CN109670876A - The price data prediction technique and device of virtual objects in a kind of game - Google Patents
The price data prediction technique and device of virtual objects in a kind of game Download PDFInfo
- Publication number
- CN109670876A CN109670876A CN201910002634.3A CN201910002634A CN109670876A CN 109670876 A CN109670876 A CN 109670876A CN 201910002634 A CN201910002634 A CN 201910002634A CN 109670876 A CN109670876 A CN 109670876A
- Authority
- CN
- China
- Prior art keywords
- price
- model
- attribute
- training
- build
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0202—Market predictions or forecasting for commercial activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0278—Product appraisal
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/131—Protocols for games, networked simulations or virtual reality
Abstract
The embodiment of the invention provides a kind of price datas of virtual objects in game to generate prediction technique and device, wherein the described method includes: obtaining the attribute information of virtual objects to be predicted;Using the attribute information of the virtual objects to be predicted, attributive character to be predicted is generated;The attributive character to be predicted is input to preset price expectation model;Receive the price data that price expectation model generates.The embodiment of the present invention may be implemented to construct price expectation model according to dimensions such as the concluded price of historical trading virtual objects, build-in attribute, random attribute, subordinate servers, and price expectation model is enabled to carry out price expectation according to the build-in attribute of virtual objects, random attribute, subordinate server.
Description
Technical field
The present invention relates to game technical fields, more particularly to a kind of price data prediction side of virtual objects in game
Method, a kind of price expectation model building method, the price data prediction meanss of virtual objects, a kind of price expectation in a kind of game
Model construction device, electronic equipment and storage medium.
Background technique
With the development of intelligent terminal and Internet technology, more and more people are using game as a kind of daily routines.Trip
The player that plays can sell or buy game articles on some game virtual article trading platforms.And the game articles traded
It can be roughly divided into two types, one kind is " mark product ", i.e., there are more identical game articles in game, in addition one kind is " non-
Mark product ", i.e., more rare or even unique game articles in game.
In the prior art, it for the price expectation of " nonstandard product ", is usually used and thinks setting rule, rough estimation is " non-
The price of mark product ", this prediction mode may with there is large error in practice, and rule needs manual maintenance, cost
It is higher.
Summary of the invention
In view of the above problems, it proposes the embodiment of the present invention and overcomes the above problem or at least partly in order to provide one kind
Price data prediction technique, device, electronic equipment and the storage medium of virtual objects in a kind of game to solve the above problems.
To solve the above-mentioned problems, the embodiment of the invention discloses a kind of price data prediction sides of virtual objects in game
Method, comprising:
Obtain the attribute information of virtual objects to be predicted;
Using the attribute information of the virtual objects to be predicted, attributive character to be predicted is generated;
The attributive character to be predicted is input to preset price expectation model;
Receive the price data that price expectation model generates.
Preferably, the attributive character to be predicted is corresponding with destination server mark;The price expectation model includes: solid
There are attributes price prediction model, random attribute price expectation model and server coefficient;The price data is by the following method
It generates:
Build-in attribute forecast price is generated using the attributive character to be predicted and the build-in attribute price expectation model;
Random attribute forecast price is generated using the attributive character to be predicted and the random attribute price expectation model;
It is determined using destination server mark, the build-in attribute forecast price and the random attribute forecast price
Destination server coefficient;
Calculating the sum of the build-in attribute forecast price and the random attribute forecast price is fundamentals of forecasting price;
The product for calculating the basic forecast price and the destination server coefficient is the price data.
Preferably, the price expectation model generates by the following method:
Obtain multiple target histories transaction records corresponding with virtual objects and initial model;The target histories are handed over
Easily record includes concluded price and attribute information;
According to the attribute information of target histories transaction record, the aspect of model is generated;
Using the concluded price and the aspect of model, the training initial model;
The multiple loss functions for calculating initial model after training, when multiple loss letters of initial model after training
When number all minimizes, initial model described in deconditioning;
Price expectation model is generated according to the initial model trained.
The embodiment of the invention also discloses a kind of price expectation model building methods, comprising:
Obtain multiple target histories transaction records corresponding with virtual objects and initial model;The target histories are handed over
Easily record includes concluded price and attribute information;
According to the attribute information, the aspect of model is generated;
Using the concluded price and the aspect of model, the training initial model;
The multiple loss functions for calculating initial model after training, when multiple loss letters of initial model after training
When number all minimizes, initial model described in deconditioning;
Price expectation model is generated according to the initial model trained.
Preferably, the attribute information includes build-in attribute and/or random attribute;It is described to use the concluded price and institute
The step of stating the aspect of model, training the initial model, comprising:
Determination only includes that the target histories transaction record of the virtual goods of build-in attribute is the first training record;
Determine that the aspect of model corresponding with first training record is the first training characteristics;
Determination only includes that the target histories transaction record of the virtual goods of a random attribute is the second training record;
Determine that the aspect of model corresponding with second training record is the second training characteristics;
According to first training characteristics and second training characteristics, the training initial module;
Wherein, the random attribute in second training record is opposite with the build-in attribute in first training record
It answers.
Preferably, the initial model includes: build-in attribute price model;It is described according to first training characteristics and institute
The step of stating the second training characteristics, training the initial module, comprising:
Determine that the concluded price in the first training record is build-in attribute price;
Determine first training characteristics and with the build-in attribute price be the first training data;
Using first training data, the training build-in attribute price model.
Preferably, the initial model further includes random attribute price model;It is described according to first training characteristics and
The step of second training characteristics, the training initial module, comprising:
Determine the first training characteristics of target corresponding with second training characteristics;
Calculate random attribute price;The random attribute price is that the corresponding price of the second training characteristics and target first are instructed
Practice the difference between the corresponding price of feature;
It determines second training characteristics and the random attribute price is the second training data;
Using second training data, the training random attribute price model.
Preferably, described the step of price expectation model is generated using the initial model trained, comprising:
Determine that the build-in attribute price model trained is build-in attribute price expectation model;
Determine that the random attribute price model trained is random attribute price expectation model;
Using the aspect of model, build-in attribute price expectation model, random attribute price expectation model, server is generated
Coefficient;
The build-in attribute price expectation model, random attribute price expectation model and the server coefficient are combined, it is raw
At the price expectation model.
Preferably, the target histories transaction record further includes server identification;The server coefficient and the service
Device mark is unique corresponding;It is described to use the aspect of model, build-in attribute price expectation model, random attribute price expectation mould
The step of type, generation server coefficient, comprising:
The aspect of model is input to the build-in attribute price expectation model;
Determine that the data that institute's build-in attribute price expectation mould returns are the first forecast price;
The aspect of model is input to the random attribute price expectation model;
Determine that the data that institute's random attribute price expectation mould returns are the second forecast price;
Calculating the sum of first forecast price and second forecast price is initial predicted price;
The ratio for determining the concluded price and the initial predicted price is price ratio;
Multiple initial predicted prices are divided at least one price range;
For same server identification, determine that price in one price section than average value is the server coefficient.
