CN107807908A - Variety result of the match Forecasting Methodology, device and storage medium - Google Patents
Variety result of the match Forecasting Methodology, device and storage medium Download PDFInfo
- Publication number
- CN107807908A CN107807908A CN201710876238.4A CN201710876238A CN107807908A CN 107807908 A CN107807908 A CN 107807908A CN 201710876238 A CN201710876238 A CN 201710876238A CN 107807908 A CN107807908 A CN 107807908A
- Authority
- CN
- China
- Prior art keywords
- match
- contestant
- variety
- result
- feature
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Operations Research (AREA)
- Economics (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Physics (AREA)
- Game Theory and Decision Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Algebra (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Development Economics (AREA)
- Evolutionary Biology (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of variety result of the match Forecasting Methodology, device and storage medium, this method to include:From preset data source, the history cycle data of a variety match are extracted, and the feature of each contestant is extracted from the history cycle data;Pairing conversion is carried out to the feature of each contestant in the history cycle, obtains the result label of the feature pairing conversion generation of multiple pairing contestants;Using the result label training Logic Regression Models of generation, the anticipation function of model coefficient and model is determined;The feature for each contestant of current period and current period that the variety is competed substitutes into predetermined anticipation function respectively, obtains the functional value of each contestant, and carry out ranking to the result of the match of each contestant according to the size order of functional value.By the calculating of competed to variety cycle and each contestant's feature, variety result of the match is predicted, improves the accuracy of prediction result.
Description
Technical field
The present invention relates to computer data excavation applications, more particularly to a kind of variety result of the match Forecasting Methodology, device and
Computer-readable recording medium.
Background technology
With social network sites application more and more common in life, many variety class matches, such as:I be singer or in
The programs such as the good sound of state, can all be promoted to obtain the concern of more spectators in social platform.This is attracting spectators' powder
While silk concern, also there are many spectators' beans vermicelli to deliver oneself support and view to entrant in social platform.
At present, on the problem of predicting result of the match, some mechanisms are using spectators' bean vermicelli in social platform to each competition
The supporting rate of person predicts result of the match, however, these methods are all didactic modes, not from historical data learning and
Come, accuracy is relatively low, it is impossible to accurately predicts result of the match.
The content of the invention
The present invention provides a kind of variety result of the match Forecasting Methodology, device and computer-readable recording medium, can basis
The result that the historical data of variety match is competed to current variety carries out Accurate Prediction.
To achieve the above object, the present invention provides a kind of variety result of the match Forecasting Methodology, and this method includes:
Data pick-up step:From preset data source, the history cycle data of variety match are extracted, and from the history
The feature of each contestant is extracted in cycle data;
Switch process:Pairing conversion is carried out to the feature of each contestant in the history cycle, obtains multiple match somebody with somebody
The result label of feature pairing conversion generation to contestant;
Model training step:Using the result label training Logic Regression Models of generation, model coefficient and model are determined
Anticipation function;
First prediction steps:The feature for each contestant of current period and current period that the variety is competed substitutes into respectively
In the anticipation function, the functional value of each contestant is obtained, and according to the size order of functional value to each contestant's
Result of the match carries out ranking.
Preferably, the expression formula of the anticipation function is:
S=f (x)=WTX
Wherein T represents the cycle of variety match, WTThe model coefficient of the anticipation function is represented, X represents the spy of contestant
Sign, S represent the functional value of the contestant.
Preferably, the model training step includes:
Data collection steps:Obtain the history cycle of variety match and the feature pairing transformation result label of contestant;
Detecting step:The feature pairing transformation result label of contestant is substituted into the type model one by one and trained, profit
The quality of the forecast model is detected with loss function;
Judgment step:The value of loss function is counted using empirical risk function, judges that empirical risk function value is
It is no to be less than the first preset value;
If empirical risk function value is more than or equal to the first preset value, then it represents that the Logic Regression Models are not optimal moulds
Type, model coefficient is adjusted, and return to training step until empirical risk function value is less than the first preset value;
If empirical risk function value is less than the first preset value, then it represents that the Logic Regression Models are optimal models.
Preferably, this method also includes:
Second prediction steps:By the current period of variety match and feature generation respectively of predetermined two contestants
Enter the anticipation function, predict the result of the match between two contestants.
