CN112053198B

CN112053198B - Game data processing method, device, equipment and medium

Info

Publication number: CN112053198B
Application number: CN202010992976.7A
Authority: CN
Inventors: 马捷思; 张涵宇; 谢思发
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2020-09-21
Filing date: 2020-09-21
Publication date: 2024-04-16
Anticipated expiration: 2040-09-21
Also published as: CN112053198A

Abstract

The invention discloses a game data processing method, a game data processing device, a game data processing equipment and a game data processing medium. The method comprises the following steps: obtaining target feature data from a game data group of a game object to be predicted, wherein the game data group comprises at least one of the following: game base data, game community data, and game derivative data; inputting the target characteristic data into a game sales prediction model; obtaining sales predicting results of the game objects to be predicted by using the game sales predicting model; the game sales prediction model is determined by machine learning training based on a plurality of sample feature data, and each sample feature data carries sales information labels. The invention can improve the adaptability of predicting sales volume data of the game object to be predicted, and can greatly improve the accuracy of predicting sales volume data.

Description

Game data processing method, device, equipment and medium

Technical Field

The present application relates to the field of internet communications technologies, and in particular, to a game data processing method, apparatus, device, and medium.

Background

An electronic game refers to all interactive games that run on the platform of an electronic device. There are five types of media: host games (or home games, video games), palm games, computer games, arcade games and mobile games (mainly mobile games). The perfect electronic game appears at the end of the 20 th century, changes the behavior mode of playing games by human and defines a word of the game, and belongs to a cultural activity which is born with the development of science and technology. The revenues generated by payment of electronic games can conceptually judge the financial status of related companies, the preference of markets, etc., so that the determination of the revenues is also an important link for third parties.

In the related art, a data company may conduct game sales statistics through physical retailers that are covered offline. For retailers who sign up to share data, the bar codes corresponding to each sales of the retailer are tracked for game sales statistics. The sales data obtained in this way are limited to off-line channels, even to off-line channels in a specific area, and the sales statistics of on-line channels are lacking, so that the availability of the data is poor and the accuracy is lacking. The data company may also make game sales statistics based on the user's public information in the relevant game development platform, and such sales statistics depend on the user protocol of the relevant game development platform (e.g. whether the user's information is disclosed or not, the degree to which the user's information is disclosed), so that it cannot be guaranteed that accurate sales data can be provided. Accordingly, there is a need to provide a more accurate and efficient determination scheme for sales data.

Disclosure of Invention

In order to solve the problems of low accuracy and the like when the prior art is applied to determining sales data of games, the application provides a game data processing method, a device, equipment and a medium:

according to a first aspect of the present application, there is provided a game data processing method, the method comprising:

Obtaining target feature data from a game data group of a game object to be predicted, wherein the game data group comprises at least one of the following: game base data, game community data, and game derivative data;

inputting the target characteristic data into a game sales prediction model;

obtaining sales predicting results of the game objects to be predicted by using the game sales predicting model;

the game sales prediction model is determined by machine learning training based on a plurality of sample feature data, and each sample feature data carries sales information labels.

According to a second aspect of the present application, there is provided a game data processing apparatus, the apparatus comprising:

and a data acquisition module: for obtaining target feature data from a game data group of game objects to be predicted, the game data group comprising at least one of: game base data, game community data, and game derivative data;

and a data input module: for inputting the target feature data into a game sales prediction model;

and a result prediction module: the game sales predicting model is used for obtaining sales predicting results of the game objects to be predicted;

According to a third aspect of the present application, there is provided an electronic device comprising a processor and a memory having stored therein at least one instruction or at least one program loaded and executed by the processor to implement the game data processing method according to the first aspect.

According to a fourth aspect of the present application, there is provided a computer readable storage medium having stored therein at least one instruction or at least one program, the at least one instruction or the at least one program being loaded and executed by a processor to implement the game data processing method according to the first aspect.

According to a fifth aspect of the present application, there is provided a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions so that the computer device performs the game data processing method according to the first aspect.

The game data processing method, device, equipment and medium provided by the application have the following technical effects:

the game sales predicting method comprises the steps of training a game sales predicting model by means of a machine learning method, and outputting sales predicting results of game objects to be predicted based on input target feature data by means of the game sales predicting model. The game sales prediction model obtained through training has high generalization capability, can improve the adaptability of predicting sales data of a game object to be predicted, and can greatly improve the prediction accuracy of the sales data.

Drawings

In order to more clearly illustrate the embodiments of the invention or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic diagram of an application environment provided by an embodiment of the present invention;

FIG. 2 is a schematic flow chart of a game data processing method according to an embodiment of the present invention;

FIG. 3 is a schematic flow chart of training a model for predicting game sales according to an embodiment of the present invention;

FIG. 4 is a schematic flow chart of training to obtain a plurality of game sales prediction models according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of an application scenario of a game sales prediction model provided by an embodiment of the present invention;

FIG. 6 is a schematic flow chart of a game data processing method according to an embodiment of the present invention;

FIG. 7 is an interface diagram showing a game sales interval provided by an embodiment of the present invention;

FIG. 8 is a block diagram showing a game data processing apparatus according to an embodiment of the present invention;

fig. 9 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.

Detailed Description

The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

It is noted that the terms "comprises" and "comprising," and any variations thereof, in the description and claims of the present invention and in the foregoing figures, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or server comprising a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus.

Before describing embodiments of the present invention in further detail, the terms and terminology involved in the embodiments of the present invention will be described, and the terms and terminology involved in the embodiments of the present invention will be used in the following explanation.

The Buy out game (B2P) is distinguished from pay-per-view (P2P) and free play object (F2P), and players can experience the complete content of the game after buying.

PS4 (PlayStation 4), home game machine proposed by sony corporation.

NS (Nintendo Switch ), a host issued by Nintendo corporation in month 3 of 2017, adopts a home machine and palm machine integrated design.

