the user terminal 100, and the first target campaign is exposed to the user terminal 200 However, if it is determined to expose the second target campaign, different target campaigns may be exposed to the two user terminals 100 and 200 as shown in FIG. Of course, it is also possible that the same target campaigns are exposed to the two user terminals 100, 200 if the game users of the two user terminals 100, 200 have similar tendencies and similar levels. Of course, depending on the situation, there may be situations in which no target campaign is exposed to the user terminal.

As described above, according to the campaign exposure determination system 1000 according to the embodiment of the present invention, the campaign can be selectively provided to each game user according to the degree of interest (for example, reaction probability) of each game user. Therefore, according to the campaign exposure determination system 1000 of the present invention, it is possible to provide more efficient information or offer products. In addition, the amount of campaigns not interested in each game user can be drastically reduced, thereby meeting the needs of all game users using the game.

4 is a block diagram of a user terminal 100 according to an embodiment of the present invention. As described above, the user terminal 100 queries the exposure control server 300 whether the target campaign is exposed when the exposure time of the campaign comes, and exposes the target campaign according to the query result. The user terminal 100 according to an exemplary embodiment of the present invention includes a campaign exposure time determination unit 110, an exposure target campaign selection unit 120, an exposure query unit 130, a terminal communication unit 140, An exposure unit 150 and a terminal storage unit 160. [ Here, the above-described configurations are described for each function in order to facilitate understanding of the present invention. The remaining configurations of the above configurations, except for the terminal communication unit 140 and the terminal storage unit 160, include a single processing unit such as a CPU and an MPU It is also possible to implement it through.

The campaign exposure time determination unit 110 determines the exposure time of the campaign to be provided to the user. A campaign in a game can be provided to a game user when certain conditions are met. The conditions may include, for example, using a content more than a certain number of times during a game connection, failing an adventure over a certain number of times, losing in a game, and the like. Of course, the above conditions may include various situations besides those described above.

The exposure target campaign selection unit 120 functions to select target campaigns to be provided to a game user among a plurality of campaigns stored in the terminal storage unit 160. [

On the other hand, when the campaign reaches the point of exposure, the target campaign selected through the exposure target campaign selection unit 120 can be instantly exposed to the game user. However, as described above, this method continuously exposes campaigns to specific game users who are not interested in a particular campaign, or the campaign is exposed too frequently, . Accordingly, the user terminal 100 according to an embodiment of the present invention may determine the exposure according to each game user through the exposure control server 300, rather than immediately exposing the campaign at that time, And exposing the campaign using an exposure control signal generated based on the exposure control signal.

To this end, the user terminal 100 includes an exposure / inquiry inquiry unit 130, generates an exposure / nonexistence inquiry signal through the exposure / inquiry inquiry unit 130, and transmits it to the exposure control server 300 ). ≪ / RTI > Here, the exposure / non-contact query signal may include an identifier of the user terminal 100 and information on the type (campaign group) or identifier of the target campaign.

When the exposure control information for the target campaign is received from the exposure control server 300 after the transmission of the exposure query message, the campaign exposing unit 150 may selectively expose the target campaign based on the exposure control information. As described above, since the exposure control information considers the interest of the game user (for example, reaction probability), the user terminal 100 according to the embodiment of the present invention can provide an optional campaign according to the game user .

5 is a block diagram of an exposure control server 300 according to an embodiment of the present invention. The exposure control server 300 according to an embodiment of the present invention is characterized in that it determines how much the game user is interested in the target campaign. To this end, the exposure control server 300 may generate a reaction probability prediction model for each campaign group using past data (i.e., past game logs of a plurality of game users according to past campaigns). The exposure control server 300 determines whether to expose the target campaign in the user terminal 100 using the response probability prediction model for each campaign group when receiving the exposure / non-presence query signal from the user terminal 100, And generates exposure control information according to the generated exposure control information.

The exposure control server 300 according to an embodiment of the present invention includes a game log collecting unit 310, a server storing unit 320, a learning unit 330, a model generating unit 340, a request receiving unit 350 A server communication unit 360, a reaction probability calculation unit 370, and an exposure determination unit 380. [ The components other than the server storage unit 320 and the server communication unit 360 are connected to the CPU and the MPU 360. The components of the exposure control server 300 are the same as the components of the exposure control server 300, And the like. Now, each configuration included in the exposure control server 300 is explained.

The server storage unit 320 stores game logs corresponding to game progress of a plurality of game users. As described above, the game log means all the raw data generated in the process of installing, executing, and removing the game application on the user terminal of the game user. Also, as described below, the game log includes at least one of a campaign exposure time of past campaigns provided to each game user, feature values according to characteristics set for each game user, and whether the game user reacts to the provision of past campaigns And the like.

