CN108014496B

CN108014496B - Game record analysis method

Info

Publication number: CN108014496B
Application number: CN201710018947.9A
Authority: CN
Inventors: 金朱一; 郑荣焕
Original assignee: Coresight Co ltd
Current assignee: Coresight Co ltd
Priority date: 2016-11-01
Filing date: 2017-01-11
Publication date: 2020-11-10
Anticipated expiration: 2037-01-11
Also published as: KR101733458B1; CN108014496A

Abstract

The invention includes a game server capable of providing game content as a game client including a terminal; and a server for collecting, recording and analyzing log data related to the game content played by the game client. The record analysis server comprises a communication module which can collect log data from the client; a record storage part for storing the collected log data through the communication module; the system also comprises an analysis module which provides different weighted values for the log data analysis according to the selection of the required information and the selected information for predicting the log data record content collected and stored in the active period or the inactive period. Displaying the game content during the activity period to perform the activity; the game activity displays game state information belonging to the game content operation information. Such as: at least one of the game difficulty level, the game character level and the game speed is changed.

Description

Game record analysis method

Technical Field

The invention relates to a game record analysis method for analyzing user churn rate by using game log data. In particular, the present invention relates to a recording and analyzing method capable of predicting a loss rate in real time when an analysis target does not play or delete game software (applications, etc.).

Background

Owing to the popularization of network infrastructure, the game market mainly based on stand-alone games in the past is gradually replaced by network games.

For example, in MMO games, network game ITEMs (ITEMs) and game virtual coins play an important role for players. Therefore, sometimes as a personal asset, it becomes the subject of transactions between players.

For the above reasons, account theft, exchange article (ITEM) fraud events, and the like sometimes occur. If the player reports the account number stealing problem, the online game company analyzes the record information generated by the server in real time and provides a relevant solution for the user after finding out the reason.

Moreover, these games are characterized by various interactions that occur between the game characters operated by the virtual world players. Various interactions between these game characters can be the object of game record analysis. That is, these games require subsequent analysis based on the recorded analysis content, and such games can be subjected to subsequent analysis, which is greatly different from the past box games and from most casual games.

The MMO network game server generates a game record file for storing the actions of the player game character within a certain time period.

In general, the game record file includes player game character operation record information, operation time information, information on the player game character, and the like. But also information about various variables of the game.

Disclosure of Invention

Technical problem to be solved

The invention analyzes the user loss possibility in the game software setting and executing stage of the analysis object. The invention not only provides the analysis content to the game company (game developer) or the game service company (game publisher), but also provides the operation algorithm capable of preventing the user from losing immediately when the user loses through the analysis of the user game loss reason and the real-time analysis.

(II) technical scheme

The invention includes a game server capable of providing game content as a game client including a terminal; and a server for collecting, recording and analyzing log data related to the game content played by the game client. The record analysis server comprises a communication module which can collect log data from the client; a record storage part for storing the collected log data through the communication module; the system also comprises an analysis module which provides different weighted values for the log data analysis according to the selection of the required information and the selected information for predicting the log data record content collected and stored in the active period or the inactive period. Displaying the game content during the activity period to perform the activity; the game activity is game state information belonging to the game content operation information. Such as: at least one of the game difficulty level, the game character level and the game speed is changed.

(III) advantageous effects

By the recording and analyzing method, not only can user analysis be performed through log data collected in real time, but also the advantages of effectively analyzing information downloaded or deleted in a user game, activity information and the like can be achieved. And the method has the advantages that the game development information, the game profit mode and the like can be more easily obtained through the log data analysis result.

Drawings

FIG. 1 is a block diagram of a game record analysis system of the present invention.

Fig. 2 is a block diagram showing a record analysis server (which performs game record analysis) according to the present invention.

Fig. 3 is a flow chart illustrating a method of game record analysis according to an embodiment of the present invention.

FIG. 4 is a flow chart illustrating a process for collecting and analyzing underlying data that can be used as a criterion for analyzing game churn rates, i.e., as a benchmark, according to an embodiment of the present invention.

FIG. 5 is a flow chart illustrating a method for predicting an initial churn rate after a game download, according to an embodiment of the present invention.

FIG. 6 is a flow chart illustrating a method for predicting a game execution churn rate in real time, according to an embodiment of the present invention.