Xbox One, a home game machine sold by Microsoft corporation.

And (3) a digital game software issuing platform.

PSN (PlayStation Network), sony corporation provides a platform for PS (PlayStation) players to provide related web services.

Steamspin, a data tracking website for Steam.

Gamstat, a data tracking website for PSN.

Referring to fig. 1, fig. 1 is a schematic diagram of an application environment provided in an embodiment of the present invention, where the application environment may include a client 01 and a server 02, and the client 01 and the server 02 may be directly or indirectly connected through a wired or wireless communication manner. Target objects (such as users, user simulators; users may include general users and staff) may send target feature data of game objects to be predicted to a server through a client, and the server processes the received target feature data to perform sales data prediction. It should be noted that fig. 1 is only an example.

The client may include a smart phone, a desktop computer, a tablet computer, a notebook computer, an Augmented Reality (AR)/Virtual Reality (VR) device, a digital assistant, a smart speaker, a smart wearable device, or other type of physical device, and may also include software running in the physical device, such as a computer program. The operating systems corresponding to the client may include Android system, IOS system (which is a mobile operating system developed by apple corporation), linux (an operating system), microsoft Windows (microsoft windows operating system), and the like.

The server may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDNs (Content Delivery Network, content delivery networks), basic cloud computing services such as big data and artificial intelligent platforms. Wherein the server may comprise a network communication unit, a processor, a memory, etc. The server may provide background services for the client.

In the following, a specific embodiment of a game data processing method according to the present invention is described, and fig. 2 is a schematic flow chart of a game data processing method according to an embodiment of the present invention, and the present specification provides the steps of the method according to the embodiment or the flowchart, but may include more or less steps based on conventional or non-inventive labor. The order of steps recited in the embodiments is merely one way of performing the order of steps and does not represent a unique order of execution. When implemented in a real system or server product, the methods illustrated in the embodiments or figures may be performed sequentially or in parallel (e.g., in a parallel processor or multithreaded environment). As shown in fig. 2, the method may include:

s201: obtaining target feature data from a game data group of a game object to be predicted, wherein the game data group comprises at least one of the following: game base data, game community data, and game derivative data;

in the embodiment of the invention, the target feature data can be acquired by the server itself or acquired by the server due to the sending of the client. The target characteristic data is from a game data group of game objects to be predicted, the game data group including at least one of: game base data, game community data, and game derivative data:

1) The game base may indicate data describing game object base information and associated information. The base information may include base attributes (such as date of sale, time of interval of date of sale publication from date of sale, number of players, highest selling price, lowest selling price, platform of sale, etc.), languages (such as chinese, french, spanish, japanese, korean, russian, turkish, portuguese, polish, etc.), etc. The association information may include the nature of the vending platform (e.g., self-vending (vending a game developed by itself), vending (vending a game developed by others), self-vending + vending), the amount of users of the vending platform (e.g., the number of users on the day of vending the game, which may be obtained by fitting a financial reimbursement number to the least square method), the number of comments corresponding to the game on the vending platform, etc.

2) The game community data may indicate feedback comments of users (which may include professional users and general users) on game objects. The game community data may include a score of a professional user to a game developer, a score of a professional user to a game issuer, a score of a professional user to a game IP (Intellectual Property ), a score (professional) user amount corresponding to a game, and the like. The domain expertise of the professional user is defined according to the scoring object, and the professional user can point to the expert, the professional media and the like. The game community data may include a score of a game by a general user (which may include scoring, adding a type tag, posting a comment), a score (general) user amount corresponding to the game, a user score distribution corresponding to the game, and the like.

3) The game derivative data may be indicative of data relating to the game derivative. The game derivative includes physical objects such as toys, foods and ornaments which are mainly composed of games, products such as images, data, music, videos (such as comment videos and fun videos) which are mainly composed of games, and search statistics (such as the search times of a certain game theme in a period) which are mainly composed of games.

In one embodiment, before the target feature data is obtained from the game data group of the game object to be predicted, the method further includes: acquiring a time of sale (date of sale) of the game object to be predicted; acquiring a preset duration and determining the current time; and when the time length between the selling time and the current time meets the requirement of the preset time length, constructing the game data group.

Considering that the sales (such as the first month sales) in the prior sale period often accounts for a major part of the total sales of the game, especially for the buy-off game, the preset duration is introduced to measure whether the duration between the sale time and the current time meets the duration condition of the new-promotion game, and the preset duration corresponding to the new-promotion game can be 28 days, 29 days, 30 days, 31 days, and the like. For example, the selling time of the game object to be predicted is 15 days of 8 months in 2020, the current time is 31 days of 8 months in 2020, and the duration (16 days) between the selling time and the current time is less than the preset duration (for example, 30 days), so that the game object to be predicted belongs to a new promotion game, and then a game data group of the game object to be predicted is constructed. In addition, the preset duration can also be used for measuring whether the duration between the selling time and the current time meets the duration condition of the new selling game in the quarter, the duration condition of the new selling game in the half year, the duration condition of the new selling game in the year and the like, so that a sales list of the new promoting game on line in 30 days, the sales list of the selling game in a certain quarter, the sales list of the selling game in the half year, the sales list of the selling game in one year and the like can be created according to the sales prediction result obtained later. In practical application, for a new promotional game, sales prediction results may be routinely updated weekly for the first 30 days of its online. If the game is on line for more than 30 days, the game is transferred to a historical game collection, and retrospective ranking and analysis can be performed on the games in the historical game collection in a quarter, a half-year and a year.

Of course, filtering the game object to be predicted can judge whether the game object is a buy-off game, exclude limit-free games, give away obtained games and the like, in addition to judging whether the game object to be predicted is a new promotional game; and whether the game is a cross-platform game can be judged, and games with small platforms such as an exclusive login uplink (a game platform belongs to a Biyu corporation), an Origin (a game platform belongs to an art electric corporation), an Epic (a game platform belongs to an Epic corporation; an Epic corporation, a game development corporation) and the like are eliminated. The filtering conditions for screening out the buy-off game and the cross-platform game can also be used for determining the sample game objects.