Preferably, second prediction steps also include:
If the difference of the functional value of two contestants is less than the second preset value, the triumph of two contestants is predicted
Rate is identical;
If the difference of the functional value of two contestants is more than or equal to the second preset value, larger functional value pair is predicted
The contestant answered wins.
In addition, the present invention also provides a kind of electronic installation, the electronic installation includes:Memory, processor and it is stored in institute
The variety result of the match forecasting system stated on memory and can run on the processor, the variety result of the match prediction system
System is by the computing device, achievable following steps:
Data pick-up step:From preset data source, the history cycle data of variety match are extracted, and from the history
The feature of each contestant is extracted in cycle data;
Switch process:Pairing conversion is carried out to the feature of each contestant in the history cycle, obtains multiple match somebody with somebody
The result label of feature pairing conversion generation to contestant;
Model training step:Using the result label training Logic Regression Models of generation, model coefficient and model are determined
Anticipation function;
First prediction steps:The feature for each contestant of current period and current period that the variety is competed substitutes into respectively
In the anticipation function, the functional value of each contestant is obtained, and according to the size order of functional value to each contestant's
Result of the match carries out ranking.
Preferably, the expression formula of the anticipation function is:
S=f (x)=WTX
Wherein T represents the cycle of variety match, WTThe model coefficient of the anticipation function is represented, X represents the spy of contestant
Sign, S represent the functional value of the contestant.
Preferably, the model training step includes:
Data collection steps:Obtain the history cycle of variety match and the feature pairing transformation result label of contestant;
Detecting step:The feature pairing transformation result label of contestant is substituted into the type model one by one and trained, profit
The quality of the forecast model is detected with loss function;
Judgment step:The value of loss function is counted using empirical risk function, judges that empirical risk function value is
It is no to be less than the first preset value;
If empirical risk function value is more than or equal to the first preset value, then it represents that the Logic Regression Models are not optimal moulds
Type, model coefficient is adjusted, and return to training step until empirical risk function value is less than the first preset value;
If empirical risk function value is less than the first preset value, then it represents that the Logic Regression Models are optimal models.
Preferably, when the variety result of the match forecasting system is by the computing device, following steps are also realized:
Second prediction steps:By the current period of variety match and feature generation respectively of predetermined two contestants
Enter the anticipation function, predict the result of the match between two contestants;
If the difference of the functional value of two contestants is less than the second preset value, the triumph of two contestants is predicted
Rate is identical;
If the difference of the functional value of two contestants is more than or equal to the second preset value, larger functional value pair is predicted
The contestant answered wins.
In addition, to achieve the above object, the present invention also provides a kind of computer-readable recording medium, described computer-readable
Storage medium includes variety result of the match forecasting system, can when the variety result of the match forecasting system is executed by processor
Realize the arbitrary steps in variety result of the match Forecasting Methodology as described above.
Variety result of the match Forecasting Methodology, electronic installation and computer-readable recording medium proposed by the present invention, according to comprehensive
The history cycle of skill match and the feature generation feature pairing transformation result label of contestant, are then matched using feature and changed
As a result label training Logic Regression Models obtain anticipation function, and each competition of current variety match is predicted with the anticipation function of generation
The ranking of player, the accuracy of result of the match prediction can be improved.
Brief description of the drawings
Fig. 1 is the schematic diagram of electronic installation preferred embodiment of the present invention;
Fig. 2 is the functional block diagram of variety result of the match forecasting system preferred embodiment in Fig. 1;
Fig. 3 is the flow chart of variety result of the match Forecasting Methodology first embodiment of the present invention;
Fig. 4 is the flow chart of variety result of the match Forecasting Methodology second embodiment of the present invention.
Embodiment
It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not intended to limit the present invention.
As shown in figure 1, it is the schematic diagram of the preferred embodiment of electronic installation 1 of the present invention.
In the present embodiment, electronic installation 1 can be server, smart mobile phone, tablet personal computer, pocket computer, on table
Type computer etc. has the terminal device of calculation function.
The electronic installation 1 includes:Memory 11, processor 12, display 13, communication bus 3, and it is stored in described deposit
The variety result of the match forecasting system 10 of reservoir 11.The electronic installation 1 connects one or more servers 4 by network 2, social
Platform provides corresponding service on network by server 4 for social user.Extracted by server 4 in social platform comprehensive
The history cycle data of skill match.By communication bus 3 by the data transfer being drawn into processor 12.Network 2 can include
The network of the types such as LAN, wide area network, Metropolitan Area Network (MAN), can be cable network, or wireless network (such as WI-FI).It is logical
Letter bus 3 is used to realize the connection communication between these components.