Detailed Description

First, the game record analysis method of the present invention is a method of performing index analysis according to the duration of a game activity using a game executed from the game company client 11 and record information generated from the user client 10, which can provide meaningful information to the game company.

The "gaming activity" of the present invention is defined as follows:

"gaming activity" refers to a specific period of time specifically set by the gaming establishment's game server or record analysis server. And plan activities giving different benefits to skills, coins, match patterns, etc. of the game during the specific period.

For example, if a new game character is added to a game and the added new game character is a Raid (team copy activity) Boss, the level of the original Boss game character is lowered, the chance of acquiring a specific Item (Unique Item) is increased for a while (specific period), and the difficulty of breaking the game is also lowered, so that more users can be attracted to play the game. And these elements may all be referred to as gaming activities.

That is, the game activity is a series of actions such as changing the ITEM (ITEM) acquisition difficulty level, adjusting the game character level (level), adjusting the game progress speed, and the like during the game activity to change the game state information for executing the game.

The game state information is information required for the user to execute the game. The game state information may be ITEMs (ITEM) information required for allowing the user to more easily pass the game by setting the game level for allowing the user to execute the game, breaking (clear) the score or grade required for each level, or the like, or game virtual medals required for purchasing the ITEMs (ITEM), that is, ITEMs, game characters, and level settings required for executing the game.

Moreover, the system and the method of the invention have a vocabulary which is a user activity record and is used together with the game state information.

The user activity record is information when the user executes a game. All activity information of the user when executing the game becomes a user activity record. For example, the user activity record may be all actions performed by the user in the game or in the game character, or actions that occur when the user executes or operates the game, such as the game character, the Item (Item), the execution state of each level, and the number of times the game is executed, or game character action information that can be regarded as the user's game.

Meanwhile, the active period refers to a period in which the above-described activity is performed, and a period in which the activity is not performed becomes an inactive period. That is, the period during which the user executes the game or logs data may be during the active period or during the inactive period.

When analyzing the loss rate, the present invention provides different weighted value indicators (indicators) according to the log data collection period (active period or inactive period).

That is, generally, the user collects log data of the action information during the game, but the present invention has a feature that the display content is changed or the weight value of the content is changed according to the period of recording the log data (whether or not the game is active) when the log data recording information is used.

Because, when a game company operating the game server 20 pushes out a new game character or adds a new game level indicating the degree of progress of the game, a plurality of game promotion activities are performed, and users having low game participation participate during the game activities.

In the present invention, the game software terminal information downloaded by the analysis object and the game software download information are used together to analyze the log data. The terminal comprises all devices capable of downloading game software, such as a PC, a Mac, a tablet computer, an IOS device and an android device.

Not only mobile terminals.

The present invention analyzes the log data in a specific manner and with reference to the following figures and contents.

Meanwhile, if it is desired to know whether the game record information transmitted from the game company client 11 is recorded during the event, the user can confirm and verify the game record information in the game server 20 or the record analysis server 100.

The game server 20 is a client 11 that downloads a mobile terminal or the like to a user terminal. The server not only provides game services, but also pushes records (logs) installed, run, and deleted in game user devices to the record analysis server 100.

The game company client 11 also transmits game records, which are downloaded, executed, and deleted by the user on the device, to the record analysis server 100.

Game information obtained from the game company client 11 and the game server 20; that is, the present invention will explain a record analysis method in detail based on the above-described record analysis server 100 that receives log data.

FIG. 2 is a block diagram of a record analysis server of the present invention that performs game record analysis; FIG. 3 is a flow chart illustrating a game record analysis approach according to an embodiment of the present invention.

First, the record analysis server 100 according to the present invention of fig. 2 includes a communication module 110 receiving log data; a record storage unit 120 for storing the collected log data via the communication module 110; a classification and extraction module 130 for giving different standard values and weighted values to the collected log data according to whether the activity period is in, and classifying and extracting the log data; an analysis module 140 that analyzes the extracted or classified log data; and a schematization system 150 that performs a modular process on the log data.

In describing the present invention, the record analysis server 100 is configured by a plurality of modules. But in reality the actual structure of the server, i.e. physically the modules, need not be executed separately. According to embodiments, multidimensional functionality of modules may be performed on a single chip. Therefore, the structure and operation of the module can be regarded as an auxiliary explanation for understanding.