Preferably, described according to the attribute information, the step of generating the aspect of model, comprising:
Determine attribute variable corresponding with the attribute information;
Judge whether the attribute variable meets and preset makes rule;
If it is not, generating the aspect of model using preset conversion regime then for the attribute variable for having continuous feature;
For the attribute variable for having discrete features, the aspect of model is generated using preset coding rule.
Preferably, the step of acquisition multiple target histories transaction records corresponding with virtual objects, comprising:
Obtain multiple historical transaction records to be screened;
Using the concluded price in historical transaction record to be screened, to the multiple historical transaction record to be screened into
Row sequence;
The historical transaction record to be screened for determining preset order is the target histories transaction record.
The embodiment of the invention also discloses a kind of price data prediction meanss of virtual objects in game, comprising:
Attribute information obtains module, for obtaining the attribute information of virtual objects to be predicted;
Feature generation module to be predicted generates category to be predicted for the attribute information using the virtual objects to be predicted
Property feature;
Feature input module to be predicted, for the attributive character to be predicted to be input to preset price expectation model;
Price data receiving module, for receiving the price data of price expectation model generation.
The embodiment of the invention also discloses a kind of price expectation model construction devices, comprising:
Module is obtained, for obtaining multiple target histories transaction records corresponding with virtual objects and initial model;Institute
Stating target histories transaction record includes concluded price and attribute information;
Feature generation module, for generating the aspect of model according to the attribute information;
Training module, for using the concluded price and the aspect of model, the training initial model;
Stopping modular, for calculating multiple loss functions of initial model after training, when introductory die after training
When multiple loss functions of type all minimize, initial model described in deconditioning;
Model generation module, for generating price expectation model using the initial model trained.
The embodiment of the invention also discloses electronic equipment, including processor, memory and it is stored on the memory simultaneously
The computer program that can be run on the processor, the computer program realize institute as above when being executed by the processor
In the game stated the step of the price data prediction technique of virtual objects, and/or price expectation model construction side as described above
The step of method.
The embodiment of the invention also discloses computer readable storage medium, meter is stored on the computer readable storage medium
Calculation machine program, the computer program realize that the price data of virtual objects in game as described above is pre- when being executed by processor
The step of survey method, and/or the step of price expectation model building method as described above.
The embodiment of the present invention includes following advantages: the attribute information according to historical trading virtual objects generates the aspect of model,
The aspect of model is divided according to the build-in attribute of historical trading virtual objects and random attribute, is divided into and corresponds only to inherently
First training characteristics of attribute, and correspond only to second training characteristics an of random attribute.Then according to the first instruction after division
Practice feature, the concluded price of the second training characteristics and historical trading virtual objects, training initial model respectively obtains build-in attribute
Price expectation model and random attribute price expectation model.Aspect of model input is separately input into build-in attribute price expectation mould
Type and random attribute price expectation model are returned according to build-in attribute price expectation model and random attribute price expectation model
Data and concluded price determine one or more server coefficients corresponding with server.Combine build-in attribute price expectation mould
Type, random attribute price expectation model and server coefficient generate price expectation model.It is virtual according to historical trading to realize
The dimensions such as concluded price, build-in attribute, random attribute, the subordinate server of article construct price expectation model, so that price is pre-
Price expectation can be carried out according to the build-in attribute of virtual objects, random attribute, subordinate server by surveying model.
Detailed description of the invention
Fig. 1 is a kind of step flow chart of price expectation model building method embodiment of the invention;
Fig. 2 is the step flow chart of the price data prediction technique embodiment of virtual objects in a kind of game of the invention;
Fig. 3 is a kind of structural block diagram of price expectation model construction Installation practice of the invention;
Fig. 4 is the structural block diagram of the price data prediction meanss embodiment of virtual objects in a kind of game of the invention.
Specific embodiment
In order to make the foregoing objectives, features and advantages of the present invention clearer and more comprehensible, with reference to the accompanying drawing and specific real
Applying mode, the present invention is described in further detail.
Referring to Fig.1, a kind of step flow chart of price expectation model building method embodiment of the invention is shown, specifically
It may include steps of:
Step 101, multiple target histories transaction records corresponding with virtual objects and initial model are obtained;The mesh
Marking historical transaction record includes concluded price and attribute information;
Crawler technology can be used, and from specified address (such as: game virtual article trading platform) obtain multiple void
The transaction record of quasi- article, and transaction record is screened, determine target histories transaction record.Virtual objects are same game
In virtual item.
In one preferred embodiment of the invention, step 101 may include:
Sub-step S11 obtains multiple historical transaction records to be screened;
From specified address, the historical trading note to be screened of multiple virtual objects in certain time section is obtained
Record, historical transaction record to be screened are text information format.
Sub-step S12, using the concluded price in historical transaction record to be screened, to the multiple history to be screened
Transaction record is ranked up;
The information in historical transaction record to be screened is extracted using regular expression, extracts history to be screened
Concluded price and attribute information in transaction record etc..To each concluded price in certain sequence (such as: from big to small suitable
Sequence) it is ranked up.
Sub-step S13 determines that the historical transaction record to be screened of preset order is the target histories transaction record.
According to the sequence of concluded price, historical transaction record to be screened is screened, determine a graded wait sieve
The historical transaction record of choosing is target histories transaction record.Such as: after historical transaction record sequence to be screened, reject sequence
In most preceding 1% and last 1% historical transaction record to be screened, remaining historical transaction record to be screened is gone through for target
History transaction record.
In practical applications, over-sampling processing can be carried out to target histories transaction record using the sequence of concluded price,
It carries out repeating record for the transaction record of preset proportion most preceding and last in sequence.Such as: in target histories transaction record
In, concluded price is replicated in the target histories transaction record of most preceding 1% and last 1%, generates 3 parts of identical target histories
Transaction record.
Step 102, according to the attribute information, the aspect of model is generated;
It can be directed to the continuity of attribute information, using a variety of generating modes, according to the category in target histories transaction record
Property information is converted to the aspect of model.Wherein, the aspect of model can be feature vector.
In one preferred embodiment of the invention, step 102 may include:
Sub-step S21 determines attribute variable corresponding with the attribute information;
Attribute information include a variety of different attributes and, and attribute variable corresponding with attribute.Such as: a virtual object
The attribute information of product includes: 2 strength and 5 attacks, then attribute information includes two attribute: strength and attack, and and strength
Corresponding attribute variable is 2, and attribute variable corresponding with attack is 5.
Sub-step S22 judges whether the attribute variable meets and preset makes rule;
Preset rule of making can be to make virtual objects using specified virtual item (such as: ideal money).