Server 4 can be file server, database server, apps server or other can pass through network
2 terminal devices to be communicated with electronic installation 1.
Memory 11 comprises at least a type of readable storage medium storing program for executing.The readable storage medium storing program for executing of at least one type
Can be such as flash memory, hard disk, multimedia card, the non-volatile memory medium of card-type memory 11.In certain embodiments, it is described
Memory 11 can be the internal storage unit of the electronic installation 1, such as the hard disk of the electronic installation 1.In other implementations
In example, the memory 11 can also be the external memory unit of the electronic installation 1, such as be equipped with the electronic installation 1
Plug-in type hard disk, intelligent memory card (Smart Media Card, SMC), secure digital (Secure Digital, SD) card,
Flash card (Flash Card) etc..
In the present embodiment, the memory 11 can be not only used for storage be installed on the electronic installation 1 application it is soft
Part and Various types of data, such as variety result of the match forecasting system 10, preset kind Text Information Data and Logic Regression Models
Training, can be also used for temporarily storing the data that has exported or will export.
Processor 12 can be in certain embodiments a central processing unit (Central Processing Unit,
CPU), microprocessor or other data processing chips, for the program code or processing data stored in run memory 11, example
Such as perform the function of variety result of the match forecasting system 10.
Alternatively, the electronic installation 1 can also include user interface, and user interface can include input block such as keyboard
(Keyboard), instantaneous speech power such as sound equipment, earphone etc., alternatively user interface can also be connect including the wired of standard
Mouth, wave point.
Display 13 can be suitably be referred to as display screen or display unit.Display 13 can be in certain embodiments
(Organic Light-Emitting Diode, have by light-emitting diode display, liquid crystal display, touch-control liquid crystal display and OLED
Machine light emitting diode) touch device etc..Display is used to show the information handled in the electronic apparatus 1 and visualized for showing
User interface, such as:The ranking of each contestant of display prediction variety match.
In the device embodiment shown in Fig. 1, as storage variety ratio in a kind of memory 11 of computer-readable storage medium
The program code of prediction of result system 10 is matched, it is real when processor 12 performs the program code of variety result of the match forecasting system 10
Now following function:
From preset data source, the history cycle of a variety match is extracted, and is extracted respectively from the history cycle data
The feature of contestant;
Pairing conversion is carried out to the feature of each contestant in the history cycle, obtains multiple pairing contestants
Feature pairing transformation result label be stored in memory 11;
Transformation result label training Logic Regression Models are matched using the feature of each pairing contestant, determine model system
Number, and the anticipation function of model;
The feature for each contestant of current period and current period that the variety is competed substitutes into predetermined pre- respectively
Survey in function, obtain the functional value of each contestant, and according to the match knot of the size order of functional value to each contestant
Fruit carries out ranking, predicts the variety result of the match.
It is specific to introduce the detailed description that join lower section on the functional block diagram of variety result of the match forecasting system 10.
As shown in Fig. 2 it is the functional block diagram of the preferred embodiment of variety result of the match forecasting system 10 in Fig. 1.The present invention
Alleged module is the series of computation machine programmed instruction section for referring to complete specific function.
In the present embodiment, variety result of the match forecasting system 10 includes:Data extraction module 110, pairing modular converter
120th, model training module 130 and prediction module 140.
Data extraction module 110, for from preset data source, extracting the history cycle data of a variety match, and
The feature of each contestant is extracted from the history cycle data.Wherein, the preset data source is micro- including Sina weibo, Tengxun
The social platforms such as rich and Sohu's microblogging.The variety match can be " I is singer ", " the good sound of China " or " the new song of China
The variety shows such as sound ".For example, when variety result of the match forecasting system 10 needs to carry out " the new song of China " variety result of the match
During prediction, processor 12 extracts " the new song of China " variety from the microbloggings such as Sina, Tengxun and Sohu and competed a upper phase or upper several
The relevant historical data for holding cycle T of phase, while the microblogging data of contestant is also extracted from various microblogs.So
The feature X of contestant is extracted in the microblogging of contestant afterwards:X={ x1,x2,……,xn, wherein feature X includes microblogging
Number, the concern information such as number and bean vermicelli number.So as to obtain the identification information of this suite of song hand (such as name or competition numbering) composition
Set SINGER={ singer1,singer2,…,singern, and its corresponding feature X={ x1,x2,……,xn}.Assuming that
Singer singeri, determine singer singeriAfter feature, singer singeriCharacter representation be Xi={ xi1,xi2,…xin}.Its
Described in feature xi1Refer to singer singeriMicroblogging number, feature xi2Refer to singer singeriConcern number, feature xi3Refer to singer
singeriBean vermicelli number etc., other singer's situations are identical, repeat no more here.