Here, the game log may be stored in the server storage unit 320 through communication with the user terminal 100 through the server communication unit 360. For example, the user terminal 100 may transmit a game log generated periodically or whenever a specific situation occurs to the exposure control server 300, and the exposure control server 300 may transmit the received game log to the game user May be stored in the server storage unit 320. Alternatively, when the game log of the game user is separately managed by the game server, the game log may be collected through communication with the game server.

The game log collecting unit 310 collects game logs stored in the server storage unit 320. [ Specifically, the game log collecting unit 310 collects past game logs of a plurality of game users according to provision of past campaigns classified by campaign group. Here, past campaigns are defined as campaigns that have been provided at least once to a plurality of game users to help understand the present invention. In addition, past game logs are defined as past game logs generated according to the past campaign.

As described below, the exposure control server 300 according to an embodiment of the present invention generates a response probability prediction model for each campaign group. Here, the reaction probability prediction model can be largely generated by using one of three methods. Specifically, the first method is a method using a campaign exposure time, the second method is a method using a reaction of a game user according to a suggestion of a campaign, and the third method is a method using both a first method and a second method . The exposure control server 300 may include a learning unit 330 and a model generation unit 340 to generate a response probability prediction model for each campaign group in consideration of the above methods.

The learning unit 330 functions to learn past game logs, and the model generation unit 340 generates a reaction probability prediction model for each campaign group based on the learning results.

In the case of the first scheme, the learning unit 330 can extract the campaign exposure times measured when the past campaigns are provided to the game user for each campaign group by analyzing past game logs, and learn the campaign exposure times. Here, the reason why the campaign exposure times are utilized is that when a game user is interested in the corresponding campaign when one of the campaigns is provided to the game user, the exposure time of the campaign is long. On the contrary, Time is short.

For example, suppose that we have provided a guide campaign for information delivery to game users. In this case, it is possible to measure the campaign exposure time using the difference between the time of popping up the campaign window and the time of closing the campaign window to the game user, and such measurement can be performed for each game user. By applying the measured campaign exposure time per game user to the sigmoid function, it is possible to derive an interest degree (for example, reaction probability) according to the campaign exposure time.

For example, we measure the exposure time of the proposed campaign for each game user, sort the campaign exposure times in descending order, allocate label 1 to the top 10-20% of all game users, You can assign label 0 to% users. And data that is less than the upper 10% and higher than the upper 90% can all be treated as noise. This allows the campaign exposure time and label to be defined, which can be solved by a logistic regression problem.

The graph of FIG. 6 shows an example of the results through logistic regression. In the graph of FIG. 6, the x-axis represents the campaign exposure time and the y-axis represents the response probability. 6, a plurality of points represented by the region bkd represent the campaign exposure time of the entire game user, a plurality of points represented by the region rd represent users corresponding to the upper 80 to 90% (bld) represents a user corresponding to the upper 10 to 20%. Also, in the graph of Fig. 6, the S-shaped curve cl shows the results obtained by using logistic regression using only the points included in the region rd and the region bld.

According to the graph of FIG. 6, a game user having a campaign exposure time of 3 seconds is assigned a response probability of 0.05, a game user having a campaign exposure time of 5 seconds is assigned a response probability of 0.98, and a game user having a campaign exposure time of 4 seconds The response probability can be assigned to 0.5. Accordingly, the model generating unit 340 can generate a reaction probability prediction model that can represent the degree of interest, that is, the reaction probability from 0 to 1, for all game users through the above method. Such generation can be performed separately for each campaign group, so that the model generation unit 340 can generate a reaction probability prediction model for each campaign group.

In the above description, it is assumed that a game user has a high interest rate, that is, a response probability, in proportion to the exposure time of the proposed campaign. However, this assumption may not be true due to the characteristics of mobile games. For example, there may be a situation in which game users proceed to a game without viewing a display unit (i.e., a screen) of the user terminal. In this case, the time for closing pop-ups of game users may be relatively long. Conversely, quickly closing a pop-up in a campaign can mean less interest. Therefore, users who close the campaign too soon or close it too late may be less interested.

Exposure control server 300 according to one embodiment of the present invention includes a first time by analytical methods or heuristics (heuristics) a method for the campaign, the exposure time as shown in FIG. 7 in order to solve the above problems (t ₁ ) And the second time (t ₂ ). That is, the learning unit 330 may determine the first time t ₁ and the second time t ₂ by sorting the campaign exposure times of the past campaigns according to the size of the time and analyzing the sorting result. Here, the second time is greater than the first time.

Then, the learning unit 330 may be by filtering campaign exposure time to the less than 1 hour (t ₁₎ than the second time (t _2), removing the reliability falling area. In addition, the learning unit 330 may reduce the amount of processing used in the study according to proceed with the study only for campaign exposure time between the first time (t ₁₎ and the second time (t _2).

The model generation unit 340 may generate a response probability prediction model for each campaign group in consideration of the characteristics. That is, the model generating unit 340 can generate the response probability prediction model such that the response probability is close to 0 or 0 when the campaign exposure time is less than the first time or exceeds the second time. In addition, the model generating unit 340 generates a reaction probability prediction model so that the response probability becomes closer to zero as the campaign exposure time approaches the first time, and the response probability approaches 1 as the campaign exposure time approaches the second time can do.