Sub-step S23, if the attribute variable be unsatisfactory for it is preset make rule, for having the attribute of continuous feature
Variable generates the aspect of model using preset conversion regime;For the attribute variable for having discrete features, use is preset
Coding rule generates the aspect of model.
When virtual objects and meet it is preset make rule when, then can be directed to attribute variable mathematical characteristic, using phase
The generating mode answered generates the aspect of model.Specifically, then generating mould using standardization formula when attribute variable is continuous variable
Type feature;When attribute variable is discrete variable, then the aspect of model is generated using one-hot coding.Wherein it is possible to using attribute
The difference of variable current value predetermined minimum corresponding with attribute variable is the first difference, is preset most using attribute variable is corresponding
Second difference of the big difference for being worth predetermined minimum corresponding with attribute variable, standardization formula are the first difference and the second difference
Ratio.
In practical applications, derivative variable can be generated using continuous variable, using derivative variable portray continuous variable it
Between non-linear relation
Step 103, using the concluded price and the aspect of model, the training initial model;
Nuclear model feature can be disliked using concluded price and generate training data, training data is input to initial model to instruct
Practice initial model, until loss function corresponding with initial model reaches most little finger of toe.
In one preferred embodiment of the invention, the attribute information includes build-in attribute and/or random attribute;Step
103 may include:
Sub-step S31, determination only include that the target histories transaction record of the virtual goods of build-in attribute is the first training
Record;
Sub-step S32 determines that the aspect of model corresponding with first training record is the first training characteristics;
Sub-step S31, determination only include that the target histories transaction record of the virtual goods of a random attribute is second
Training record;
Sub-step S33 determines that the aspect of model corresponding with second training record is the second training characteristics;
Sub-step S34, according to first training characteristics and second training characteristics, the training initial module;
Wherein, the random attribute in second training record is opposite with the build-in attribute in first training record
It answers.
Same virtual objects may include one or more build-in attributes and one or more random attributes.
Target histories transaction record is screened, virtual objects are filtered out only and include the virtual goods of build-in attribute
Target histories transaction record is the first training record, and determines that the aspect of model corresponding with the first training record is that the first training is special
It levies, is i.e. only includes build-in attribute in the corresponding attribute information of the first training characteristics;Filter out only includes one in virtual objects
The target histories transaction record of the virtual goods of a random attribute is the second training record, and determination is corresponding with the second training record
The aspect of model be the second training characteristics, i.e. only include a random attribute in the corresponding attribute information of the second training characteristics,
And with only include the corresponding aspect of model of the virtual objects of build-in attribute and with the virtual objects that only include a random attribute
Corresponding aspect of model training initial model.Wherein, build-in attribute is the primary attribute centainly having in virtual objects, random to belong to
Property on primary attribute probability generation attribute
In one preferred embodiment of the invention, the initial model includes: build-in attribute price model;Sub-step
S34 may include:
Sub-step S341 determines that the concluded price in the first training record is build-in attribute price;
Sub-step S342, determine first training characteristics and with the build-in attribute price be the first training data;
Sub-step S343, using first training data, the training build-in attribute price model.
First training data can be input to build-in attribute price model, to train build-in attribute price model, determined
Model coefficient in build-in attribute price model.
Specifically, build-in attribute price model can be linear regression model (LRM):
hθ(x(i))=∑iθi*xi;
Wherein, xiFor first training characteristics, θiFor the first model coefficient of build-in attribute price model.
The above-mentioned linear regression model (LRM) training build-in attribute price model is input to using the first training data, until
Loss function corresponding with build-in attribute price model minimizes.It determines when loss function corresponding with build-in attribute price model
The first model coefficient when minimum is the first model coefficient of target.
In one preferred embodiment of the invention, the initial model further includes random attribute price model;Step S34
It can be with further include:
Sub-step S344 determines the first training characteristics of target corresponding with second training characteristics;
Since random attribute is generated based on build-in attribute, all random attributes have related to one of build-in attribute
Property.In multiple first training characteristics, the first training characteristics corresponding with the second training characteristic are filtered out as the training of target first
Feature, wherein the corresponding build-in attribute of the first training characteristics of target random attribute corresponding to the first training characteristics has related
Property.
Sub-step S345 calculates random attribute price;The random attribute price is the corresponding price of the second training characteristics
Difference between price corresponding with the first training characteristics of target;
Calculate the price difference between the corresponding price of the second training characteristics price corresponding with the first training characteristics of target
For random attribute price.Specifically, can be using the corresponding price of multiple second training characteristics and the training of multiple targets first
The average value of difference between the corresponding price of feature is the random attribute price.
Sub-step S346, determines second training characteristics and the random attribute price is the second training data;
Sub-step S347, using second training data, the training random attribute price model.
It is the second training data by second training characteristics and the random attribute price, and the second data is input to
Random attribute price model determines the model coefficient in random attribute price model to train random attribute price model.
Specifically, random attribute price model can be with are as follows:
Wherein, xiIt is the first training characteristics of target, kiFor corresponding second model coefficient of random attribute price model, ysmiFor
Random attribute price.
Above-mentioned random attribute price model is input to using the second training data, until with random attribute price model pair
The loss function answered minimizes.Determine the second model system when loss function corresponding with random attribute price model minimizes
Number is the second model coefficient of target.
Step 104, the multiple loss functions for calculating initial model after training are more when initial model after training
When a loss function all minimizes, initial model described in deconditioning;
In practical applications, when judging whether the first model coefficient of target meets pre-set business rules, if target first
Model coefficient does not meet business rule, then rejects first model coefficient of target and re -training build-in attribute price model.Sentence
Whether disconnected the second model coefficient of target meets pre-set business rules, if the second model coefficient of target does not meet business rule,
Reject second model coefficient of target and re -training random attribute price model.
Step 105, price expectation model is generated according to the initial model trained.
Determining the second model coefficient of the first model coefficient of target and target, and the first model coefficient of target and target second
When model coefficient meets business rule, valence is generated according to the build-in attribute price model and random attribute price model trained
Lattice prediction model.
In one preferred embodiment of the invention, step 105 may include:
Sub-step S51 determines that the build-in attribute price model trained is build-in attribute price expectation model;
Sub-step S52 determines that the random attribute price model trained is random attribute price expectation model;
Sub-step S53, using the aspect of model, build-in attribute price expectation model, random attribute price expectation model,
Generate server coefficient;
Sub-step S54 combines the build-in attribute price expectation model, random attribute price expectation model and the service
Device coefficient generates the price expectation model.
Price expectation model are as follows:
y′all=(y 'w+y′s)*w′s;
Wherein, y 'wFor build-in attribute price expectation model, y 'sRandom attribute price expectation model, w 'sFor server system
Number.
Different virtual objects can correspond to different server, the same alike result information from different server it is virtual
The concluded price of article would also vary from, by introducing server coefficient, to increase server in price expectation model
The dimension of factor, so that the data of price expectation model output have more referential.