The operation of each modular structure is described with reference to fig. 3.

First, log data will be collected (S11) from the client terminal 11 and the game server 20 through the communication module 110.

Here, the collected log data includes time information for recording log data, the user activity record indicating game user activity information, and game state information (terminal state information) downloaded at the terminal. And the log data may also include user information (account, mailbox, etc.).

In the present invention, in order to predict the user churn rate of the game software to be participated in and executed for analysis in advance, churn records and terminal-related information such as the game are collected and deleted, and after log data collected from the user and the terminal is analyzed, the related information required for predicting the churn rate is read.

In order to perform the above analysis, the log data information collected by the log storage unit 120 may include the following items, and these may be user activity information and terminal status information.

If the information is divided into pre-game deletion and post-game deletion, the pre-game user activity information, post-game user activity information, pre-game terminal state information, and post-game terminal state information are deleted. The terminal downloading the game may transmit the log data recorded before the log data is downloaded to the record analysis server 100 and the log data recorded after the log data is deleted to the record analysis server 100 for a predetermined period.

For example, the user activity information before the game is deleted may be information such as a game settlement failure record, a record of a dozen days of stay at a special level but not a breakthrough, and a record of a dozen days of non-execution of the game by the user even if the user performs the activity but does not a breakthrough for a special level.

Also, the terminal state information before the game is deleted may be, for example: when a network failure occurs, the game user account is recorded when another terminal is connected (such as a terminal is replaced), and the game is executed, the recording of the game is stopped by a telephone or other software (application) being executed.

These log data are transmitted from the game company client 11 and the game server 20 to the communication module 110. As described above, the game company client 11 includes a terminal or the like that has downloaded the analysis target game software.

According to the embodiment of the invention, in order to analyze and utilize the log data more effectively, the specification management can be carried out according to the user action.

For example, the analysis object may collect log data by downloading the screen information of the user terminal where the game was located before the game. When the UI is downloaded from the specific application (Facebook) connected to the object game, the specific application screen information is stored and transmitted.

Thus, the collected log data can be stored (S12) in the record storage unit 120. The stored file will execute the storage protocol and thus may also be stored in the Hadoop Ecosystem. Here, the Hadoop ecosys is a file storage system composed of a plurality of subitems of HDFS/masprreduce having a storage function in common. Collectively referred to as large data related items such as flash, HBase, Hive, Pig, Sqoop, Zookeeper, Storm, Kafka, etc.

For example, the collected log data is stored in the record storage 120 in the format of 20151010120000_ connectionlog _0001.log, 20151010120100_ itemllog _001.log, 20151010120100_ connectionlog _0001.log, 20151010120000_ virtualmonolylog _0001.log, and the like.

Moreover, the log data stored in the format automatically executes the storage protocol and stores the information to Hadoop Ecosystem.

The storage protocol engine may use Flafka, a service that mixes the Flume and Fafka used by Hadoop Ecosystem. The data is stored in three forms, Spark Streaming, Hive, which can analyze the data in real time, or HBase which can provide the original record. After the above-mentioned operation is completed, the analysis program is entered.

Moreover, Spark Streaming only analyzes game information that needs to be analyzed in real time, such as: game settlement (including virtual money), game item-related information, user access information, content execution information, and the like.

Here, since the task (task) that can be completed by a Spark Streaming solution by various platforms has been repeated in the past and distributed analysis can be performed by a plurality of servers, it is a platform suitable for large data analysis. The user can analyze the object information in real time by using Spark Streaming, such as: the analysis object downloads the initial churn rate of the game user and the real-time churn rate when the game is executed, and the like.

Further, Spark Streaming, for example, is presented in a scalar format, but has an SDK that supports a plurality of programming languages such as Python and JAVA, and thus provides a storage database for storing a plurality of data such as Hadoop, amazon S3, and Cassandra.

According to the above storage protocol, the files stored in the Hadoop Ecosystem can be subjected to a record analysis program by the analysis module 140.

Then, the log data stored by the classification and extraction module 130 according to the present invention is classified, extracted, and the like. If so, the operation of extracting statistical indexes is performed (S13).

The operation of extracting the statistical indexes is a preprocessing process of log data analysis operation at the later stage. If the log data processed by Hive is loaded to Spark in order to use the stored log data statistics, the loaded data will extract the statistical index through Spark SQL.

