AI-based user gameplay prediction system

Abstract

The user game play prediction system includes a collection module that collects play information provided from a user terminal, which is information related to the user's game play, a processing module that processes the collected play information into a form of data required for analysis, and an AI analysis module that analyzes the processed play information through machine learning using an AI algorithm to derive a predicted value of the user's future game-related behavior. By applying the AI ​​algorithm to the user's game play information, a predicted value of the user's future play can be derived. Such a predicted value is used for target marketing, etc.

Description

[Technical field]

[0001] The present invention relates to an AI-based user gameplay prediction system, and more particularly to a system that predicts a user's future gameplay-related behavior based on play information such as the user's log records. [Background technology]

[0002] With the development of IT technology, various computer games have become popular, and many users enjoy online games. Many game developers are competitively releasing various games, and it is necessary to market the games to users who enjoy computer games. Game developers who wish to market have difficulty in understanding the play preferences of an unspecified number of users, and therefore have difficulty in targeting the main demand group suitable for a particular game. This results in a situation where sufficient marketing results cannot be expected in comparison with the effort and cost involved in marketing. Moreover, fraudulent manipulation, such as artificially increasing advertising inquiry traffic, is common in the online advertising market, which leads to increased costs for game developers and, as a result, to an increase in game usage fees charged to users.

[0003] A basic requirement for target marketing is a user's tendency analysis. That is, if information such as what kind of user plays games, how often and for how long, etc., can be grasped, advertising can be concentrated on that user, thereby increasing the effectiveness of advertising. However, such tendency analysis alone still has limitations in the effectiveness of targeted advertising. If future playing behavior of a certain user can be predicted based on past playing patterns, it will be possible to perform targeted advertising that is more suited to that user and more effective based on the predicted value for that user. For example, if a certain user is expected to play games for two hours from 9 o'clock tonight, and there is a new game that is expected to be enjoyed by this user, a coupon that allows the user to try the new game for free can be sent based on such predicted value just before the user actually starts playing the game. When target advertising is performed based on predicted values ​​in this way, it is expected that the effectiveness of advertising will be far greater than that based on simple tendency analysis. Summary of the Invention [Problem to be solved by the invention]

[0004] The present invention has been devised to solve the above problems, and an object of the present invention is to provide a method for applying an artificial intelligence algorithm to a user's game play information and deriving a prediction value for the user's future play.

[0005] Another object of the present invention is to utilize such predicted values ​​for, for example, targeted marketing, thereby enabling game developers to target advertise newly released games to users who are predicted to have a high interest in the game. [Means for solving the problem]

[0006] In order to achieve the above object, the present invention provides a user game play prediction system comprising: a collection module that collects play information, which is information related to a user's game play, provided by a user terminal device used by the user; a processing module that processes the play information collected by the collection module into a data form required for analysis; and an AI analysis module that analyzes the play information processed by the processing module by machine learning using an artificial intelligence algorithm and derives a predicted value of the user's future game-related behavior.

[0007] Here, the play information includes one or more of an input signal input to the user terminal on which the game is played for playing the game, an output signal output by the user terminal on which the game is played, a log record of the user's game, a history of installing and deleting a program for playing the game on the user terminal, the user's play grade for the game, and a payment history related to the user's game. And the predicted value includes one or more of a time when the user plays a game, whether the user plays a game at a specific time, the game the user plays, whether the user has made a payment for the game and / or the payment amount, a change in the number of users for a specific game, and whether the user has dropped out of a specific game.

[0008] The AI ​​analysis module derives a predetermined discrimination value related to the play of the game together with the predicted value, the discrimination value including one or more of a game play operation mode of the user, whether the user is a robot, game graphics and / or sounds preferred by the user, a game genre preferred by the user, and whether cheating has occurred in the play of the user.

[0009] The processing module extracts features for each event from the play information, and vectorizes each extracted feature by word embedding.