In one preferred embodiment of the invention, the target histories transaction record further includes server identification;It is described
Server coefficient and the server identification are uniquely corresponding;Sub-step S53 may include:
The aspect of model is input to the build-in attribute price expectation model by sub-step S531;
Sub-step S532 determines that the data that institute's build-in attribute price expectation mould returns are the first forecast price;
The aspect of model that step 102 can be generated is input to build-in attribute price expectation model, and build-in attribute price is pre-
Model is surveyed for generating and matched first forecast price of the aspect of model.
The aspect of model is input to the random attribute price expectation model by sub-step S533;
Sub-step S534 determines that the data that institute's random attribute price expectation mould returns are the second forecast price;
The aspect of model that step 102 can be generated is input to random attribute price expectation model, and random attribute price is pre-
Model is surveyed for generating and matched second forecast price of the aspect of model.
Sub-step S535, calculating the sum of first forecast price and second forecast price is initial predicted price;
Sub-step S536 determines that the ratio of the concluded price and the initial predicted price is price ratio;
Concluded price is the knock-down price of virtual objects corresponding with the aspect of model in sub-step S531 and sub-step S533
Lattice.
Multiple initial predicted prices are divided at least one price range by sub-step S537;
According to the different aspect of model, sub-step S531-S535, available multiple initial predicted prices are repeated.It can be with
Obtained multiple initial predicted prices are divided at least one price range.
In a kind of example, obtained multiple initial predicted prices are divided into 5 price ranges.
Sub-step S538 determines that the price in one price section than average value is described for same server identification
Server coefficient.
Difference can be found out respectively to identify relative to different server, and relative to the server system of different price ranges
Number.It is directed to identical server identification, each price range both corresponds to a server coefficient.
In practical applications, new transaction record can be acquired, and remember according to new transaction after at regular intervals
The accuracy of record detection price expectation model is remembered when the exactness of price expectation model is less than preset threshold using new transaction
Record updates price expectation model.
In embodiments of the present invention, the aspect of model is generated according to the attribute information of historical trading virtual objects, according to history
The build-in attribute and random attribute of trade virtual article divide the aspect of model, are divided into and correspond only to the of build-in attribute
One training characteristics, and correspond only to second training characteristics an of random attribute.Then according to the first training characteristics, after division
The concluded price of two training characteristics and historical trading virtual objects, training initial model, respectively obtains build-in attribute price expectation
Model and random attribute price expectation model.Aspect of model input is separately input into build-in attribute price expectation model and random
Attributes price prediction model, the data returned according to build-in attribute price expectation model and random attribute price expectation model and at
Price is handed over, determines one or more server coefficients corresponding with server.Combine build-in attribute price expectation model, random category
Sexual valence lattice prediction model and server coefficient generate price expectation model.To realize according to historical trading virtual objects at
It hands over price, build-in attribute, random attribute, the dimensions such as subordinate server to construct price expectation model, enables price expectation model
It is enough that price expectation is carried out according to the build-in attribute of virtual objects, random attribute, subordinate server.
Referring to Fig. 2, a kind of step flow chart of price data embodiment of the method for the invention is shown, can specifically include
Following steps:
Step 201, the attribute information of virtual objects to be predicted is obtained;
The attribute information of virtual objects to be predicted may include build-in attribute and/or random attribute, and attribute information can be
Text information.
Step 202, using the attribute information of the virtual objects to be predicted, attributive character to be predicted is generated;
Attribute to be predicted is converted into using the attribute information of predetermined manner (such as: regular expression) virtual objects to be predicted
Feature.
Step 203, the attributive character to be predicted is input to preset price expectation model;
Attributive character to be predicted is input to the price expectation model trained.
Step 204, the price data that price expectation model generates is received.
The price data that price expectation model is returned is the forecast price corresponding to virtual objects to be predicted.
In one preferred embodiment of the invention, it is only necessary to which the attribute information of virtual objects to be predicted is converted into pre-
Attributive character is surveyed, and attributive character to be predicted is input to and has trained price expectation model, can be obtained corresponding to void to be predicted
The forecast price of quasi- article avoids artificial prediction virtual object to realize the price expectation for carrying out virtual objects using model
The drawbacks such as the reliability of product price is lower, prediction steps are complicated.
In one preferred embodiment of the invention, the attributive character to be predicted is corresponding with destination server mark;Institute
Stating price expectation model includes: build-in attribute price expectation model, random attribute price expectation model and server coefficient;It is described
Price data generates by the following method: being generated using the attributive character to be predicted and the build-in attribute price expectation model
Build-in attribute forecast price;Random attribute is generated using the attributive character to be predicted and the random attribute price expectation model
Forecast price;It is true using destination server mark, the build-in attribute forecast price and the random attribute forecast price
Set the goal server coefficient;Calculating the sum of the build-in attribute forecast price and the random attribute forecast price is fundamentals of forecasting
Price;The product for calculating the basic forecast price and the destination server coefficient is the price data.
In one preferred embodiment of the invention, the price expectation model generates by the following method:
Obtain initial model, and target histories transaction record corresponding with multiple virtual objects;The target histories are handed over
Easily record includes concluded price and attribute information;
According to the attribute information of target histories transaction record, the aspect of model is generated;
Using the concluded price and the aspect of model, the training initial model;
The multiple loss functions for calculating initial model after training, when multiple loss letters of initial model after training
When number all minimizes, initial model described in deconditioning;
Price expectation model is generated using the initial model trained.
In one preferred embodiment of the invention, described to use the concluded price and the aspect of model, training institute
The step of stating initial model, comprising:
Determination only includes that the target histories transaction record of the virtual goods of build-in attribute is the first training record;
Determine that the aspect of model corresponding with first training record is the first training characteristics;
Determination only includes that the target histories transaction record of the virtual goods of a random attribute is the second training record;
Determine that the aspect of model corresponding with second training record is the second training characteristics;
According to first training characteristics and second training characteristics, the training initial module;
Wherein, the random attribute in second training record is opposite with the build-in attribute in first training record
It answers.
In one preferred embodiment of the invention, the initial model includes: build-in attribute price model;The basis
First training characteristics and second training characteristics, the training initial module;The step of, comprising:
Determine that the concluded price in the first training record is build-in attribute price;
Determine first training characteristics and with the build-in attribute price be the first training data;
Using first training data, the training build-in attribute price model.
In one preferred embodiment of the invention, the initial model further includes random attribute price model;Described
According to first training characteristics and second training characteristics, the step of the training initial module, comprising:
Determine the first training characteristics of target corresponding with second training characteristics;
Calculate random attribute price;The random attribute price is that the corresponding price of the second training characteristics and target first are instructed
Practice the difference between the corresponding price of feature;
It determines second training characteristics and the random attribute price is the second training data;
Using second training data, the training random attribute price model.