Spark SQL refers to processing structured or unstructured data. Not only provides SQL related basic functions, but also simplifies the data query times stored in external programs, thereby eliminating the model of database structure relationship.

The statistical indexes using Spark SQL may be in game software, such as: information such as the number of active users per day, a new sales amount per day, a repeated purchase sales amount per day, a cumulative sales amount per day, users who use less than 10 minutes per day, new registered members who naturally flow into each day, new registered members who flow in through advertisements per day, active users per day, users who newly purchase game items per day, time required for newly purchasing game items per day, repeated purchase game item users per day, users who purchase entire game items per day, users who purchase game items 5 kilo won or less per day, users who purchase game items 1 to 1.5 kilo won per day, the number of game downloads per day, the number of game deletions per day, the number of game downloads per day, and the number of game executions per day.

The extracted statistical indexes may be stored in a general database system or in the record storage unit 120, and may be used as needed by the analysis module 140. The statistical index is an index that represents various information and enables selection of a plurality of options according to the embodiment. The above statistical index is an example.

Next, in order to extract a meaningful game software analysis result, an extraction work is performed using the extracted statistical index and game log data (S14).

That is, after data analysis (record analysis) is performed on the collected game log data and the statistical index extracted by the game log data processing, the analysis module 140 performs the log data analysis again.

The analysis module 140 performs analysis by using the distribution Computing technology of Spark Core. Moreover, to evaluate the performance (or effectiveness) of the game, the changes of the statistical indicators before the game, during the game, and at the end of the game can be tracked.

The user can use the statistical indexes extracted by the classification extraction module 130 to confirm the increase range of the statistical indexes before, during and after the activity, and maintain the existence of the certain state indexes and the existence of the statistical indexes greatly increased during the activity. The data before and after the activity can be compared in detail.

When the game log data and the statistical index analysis work are performed (S104), the initial churn rate after the analysis object downloads the game and the churn rate when the game is executed are analyzed according to the embodiment of the present invention.

The following is a detailed description of prediction and analysis of the churn rate using the game log data.

First, according to the embodiment of the present invention, basic data needs to be analyzed before analyzing the initial churn rate after downloading the game and the churn rate when executing the game.

That is, the present invention analysis module 140 may predict the likelihood that a game software user will continue to execute a game, as well as the likelihood that the game will be deleted or not executed for a long period of time. Here, the terminal and the user who generate the log data belonging to the basic data may be regarded as the first user, and the terminal and the user who have a possibility of losing the newly downloaded game may be regarded as the second user by extracting the index based on the basic data.

According to the embodiment of the invention, the game loss and the user loss refer to actions of deleting the game, reconnecting and executing the game in a set period and the like by the user. Both game churn and user churn rates have the above-mentioned potential for churn.

In the future, the related work of specific standard indexes is carried out for predicting the loss rate.

The analysis module 140 collects and classifies (S101) the log data related to the game in order to analyze and predict the loss rate of the log data stored in the record storage unit 120. That is, the record storage unit 120 stores big data information related to the game log data. The analysis module 140 extracts game-related log data that can predict the churn rate from the record storage unit 120.

The extracted log data can be divided into the user activity information and the terminal state information described above by the analysis module 140, which can analyze and predict the attrition rate of the game object.

The analysis module 140 also performs a sorting operation according to time with respect to the log data determined to be related to the game (S102). That is, the log data generated by the user and the terminal is classified according to the user information (such as an account number) having a deletion record after the game download and the terminal information (such as a MAC address) having a deletion record after the game download. Then, the classified log data is divided according to the recording and generating time.

Here, the classified log data recording time cannot be a criterion for determining whether or not the game is recorded during the activity. In other words, the classified log data cannot be a criterion for determining whether the recording time is during an active period or an inactive period.

The analysis tables may be consistent during active and inactive periods. However, a different weight may be given to specific information (index information) included in log data generated by users or terminals who actually run away.

For example, when the attrition rate (P) is defined bY the formula aX + bY + cZ ═ P, X, Y and Z become index information and numerical values included in the log data, and a, b, and c become weighted values of information of each index. A specific example of calculating the loss rate, that is, the loss rate (P) ═ a (Raid ratio) + b (Raid contribution) + c (user rank) + d (attack force + defense force)/2.