[0010] The AI ​​analysis module may include a CNN module that applies a Convolutional Neural Network (CNN) algorithm to the vectorized play information to derive the predicted value, a Long-Short Term Memory (LSTM) module that applies a LSTM algorithm to the vectorized play information to derive the predicted value, and a static classification module that applies a static classification algorithm including a Random Forest and / or a Decision Tree to the vectorized play information to derive the predicted value.

[0011] In addition, the AI ​​analysis module may further include a neural network module that derives the predicted value based on a result of combining at least two of the CNN module, the LSTM module, and the static classification module using a neural network algorithm.

[0012] The user game play prediction system according to the present invention may further include a transmission module that transmits the predicted value derived by the AI ​​analysis module to a game providing server operated by a game provider.

[0013] In addition, the user game play prediction system according to the present invention may further include a promotion management module that receives promotion information, which is promotion-related information that the game provider wishes to provide to the user, from the game providing server and provides the promotion information to the user terminal. Effect of the Invention

[0014] According to the present invention, an artificial intelligence algorithm is applied to information on a user's game play to derive a predicted value for the user's future play. Such a predicted value can be used, for example, for targeted marketing, to target advertisements of newly released games to users who are predicted to have a high interest in the game. [Brief description of the drawings]

[0015] [Figure 1] 1 is a diagram showing an overall configuration including a user game play prediction system according to the present invention. [Diagram 2] FIG. 2 is a diagram showing a detailed configuration of the AI ​​analysis module in FIG. 1. [Diagram 3] FIG. 3 is a diagram illustrating an example of the form of data input to the CNN module of FIG. 2. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0016] The present invention will now be described in more detail with reference to the drawings.

[0017] In the following description, the components described as "system", "module", etc. may be embodied as a single computer or server, or may be configured to be provided separately as functional modules within a single computer or server. In the description of the present invention, detailed illustrations and descriptions of general components that are not directly related to the gist of the present invention or that are normally clearly provided will be omitted unless they are directly related to the characteristic components of the present invention.

[0018] FIG. 1 is a diagram showing an overall configuration including a user game play prediction system according to the present invention.

[0019] The system 100 of the present invention receives play information from a user playing a game, and derives a predicted value of the user's behavior based on the play information. To this end, a user terminal 70 used by the user is connected to the system 100 of the present invention via the Internet, and they can communicate with each other.

[0020] In addition, the system 100 of the present invention provides such predicted values ​​to game developers, and generates or transmits promotion information required for game promotion, or receives a promotion request from the game developer and provides promotion information to users whose behavior is predicted to match the game based on the predicted values. To this end, a game providing server 50 operated by a game developer is connected to the system 100 of the present invention via the Internet, and they can communicate with each other.

[0021] The system 100 of the present invention may be communicatively connected to the user terminal 70 and receive play information from the user terminal 70, which will be described later, in various ways. However, since it is not suitable for the collection of various play information if the user terminal 70 is made to connect to the system 100 voluntarily, it is preferable to install a separate application in the user terminal 70 and transmit play information from the user terminal 70 to the system 100 by the operation of the application every time the user plays a game. For this purpose, for example, a method may be used in which an application for transmitting play information is installed with the user's consent when a security program that is generally installed by the user is installed.

[0022] As shown in FIG. 1, the system 100 of the present invention includes a collection module 110, a processing module 120, an AI analysis module 130, a transmission module 140, and a promotion management module 150.