In another embodiment, when the game data group includes the game base data and the game community data, the method further includes building the game data group, including: determining a target game dispensing site and obtaining the game base data from the target game dispensing site; determining a target game community site and acquiring the game community data from the target game community site; the game data group is constructed based on the game base data and the game community data.

For the game base, a target game selling site may be determined first, and the game selling site may be a site for providing a game selling service, where the target game selling site may be selected from Nintendo reshop (a game selling site belonging to Nintendo), XBOX Store (a game selling site belonging to microsoft corporation), stem and PSN. Then, game basic data are acquired from the target game community site, and a crawler can be utilized in a mode of acquiring the game basic data. The game base data here indicates data describing the game object base information to be predicted and the associated information, and reference may be made to the foregoing description about the game base data, which is not repeated here.

For the game community data, a target game community site may be determined first, where the game community site may be a website for providing a game community service, and the target game community site may be a bean-net (a community website), rawg.io (a full-platform game database for providing a game exploration service), VGtime (a game time, a game platform), IMDb (Internet Movie Database, an internet movie database), stepdb (a database website for stepm), and Metacritic (a website for specifically collecting comments for movies, television programs, music albums, games). Then, game community data is acquired from the target game community site, and a crawler can be utilized in a mode of acquiring the game community data. Here, the game community data indicates feedback comments of users (which may include professional users and general users) to be predicted game objects, and the description of the game community data may be referred to as the foregoing description, and will not be repeated here.

In practical applications, the user rating number or possession number corresponding to the game object to be predicted may be obtained from the bean net, rawg.io, VGtime, and IMDb, the rating number, the attention number, and the highest simultaneous online player number (PCU, peak concurrent users) corresponding to the game object to be predicted may be obtained from stepdb, and the MC user rating, the MC media rating, the number of MC user ratings, the number of MC media ratings, and the MC game type tag corresponding to the game object to be predicted may be obtained from meta (the feature may be from meta detail page). The game type tag may include high-weight tags such as ACT (action), adventure, RPG (Role-playing game), narrative, strategy, FPS (First-person shooter), combat, puzzle solution, street machine, science fiction, open world, survival, etc.

Because the labels of the game types are more chaotic on each website, each game has more than ten labels, for example, the intersection of 'first person shooting' and 'action' exists, the inclusion relationship of 'shooting' and 'first person shooting' exists, the 'first person puzzle' and 'first person shooting' are easy to be confused, and the unclear labels can increase the processing difficulty for the corresponding character type features. So that confusing tags can be corrected by manual intervention as needed.

In another embodiment, when the group of game data includes the game derived data, the method further includes building the group of game data, including: determining a target release site, wherein the target release site is a site for releasing a derivative object corresponding to the game object to be predicted, and the derivative object comprises at least one selected from the group consisting of a media object and a statistical object; calling a corresponding application program interface to acquire the game derivative data from the target issuing site, wherein the game derivative data is associated data of the derivative object; the game data group is constructed based on the game derived data.

The game derivative data herein indicates relevant data of the game derivative corresponding to the game object to be predicted, and reference may be made to the foregoing description of the game derivative data, which is not repeated herein. In practical applications, the sites on which the derivative objects corresponding to the game objects to be predicted are published may be sites providing live broadcast services, sites providing video services, and sites providing search services. When the target publishing site is a website for providing live broadcast service, the derivative object may be live broadcast content corresponding to the game object to be predicted (such as a host playing in a live broadcast room to explain game operation skills), and the game derivative data may be associated data (such as watching times) of the live broadcast content. When the target posting site is a website for providing video services, the derivative object may be video content (such as a game promotion video and a fun video made based on game materials) corresponding to the game object to be predicted, and the game derivative data may be associated data (such as play amount) of the video content. When the target posting site is a website providing a search service, the derivative object may be search content corresponding to the game object to be predicted (for example, searching is performed using a game as a keyword), and the game derivative data may be association data of the search content (for example, a search variation trend of the game as a keyword in a certain period based on the number of searches corresponding to each period in the certain period) and a common association word when searching is performed using the game as a keyword.

A Youtube (a video website) can be used as a target release site, and a corresponding application program interface is called to acquire Youtube video data which corresponds to a game object to be predicted and is sold for 48 hours as game derivative data. Twitch (a real-time streaming media video platform facing video games) can be used as a target release site, and a corresponding application program interface is called to acquire the highest number of watching people of the Twitch history corresponding to the game object to be predicted as game derivative data. The Google search engine can be used as a target release site, and a corresponding application program interface is called to acquire a Google trend (Google Trends) relative value corresponding to the game object to be predicted as game derivative data.

In another embodiment, the target feature data may also be from a data analysis platform based on massive user behavior data, and target feature data corresponding to the game object to be predicted may be obtained from such a data analysis platform. Such data analysis platforms often display analysis data in a form of a graph, and the displayed analysis data often corresponds to numerical characteristics, so that a graph recognition mode can be adopted when target characteristic data are acquired from such data analysis platforms.

S202: inputting the target characteristic data into a game sales prediction model;

in the embodiment of the invention, the game sales prediction model is determined by machine learning training based on a plurality of sample feature data, and each sample feature data carries sales information labels. The sample feature data is derived from the game data group of the sample game object, and the sample feature data and the game data group of the sample game object may refer to the description of the target feature data and the game data group of the game object to be predicted in step S201, which is not described in detail. It should be noted that, the "plurality of sample feature data" herein is from a game data group of a plurality of sample game objects, and the number of sample feature data corresponding to each sample game object may be a plurality. Multiple sample feature data corresponding to the same sample game object may be incorporated into one sample feature data set. In practice, the sample game object may indicate games for sale in 2013 and later, because both PS4 and Xbox One were published in 2013. Further, the sample game correspondence may select a buy-off game.