In one preferred embodiment of the invention, described that price expectation model is generated using the initial model trained
Step, comprising:
Determine that the build-in attribute price model trained is build-in attribute price expectation model;
Determine that the random attribute price model trained is random attribute price expectation model;
Using the aspect of model, build-in attribute price expectation model, random attribute price expectation model, server is generated
Coefficient;
The build-in attribute price expectation model, random attribute price expectation model and the server coefficient are combined, it is raw
At the price expectation model.
In one preferred embodiment of the invention, the target histories transaction record further includes server identification;It is described
Server coefficient and the server identification are uniquely corresponding;It is described using the aspect of model, build-in attribute price expectation model,
The step of random attribute price expectation model, generation server coefficient, comprising:
The aspect of model is input to the build-in attribute price expectation mould;
Determine that the data that institute's build-in attribute price expectation mould returns are the first forecast price;
The aspect of model is input to the random attribute price expectation mould;
Determine that the data that institute's random attribute price expectation mould returns are the second forecast price;
Calculating the sum of first forecast price and second forecast price is initial predicted price;
The ratio for determining the concluded price and the initial predicted price is price ratio;
Multiple initial predicted prices are divided at least one price range;
For same server identification, determine that price in one price section than average value is the server coefficient.
In one preferred embodiment of the invention, the attribute information according to target histories transaction record, generates mould
The step of type feature, comprising:
Determine attribute variable corresponding with the attribute information of target histories transaction record;
Judge whether the attribute variable meets and preset makes rule;
If the attribute variable be unsatisfactory for it is preset make rule, for having the attribute variable of continuous feature, use
Preset conversion regime generates the aspect of model;For the attribute variable for having discrete features, using preset coding rule
Generate the aspect of model.
In one preferred embodiment of the invention, described to obtain multiple target histories transaction notes corresponding with virtual objects
The step of record, comprising:
Obtain multiple historical transaction records to be screened;
Using the concluded price in historical transaction record to be screened, to the multiple historical transaction record to be screened into
Row sequence;
The historical transaction record to be screened for determining preset order is the target histories transaction record.
It should be noted that for simple description, therefore, it is stated as a series of action groups for embodiment of the method
It closes, but those skilled in the art should understand that, embodiment of that present invention are not limited by the describe sequence of actions, because according to
According to the embodiment of the present invention, some steps may be performed in other sequences or simultaneously.Secondly, those skilled in the art also should
Know, the embodiments described in the specification are all preferred embodiments, and the related movement not necessarily present invention is implemented
Necessary to example.
Referring to Fig. 3, a kind of structural block diagram of price expectation model construction Installation practice of the invention is shown, specifically may be used
To include following module:
Module 301 is obtained, for obtaining multiple target histories transaction records corresponding with virtual objects and introductory die
Type;The target histories transaction record includes concluded price and attribute information;
Feature generation module 302, for generating the aspect of model according to the attribute information;
Training module 303, for using the concluded price and the aspect of model, the training initial model;
Stopping modular 304, it is initial when after training for calculating multiple loss functions of initial model after training
When multiple loss functions of model all minimize, initial model described in deconditioning;
Model generation module 305, for generating price expectation model using the initial model trained.
In one preferred embodiment of the invention, the attribute information includes build-in attribute and/or random attribute;Training
Module 303 includes:
First training record submodule only includes the target histories transaction note of the virtual goods of build-in attribute for determination
Record is the first training record;
First training characteristics submodule, for determining that corresponding with first training record aspect of model is
First training characteristics;
Second training record submodule is handed over for the target histories that determination only includes the virtual goods of a random attribute
Easily it is recorded as the second training record;
Second training characteristics submodule, for determining that the aspect of model corresponding with second training record is the second training
Feature;
Training submodule, for according to first training characteristics and second training characteristics, the training introductory die
Block;
Wherein, the random attribute in second training record is opposite with the build-in attribute in first training record
It answers.
In one preferred embodiment of the invention, the initial model includes: build-in attribute price model;The training
Submodule includes:
Intrinsic price unit, for determining that the concluded price in the first training record is build-in attribute price;
First training data unit is first for determining first training characteristics and with the build-in attribute price
Training data;
First training unit, for using first training data, the training build-in attribute price model.
In one preferred embodiment of the invention, the initial model further includes random attribute price model;The instruction
Practicing submodule includes:
Target the first training characteristics unit, for determining that the training of target first corresponding with second training characteristics is special
Sign;
Random price unit, for calculating random attribute price;The random attribute price is corresponding for the second training characteristics
Price price corresponding with the first training characteristics of target between difference;
Second training data unit, for determining second training characteristics and the random attribute price for the second training
Data;
Second training unit, for using second training data, the training random attribute price model.
In one preferred embodiment of the invention, the model generation module 305 includes:
First prediction submodule, for determining that the build-in attribute price model trained is build-in attribute price expectation mould
Type;
Second prediction submodule, for determining that the random attribute price model trained is random attribute price expectation mould
Type;
Service factor generates submodule, for using the aspect of model, build-in attribute price expectation model, random attribute
Price expectation model generates server coefficient;
Submodule is combined, for combining the build-in attribute price expectation model, random attribute price expectation model and institute
Server coefficient is stated, the price expectation model is generated.
In one preferred embodiment of the invention, the target histories transaction record further includes server identification;It is described
Server coefficient and the server identification are uniquely corresponding;The service factor generates submodule
First input unit, for the aspect of model to be input to the build-in attribute price expectation model;
First return unit, for determining that the data that institute's build-in attribute price expectation mould returns are the first prediction valence
Lattice;
Second input unit, for the aspect of model to be input to the random attribute price expectation model;
Second return unit, for determining that the data that institute's random attribute price expectation mould returns are the second prediction valence
Lattice;
Initial predicted calculation of price unit is for calculating the sum of first forecast price and second forecast price
Initial predicted price;
Price is than unit, for determining that the ratio of the concluded price and the initial predicted price is price ratio;
Division unit, for multiple initial predicted prices to be divided at least one price range;
Coefficient generation unit determines that the price in one price section compares average value for being directed to same server identification
For the server coefficient.
In one preferred embodiment of the invention, feature generation module 302 includes:
Variable determines submodule, for determining attribute variable corresponding with the attribute information;
Variable judging submodule preset makes rule for judging whether the attribute variable meets;
Variable transform subblock, if for the attribute variable be unsatisfactory for it is preset make rule, it is continuous for having
The attribute variable of feature generates the aspect of model using preset conversion regime;For the attribute variable for having discrete features,
The aspect of model is generated using preset coding rule.