Also, instead of the campaign exposure time, the learning unit 330 may perform learning using the information on whether or not the game users accept the proposal according to the provision of the past campaign, as mentioned in the second scheme. Here, the information on whether or not the proposal is acceptable can be expressed as 0 and 1. For example, when a game proposes an action or a purchase to a game user in the game, the game user can judge that the proposal is accepted when he / she conducts or purchases the game.

If the concept is extended to a plurality of game users and machine learning is performed using the user status of each game user as a feature vector, it is possible to generate a reaction probability prediction model for each campaign group described above. Here, the machine learning can be performed through, for example, SVM (Support Vector Machine), boosting, and random forest.

Specifically, the learning unit 330 defines at least one of the features and, for the plurality of game users in the server storage unit 320, the feature values according to the defined features, The information can be extracted and the table can be configured with the generated label. For example, the characteristics defined through the learning unit 330 may be game user level, in-game amount, in-game use amount, purchase probability, user status information, campaign exposure time, game play pattern, The characteristic defined through the learning unit 330 is not limited thereto.

Thereafter, the learning unit 330 performs learning using SVM, boosting, or random forest machine learning, and the model generating unit 340 can generate a response probability prediction model using the learning results.

For example, assume that at least one of the campaigns belonging to the same campaign group as shown in FIG. 8A is provided to five game users (A1, B2, C1, A2, D2) through the learning unit 330 . At this time, the learning unit 330 extracts feature values of the features set in the game logs of the respective game users A1, B2, C1, A2, and D2. Of course, the number of game users shown in FIG. 8A is only five for explanation, and in reality, the same process can be performed for a larger number of game users.

Learning performed through the learning unit 330 is characterized by using a game log of the past time point. Among these characteristics showing a state of the game user information may be obtained from gaming log for a period of time of the log in the game, or t _-n, or the point of time t _-2 from time t _-2. Then, the proposed acceptance information may be obtained at the time t _-1 or _{+ 1} or t _-n point later than the time t _-2 from the game log for the time period up to the time t _-1.

It is preferable that the game logs that are too old among the game logs of the past time used in the learning unit 330 are not used because the preference of the game users may be rapidly changed due to the characteristics of the mobile game. Therefore, it is preferable that the game log of the past time used in the learning unit 330 is a game log in a predetermined period based on the current time point.

In the present example, the number of features is assumed to be three for illustrative purposes, but the number of features may be applied in various other than three. Then, the learning unit 330 can extract the proposal acceptance information of the game user when the campaign is provided to the corresponding game users A1, B2, C1, A2, and D2. In this example, it is assumed that the game users A1, B2, and D2 accepted the proposal, while the game users C1 and A2 did not accept the proposal.

Thereafter, the learning unit 330 may perform supervised learning using the above-described machine learning (ml). Thereafter, the model generating unit 340 can generate a reaction probability prediction model of the corresponding campaign group using the learning results.

In the above description, the learning unit 330 and the model generating unit 340 generate the reaction probability prediction model of the campaign group using the campaign exposure time or the acceptance of the proposal. However, this is merely an example, and the reaction probability prediction model may be generated using all two parameters as mentioned in the third method above.

In this way, the learning unit 330 and the model generating unit 340 can generate a response probability prediction model for one campaign group, and apply the method to other campaign groups, thereby generating a response probability prediction model for the entire campaign group can do.

Referring again to FIG. 5, the request receiving unit 350 receives an exposure / nonexistence query signal from the user terminal 100 through communication with the user terminal 100. [ As described with reference to FIG. 4, the exposure / inquiry signal may include an identifier of the user terminal 100 and an identifier of the target campaign. At this time, the game log collecting unit 310 collects the past game log information of the corresponding game user in the server storage unit 320 for performing the control process described below, (370). Here, the past-time log information of the game user may be log information within a predetermined period based on the current time point.

The game user's past point log information may include the campaign exposure time as described below. The past game log information of the game user may include the feature values derived according to the predetermined features and the proposal acceptance information according to the proposal of the campaign. The past game log information of the game user may further include the latest exposure time information of each campaign group exposed to the game user.

The reaction probability calculation unit 370 calculates the reaction probability of the game user of the user terminal 100 that has transmitted the exposure / non-contact query signal. To this end, the reaction probability calculation unit 370 extracts a reaction probability determination factor for the target campaign from the past game log of the user of the user terminal 100 with respect to the game user. Then, the response probability calculation unit 370 calls the response probability prediction model of the campaign group to which the target campaign belongs. Here, the campaign group to which the target campaign belongs can be searched using the identifier of the target campaign.