Modular converter 120, for carrying out pairing conversion to the feature of each contestant in the history cycle, obtain
The result label of the feature pairing conversion generation of multiple pairing contestants.For example, singer singeriWith singer singerjIt is two
The contestant of individual contest on the same stage, processor 12 is by singer singeriWith singer singerjThe feature of two singers is matched
Conversion, obtains singer to (singeri,singerj) result label, if singer singeriSinger singer is wonj, then
(singeri,singerj) result label Y (Xi,Xj) it is 1, whereas if singer singerjSinger singer is woni, then
(singeri,singerj) result label Y (Xi,Xj) it is 0.Assuming that variety match has n contestant, then n* (n-1) is obtained
Bar result label, the label of victory or defeat between the label of no single one of which singer, only singer.If singer singeriSong is won
Hand singerj, then two result labels can be produced herein:First strip label Y (Xi,Xj) it is 1, characteristic value is equal to Xi-Xj.It is another
Bar record label Y (Xj,Xi) it is 0, characteristic value is equal to Xj-Xi。
Model training module 130, for the result label training Logic Regression Models of generation, determine model coefficient and model
Anticipation function.After matching modular converter 120 and utilizing the feature calculation of contestant to generate result label, model training module
130 will train in result label substitution Logic Regression Models, obtain the anticipation function of model coefficient and model.For example, modulus of conversion
After block 120 generates result label, it is assumed that anticipation function:S=f (x)=WTX, make S=f (x)=WTX meets
singerj∈SINGER:
It is rightsingerj∈ SINGER, work as singeri<R singerjWhen, Si>Sj。
By Si、SiSubstitute into anticipation function S=f (x)=WTIn X:
Si-Sj=WTXi-WTXj=WT(Xi-Xj)>0
So as to obtain:
Be converted to Logic Regression Models:
So as to obtain this Logic Regression Models as Logic Regression Models.Then it is n* (n-1) bar result label is right one by one
Logic Regression Models are trained, and so as to obtain optimal Logic Regression Models, determine model coefficient WTAnd the prediction letter of model
Number S=f (x)=WTX。
Wherein, the training step of the Logic Regression Models is as follows:
Data collection steps:Obtain the history cycle of variety match and the feature pairing transformation result label of contestant.
Assuming that variety match has n contestant, then history cycle T and n* (n-1) bar result label of variety match is obtained.
Training step:The result label of contestant is substituted into type of prediction model one by one and trained, utilizes loss function
Judge the quality of forecast model.Patrolled for example, n* (n-1) the bars feature pairing transformation result label of n contestant is updated to
Collect prediction result in regression model.Assuming that loss function is 0-1 loss functions:
Wherein, Y is singer singeriWith singer singerjResult label value, y (X) is singer singeriWith singer
singerjPredicted value in Logic Regression Models.If result label value Y is equal with predicted value y (X), L (Y, y (X)) is 0,
Illustrate that Logic Regression Models prediction is accurate;If result label value Y and predicted value y (X) is unequal, L (Y, y (X)) is 1, is said
Bright Logic Regression Models prediction is not accurate enough.Wherein, loss function can be quadratic loss function, absolute loss function and right
Number loss function.
Judgment step:The value of loss function is counted using empirical risk function, judges that empirical risk function value is
It is no to be less than the first preset value.Assuming that empirical risk function:
Wherein, Eemp[L (Y, y (X))] is average loss of the model on training sample, and N is result number of tags, in this reality
Apply in example, as a result number of tags:N=n* (n-1).If empirical risk function value is more than the first preset value, the Logic Regression Models
It is not optimal models, adjustment model coefficient WT, and training step is returned to until empirical risk function value is less than the first preset value;
If empirical risk function value is less than the first preset value, the Logic Regression Models are optimal models.