In one preferred embodiment of the invention, obtaining module 301 includes:
Transaction record acquisition submodule, for obtaining multiple historical transaction records to be screened;
Transaction record sorting sub-module, for using the concluded price in historical transaction record to be screened, to described more
A historical transaction record to be screened is ranked up;
Transaction record screens submodule, for determining that the historical transaction record to be screened of preset order is that the target is gone through
History transaction record.
Referring to Fig. 4, the knot of the price data prediction meanss embodiment of virtual objects in a kind of game of the invention is shown
Structure block diagram, can specifically include following module:
Attribute information obtains module 401, for obtaining the attribute information of virtual objects to be predicted;
Feature generation module 402 to be predicted generates to be predicted for the attribute information using the virtual objects to be predicted
Attributive character;
Feature input module 403 to be predicted, for the attributive character to be predicted to be input to preset price expectation mould
Type;
Price data receiving module 404, for receiving the price data of price expectation model generation.
In one preferred embodiment of the invention, the attributive character to be predicted is corresponding with destination server mark;Institute
Stating price expectation model includes: build-in attribute price expectation model, random attribute price expectation model and server coefficient;It is described
Price data is generated by following module:
First prediction module, for being generated using the attributive character to be predicted and the build-in attribute price expectation model
Build-in attribute forecast price;
Second prediction module, for being generated using the attributive character to be predicted and the random attribute price expectation model
Random attribute forecast price;
Coefficient determination module, for using the destination server mark, the build-in attribute forecast price and it is described with
Machine attribute forecast price determines destination server coefficient;
Underlying price prediction module, for calculate the build-in attribute forecast price and the random attribute forecast price it
With for fundamentals of forecasting price;
Price data computing module, for calculating the basic forecast price and the product of the destination server coefficient is
The price data.
In one preferred embodiment of the invention, the price expectation model is generated by following module:
Module is obtained, for obtaining multiple target histories transaction records corresponding with virtual objects and initial model;Institute
Stating target histories transaction record includes concluded price and attribute information;
Feature generation module generates the aspect of model for the attribute information according to target histories transaction record;
Training module, for using the concluded price and the aspect of model, the training initial model;
Stopping modular, for calculating multiple loss functions of initial model after training, when introductory die after training
When multiple loss functions of type all minimize, initial model described in deconditioning;
Model generation module, for generating price expectation model using the initial model trained.
In one preferred embodiment of the invention, the attribute information includes build-in attribute and/or random attribute;Training
Module includes:
First training record submodule only includes the target histories transaction note of the virtual goods of build-in attribute for determination
Record is the first training record;
First training characteristics submodule, for determining that corresponding with first training record aspect of model is
First training characteristics;
Second training record submodule is handed over for the target histories that determination only includes the virtual goods of a random attribute
Easily it is recorded as the second training record;
Second training characteristics submodule, for determining that the aspect of model corresponding with second training record is the second training
Feature;
Training submodule, for according to first training characteristics and second training characteristics, the training introductory die
Block;
Wherein, the random attribute in second training record is opposite with the build-in attribute in first training record
It answers.
In one preferred embodiment of the invention, the initial model includes: build-in attribute price model;The training
Submodule includes:
Intrinsic price unit, for determining that the concluded price in the first training record is build-in attribute price;
First training data unit is first for determining first training characteristics and with the build-in attribute price
Training data;
First training unit, for using first training data, the training build-in attribute price model.
In one preferred embodiment of the invention, the initial model further includes random attribute price model;The instruction
Practicing submodule includes:
Target the first training characteristics unit, for determining that the training of target first corresponding with second training characteristics is special
Sign;
Random price unit, for calculating random attribute price;The random attribute price is corresponding for the second training characteristics
Price price corresponding with the first training characteristics of target between difference;
Second training data unit, for determining second training characteristics and the random attribute price for the second training
Data;
Second training unit, for using second training data, the training random attribute price model.
In one preferred embodiment of the invention, the model generation module includes:
First prediction submodule, for determining that the build-in attribute price model trained is build-in attribute price expectation mould
Type;
Second prediction submodule, for determining that the random attribute price model trained is random attribute price expectation mould
Type;
Service factor generates submodule, for using the aspect of model, build-in attribute price expectation model, random attribute
Price expectation model generates server coefficient;
Submodule is combined, for combining the build-in attribute price expectation model, random attribute price expectation model and institute
Server coefficient is stated, the price expectation model is generated.
In one preferred embodiment of the invention, the target histories transaction record further includes server identification;It is described
Server coefficient and the server identification are uniquely corresponding;The service factor generates submodule
First input unit, for the aspect of model to be input to the build-in attribute price expectation model;
First return unit, for determining that the data that institute's build-in attribute price expectation mould returns are the first prediction valence
Lattice;
Second input unit, for the aspect of model to be input to the random attribute price expectation model;
Second return unit, for determining that the data that institute's random attribute price expectation mould returns are the second prediction valence
Lattice;
Initial predicted calculation of price unit is for calculating the sum of first forecast price and second forecast price
Initial predicted price;
Price is than unit, for determining that the ratio of the concluded price and the initial predicted price is price ratio;
Division unit, for multiple initial predicted prices to be divided at least one price range;
Coefficient generation unit determines that the price in one price section compares average value for being directed to same server identification
For the server coefficient.
In one preferred embodiment of the invention, feature generation module includes:
Variable determines submodule, for determining attribute variable corresponding with the attribute information of target histories transaction record;
Variable judging submodule preset makes rule for judging whether the attribute variable meets;
Variable transform subblock, if for the attribute variable be unsatisfactory for it is preset make rule, it is continuous for having
The attribute variable of feature generates the aspect of model using preset conversion regime;For the attribute variable for having discrete features,
The aspect of model is generated using preset coding rule.
In one preferred embodiment of the invention, obtaining module includes:
Transaction record acquisition submodule, for obtaining multiple historical transaction records to be screened;
Transaction record sorting sub-module, for using the concluded price in historical transaction record to be screened, to described more
A historical transaction record to be screened is ranked up;
Transaction record screens submodule, for determining that the historical transaction record to be screened of preset order is that the target is gone through
History transaction record.
For device embodiment, since it is basically similar to the method embodiment, related so being described relatively simple
Place illustrates referring to the part of embodiment of the method.
The embodiment of the invention also discloses electronic equipment, including processor, memory and it is stored on the memory simultaneously
The computer program that can be run on the processor, the computer program realize institute as above when being executed by the processor
The price data of virtual objects generates the step of prediction technique, and/or price expectation model structure as described above in the game stated
The step of construction method.
The embodiment of the invention also discloses computer readable storage medium, meter is stored on the computer readable storage medium
Calculation machine program, the computer program realize that the price data of virtual objects in game as described above is raw when being executed by processor
The step of at prediction technique, and/or the step of price expectation model building method as described above.
All the embodiments in this specification are described in a progressive manner, the highlights of each of the examples are with
The difference of other embodiments, the same or similar parts between the embodiments can be referred to each other.