[0023] The collection module 110 collects play information related to a user's game play provided from the user terminal 70. Here, the play information is information that is a basis for deriving a predicted value of a user's future game-related behavior, and is mainly information related to a user's play pattern. The play information may include an input signal input to the user terminal 70 where the game is played for playing the game, and an output signal output by the user terminal where the game is played. Such input and output signals may be data for deriving a user's play orientation or a type of game that the user likes by, for example, analyzing the number, frequency, speed, etc. of inputs or analyzing sounds output from the game during play. The play information may include a log record of the user's game. The log record may be data for deriving the frequency of play, the duration of one play, the time period of play, etc. for each game using data on the start and end times of connection to a certain game. The play information may include a history of installation and deletion of a program for playing a game in the user terminal 70. Such history may be used to derive the duration of a user's enjoyment of a particular game, the types of games enjoyed, and changes in the games enjoyed. The play information may include a user's play grade for a game. The play grade may be used to derive whether the user is an aficionado depending on the user's game proficiency. The play information may include a payment history related to the user's game. The payment history may be used to derive the amount a user is willing to pay to enjoy a game.

[0024] The processing module 120 processes the play information collected by the collection module 110 into a data form required for the AI ​​analysis module 130 to perform analysis. In this case, the processing module 120 may process the play information by extracting features for each event and vectorizing each extracted feature by word embedding. Specifically, the play information of the user is classified into each event and classified into words that match the events. For example, the events may be classified into events such as ligin, signup, levelup, etc. Each event classified in this way is changed into a vector by word embedding. The vector of each event becomes a vector having a similar value when the correlation between the events is high, and becomes a vector having a dissimilar value when the correlation is low. Features are classified based on the data values ​​changed into vectors in this way. The features classified in this way are used as input data to be input to the learning algorithm in the AI ​​analysis module 130.

[0025] The AI ​​analysis module 130 analyzes the play information processed by the processing module 120 by machine learning using an artificial intelligence algorithm, and derives a predicted value of the game-related behavior expected for each user in the future. Here, the predicted value is a result that the system 100 of the present invention ultimately seeks, and may include various types. That is, the predicted value may include one or more items of the time when the user plays the game, whether the user plays the game at a specific time, the game the user plays, whether the user makes a payment for the game and / or the payment amount, the change in the number of users for the specific game, and whether the user has dropped out of the specific game. Each of these predicted value items may be constantly derived by the system 100 of the present invention, or when a predicted value of a specific item is requested by the game providing server 50, the system 100 may derive the predicted value of the item.

[0026] For example, if the game providing server 50 wishes to send specific advertisement content to a user who is predicted to play a game between 9:00 and 10:00 tonight, the game providing server 50 sends such a request to the system 100 of the present invention. The AI ​​analysis module 130 of the system 100 executes an analysis algorithm to derive a predicted value for that item. Based on the derived result, the system 100 transmits the advertisement content received from the game providing server 50 to the user terminal 70 of the predicted user. Preferably, the AI ​​analysis module 130 derives each predicted value item as a result of dividing each user into 1 or 0. That is, in the above example, a predicted value is determined for all users, such as "User1=0, User2=0, User3=1, ..." for each user.

[0027] Meanwhile, the AI ​​analysis module 130 of the present invention is configured to derive a predetermined discriminant value related to the play of the game in addition to the above-mentioned predicted value. The discriminant value is also derived by applying an artificial intelligence algorithm.

[0028] The discriminant value is a concept distinct from a predicted value of future behavior of a specific user, and refers to a value determined based on various statistics related to the user's game or on abnormal behavior related to the user's overall play. For example, the discriminant value may include the user's game play operation method, whether the user is a robot (i.e., whether the game is played by a machine developed for playing games while pretending to be a human user), the graphics and / or sounds of the game preferred by the user, the genre of the game preferred by the user, and whether or not the user has cheated in play. If a determination based on statistics related to play or a determination based on the presence or absence of a phenomenon such as abnormal behavior is simply determined based on the occurrence of a specific event, the accuracy of the determination is reduced. Therefore, by applying the artificial intelligence algorithm of the present invention, the accuracy of the determination may be adaptively improved by machine learning.

[0029] The transmission module 140 transmits the predicted value and / or the discriminant value derived by the AI ​​analysis module 130 to the game providing server 50. By receiving the predicted value and / or the discriminant value from the transmission module 140, the game providing server 50 can predict the behavior of the user for advertising purposes or refer to the discriminant value for setting a direction for game development. It is preferable that the transmission of the transmission module 140 is performed only when there is a separate request from the game providing server 50.