Thereafter, the reaction probability calculation unit 370 may calculate the reaction probability of the game user for the target campaign by applying the reaction probability determination factor of the game user to the reaction probability prediction model of the target campaign group to which the target campaign belongs ( 8B). For example, when the reaction probability prediction model is generated based on the campaign exposure time, the response probability calculation unit 370 calls the campaign exposure times of past campaigns in the past game log of the game user, The probability of the response to the target campaign can be calculated. For example, when the reaction probability prediction model is generated using the campaign exposure time, the input value input to the reaction probability prediction model may be the average time of the campaign exposure times according to the past user's past campaigns. Of course, such input values can be changed and applied in various ways.

The exposure determination unit 380 determines whether to expose the target campaign using the response probability calculated through the response probability calculation unit 370. [

For example, the exposure determination unit 380 may derive a random value between 0 and 1 through a random function, and may determine to expose a target campaign when a game user's response probability exceeds a random value. Here, the random function may be a function provided in a programming language (for example, c language, Java), and rand () and srand () may be used for c language. Random class for Java, Math.random () May be used.

For example, it is possible to determine that the response probability of the game user's target campaign is 0.7, and if the random value derived through the exposure determination unit 380 is 0.5, the target campaign is exposed. Conversely, if the random value derived through the exposure determination unit 380 is 0.8, it can be determined that the exposure is not to be exposed.

However, although this approach may accomplish the object of the present invention in reducing the frequency of exposure of a campaign with a low degree of interest (e. G., Response probability), if the response probability for a particular campaign group is too low , And the probability of the reaction is 0.1 or less), there is a problem that most of the game users are not exposed to the campaign at all.

Accordingly, in one embodiment of the present invention, the exposure determination unit 380 may determine the exposure by determining the exposure determination value and comparing the response probability with the exposure determination value, rather than using the random value. In other words, the exposure determination unit 380 is independent of whether the response probability is greater than or equal to a certain size, but when the response probability is less than a predetermined value, the exposure determination value is determined and the exposure determination unit 380 can determine the exposure. For example, the exposure determination unit 380 may determine to expose the target campaign when the response probability of the game user exceeds the exposure determination value.

Here, the exposure determination value can be determined in various ways. For example, the determination of the exposure determination value may be made by determining a reaction probability of a game user corresponding to a predetermined upper percentage of all game users as the exposure determination value.

For example, it is assumed that the exposure control server 300 transmits exposure query signals for two target campaigns from three user terminals, as shown in Table 1. In Table 1, the User Key represents the identifier of each game user for the user terminal, the response probability A represents the response probability for the first target campaign, the response probability B represents the response probability for the second target campaign, Whether or not A represents the exposure to the first target campaign, and B represents the exposure to the second target campaign. Also, in the column of exposure, 1 indicates a situation determined to expose the target campaign, and 0 indicates a situation determined not to expose the target campaign.

The reaction probabilities for the first target campaign and the second target campaign can be calculated for the three game users through the reaction probability calculation unit 370 as described above. When the calculation of the reaction probability is completed, the first target campaign and the second target campaign are determined for the three game users through the determination (for example, the random value or the exposure determination value) through the exposure determination unit 380 And the result can be recorded.

In addition, in another embodiment of the present invention, the exposure determination unit 380 may perform an additional determination process in addition to the exposure determination method through the above-described determination, thereby performing the exposure determination. This is to customize the game user to determine whether the target campaign is exposed. In other words, in addition to the above-described determination method, the exposure determination unit 380 according to another embodiment of the present invention considers the latest exposure time of the campaign last exposed to the game user among the same campaign group as the target campaign .

Specifically, the exposure determination unit 380 extracts the latest exposure time of the last exposed campaign in the past game log of the game user. Thereafter, if the exposure determination unit 380 determines to expose the target campaign in the previous determination, it may further determine whether the difference between the current time and the latest exposure time exceeds the minimum exposure time. For example, a comparison is made between the difference between the current time and the latest exposure time and the minimum exposure frequency time when the probability of reaction exceeds the random value, or when the probability of reaction exceeds the exposure determination value, If you exceed the minimum impression frequency time, you can be sure to expose the target campaign.

In addition, in the above judgment, when it is determined that the target campaign is not to be exposed (that is, the response probability is less than the random value or the response probability is less than the exposure determination value), the campaign is not provided to the game user for too long Is not desirable. Accordingly, the exposure determination unit 380 may expose the target campaign if the difference between the current time and the latest exposure time exceeds the maximum exposure time even if the response probability is less than the random value or the response probability is less than the exposure determination value Can be confirmed.

The process of determining whether or not to be exposed through the exposure determination unit 380 will be further described with reference to Table 2 below.

As described with reference to Table 1, Table 2 assumes a situation in which there are three game users, and the exposure probability for each target campaign is determined using the reaction probability A and the reaction probability B representing different campaign groups. That is, as a result of the determination using the reaction probability A and the reaction probability B, it is assumed that the first game user is exposed to the first target campaign but not the second target campaign. It is assumed that the second game user is determined to expose the first target campaign but not the second campaign. And that it is determined to expose the first and second target campaigns to the third game user.