It should be understood by those skilled in the art that, the embodiment of the embodiment of the present invention can provide as method, apparatus or calculate
Machine program product.Therefore, the embodiment of the present invention can be used complete hardware embodiment, complete software embodiment or combine software and
The form of the embodiment of hardware aspect.Moreover, the embodiment of the present invention can be used one or more wherein include computer can
With in the computer-usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) of program code
The form of the computer program product of implementation.
The embodiment of the present invention be referring to according to the method for the embodiment of the present invention, terminal device (system) and computer program
The flowchart and/or the block diagram of product describes.It should be understood that flowchart and/or the block diagram can be realized by computer program instructions
In each flow and/or block and flowchart and/or the block diagram in process and/or box combination.It can provide these
Computer program instructions are set to general purpose computer, special purpose computer, Embedded Processor or other programmable data processing terminals
Standby processor is to generate a machine, so that being held by the processor of computer or other programmable data processing terminal devices
Capable instruction generates for realizing in one or more flows of the flowchart and/or one or more blocks of the block diagram
The device of specified function.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing terminal devices
In computer-readable memory operate in a specific manner, so that instruction stored in the computer readable memory generates packet
The manufacture of command device is included, which realizes in one side of one or more flows of the flowchart and/or block diagram
The function of being specified in frame or multiple boxes.
These computer program instructions can also be loaded into computer or other programmable data processing terminal devices, so that
Series of operation steps are executed on computer or other programmable terminal equipments to generate computer implemented processing, thus
The instruction executed on computer or other programmable terminal equipments is provided for realizing in one or more flows of the flowchart
And/or in one or more blocks of the block diagram specify function the step of.
Although the preferred embodiment of the embodiment of the present invention has been described, once a person skilled in the art knows bases
This creative concept, then additional changes and modifications can be made to these embodiments.So the following claims are intended to be interpreted as
Including preferred embodiment and fall into all change and modification of range of embodiment of the invention.
Finally, it is to be noted that, herein, relational terms such as first and second and the like be used merely to by
One entity or operation are distinguished with another entity or operation, without necessarily requiring or implying these entities or operation
Between there are any actual relationship or orders.Moreover, the terms "include", "comprise" or its any other variant meaning
Covering non-exclusive inclusion, so that process, method, article or terminal device including a series of elements not only wrap
Those elements are included, but also including other elements that are not explicitly listed, or further includes for this process, method, article
Or the element that terminal device is intrinsic.In the absence of more restrictions, being wanted by what sentence "including a ..." limited
Element, it is not excluded that there is also other identical elements in process, method, article or the terminal device for including the element.
Above to the price data prediction technique of virtual objects, a kind of price expectation in a kind of game provided by the present invention
The price data prediction meanss of virtual objects, a kind of price expectation model construction device, electricity in model building method, a kind of game
Sub- equipment and storage medium, are described in detail, and specific case used herein is to the principle of the present invention and embodiment
It is expounded, the above description of the embodiment is only used to help understand the method for the present invention and its core ideas;Meanwhile for
Those of ordinary skill in the art have change according to the thought of the present invention in specific embodiments and applications
Place, in conclusion the contents of this specification are not to be construed as limiting the invention.
Claims (15)
1. the price data prediction technique of virtual objects in a kind of game characterized by comprising
Obtain the attribute information of virtual objects to be predicted;
Using the attribute information of the virtual objects to be predicted, attributive character to be predicted is generated;
The attributive character to be predicted is input to preset price expectation model;
Receive the price data that price expectation model generates.
2. according to method described in claim 1, which is characterized in that the attributive character to be predicted is corresponding with destination server mark
Know;The price expectation model includes: build-in attribute price expectation model, random attribute price expectation model and server system
Number;The price data generates by the following method:
Build-in attribute forecast price is generated using the attributive character to be predicted and the build-in attribute price expectation model;
Random attribute forecast price is generated using the attributive character to be predicted and the random attribute price expectation model;
Target is determined using destination server mark, the build-in attribute forecast price and the random attribute forecast price
Server coefficient;
Calculating the sum of the build-in attribute forecast price and the random attribute forecast price is fundamentals of forecasting price;
The product for calculating the basic forecast price and the destination server coefficient is the price data.
3. according to the method described in claim 2, it is characterized in that, the price expectation model generates by the following method:
Obtain multiple target histories transaction records corresponding with virtual objects and initial model;The target histories transaction note
Record includes concluded price and attribute information;
According to the attribute information of target histories transaction record, the aspect of model is generated;
Using the concluded price and the aspect of model, the training initial model;
The multiple loss functions for calculating initial model after training, when initial model after training multiple loss functions all
When minimum, initial model described in deconditioning;
Price expectation model is generated according to the initial model trained.
4. a kind of price expectation model building method characterized by comprising
Obtain multiple target histories transaction records corresponding with virtual objects and initial model;The target histories transaction note
Record includes concluded price and attribute information;
According to the attribute information, the aspect of model is generated;
Using the concluded price and the aspect of model, the training initial model;
The multiple loss functions for calculating initial model after training, when initial model after training multiple loss functions all
When minimum, initial model described in deconditioning;
Price expectation model is generated according to the initial model trained.
5. according to the method described in claim 4, it is characterized in that, the attribute information includes build-in attribute and/or random category
Property;The step of use concluded price and aspect of model, the training initial model, comprising:
Determination only includes that the target histories transaction record of the virtual goods of build-in attribute is the first training record;
Determine that the aspect of model corresponding with first training record is the first training characteristics;
Determination only includes that the target histories transaction record of the virtual goods of a random attribute is the second training record;
Determine that the aspect of model corresponding with second training record is the second training characteristics;
According to first training characteristics and second training characteristics, the training initial module;
Wherein, the random attribute in second training record is corresponding with the build-in attribute in first training record.
6. according to the method described in claim 5, it is characterized in that, the initial model includes: build-in attribute price model;Institute
The step of stating according to first training characteristics and second training characteristics, training the initial module, comprising:
Determine that the concluded price in the first training record is build-in attribute price;
Determine first training characteristics and with the build-in attribute price be the first training data;
Using first training data, the training build-in attribute price model.
7. according to the method described in claim 6, it is characterized in that, the initial model further includes random attribute price model;
It is described according to first training characteristics and second training characteristics, the step of the training initial module, comprising:
Determine the first training characteristics of target corresponding with second training characteristics;
Calculate random attribute price;The random attribute price is that the corresponding price of the second training characteristics and the training of target first are special
Levy the difference between corresponding price;
It determines second training characteristics and the random attribute price is the second training data;
Using second training data, the training random attribute price model.