[0030] The promotion management module 150 is provided with promotion information, which is promotion-related information that the game provider wishes to provide to the user, from the game providing server 50, and provides the promotion information to the user terminal. The promotion information may include, for example, advertisements for the games provided by the game providing server 50, rewards for the games, coupons for free experience of the games, etc., and may also be created as individual contents that match only users who are predicted to match the characteristics of the games provided by the game providing server 50 among all users.

[0031] The configuration and operation of the AI ​​analysis module 130 in the system 100 of the present invention will be described in more detail below.

[0032] Fig. 2 is a diagram showing a detailed configuration of the AI ​​analysis module 130 of Fig. 1. The AI ​​analysis module 130 of the present invention includes a CNN module 131, an LSTM module 132, a static classification module 133, and a neural network module 135.

[0033] The CNN module 131 derives a predicted value by applying a CNN (Convolutional Neural Network) algorithm to the play information vectorized by the processing module 120. The CNN algorithm is an artificial intelligence algorithm optimized for image classification processing, and is mainly used to find peculiarities by operating a filter from visualized data. FIG. 3 is a diagram illustrating the form of data input to the CNN module 131 of FIG. 2, and as described above, the data of each user vectorized by the processing module 120 and extracted by feature is configured in the form of image data. In FIG. 3, the play information for each user (User1, User2, .....) is divided into rows by each date (Day1, Day2, .....), and the play information for each date is arranged in columns by each vectorized feature (Feature1, Feature2, .....), and the play information for one user (User1) is arranged in the form of image data. In this way, the play information for each user is formed by the number of users, and the play information to be input to the CNN module 131 is generated. The CNN module 131 applies a CNN algorithm to the classified play information to derive a prediction value for a specific item. The CNN algorithm applies machine learning to the play information in the visualized form, providing an optimized algorithm.

[0034] The LSTM module 132 applies a Long-Short Term Memory (LSTM) algorithm to the vectorized play information to derive a prediction value. The LSTM algorithm is an algorithm that improves the short-term persistence of the RNN algorithm that stores and analyzes previous input data, and is an algorithm optimized for persistent prediction of data over a long period of time. For example, when analyzing a log record of a user's game connection to play information, analysis of data on a log record on a certain day and data on log records across multiple dates is required, and other data, such as purchase records of input and output signals, is also mixed in while data on these log records is input. When only analysis of log records is required among various play information input in this way, the LSTM algorithm provides an effective analysis tool with fewer errors.

[0035] The static classification module 133 derives a prediction value by applying a static classification algorithm including a random forest and / or a decision tree to the vectorized play information. The static classification method is a traditional method that uses a formula to classify before research on deep learning methods became active, and has been widely used for a long time in various fields such as spam classification, malicious code detection, image classification, and compound classification. When applying a static classification algorithm, it is suitable for deriving a very stable and balanced result. Therefore, when the static classification algorithm is combined with the above-mentioned CNN algorithm and LSTM algorithm, a stable result value is combined with a result value specialized for the CNN algorithm and the LSTM algorithm, and the accuracy of the prediction can be increased.

[0036] The neural network module 135 derives a predicted value based on a result of combining the output results of the CNN module 131, the LSTM module 132, and the static classification module 133 using a neural network algorithm. In general, a machine learning model is specialized for searching for a solution in a specific area. When utilizing machine learning, a stronger model can be constructed by combining the individual features of each model. For example, the result of model A can be combined as an input of model B, or an ensemble method can be applied to combine the results of different models. The ensemble method can more effectively derive a final result value by applying a method of combining the results of each model by applying a weighted value. This allows more accurate results to be obtained when there is diversity between the individual models. In the present invention, a more accurate predicted value can be obtained by combining the above-mentioned modules 131, 132, and 133 using a neural network algorithm. On the other hand, the neural network module 135 may be configured to derive a result of combining only any two of the CNN module 131, the LSTM module 132, and the static classification module 133, rather than all of these modules.