Referring to Table 2, the latest exposure time for the first campaign group of the first game user whose identifier is a9d98afb2 is 12:29:51 on February 7, 2017, and the latest exposure time for the second campaign group is 2017 It is assumed that February 1 is 14:52:1. The latest exposure time for the first campaign group of the second game user with the identifier bc98dnd18 is 18:52:35 on February 6, 2017, and the latest exposure time for the second campaign group is February 15, 2017 It is assumed that the hour is 35 minutes and 23 seconds. The latest exposure time for the first campaign group of the third game user with an identifier of c972gfk2a is 9:23:24 on February 7, 2017, and the latest exposure time for the second campaign group is February 7, 2017 Hour and 18 minutes and 33 seconds.

Also, in this example, it is assumed that the current time is 0:00:00 on February 8, 2017, the minimum exposure time is 1 day, and the maximum exposure time is 3 days.

In this case, in the original judgment, it may be determined that the first target campaign is exposed but the second target campaign is not exposed for the first game user. If the campaign is provided according to the determination result without considering the minimum exposure frequency time and the maximum exposure frequency time, the first game user will receive another campaign or campaign similar to the one already received yesterday, You can leave your experience. In addition, since the first game user has not received the campaigns belonging to the second campaign group for a long time, but the second target campaign is also not received, this may be an undesirable situation. Accordingly, it may be desirable to provide the game user with a campaign belonging to a second campaign group that has not been served for a relatively long period of time.

On the other hand, if additional judgment using the above-described minimum exposure frequency time and maximum exposure frequency time is applied, the first game user can receive the second target campaign that has not been received for a relatively long period of time instead of the first target campaign provided before a comparatively short period . If the current time in this example is February 0, 2017 0:00 0:00, the first game user can be changed to receive all of the two target campaigns.

If the current time is 0:00:00 on February 8, 2017, the difference between the current time and the latest exposure time of the first campaign group exceeds the minimum exposure time, Exposure can be confirmed. In addition, since the difference between the current time and the latest exposure time of the second campaign group is less than the maximum exposure frequency time, the second game user can be determined not to be exposed.

In addition, since the difference between the current time and the latest exposure time of the first campaign group and the second campaign group is less than the minimum exposure time, the first and second target campaigns can be changed to not be exposed.

As described above, the exposure determination unit 380 takes into account not only the response probability but also the latest exposure time of the campaign group provided to each game user, so that the frequency of the more customized campaign can be adjusted to the game user.

In the above description, the exposure control server 300 determines that the target campaign is exposed. However, it is also conceivable that the user terminal 100 performs the exposure determination process through the exposure control server 300 only. However, the load of the user terminal is high due to the characteristics of the mobile game in which the complexity of graphics or coding is high. Here, if an additional process is performed in the user terminal to determine whether or not to control the exposure of the campaign, the user terminal may be overloaded, so that the above-described determination process may be performed in the exposure control server 300.

9 is a flowchart illustrating a method of determining whether a campaign is exposed through the exposure control server according to an exemplary embodiment of the present invention. Figures 10 and 11 are flow charts of the learning steps of the present invention. 12 and 13 are flow charts for determining whether to expose the subject campaign of the present invention.

As described above, the method for determining whether or not a campaign is exposed according to an exemplary embodiment of the present invention includes generating a response probability prediction model for each campaign group, and when receiving an exposure / (For example, a reaction probability) of the corresponding game user according to the provision of the game player. According to an embodiment of the present invention, a method for determining whether or not to expose a user is provided to control a frequency of a more customized campaign to a corresponding game user in consideration of the calculated reaction probability and various information included in the game log of the corresponding game user . Now, referring to FIG. 9, a description will be given of a method of determining whether or not to expose according to an embodiment of the present invention. In the following, the duplication of the above-mentioned parts is omitted and the description is made.

Step S110 is a step performed by the game log collecting unit, which collects game logs stored in the server storage unit. Specifically, step S110 is a step of collecting past game logs of a plurality of game users according to the provision of past campaigns for each campaign group. As described above, the response probability prediction model in the campaign exposure determination method according to an exemplary embodiment of the present invention is generated using at least one of the exposure time of the campaign and the proposal acceptance information according to the provision of the past campaign. Accordingly, the past game log may include at least one of the campaign exposure time of the past campaigns provided to each game user and the proposal acceptance information according to past campaigns provided to each game user.

In addition, as described above, the past game log collected through S110 is a past game log within a predetermined period based on the current time. Because too old game logs are unreliable, it is desirable not to collect game logs beyond a predetermined period of time. Thus, the past time game logs gathered from the step S110 may be, for example, a game log of the period between the time t _-2, or t _-n t _-2 point in time.