Prediction module 140, for will currently hold variety match cycle and each contestant feature substitute into respectively it is pre-
The anticipation function first determined, predict the result of the match between each contestant.
Preferably, the prediction module 140 includes the first predicting unit 141 and the second predicting unit 142.
First predicting unit 141, for the feature for each contestant of current period and current period that the variety is competed
Predetermined anticipation function is substituted into respectively, obtains the functional value of each contestant, and it is right according to the order of functional value size
Each contestant carries out ranking.For example, in " the new song of China " variety match, processor 12 extracts the current period of the match
Feature X corresponding to T and all contestants of current period, and it is updated to anticipation function S=f (x)=WTX, obtain all
The functional value W of contestantTX.Finally by the functional value W of all contestantsTX is ranked up, and selects 3 functional values of maximum
Corresponding contestant shows result as champion's candidate and by display 13.
Second predicting unit 142, for the current period that variety is competed and the spy of predetermined two contestants
Sign substitutes into the anticipation function respectively, predicts the result of the match between two contestants.For example, " the new song of China " variety
Match proceeds to singer singer1With singer singer2The link of champion and runner-up is determined, the second predicting unit 142 needs to predict singer
singer1With singer singer2Between result of the match.Processor 12 extracts the current period T of " the new song of China ", singer
singer1Characteristic information X1And singer singer2Characteristic information X2, and it is updated to anticipation function S=f (x)=WTX
In, singer singer is obtained respectively1Functional value WTX1With singer singer2Functional value WTX2.Finally by singer singer1's
Functional value WTX1With singer singer2Functional value WTX2It is poor to make.If the difference of the functional value of two contestants is less than second
Preset value, then show that the winning probability of two contestants is identical on the monitor 13;If the functional value of two contestants
Difference be more than or equal to the second preset value, then the larger contestant of explicit function value wins.
As shown in figure 3, it is the flow chart of variety result of the match Forecasting Methodology first embodiment of the present invention.
In the present embodiment, variety result of the match Forecasting Methodology includes:Step S10- steps S40.
Step S10, from preset data source, extract the history cycle data of variety match, and from the history cycle
The feature of each contestant is extracted in data.Wherein, the preset data source can be that Sina weibo, Tengxun's microblogging and Sohu are micro-
The social platform such as rich.The variety match can be variety sections such as " I are singer ", " the good sound of China " or " the new song of China "
Mesh.For example, when variety result of the match forecasting system 10 needs to be predicted " the new song of China " variety result of the match, processing
What device 12 extracted that " China new song " variety competed a upper phase or upper a few phases from the microbloggings such as Sina, Tengxun and Sohu holds the cycle
T relevant historical data, while the microblogging data of contestant is also filtered out from various microblogs.Then from the pre- of microblogging
If the feature X of contestant is extracted in type text message:X={ x1,x2,……,xn, wherein feature X include microblogging number,
Pay close attention to the information such as number and bean vermicelli number.So as to obtain the set of the identification information of this suite of song hand (such as name or competition numbering) composition
SINGER={ singer1,singer2,…,singern, and its corresponding feature X={ x1,x2,……,xn}.Assuming that need
Extract singer singeriFeature, then singer singer can be found in Sina weiboiMicroblogging, and extract singer
singeriMicroblogging data, it is then determined that singer singeriFeature Xi, singer singeriCharacter representation be Xi={ xi1,
xi2,…xin, other singers are also by similar method extraction feature information.
Step S20, pairing conversion is carried out to the feature of each contestant in the history cycle, obtains multiple pairings
The feature pairing transformation result label of contestant.For example, singer singeriWith singer singerjIt is the ginseng of two contests on the same stage
Player is matched, processor 12 is by singer singeriWith singer singerjThe feature of two singers carries out pairing conversion, obtains singer couple
(singeri,singerj) result label, if singer singeriSinger singer is wonj, then (singeri,singerj)
Result label Y (Xi,Xj) it is 1, whereas if singer singerjSinger singer is woni, then (singeri,singerj)
As a result label Y (Xi,Xj) it is 0.Assuming that variety match has n contestant, then n* (n-1) bars feature pairing transformation result is obtained
Label record.If singer singeriSinger singer is wonj, then two result labels can be produced herein:First strip label Y
(Xi,Xj) it is 1, characteristic value is equal to Xi-Xj.Another record label Y (Xj,Xi) it is 0, characteristic value is equal to Xj-Xi.Without list
The label of victory or defeat between the label of individual singer, only singer.Finally, the feature pairing transformation result label of generation is stored in storage
In device 11.