The sales information label carried by the sample feature data can be determined based on the sales revealed by stem 8 in 2018, PSN 12 in 2018, the published sales of the financial newspaper of each gaming company, and news-revealed messages. In practical application, the sample game object corresponds to the games of each platform of 1200 remaining money in 2013. In determining sales information labels based on the sales leaked by stem and the sales leaked by PSN, the following method may be used:

1) Random sampling method: randomly acquiring public information of part of users, further presuming the possession of a game corresponding to a platform and expressing sales volume by the possession, and then optimizing parameters used by the presumption based on the leaked sales volume data, so that the optimized parameters can be utilized in subsequent sales volume determination;

2) Comment number estimation algorithm: the sales of a game on a corresponding platform are estimated by multiplying the number of comments by a coefficient (median 77, average 82) between 30 and 150. The earlier the time of sale game, the higher the coefficient may be selected, e.g., game sold in 2014 selects 110 as the coefficient and game sold in 2017 selects 65 as the coefficient. Of course, some game type correction parameters may be added when the sales are inferred by the number of comments, but in general, the calculation of the correction parameters is introduced to be highly correlated with the number of comments multiplied by 65. The parameters used for the above-described speculation are optimized based on the leaked sales data, so that the optimized parameters can be utilized in subsequent sales determinations.

Of course, for the application of the reference player number corresponding to the achievement trophy leaked by the PSN, it may be determined based on the disclosure information of the user how many proportion of the players play a certain game, that is, the achievement trophy (such as MyPS4Life achievement trophy; myPS4Life, my PS4Life, a community media planning activity of sony corporation) of the game is obtained, and then the actual player number (=reference player number) of the game may be known based on the reference player number corresponding to the leaked achievement trophy and the sales may be represented by the actual player number.

For sales information labels carried by sample feature data, the sales information label indication may include at least one of: the total sales of all platforms (when only the current total sales of a certain platform is used as the current total sales of all platforms, otherwise, the sum of the current total sales of at least two certain platforms is used as the current total sales of all platforms), and the total sales of all platforms corresponding to a certain historical time (when only the total sales of a certain platform corresponding to a certain historical time is used as the total sales of all platforms corresponding to a certain historical time, otherwise, the sum of the total sales of at least two certain platforms corresponding to a certain historical time is used as the total sales of all platforms corresponding to a certain historical time).

In one embodiment, as shown in FIG. 3, the method further includes training the process of deriving the game sales prediction model:

s301: acquiring the plurality of sample feature data;

before the sample feature data is input into the preset machine learning model, it may be determined whether to further process the sample feature data according to its type. For numerical features, it may be used without treatment. For literal features (such as language, labels, etc.), one-hot encoding may be used after encoding.

Before the sample characteristic data is input into a preset machine learning model, the characteristic data to be input can be optimized through outlier rejection and missing value filling. Outliers may indicate sample feature data that is clearly undesirable for practical situations. For example, a game title has only one common word, and when a Youtube API is called, abnormal return data may occur, and when data crawling from stem, return data of the same-name game may occur, so that screening is required. The usual optimization is to discard the sample feature data of these anomalies as missing values. In view of the application of the limit gradient lifting (XGBoost, eXtreme Gradient Boosting) model (algorithm), or the adoption of a regression method (such as the squarederror method) in XGBoos toolkit, the missing values (XGBoost model (algorithm) or XGBoos toolkit) can be treated as a sparse matrix, and the missing values are not considered when the nodes are split.

The feature dimension of the sample feature data and the accuracy of the model output data are comprehensively considered to determine whether to compress the feature dimension of the sample feature data by using PCA (Principle Component Analysis, a compression method). Under the condition of few feature dimensions, the feature dimensions can not be compressed so as to ensure the accuracy of model output data. Accordingly, whether to compress the feature dimension of the target feature data can be determined according to whether the sample feature data is compressed in the feature dimension.

In addition, the process of training to obtain a game sales prediction model can be regarded as a regression process, and the target of regression is sales. Considering that sales for a period do not fit a normal distribution, and that all sales are greater than zero and have a relatively strong right deviation (the higher the sales, the stronger the deviation), the sales indicated by the sales information can be annotated with natural logarithms to reduce the deviation. That is, the regression target may be set to the natural logarithm of sales first, and then later restored to sales figures. And marking the indicated sales volume by combining the sales volume information, and preferentially selecting the current total sales volume of all platforms by the regression target.

S302: based on the sample feature data, performing sales volume data prediction training by using a preset machine learning model, and adjusting model parameters of the preset machine learning model in training until sales volume prediction results output by the preset machine learning model are matched with sales volume information labels carried by the input sample feature data;

the preset machine learning model can adopt an XGBoost model, the XGBoost model is a model based on the construction of a decision tree, whether characteristic data of the XGBoost model is subjected to standardized, normalized and other characteristic scaling treatment has small influence on model training, and the accuracy of model output data cannot be effectively improved. In view of the efficiency of model training, feature scaling processing may not be performed on feature data input to the model.

The preset machine learning model can be an initial model or an intermediate model. And inputting sample characteristic data carrying sales information labels into a preset machine learning model to conduct sales data prediction training. In training, model parameters may be adjusted based on differences between intermediate results (sales volume data of a sample game object corresponding to sample feature data) output by the model and sales volume information labels carried by the sample feature data.

S303: and taking a preset machine learning model corresponding to the adjusted model parameters as the game sales predicting model.

The game sales predicting model with high generalization capability is obtained by training a machine learning model, and the adaptability of the game sales predicting can be improved when the sales data is predicted by using the game sales predicting model, so that the reliability and the effectiveness of the game sales predicting can be greatly improved.