Step S120 is a step performed by the learning unit, which learns past game logs. As described above, the learning performed in step S120 may be performed using at least one of the campaign exposure time and the information on whether the game user can accept the proposal according to the past campaign. Here, if step S 120 is performed using the campaign exposure time, it may proceed according to the flow shown in FIG. 10.

10, step S121 is a step of sorting campaign exposure times of past campaigns according to the size of time, and step S122 is a step of determining first and second times by analyzing the sorting result.

As mentioned above, the shorter the campaign exposure time is, the lower the interest in the campaign, and the longer the campaign exposure time, the more interested in the campaign. However, from the viewpoint of the mobile game, the game user can perform actions other than the operation of the user terminal after operating the game application. Accordingly, although the campaign exposure time gradually increases, since the game user performs an action other than viewing the campaign, if such data is applied, the reliability of the determination of the exposure of the campaign described below may be adversely affected.

Accordingly, the method of determining whether or not the campaign is exposed according to the embodiment of the present invention analyzes the campaign exposure times through step S121 and determines the first time and the second time based on the analysis result in step S122. Here, the second time is greater than the first time.

Step S123 is a step of filtering campaign exposures of past campaigns that are less than the first time or exceed the second time. That is, the step S123 is a step of setting the first time and the second time through step S122, and filtering the campaign exposure times out of the range between the first time and the second time.

Step S124 is a step of learning campaign exposure times according to provision of past campaigns between the first time and the second time.

As described above, in addition to the method using the campaign exposure time, step S120 may be performed using the acceptance of the proposal by the game user according to the provision of the past campaign. The operation when the step S120 is performed by using the game acceptance information of the game user is shown in FIG.

Step S221 is a step of defining the features for each game user. Here, the user-specific characteristics may include at least one of a game user level, an in-game good, an in-game good, a purchase probability, a user's status information, a campaign exposure time, and a game play pattern.

Step S222 is a step of extracting feature values and proposal acceptance information of each game user in past game logs, and step S223 is a step of learning feature values and proposal acceptance information of each game user. Here, the description of the steps S222 and S223 has been described in detail with reference to FIG. 8A, and a duplicated description will be omitted.

Step S130 is a step of generating a response probability prediction model for each campaign group based on the learning result. As described above, the response probability prediction model can be generated through at least one of the campaign exposure time and the proposal acceptance information according to the learning method through step S120, and outputs a different response probability according to the input value.

Also, when the response probability prediction model is generated using the campaign exposure time as described above, the response probability in the response probability prediction model becomes smaller as the exposure time of the campaign approaches the first time, As shown in FIG. Also, when the response probability prediction model is generated using the campaign exposure time, the response probability with respect to the campaign exposure time that is less than the first time or exceeds the second time in the response probability prediction model is 0 or is modeled as a value close to 0 .

Step S140 is a step performed by the request reception unit, which receives the exposure / non-availability query signal from the user terminal.

Step S150 is a step performed by the game log collecting unit or the response probability calculating unit. The collected game log collects past game logs for the game user of the user terminal, and determines reaction probability determination factors for the target campaign in the collected past game logs Respectively. As described above, the game log for the game user of the user terminal exists in the server storage unit. When receiving the exposure / non-availability query signal through step S140, the game log is stored in the server storage unit based on the identifier of the game user included in the exposure / Lt; / RTI >

In addition, the extraction of the response probability determination factor through step S150 may be performed for at least one of the campaign exposure time and the proposal acceptance information depending on which information is used to generate the response probability prediction model.

Step S160 is a step performed by the reaction probability calculation unit, which calculates a reaction probability of the game user for the target campaign. Specifically, step S160 is a step of calculating a reaction probability of the game user with respect to the target campaign by applying a reaction probability determination factor of the game user to the reaction probability prediction model of the campaign group to which the target campaign belongs.

If the response probability prediction model is generated based on the campaign exposure times, the reaction probability determination factor, which is an input value to the reaction probability prediction model, is derived based on the campaign exposure times of the past campaigns provided to the game user of the user terminal . For example, the response probability judgment factor input into the reaction probability prediction model may be an average value of the campaign exposure times of the corresponding game user, or a value calculated through another method.

Step S170 is a step performed by the exposure determination unit, which determines whether the target campaign is exposed using the response probability. As described above, step S170 can be largely performed in two ways. Of these methods, the first method is a method using a random value, and the second method is a method using an exposure determination value. First, with reference to FIG. 12, a step of determining whether to expose a target campaign according to the first scheme is explained.

Step S171 is a step of deriving a random value. Specifically, step S171 is a step of deriving a random value between 0 and 1 through a random function. Here, the random value can be derived as a prime number such as 0.xxx, and the prime number of the prime number is not limited to a specific number.

Step S172 is a step of comparing the response probability of the game user with respect to the target campaign calculated through step S160 and the random value calculated through step S171. Specifically, step S172 is a step of determining whether the probability of reaction exceeds a random value. That is, in step S172, it is determined that the response probability derived through step S160 is a reference size, and when the random value falls within the reference size, the target campaign is exposed (step S173). On the contrary, It is determined that the target campaign will not be exposed (step S176).