[0037] According to the present invention as described above, an artificial intelligence algorithm is applied to information on a user's game play, and a prediction value for the user's future play can be derived. By utilizing such a prediction value for, for example, targeted marketing, a game developer can target advertisements of a newly released game to users who are predicted to have a high interest in the game.

[0038] In addition, according to the present invention, various prediction values ​​and discrimination values ​​can be obtained, and based on these, user needs related to the types and characteristics of new games can be understood. Therefore, the present invention can be used to set the direction of game development that will be popular with many users.

[0039] Although the present invention has been described above with reference to the illustrated embodiment, it is merely illustrative, and those skilled in the art will appreciate that various modifications and equivalent embodiments are possible therefrom. Therefore, the true technical scope of the present invention should be determined by the technical ideas of the appended claims. [Explanation of symbols]

[0040] 50 Game Servers 70 User terminal 100 Systems 110 Collection Module 120 Processing Module 130 AI Analysis Module 131 CNN Module 132 LSTM Module 133 Static Classification Module 135 Neural Network Module 140 Transmitting Module 150 Promotion Management Module

Claims

1. A user game play prediction system, comprising: a collection module for collecting play information, which is information related to a user's play of a game, provided from a user terminal used by the user; a processing module that processes the play information collected by the collection module into a form of data required for analysis; an AI analysis module that analyzes the play information processed by the processing module through machine learning using an artificial intelligence algorithm and derives a predicted value of the user's future game-related behavior; The present invention relates to a method for manufacturing a computer-implemented ... The play information includes at least one of an input signal input to the user terminal device on which the game is played for playing the game, an output signal output by the user terminal device on which the game is played, a log record of the user's game, a history of installing and deleting a program for playing the game from the user terminal device, the user's play grade for the game, and a payment history related to the user's game, The predicted value includes one or more of a time when the user plays a game, whether the user plays a game at a specific time, a game played by the user, whether the user makes a payment for the game and / or the payment amount, a change in the number of users for a specific game, and whether the user has dropped out of a specific game. A system for predicting a user's game play.

2. the AI ​​analysis module derives, together with the predicted value, a predetermined discriminant value associated with play of the game; 2. The user game play prediction system of claim 1, wherein the discrimination value includes one or more of the user's game play operation method, whether the user is a robot, the user's preferred game graphics and / or sounds, the user's preferred game genre, and whether or not the user has engaged in cheating in his / her game play.

3. 3. The system of claim 2, wherein the processing module extracts a feature for each event from the play information, and vectorizes each extracted feature by word embedding.

4. The system of claim 3, wherein the AI ​​analysis module includes a CNN module that applies a CNN (Convolutional Neural Network) algorithm to the vectorized play information to derive the predicted value.

5. The system of claim 4, wherein the AI ​​analysis module further comprises a Long-Short Term Memory (LSTM) module that applies an LSTM algorithm to the vectorized play information to derive the predicted value.

6. 6. The system of claim 5, wherein the AI ​​analysis module further comprises a static classification module that applies a static classification algorithm, including a Random Forest and / or a Decision Tree, to the vectorized play information to derive the predicted value.

7. The user's game play prediction system of claim 6, wherein the AI ​​analysis module further includes a neural network module that derives the prediction value based on a result of combining at least two of the CNN module, the LSTM module, and the static classification module using a neural network algorithm.

8. The user's game play prediction system according to any one of claims 1 to 7, further comprising a transmission module that transmits the predicted value derived by the AI ​​analysis module to a game providing server operated by a game provider.

9. 9. The user game play prediction system of claim 8, further comprising a promotion management module that receives promotion information, which is promotion-related information that the game provider wishes to provide to the user, from the game providing server and provides the promotion information to the user terminal.

Images