As shown in fig. 5, fig. 5 is a schematic diagram of an application scenario of a game sales prediction model provided by an embodiment of the present invention. In fig. 5, the training data is sample feature data, and each sample feature data carries sales information labels; correspondingly, the game sales prediction model trained later can predict sales data of the target characteristic data. In fig. 5, target feature data is input into the game sales prediction model, and the sales prediction result of the game object to be predicted corresponding to the target feature data is output through the game sales prediction model.

In another embodiment, as shown in FIG. 4, the method further includes the process of training a plurality of game sales prediction models:

s401: acquiring the plurality of sample feature data;

Reference is made to the processing content of the sample feature data in the foregoing step S301, and details are not repeated here.

S402: acquiring a plurality of random numbers, wherein the random numbers indicate the same proportion of dividing the characteristic data of the samples, and the random numbers indicate different sample sets to form a data combination;

for example, the same division ratio indicated by the plurality of random numbers is 8:2, the plurality of sample feature data includes data 1-10, and the sample set indicated by the random number a is different from the random numbers b and c because the plurality of random numbers indicate different sample sets to form a data combination.

S403: dividing the plurality of sample feature data into corresponding first and second sample sets based on sample set composition data combinations indicated by each of the random numbers, respectively;

correspondingly, the sample sets indicated by the random numbers form data combinations to divide the sample characteristic data into different sample sets: for the random number a, the data ratio of the first sample set to the second sample set is 8:2, the composition data of the first sample set is 1-8, and the composition data of the second sample set is 9-10; for the random number b, the data ratio of the first sample set to the second sample set is 8:2, the constituent data of the first sample set are data 1-3, 5 and 7-10, and the constituent data of the second sample set are data 4 and 6; for the random number c, the data ratio of the first sample set to the second sample set is 8:2, the constituent data of the first sample set are data 2-4 and 6-10, and the constituent data of the second sample set are data 1 and 5.

Of course, a plurality of random numbers may also indicate different division ratios. For example, the dividing ratio indicated by the random number a is 8:2, the dividing ratio indicated by the random number b is 7:3, and the dividing ratio indicated by the random number c is 9:1. Correspondingly, the plurality of sample characteristic data are divided into different sample sets according to the division proportion indicated by the random number. For example, the division ratio indicated by the random number a is 8:2, and 80% of the plurality of sample feature data may be divided into the first sample set and 20% of the plurality of sample feature data may be divided into the second sample set based on the number dimension. Since the "plurality of sample feature data" is from a game data group of a plurality of sample game objects, and the number of sample feature data corresponding to each sample game object may be plural, a plurality of sample feature data corresponding to the same sample game object may be incorporated into one sample feature data group. Then 80% of the plurality of sample feature data sets may be partitioned into the first sample set and 20% of the plurality of sample feature data sets may be partitioned into the second sample set based on the group dimension when the partition ratio indicated by the random number a is 8:2.

S404: constructing a reference model corresponding to each random number based on a first sample set corresponding to each random number and a preset machine learning algorithm;

The first sample set may be considered as a training set, since each random number corresponds to a training set, here a reference model is built based on each training set and a pre-set machine learning algorithm, respectively. For example, the random number a corresponds to a first sample set a and a second sample set a, and a reference model a is built based on the first sample set a and a preset machine learning algorithm; the random number b corresponds to the first sample set b and the second sample set b, and a reference model b is built based on the first sample set b and a preset machine learning algorithm; the random number c corresponds to the first sample set c and the second sample set c, and a reference model c is built based on the first sample set c and a preset machine learning algorithm.

S405: and optimizing a corresponding reference model by using a corresponding second sample set based on each random number to obtain a game sales prediction model corresponding to each random number.

Combining the reference model a, the reference model b, and the reference model c obtained in step S404. The second sample set a can be utilized to optimize the reference model a, and the optimized reference model a is used as a game sales predicting model a; the second sample set b can be utilized to optimize the reference model b, and the optimized reference model b is used as a game sales predicting model b; the reference model c may be optimized using the second sample set c, and the optimized reference model c may be used as the game sales prediction model c.

Correspondingly, the inputting the target feature data into the game sales prediction model comprises the following steps: the target feature data is input into a plurality of the game sales prediction models, respectively, and the difference between the plurality of the game sales prediction models is a composition data combination divided into a first sample set and a second sample set in the same ratio, or a ratio of the plurality of sample feature data divided into the first sample set and the second sample set.

The process of optimizing the corresponding reference model using the corresponding second sample set may be:

first kind: firstly, acquiring a preset fitting goodness; and then, based on the corresponding second sample set, carrying out sales volume data prediction training by using a corresponding reference model, and adjusting model parameters of the corresponding reference model in the training until sales volume prediction results output by the corresponding reference model and sales volume information labels carried by input sample characteristic data meet the requirement of the preset fitting goodness.

The preset goodness of fit is used as a model evaluation index of the reference model, the preset goodness of fit corresponds to an R2 score (namely a decision coefficient in statistics and also translated into a decision coefficient), and the R2 score can reflect the proportion that all variations of the dependent variable can be interpreted by the independent variable through a regression relation. For example, if the R2 score is 0.8, it means that the variance of 80% of the dependent variable can be explained based on the regression relationship. The larger the value of the R2 score, the higher the degree of interpretation of the dependent variable by the independent variable, the higher the percentage of variation due to the independent variable to the total variation, and the denser the observation points are near the regression line. The preset fitting goodness can be 0.8, 0.88, 0.9 and the like, and can be flexibly adjusted according to the needs. This optimization can be seen as the second sample set both as a test set for testing the reference model and as another training set for guiding the parameter adjustment of the reference model.

Second kind: the second sample set can be regarded as a test set to test the reference model, the preset goodness of fit (the related record of the preset goodness of fit can be referred to) is regarded as a model evaluation index of the reference model, and whether the reference model meets the requirement of the model evaluation index is determined according to the sales prediction result output by the model. When not satisfied, continuing to adjust parameters of the reference model using the first sample set.