8. the method according to the description of claim 7 is characterized in that described generate price expectation using the initial model trained
The step of model, comprising:
Determine that the build-in attribute price model trained is build-in attribute price expectation model;
Determine that the random attribute price model trained is random attribute price expectation model;
Using the aspect of model, build-in attribute price expectation model, random attribute price expectation model, server system is generated
Number;
The build-in attribute price expectation model, random attribute price expectation model and the server coefficient are combined, institute is generated
State price expectation model.
9. according to the method described in claim 8, it is characterized in that, the target histories transaction record further includes server mark
Know;The server coefficient and the server identification are uniquely corresponding;It is described pre- using the aspect of model, build-in attribute price
The step of surveying model, random attribute price expectation model, generating server coefficient, comprising:
The aspect of model is input to the build-in attribute price expectation model;
Determine that the data that institute's build-in attribute price expectation mould returns are the first forecast price;
The aspect of model is input to the random attribute price expectation model;
Determine that the data that institute's random attribute price expectation mould returns are the second forecast price;
Calculating the sum of first forecast price and second forecast price is initial predicted price;
The ratio for determining the concluded price and the initial predicted price is price ratio;
Multiple initial predicted prices are divided at least one price range;
For same server identification, determine that price in one price section than average value is the server coefficient.
10. according to the method described in claim 4, generating the aspect of model it is characterized in that, described according to the attribute information
Step, comprising:
Determine attribute variable corresponding with the attribute information;
Judge whether the attribute variable meets and preset makes rule;
If it is not, generating the aspect of model using preset conversion regime then for the attribute variable for having continuous feature;For
Have the attribute variable of discrete features, the aspect of model is generated using preset coding rule.
11. according to the method described in claim 4, it is characterized in that, the multiple targets corresponding with virtual objects of acquisition are gone through
The step of history transaction record, comprising:
Obtain multiple historical transaction records to be screened;
Using the concluded price in historical transaction record to be screened, the multiple historical transaction record to be screened is arranged
Sequence;
The historical transaction record to be screened for determining preset order is the target histories transaction record.
12. the price data prediction meanss of virtual objects in a kind of game characterized by comprising
Attribute information obtains module, for obtaining the attribute information of virtual objects to be predicted;
It is special to generate attribute to be predicted for the attribute information using the virtual objects to be predicted for feature generation module to be predicted
Sign;
Feature input module to be predicted, for the attributive character to be predicted to be input to preset price expectation model;
Price data receiving module, for receiving the price data of price expectation model generation.
13. a kind of price expectation model construction device characterized by comprising
Module is obtained, for obtaining multiple target histories transaction records corresponding with virtual objects and initial model;The mesh
Marking historical transaction record includes concluded price and attribute information;
Feature generation module, for generating the aspect of model according to the attribute information;
Training module, for using the concluded price and the aspect of model, the training initial model;
Stopping modular, for calculating multiple loss functions of initial model after training, when initial model after training
When multiple loss functions all minimize, initial model described in deconditioning;
Model generation module, for generating price expectation model using the initial model trained.
14. electronic equipment, which is characterized in that including processor, memory and be stored on the memory and can be described
The computer program run on processor is realized when the computer program is executed by the processor as in claims 1 to 3
In described in any item game the step of the price data prediction technique of virtual objects, and/or such as any one of claim 4 to 11
The step of described price expectation model building method.
15. computer readable storage medium, which is characterized in that computer program is stored on the computer readable storage medium,
The computer program realizes virtual objects in game as claimed any one in claims 1 to 3 when being executed by processor
The step of price data prediction technique, and/or such as the described in any item price expectation model building methods of claim 4 to 11
Step.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910002634.3A CN109670876A (en) | 2019-01-02 | 2019-01-02 | The price data prediction technique and device of virtual objects in a kind of game |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910002634.3A CN109670876A (en) | 2019-01-02 | 2019-01-02 | The price data prediction technique and device of virtual objects in a kind of game |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109670876A true CN109670876A (en) | 2019-04-23 |
Family
ID=66146774
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910002634.3A Pending CN109670876A (en) | 2019-01-02 | 2019-01-02 | The price data prediction technique and device of virtual objects in a kind of game |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109670876A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111737317A (en) * | 2020-06-23 | 2020-10-02 | 广联达科技股份有限公司 | Measuring and calculating method and device |
CN113344628A (en) * | 2021-06-04 | 2021-09-03 | 网易(杭州)网络有限公司 | Information processing method and device, computer equipment and storage medium |
-
2019
- 2019-01-02 CN CN201910002634.3A patent/CN109670876A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111737317A (en) * | 2020-06-23 | 2020-10-02 | 广联达科技股份有限公司 | Measuring and calculating method and device |
CN113344628A (en) * | 2021-06-04 | 2021-09-03 | 网易(杭州)网络有限公司 | Information processing method and device, computer equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US20070293959A1 (en) | Apparatus, method and computer product for predicting a price of an object | |
Singh et al. | Sales forecast for amazon sales with time series modeling | |
CN110400010A (en) | Prediction technique, device, electronic equipment and computer readable storage medium | |
CN109670876A (en) | The price data prediction technique and device of virtual objects in a kind of game | |
Pang et al. | Forecasting container throughput using aggregate or terminal-specific data? The case of Tanjung Priok Port, Indonesia | |
CN108830417B (en) | ARMA (autoregressive moving average) and regression analysis based life energy consumption prediction method and system | |
CN107132762A (en) | Online Static security assessment method based on forecast failure collection automatic screening | |
CN110008397A (en) | A kind of recommended models training method and device | |
Panda et al. | Long term electrical load forecasting: an empirical study across techniques and domains | |
CN112036737B (en) | Method and device for calculating regional electric quantity deviation | |
Navarro et al. | Optimal short-term forecasting using ga-based holt-winters method | |
CN113361810B (en) | Passenger flow volume prediction method, device, equipment and storage medium | |
CN107945079A (en) | A kind of poverty alleviation object selection method and device | |
Hendrawan et al. | Testing methods on system dynamics: a model of reliability, average reliability, and demand of service | |
CN116957640A (en) | Product sales prediction method and system | |
CN111401787A (en) | Site selection method, device and system | |
Khusuwan et al. | EBITDA time series forecasting case study: Provincial Waterworks Authority | |
Voumik et al. | Forecasting GDP growth rates of Bangladesh: an empirical study | |
Hurley et al. | New policies for stochastic inventory control models–theoretical and computational results | |
Dong et al. | Adaptive neural network short term load forecasting with wavelet decompositions | |
CN117216081B (en) | Automatic question bank updating method and device, electronic equipment and storage medium | |
Banditvilai et al. | Forecasting Models for Thailand’s Electrical Appliances Export Values | |
CN110490368A (en) | A kind of combination forecasting method being fitted based on least square and combining weights | |
Kesumawati et al. | Forecasting Bank Indonesia Rate based on Fuzzy Time Series with Higher Forecast Accuracy Rate | |
Awajan et al. | Stock market forecasting using empirical mode decomposition with holt-winter |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190423 |