The present invention gives a different weight value to specific information included in the pieces of information of the classified log data depending on whether the period of recording the log data is during an active period or an inactive period (S103).

During the activity, there is a change in the game state information. That is, there may be a change in the price of game items or a decrease in the number of crime points, etc., in order to allow more users to join the game. For example, the winning rate of Raid may be higher than the past average value, and the average contribution degree of the game user may also be changed. Therefore, the number of users in the active period and the inactive period may be greatly different. In addition, since the difficulty level of the game and the contribution level of the user may vary, the attrition rate is predicted by dividing the active period and the inactive period and giving different weights to the selected information.

Secondly, the user group with the highest loss rate can be judged by utilizing different weighted value information (index information) adopted in each log data and a set mathematical formula.

That is, after different weighted values are applied to the log data information collected by the actual churn users and the terminal according to the recording time, the information (index) with the highest churn rate is extracted and generated (S104) by using the set formula.

Since this is the log data with loss, in order to generate a meaningful index in the log data, after different weights are given to the preset mathematical formula, an index with a loss rate close to 100% needs to be obtained through calculation. In this case, the generated index can be an important reference index for predicting the slip ratio.

When log data of game users and terminals actually having a loss is newly collected and classified, information used for predicting the loss rate information included in the log data may be variously changed and modified according to the embodiment.

When the target game software for predicting the churn rate is a game like the embodiment, the following information can be applied to predict the churn rate. However, these items are merely one embodiment of the present invention, including but not limited to.

For example, the risk-breaking game may be used to predict the attrition rate, such as the ability to break customs, the time to break customs, and the percentage of content ending.

Also, for example, for the PVP (Player Vs Player) game, the PVP win rate information and PVP contribution information can be used to predict the attrition rate.

In addition, Raid (team copy activity) is taken as an example, and Raid winning rate information, Raid contribution information, Raid obtained item information, and the like can be used for predicting the attrition rate.

That is, the index used is selected to predict and analyze the loss rate of the game in real time.

And then, carrying out loss rate prediction according to the selected indexes. The erosion rate prediction is divided into the processes of initial erosion rate analysis after the analysis object downloads the game, and real-time erosion rate analysis occurring in the execution after the game is downloaded.

First, the status information of the user and the terminal downloading the game will be collected (S201). That is, the analysis module 140 predicts the churn rate in advance for increasing the usage of the game software. Therefore, personal information such as the user account, sex, age, etc. of the downloaded game software, and terminal status information of the downloaded game such as the terminal type, etc. are extracted.

Further, to predict the churn rate, basic data (journal data of churn users and terminals) index information is extracted (S202) from the journal data.

At this time, in order to predict the churn rate from the game download stage, the analysis module 140 extracts game download channel information from each log data. Although the loss rate is predicted by the log data, it is possible to give different weights to the game during the activity period according to the recording time, but the downloading method and channel of the game also play an important role in selecting the weight of the index.

In the present invention, a user who downloads or executes a game through a reward advertisement is called a non-organic user, and a user who searches and selects a download game directly from a software shop or the like without a channel of a reward advertisement or the like is called an organic user.

The return advertisement is typically presented in a manner that gives an offer to download money, items, etc. at a web link or other application link, etc.

Among these users, non-organic users generally only download games or delete games without executing the features of the games after breaking the target level. Therefore, the invention needs to record the relevant information of the game downloading channel in the log data.

The analysis module 140 predicts the churn rate of the organic user and the non-organic user, and analyzes the game use time, the content end rate, the content execution time, and the like, using the log data of the users.

Next, the extraction index is formulated to be given different weight values according to whether the collected log data is during the activity (S203).

Thereafter, the loss rate in the game download phase is predicted (S204) by comparing the weighted result value with the index standard value extracted from the base data.

Information such as downloading a game channel or mode and streaming through a return advertisement is included in the log data. When the loss rate is predicted by using the data, the loss rate of the organic user and the loss rate of the non-organic user can be predicted and analyzed in advance. Moreover, only organic user having significance of predicting the loss rate can be analyzed.

Log data is collected (S301) from a game client (terminal) in real time based on analysis target game user and terminal information.