In practical applications, sklearn (an open-source python language-based machine learning tool package) as a third party module commonly used in machine learning, the training set and the test set may be divided based on the train_test_split implementation of the sklearn tool package (see the following codes), and the R2 score may be calculated based on the r2_score of the sklearn tool package:

X_train，X_test，y_train，y_test＝train_test_split(X，y，test_size＝0.2，random_state＝1)

regress_model＝xgb.XGBRegressor(objective＝”reg：squarederror”，nthread＝-1，seed＝1234)

the parameter random_state and seed can jointly determine generation of random numbers, or can be regarded as a label of the random numbers, and different random numbers can be positioned through different assignment combinations of the random numbers. The parameter test_size may indicate a division ratio. The number of random numbers can be 1000, and correspondingly, 1000 game sales prediction models can be obtained.

The model r2 score after multiple tests is about 0.88, which can indicate that the calculation of the model is relatively accurate. Of course, the model may be updated and retrained based on the amount of sales published by the new financial newspaper, the new amount of sales disclosed by the news, and the amount of sales of subsequent platform leaks.

S203: obtaining sales predicting results of the game objects to be predicted by using the game sales predicting model;

in the embodiment of the invention, the sales predicting result of the game object to be predicted can be output by using the game sales predicting model. In combination with the description in the foregoing steps S401 to S405, the target feature data may be input into a plurality of game sales prediction models, and then each game sales prediction model may be used to obtain a corresponding candidate result, and further, a sales prediction result may be obtained based on the candidate results corresponding to the plurality of game sales prediction models. Statistical analysis can be performed on a plurality of candidate results, and sales intervals for the game object to be predicted can be obtained by calculating the corresponding average value mu and standard deviation sigma. Instead of a single sales number, a more representative data can be given by using sales intervals. For each game, 1000 randomized models are substituted for calculation, so that the average value mu and standard deviation sigma of sales of the corresponding game can be obtained. According to the Laida criterion, the actual sales corresponding to the game object to be predicted fall within (μ - σ, μ+σ) with a probability of 65%, with a probability of 95% falling within (μ -2σ, μ+2σ), and with a probability of 99% falling within (μ -3σ, μ+3σ), also referred to as the 3σ principle. In practical applications, when the front-end presentation is performed using the sales prediction result of the game object to be predicted, the 6 σ interval is considered to be too large, sometimes even more than an order of magnitude, because (μ - σ, μ+σ) is selected to be used in determining the sales interval for the game object to be predicted at the time of the front-end presentation.

In one embodiment, based on the idea that the regression target can be set to be the natural logarithm of sales first, and then restored to be sales figures later, the sales prediction result of the game object to be predicted, which is output by the game sales prediction model, is the natural logarithm of sales, and needs to be restored to be sales figures.

Referring to fig. 6 and 7, the game sales interval obtained after calculation of the game sales prediction model will be shown on a specific page. The embodiment of the invention can develop a set of models for calculating sales of any buying and breaking PC or host game by utilizing machine learning and data mining technologies, and integrate the models into the background of a small program in instant messaging application so as to display the sales of the game through the small program. Therefore, the automatic flow of data mining, sales calculation and display can be realized every time a new online game is found. The small program product has the characteristics of high visualization degree, strong usability, high data precision and the like, and has irreplaceable huge advantages in calculation of cross-platform game sales.

Considering that the sales (such as the first month sales) in the pre-sales period often account for a major portion of the total sales of games, especially for the buy-out games, the sales interval for games sold for nearly 30 days can be presented and ranked from high to low on the main page of the applet product. After the selling time of the game exceeds 30 days, the sales volume interval of the game can be classified into historical archives, and further summarized list review is carried out in quarters, half-years and years, so that the method is convenient to check and correct with the financial newspaper numbers published by various manufacturers, and the sales volume published by the financial newspaper, the sales volume disclosed by news and the sales volume disclosed by a follow-up platform are convenient to compare to update a model. Besides the list updated in real time, the game icons on the right side of the figure 7 can be triggered to jump into game detail pages of the corresponding games, and the game detail pages integrate sales intervals calculated through the model, so that visual comprehensive information display is conveniently carried out for users.

In practical application, the embodiment of the invention adopts a machine learning method, and can train and model each feature of the PC+host platform buying and breaking game in the global market by utilizing known real sales figures (sources of financial accounting, news, data leakage and the like) and calculate sales of other games. The model obtained by the embodiment of the invention can give out sales figures summarized by each platform in the global market, has no limitation on regions and platforms, can also calculate sales for some platforms (such as NS) without API support, and makes up the blank of the existing game world calculation tool. Since most games can log on to several platforms, it is difficult to give the overall performance of the game considering only the sales of the PC. The sales awareness of host platforms such as PS4 and NS is not known for a long time, so that the model obtained by the embodiment of the invention can realize sales calculation tasks of a plurality of platforms such as PC, PS4, XBOXONE, NS and the like. Meanwhile, since the sum of the financial reports of each game company is the total sales of all the platforms, sales analysis tools for verifying the data are lacking at present. If the sales of any single game on a plurality of platforms can be calculated, the income of the game and the financial situation of the corresponding company can be calculated to a certain degree.

The game sales prediction model provided by the embodiment of the invention not only can provide the cross-platform sales corresponding to the game to be predicted, but also can provide a sales interval with smaller range and higher accuracy. The game sales prediction model provided by the embodiment of the invention can ensure that the actual sales of most games fall in a section with the standard deviation of plus or minus three times of the sales average value calculated by using the model, even the actual sales of a plurality of games fall in a narrower section. Compared with wide and fixed intervals such as 50-100 ten thousand and 200-500 ten thousand which can be given in the related art, the sales intervals (such as 17-27 ten thousand, 22-34 ten thousand, 51-75 ten thousand and 419-623 ten thousand) obtained by using the game sales prediction model provided by the embodiment of the invention are smaller and more flexible, and the average value is closer to the real sales.