However, when determining whether to expose the target campaign by only the determination in step S172, only the characteristics of the game user and other game users having a similar tendency to the game user are reflected, so that it is less customized can see. In addition, as described above, it is determined that each game user has a high probability of response (for example, interest level) only in a specific campaign group, so that only the campaigns included in the corresponding campaign groups may be exposed to the game user with frequent frequency And the probability of the other campaign group is too low, the campaigns included in the campaign group may not be exposed to the corresponding game user.

Accordingly, when determining whether to expose a target campaign according to an exemplary embodiment of the present invention, a method of determining whether a target campaign is exposed may include not only a response probability but also the latest exposure of the last exposed campaign among the campaign groups to which the target campaign belongs And further considers time.

Accordingly, even if it is determined in step S173 that the target campaign is to be exposed, the process of determining whether the difference between the current time and the latest exposure time exceeds the minimum exposure time through step S174 may be performed. Likewise, even if it is determined in step S176 that the target campaign is not to be exposed, the process of determining whether the difference between the current time and the latest exposure time exceeds the maximum exposure frequency time through step S177 may be performed.

By further performing the determination process through steps S174 and S177, the exposure of the target campaign can be further customized by considering the tendency of the game user and the latest exposure time of the campaign provided to the game user.

That is, even if it is determined in step S173 that the target campaign is to be exposed, if it is determined in step S174 that the difference between the current time and the latest exposure time is less than the minimum exposure time, If it is provided within a specific period), step S175b may be performed to change the target campaign to not be exposed. Of course, it is determined in step S173 that the target campaign is to be exposed. If the difference between the current time and the latest exposure time exceeds the minimum exposure time in step S174, step S175a is performed and the target campaign is determined to be exposed.

Similarly, even if it is determined in step S176 that the target campaign is not to be exposed, if it is determined in step S177 that the difference between the current time and the latest exposure time exceeds the maximum exposure frequency time (i.e., If the campaign has not been provided beyond a certain period of time), step S178a may be performed to change the target campaign to be exposed. If it is determined in step S176 that the target campaign is not to be exposed in step S176, and if the difference between the current time and the latest exposure time is less than the maximum exposure frequency time in step S177, step S178b is performed to determine that the target campaign is not to be exposed have.

As described with reference to FIG. 5, the method using a random value can achieve the object of the present invention in that a campaign with a low degree of interest (for example, a response probability) can lower its frequency. However, if the response probability for a particular campaign group is too low, the problem may occur that almost all the game users are not exposed to the campaign at all.

Therefore, in step S170, it is preferable that the exposure determination value is determined instead of the random value, the response probability is compared with the exposure determination value, and the comparison result is used. For this, step S170 may include determining the exposure determination value (step S271). In step S271, the exposure determination value may be determined in various manners. For example, in step S271, a response probability of a game user corresponding to a predetermined upper percentage of all game users may be determined as an exposure determination value.

The steps S272 through S278 differ substantially from the steps S172 through S178 shown in FIG. 12, but substantially the same operations are performed. Accordingly, the description of steps S272 through S278 will be omitted.

Referring again to FIG. 9, in step S180, the exposure control information is generated according to whether the target campaign determined through step S170 is exposed, and is transmitted to the user terminal.

As described above, by using the method and server for judging whether or not the campaign is exposed according to the embodiment of the present invention, it is possible to control the frequency of the campaign exposure customized to the game user for each game user. The frequency of the campaigns is adjusted according to the probability of the reaction of each game user, so that each game user can be provided with the campaigns he is interested in, and the situation of annoying the game users can be avoided.

In addition, according to the method and server for determining whether or not a campaign is exposed according to an embodiment of the present invention, campaigns belonging to the same group are frequently exposed using frequent exposure time information of the campaign for each campaign group provided to each game user Situations, and situations in which campaigns in a particular campaign group are not exposed for too long. As a result, it is expected that a game which satisfies both mobile game providers and game users will be possible.

In the above description, the method of determining whether or not the campaign is exposed according to an embodiment of the present invention and the server utilizes game logs according to a plurality of campaigns included in the campaign group. However, this is only an example, and can be expanded and operated on a campaign basis rather than a campaign group basis.

The above exemplary methods according to the present invention may be implemented in a computer program product, a computer program product recorded on a recording medium readable by a computer (including all devices having an information processing function) May be implemented in a variety of ways including, but not limited to, applications, logic circuits, custom semiconductors, or firmware. Examples of the computer-readable recording medium include, but are not limited to, ROM, RAM, CD, DVD, magnetic tape, hard disk, floppy disk, hard disk and optical data storage. In addition, the computer-readable recording medium may be distributed over network-connected computer systems so that computer readable codes can be stored and executed in a distributed manner.

The foregoing description is merely illustrative of the present invention, and various modifications may be made by those skilled in the art without departing from the spirit of the present invention.