Furthermore, to predict the attrition rate of each log data, information is extracted from the indicators that are determined to be selected. Further, the base index changes with the time (whether or not during the activity period) at which the log data is recorded (S302).

For example, log data collected during an active period is taken as an example, indexes are extracted based on a information, B information, and C information, and indexes are extracted based on B information, C information, and D information during an inactive period.

Further, a different weighting value is applied to the selection index information included in the log data with the collection period (active period or inactive period) (S303).

To predict the churn rate of the game, the user game churn rate is predicted based on log data collected in real time using the selection criteria index (S304).

Also, the user or the terminal whose loss rate is close to or breaks through its setting value may adopt the set loss prevention algorithm (S305).

The loss prevention algorithm can be adopted for users or terminals with high log data loss rate prediction values. Moreover, different algorithms may be used depending on the user gaming activity information or terminal state information.

For example, the risk game has the following conditions that the user activity information is less than the standard value and the loss rate is high: longer group adventure than low runoff rate, lower weapons use and defense than required by current adventure ratings, or 1.3 times higher consumption of goods (liquid medicine) than usual.

In this case, the user can be provided with a suggestion of taking a dangerous area, a suggestion of discounting an article (liquid medicine), or a free article, etc. in accordance with the game character level by the loss prevention algorithm, and various algorithms can be adopted.

Further, the user game execution information may have a high game churn rate due to user technical shortage.

That is, users with insufficient technology have a lot of information such as long time to break one checkpoint, low content ending rate, and low PVP.

In this case, according to the embodiment, it is possible to improve the offensive power and defensive power of the user's game character by the anti-churn algorithm, transmit item purchase prompt information capable of enhancing the power, provide free items or discount sales prompts capable of being easily operated, and the like by using different algorithms.

In other cases, for example, the user churn rate is high according to the terminal information display content. If a computer or a mobile phone with low performance is used and is frequently crashed, the judgment possibility of high loss rate exists. In this case, it is possible to provide the user with a prompt message indicating that the game is not smooth due to the game client such as a terminal or a PC.

Claims

1. The game client terminal comprises a game server capable of providing game content; and a record analysis server for collecting the log data related to the game content at the game client;

the record analysis server comprises a communication module for collecting log data from the game client, and a record storage part for storing the collected log data through the communication module; and according to the said collected or stored log data in activity or not in activity, give different weighted values to the desired selection index and selected index of the rate of loss prediction, and can analyze the analytical module of the log data;

the activity period refers to a period during which the game content performs a game activity,

the game activity refers to game state information belonging to the game content operation information, such as: at least one set value of the game difficulty, the game role level and the game speed;

the analysis module is used for extracting the user information of the game client for deleting the game content and the game client information from the log data stored in the record storage part, and analyzing the extracted log data as basic data;

the analysis module is also used for extracting basic indexes capable of predicting the attrition rate from the basic data analysis result and predicting the attrition rate of the game client through the extracted basic indexes.

2. The client of claim 1,

the analysis module is also used for extracting user activity information and terminal state information from the basic indexes extracted from the basic data;

the user activity information is user activity record information of game contents, and at least comprises more than one of game article purchase information, game level execution information, game content execution times and execution time information;

the terminal state information, that is, the game content includes game record analysis information characterized by the game client network state information before or during execution of the game client.

3. The client of claim 2,

when the game content is downloaded to any one terminal of the game client, the analysis module starts to operate in order to predict the loss rate which represents that the terminal deletes the game and the initial loss rate;

the analysis module is used for judging whether the user belongs to a first user or a second user after acquiring the game content downloading channel and the downloading mode from the log data collected by the terminal;

the analysis module may also not perform the initial churn rate analysis or give different weights to the user information depending on whether the user is the first user or the second user.

4. The client of claim 3,

the game content is downloaded to any one terminal of the game client, and when the terminal executes the game content, the analysis module performs real-time prediction of the attrition rate by comparing the log data of the terminal with the initial index;

the analysis module gives different weighted values according to whether the log data of the terminal is in the activity period or not, and the basic indexes are compared or calculated through formulation.

5. The client of claim 4,

the analysis module compares the loss rate through the basic indexes and the log data collected in real time, and then changes the game state information executed at the terminal and executes a loss prevention operation mode which can improve the user participation rate when judging that the loss rate is higher than a set value.