As can be seen from the technical solutions provided in the embodiments of the present disclosure, a game sales prediction model is trained by using a machine learning method, and a sales prediction result of a game object to be predicted is output based on input target feature data by using the game sales model. The game sales prediction model obtained through training has high generalization capability, can improve the adaptability of predicting sales data of a game object to be predicted, and can greatly improve the prediction accuracy of the sales data.

The embodiment of the invention also provides a game data processing device, as shown in fig. 8, comprising:

the data acquisition module 810: for obtaining target feature data from a game data group of game objects to be predicted, the game data group comprising at least one of: game base data, game community data, and game derivative data;

a data input module 820: for inputting the target feature data into a game sales prediction model;

the result prediction module 830: the game sales predicting model is used for obtaining sales predicting results of the game objects to be predicted;

It should be noted that the apparatus and method embodiments in the apparatus embodiments are based on the same inventive concept.

The embodiment of the invention provides an electronic device, which comprises a processor and a memory, wherein at least one instruction or at least one section of program is stored in the memory, and the at least one instruction or the at least one section of program is loaded and executed by the processor to realize the game data processing method provided by the embodiment of the method.

Further, fig. 9 shows a schematic hardware structure of an electronic device for implementing the game data processing method provided by the embodiment of the present invention, where the electronic device may participate in forming or including the game data processing apparatus provided by the embodiment of the present invention. As shown in fig. 9, the electronic device 90 may include one or more processors 902 (shown in the figures as 902a, 902b, … …,902 n) (the processor 902 may include, but is not limited to, a microprocessor MCU or a processing device such as a programmable logic device FPGA), a memory 904 for storing data, and a transmission device 906 for communication functions. In addition, the method may further include: a display, an input/output interface (I/O interface), a Universal Serial Bus (USB) port (which may be included as one of the ports of the I/O interface), a network interface, a power supply, and/or a camera. It will be appreciated by those skilled in the art that the configuration shown in fig. 9 is merely illustrative and is not intended to limit the configuration of the electronic device. For example, the electronic device 90 may also include more or fewer components than shown in FIG. 9, or have a different configuration than shown in FIG. 9.

It should be noted that the one or more processors 902 and/or other data processing circuitry described above may be referred to herein generally as "data processing circuitry. The data processing circuit may be embodied in whole or in part in software, hardware, firmware, or any other combination. Further, the data processing circuitry may be a single stand-alone processing module, or incorporated, in whole or in part, into any of the other elements in the electronic device 90 (or mobile device). As referred to in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of the path of the variable resistor termination to interface).

The memory 904 may be used to store software programs and modules of application software, such as program instructions/data storage devices corresponding to the game data processing methods described in the embodiments of the present invention, and the processor 902 executes the software programs and modules stored in the memory 94 to perform various functional applications and data processing, i.e., implement one of the game data processing methods described above. The memory 904 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 904 may further include memory remotely located relative to the processor 902, which may be connected to the electronic device 90 via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The transmission means 906 is used for receiving or transmitting data via a network. Specific examples of the network described above may include a wireless network provided by a communications provider of the electronic device 90. In one example, the transmission means 906 includes a network adapter (NetworkInterfaceController, NIC) that can be connected to other network devices through a base station to communicate with the internet. In one embodiment, the transmission device 906 may be a radio frequency (RadioFrequency, RF) module for communicating wirelessly with the internet.

The display may be, for example, a touch screen type Liquid Crystal Display (LCD) that may enable a user to interact with a user interface of the electronic device 90 (or mobile device).

Embodiments of the present invention also provide a computer readable storage medium that can be disposed in an electronic device to store at least one instruction or at least one program related to implementing a game data processing method in a method embodiment, where the at least one instruction or the at least one program is loaded and executed by the processor to implement the game data processing method provided in the above method embodiment.

Alternatively, in this embodiment, the storage medium may be located in at least one network server among a plurality of network servers of the computer network. Alternatively, in the present embodiment, the storage medium may include, but is not limited to: a U-disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a removable hard disk, a magnetic disk, or an optical disk, or other various media capable of storing program codes.

It should be noted that: the sequence of the embodiments of the present invention is only for description, and does not represent the advantages and disadvantages of the embodiments. And the foregoing description has been directed to specific embodiments of this specification. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for the apparatus and electronic device embodiments, since they are substantially similar to the method embodiments, the description is relatively simple, and references to the parts of the description of the method embodiments are only required.

It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program for instructing relevant hardware, where the program may be stored in a computer readable storage medium, and the storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.

The foregoing description of the preferred embodiments of the invention is not intended to limit the invention to the precise form disclosed, and any such modifications, equivalents, and alternatives falling within the spirit and scope of the invention are intended to be included within the scope of the invention.

Claims

1. A method of game data processing, the method comprising:

obtaining target feature data from a game data group of a game object to be predicted, wherein the game data group comprises at least one of the following: game basic data, game community data and game derivative data, wherein the game basic data indicates the highest selling price and the lowest selling price of the game object to be predicted and the comment number of the game object to be predicted at a selling site, the game community data indicates the user grading, the comment number and the attention number of the game object to be predicted at a community site, and the game derivative data indicates the number of media contents corresponding to the game object to be predicted, the watching times and the searched times of the searched contents corresponding to the game object to be predicted;