Step S30, using the result label training Logic Regression Models of each pairing contestant, determine model coefficient and model
Anticipation function.For example, it is assumed that anticipation function:S=f (x)=WTX, and cause the anticipation function S=f (x)=WTX disclosure satisfy that:singerj∈SINGER:
It is rightsingerj∈ SINGER, work as singeri<R singerjWhen, Si>Sj。
By Si、SjSubstitute into anticipation function S=f (x)=WTIn X:
Si-Sj=WTXi-WTXj=WT(Xi-Xj)>0
So as to obtain:
Be converted to Logic Regression Models:
So as to obtain this Logic Regression Models as Logic Regression Models.Then n* (n-1) bars feature is matched into Change-over knot
Fruit label is trained to Logic Regression Models one by one, so as to obtain optimal Logic Regression Models, determines model coefficient WTAnd
The anticipation function S=f (x) of model=WTX。
Step S40, for extracting the current period of variety match and feature generation respectively of each contestant of current period
Enter in predetermined anticipation function, obtain the functional value of each contestant, and according to the order of functional value size to each ginseng
Match player and carry out ranking.For example, in Chinese new song variety match, processor 12 extracts cycle T and the institute of the current match
There is feature X corresponding to contestant, and be updated to anticipation function S=f (x)=WTX, obtain the function of all contestants
Value WTX.Finally by the functional value W of all contestantsTX is ranked up from big to small, corresponding to 3 functional values for selecting maximum
Contestant shows result as champion's candidate and by display 13.For example, singer singeri, singer
singerj, singer singerkFunctional value W corresponding to respectivelyTXi、WTXj、WTXkMaximum, then display show:Singer singeri、
Singer singerj, singer singerkIt is champion's candidate.
The variety result of the match Forecasting Methodology that the present embodiment proposes, selected by using the history cycle and competition of variety match
The feature generation feature pairing transformation result label of hand, to train Logic Regression Models to obtain anticipation function, is finally substituted into current
The cycle of variety match and the feature of each contestant are held to anticipation function, contestant's ranking is determined, improves prediction match
As a result accuracy.
As shown in figure 4, it is the flow chart of variety result of the match Forecasting Methodology second embodiment of the present invention.
In the present embodiment, variety result of the match Forecasting Methodology includes:Step S10- steps S70.Wherein, step S10- is walked
Rapid S40 is roughly the same with content in first embodiment, repeats no more here.
Step S50, judges whether the difference of the anticipation function value of two contestants of current variety match is less than second
Preset value.For example, the new song variety match of China proceeds to singer singer1With singer singer2The link of champion and runner-up is determined,
Second predicting unit 142 needs to predict singer singer1With singer singer2Between result of the match.Processor 12 extracts currently
The cycle T for the new song of China held, singer singer1Characteristic information X1And singer singer2Characteristic information X2, and will
It is updated to anticipation function S=f (x)=WTIn X, singer singer is obtained respectively1Functional value WTX1With singer singer2Letter
Numerical value WTX2.Finally by singer singer1Functional value WTX1With singer singer2Functional value WTX2It is poor to make, and judges two songs
Whether the difference of the anticipation function value of hand is less than the second preset value.
Step S60, if the difference of the functional value of two contestants is less than the second preset value, two contestant's victory
Rate is 50%.For example, by singer singer1Functional value WTX1With singer singer2Functional value WTX2It is poor to make, and difference is less than
During the second preset value, display 13 shows " singer singer1With singer singer2The average of wins is identical ".
Step S70, if the difference of the functional value of two contestants is more than or equal to the second preset value, larger function
Contestant corresponding to value wins.For example, by singer singer1Functional value WTX1With singer singer2Functional value WTX2Make
Difference, when difference is more than the second preset value, it is assumed that singer singer1Functional value WTX1More than singer singer2Functional value WTX2,
Then display 13 shows " singer singer1Win ".