[Description of Symbols]

100: User terminal 110: Campaign exposure time determination unit

120: Exposure object campaign selection unit 130: Exposure object query unit

140: terminal communication unit 150:

300: Exposure control server 310: Game log collection unit

320: server storage unit 330: learning unit

340: model generation unit 350: request reception unit

360: server communication unit 370: response probability calculation unit

380: Exposure determination unit

1000: Campaign exposure determination system

Claims (17)

1. A method for an exposure control server to determine whether a target campaign to be provided to a user terminal equipped with a game application is exposed,

   Collecting past game logs of a plurality of game users for each campaign group;

   Learning the past game logs;

   Generating a response probability prediction model for each campaign group based on a result of the learning;

   A step of receiving a query signal of the target campaign from the user terminal and extracting a response probability determination factor for the target campaign from a past point game log for a game user of the user terminal;

   Calculating a response probability of the game user to the target campaign by applying a reaction probability determination factor of the game user to a reaction probability prediction model of a campaign group to which the target campaign belongs; And

   And determining whether to expose the target campaign using the response probability.
2. The method according to claim 1,

   The past game log includes a campaign exposure time of a past campaign provided to the game user,

   Wherein the learning of the past game logs is performed using campaign exposure times of past campaigns provided to the plurality of game users.
3. 3. The method of claim 2,

   The step of learning the past game logs may include:

   Arranging campaign exposures of the past campaigns according to size of time;

   Determining a first time and a second time by analyzing the result of the alignment; And

   Filtering the campaign exposure times of the past campaigns that are less than the first time or exceed the second time.
4. The method of claim 3,

   Wherein the response probability in the response probability prediction model for each campaign group is modeled such that the response probability decreases as the campaign exposure time approaches the first time and increases as the campaign exposure time approaches the second time. How to Determine If You Are Exposed.
5. The method of claim 3,

   Wherein the step of generating the response probability prediction model for each campaign group comprises:

   Wherein the response probability for the campaign exposure times is less than the first time or is greater than the second time.
6. 3. The method of claim 2,

   Wherein the response probability determination factor is derived based on campaign exposure times of past campaigns provided to a game user of the user terminal.
7. The method according to claim 1,

   The past game log includes information on whether or not to accept a proposal according to a past campaign provided to the game user,

   Wherein the step of learning the past game logs is performed using information on whether to accept the proposals of the plurality of game users.
8. 8. The method of claim 7,

   The step of learning the past game logs may include:

   Extracting feature values and proposal acceptance information of each game user in the past game logs; And

   And learning the feature values and the proposal acceptability information of each of the game users.
9. 9. The method of claim 8,

   The feature values represent values for predefined features of game users, and the features include game user level, in-game good, in-game good, purchase probability, user status information, campaign exposure time, The method comprising the steps of:
10. The method according to claim 1,

    Wherein the step of determining whether to expose the target campaign comprises:

    Determining an exposure determination value; And

    And determining that the target campaign is to be exposed when a response probability of the game user exceeds the exposure determination value.
11. 11. The method of claim 10,

    The step of determining the exposure determination value includes:

    And determining a response probability of a game user corresponding to a predetermined upper percentage of all game users as an exposure determination value.
12. 11. The method of claim 10,

    Determining that the target campaign is to be exposed when a response probability of the game user exceeds the exposure determination value and a difference between the current time and the latest exposure time exceeds a minimum exposure frequency time, Wherein the time is a presentation time of a campaign last provided to the user terminal among the campaign groups including the target campaign.
13. 11. The method of claim 10,

    Determining that the target campaign is to be exposed when the response probability of the game user is less than the exposure determination value and the difference between the current time and the latest exposure time exceeds a maximum exposure frequency time, .
14. The method according to claim 1,

    Wherein the step of determining whether to expose the target campaign comprises:

    Deriving a random value between 0 and 1 through a random function; And

    And determining to expose the target campaign if the response probability of the game user exceeds the random value.
15. A game log collecting unit collecting past game logs of a plurality of game users for each campaign group;

    A learning unit for learning the past game logs;

    A model generating unit for generating a response probability prediction model for each of the campaign groups based on a result of the learning;

    A request receiving unit for receiving an exposure / non-contact query signal from the user terminal;

    Extracting a response probability determination factor for the target campaign from the past game log of the user of the user terminal and applying a reaction probability determination factor of the game user to the reaction probability prediction model of the campaign group to which the target campaign belongs A response probability calculation unit for calculating a response probability of the game user to the target campaign; And

    And an exposure determining unit determining whether to expose the target campaign using the response probability,

    An exposure control server for executing each step of the method according to any one of claims 1 to 14.
16. 15. A computer program stored in a computer-readable medium for causing a computer to execute the method of any one of claims 1 to 14.
17. A computer-readable recording medium having recorded thereon a computer program for performing the method of any one of claims 1 to 14.

Images