Inputting the target feature data into a plurality of game sales prediction models respectively to obtain a plurality of candidate results, wherein the plurality of candidate results are in one-to-one correspondence with the candidate results output by the game sales prediction models respectively, the game sales prediction models are determined by machine learning training based on a plurality of sample feature data, each sample feature data carries sales information labels, and the difference among the game sales prediction models is that the sample feature data are divided into different composition data combinations of a first sample set and a second sample set according to the same proportion;

obtaining sales prediction results based on the plurality of candidate results, wherein the sales prediction results indicate sales intervals obtained by carrying out statistical analysis on the plurality of candidate results;

wherein the method further comprises the process of training the plurality of game sales prediction models: acquiring the plurality of sample feature data, wherein the sample feature data is acquired from a game data group of a sample game object; acquiring a plurality of random numbers, wherein the random numbers indicate the same proportion of dividing the characteristic data of the samples, and the random numbers indicate different sample sets to form a data combination; dividing the plurality of sample feature data into corresponding first and second sample sets based on sample set composition data combinations indicated by each of the random numbers, respectively; constructing a reference model corresponding to each random number based on a first sample set corresponding to each random number and a preset machine learning algorithm; based on each random number, optimizing a corresponding reference model by using a corresponding second sample set to obtain a game sales prediction model corresponding to each random number;

The optimizing the corresponding reference model using the corresponding second sample set includes: acquiring a preset fitting goodness; and based on the corresponding second sample set, carrying out sales volume data prediction training by using the corresponding reference model, and adjusting model parameters of the corresponding reference model in training until sales volume prediction results output by the corresponding reference model and sales volume information labels carried by input sample characteristic data meet the requirement of the preset fitting goodness.

2. The method of claim 1, wherein prior to the obtaining target feature data from the game data group of the game object to be predicted, the method further comprises:

acquiring the selling time of the game object to be predicted;

acquiring a preset duration and determining the current time;

and when the time length between the selling time and the current time meets the requirement of the preset time length, constructing the game data group.

3. The method of claim 1, wherein when the game data group includes the game base data and the game community data, the method further comprises constructing the game data group, comprising:

Determining a target game dispensing site and obtaining the game base data from the target game dispensing site;

determining a target game community site and acquiring the game community data from the target game community site;

the game data group is constructed based on the game base data and the game community data.

4. The method of claim 1, wherein when the group of game data includes the game derived data, the method further comprises building the group of game data, comprising:

determining a target release site, wherein the target release site is a site for releasing a derivative object corresponding to the game object to be predicted, and the derivative object comprises at least one selected from the group consisting of a media object and a statistical object;

calling a corresponding application program interface to acquire the game derivative data from the target issuing site, wherein the game derivative data is associated data of the derivative object;

the game data group is constructed based on the game derived data.

5. A game data processing device, the device comprising:

and a data acquisition module: for obtaining target feature data from a game data group of game objects to be predicted, the game data group comprising at least one of: game basic data, game community data and game derivative data, wherein the game basic data indicates the highest selling price and the lowest selling price of the game object to be predicted and the comment number of the game object to be predicted at a selling site, the game community data indicates the user grading, the comment number and the attention number of the game object to be predicted at a community site, and the game derivative data indicates the number of media contents corresponding to the game object to be predicted, the watching times and the searched times of the searched contents corresponding to the game object to be predicted;

And a data input module: the method comprises the steps that target feature data are respectively input into a plurality of game sales volume prediction models to obtain a plurality of candidate results, the plurality of candidate results are in one-to-one correspondence with candidate results output by the game sales volume prediction models, the game sales volume prediction models are determined by machine learning training based on a plurality of sample feature data, each sample feature data carries sales volume information labels, and the difference among the game sales volume prediction models is that the sample feature data are divided into different composition data combinations of a first sample set and a second sample set according to the same proportion;

and a result prediction module: the sales prediction method comprises the steps of obtaining sales prediction results based on the plurality of candidate results, wherein the sales prediction results indicate sales intervals obtained by carrying out statistical analysis on the plurality of candidate results;

the device further comprises a model training module, wherein the model training module is used for training and obtaining the game sales prediction models, and the model training module is used for: for obtaining the plurality of sample feature data, the sample feature data obtained from a game data group of sample game objects; acquiring a plurality of random numbers, wherein the random numbers indicate the same proportion of dividing the characteristic data of the samples, and the random numbers indicate different sample sets to form a data combination; dividing the plurality of sample feature data into corresponding first and second sample sets based on sample set composition data combinations indicated by each of the random numbers, respectively; constructing a reference model corresponding to each random number based on a first sample set corresponding to each random number and a preset machine learning algorithm; based on each random number, optimizing a corresponding reference model by using a corresponding second sample set to obtain a game sales prediction model corresponding to each random number;

6. The apparatus of claim 5, further comprising a first data group construction module, the first data group construction module: acquiring a selling time of the game object to be predicted; acquiring a preset duration and determining the current time; and when the time length between the selling time and the current time meets the requirement of the preset time length, constructing the game data group.

7. The apparatus of claim 5, wherein when the game data group includes the game base data and the game community data, the apparatus further comprises a second data group construction module that: for determining a target game dispensing site, and obtaining the game base data from the target game dispensing site; determining a target game community site and acquiring the game community data from the target game community site; the game data group is constructed based on the game base data and the game community data.

8. The apparatus of claim 5, wherein when the game data group includes the game-derived data, the apparatus further comprises a third data group construction module that: the target issuing station is used for issuing a derivative object corresponding to the game object to be predicted, and the derivative object comprises at least one selected from the group consisting of a media object and a statistical object; calling a corresponding application program interface to acquire the game derivative data from the target issuing site, wherein the game derivative data is associated data of the derivative object; the game data group is constructed based on the game derived data.

9. An electronic device comprising a processor and a memory, wherein the memory has stored therein at least one instruction or at least one program, the at least one instruction or the at least one program being loaded and executed by the processor to implement the game data processing method of any of claims 1-4.

10. A computer readable storage medium having stored therein at least one instruction or at least one program, the at least one instruction or the at least one program being loaded and executed by a processor to implement the game data processing method of any one of claims 1-4.