The variety result of the match Forecasting Methodology proposed compared to first embodiment, the present embodiment, gone through according to what variety was competed
The feature of history cycle and contestant generation feature pairing transformation result label, then match transformation result label instruction using feature
Practice Logic Regression Models and obtain anticipation function, can finally utilize the anticipation function of generation to predict each competition of current variety match
The ranking of player, anticipation function can also be utilized to predict during the match of current variety the match of two singers of contest on the same stage
As a result, the accuracy and comprehensive of result of the match prediction is improved.
In addition, the embodiment of the present invention also proposes a kind of computer-readable recording medium, the computer-readable recording medium
Include variety result of the match forecasting system 10, realized when the variety result of the match forecasting system 10 is executed by processor as follows
Operation:
Data pick-up step:From preset data source, the history cycle data of variety match are extracted, and from the history
The feature of each contestant is extracted in cycle data;
Switch process:, pairing conversion is carried out to the feature of each contestant in the history cycle, obtains multiple match somebody with somebody
Transformation result label is matched to the feature of contestant;
Model training step:Transformation result label training logistic regression mould is matched using the feature of each pairing contestant
Type, determine the anticipation function of model coefficient and model;
First prediction steps:The feature for each contestant of current period and current period that the variety is competed substitutes into respectively
In the anticipation function, the functional value of each contestant is obtained, and according to the order of functional value size to each contestant's
Result of the match carries out ranking.
Preferably, the expression formula of the anticipation function is:
S=f (x)=WTX
Wherein T represents the cycle of variety match, WTThe model coefficient of the anticipation function is represented, X represents the spy of contestant
Sign, S represent the functional value of the contestant.
Preferably, when the variety result of the match forecasting system is by the computing device, following steps are also realized:
Second prediction steps:By the current period of variety match and feature generation respectively of predetermined two contestants
Enter the anticipation function, predict the result of the match between two contestants;
Preferably, second prediction steps also include:
If the difference of the functional value of two contestants is less than the second preset value, the triumph of two contestants is predicted
Rate is identical;
If the difference of the functional value of two contestants is more than or equal to the second preset value, larger functional value pair is predicted
The contestant answered wins.
The embodiment of the computer-readable recording medium of the present invention and above-mentioned variety result of the match Forecasting Methodology
Embodiment is roughly the same, will not be repeated here.
The embodiments of the present invention are for illustration only, do not represent the quality of embodiment.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can add the mode of required general hardware platform to realize by software, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Based on such understanding, technical scheme is substantially done to prior art in other words
Going out the part of contribution can be embodied in the form of software product, and the computer software product is stored in one as described above
In storage medium (such as ROM/RAM, magnetic disc, CD), including some instructions to cause a station terminal equipment (can be mobile phone,
Computer, server, or network equipment etc.) perform method described in each embodiment of the present invention.
The preferred embodiments of the present invention are these are only, are not intended to limit the scope of the invention, it is every to utilize this hair
The equivalent structure or equivalent flow conversion that bright specification and accompanying drawing content are made, or directly or indirectly it is used in other related skills
Art field, is included within the scope of the present invention.
Claims (10)
1. a kind of variety result of the match Forecasting Methodology, it is characterised in that methods described includes:
Data pick-up step:From preset data source, the history cycle data of variety match are extracted, and from the history cycle
The feature of each contestant is extracted in data;
Switch process:Pairing conversion is carried out to the feature of each contestant in the history cycle, obtains multiple pairing ginsengs
Match the result label of the feature pairing conversion generation of player;
Model training step:Using the result label training Logic Regression Models of generation, the prediction of model coefficient and model is determined
Function;
First prediction steps:The feature for each contestant of current period and current period that the variety is competed substitutes into described respectively
In anticipation function, the functional value of each contestant is obtained, and according to the match of the size order of functional value to each contestant
As a result ranking is carried out.
2. variety result of the match Forecasting Methodology according to claim 1, it is characterised in that the anticipation function of the model
Expression formula is:
S=f (x)=WTX
Wherein T represents the cycle of variety match, WTThe model coefficient of the anticipation function is represented, X represents the feature of contestant, S generations
The functional value of the table contestant.
3. variety result of the match Forecasting Methodology according to claim 1, it is characterised in that the model training step bag
Include:
Data collection steps:Obtain the history cycle of variety match and the feature pairing transformation result label of contestant;
Detecting step:The feature pairing transformation result label of contestant is substituted into the type model one by one and trained, utilizes damage
Lose the quality of the function check forecast model;
Judgment step:The value of loss function is counted using empirical risk function, judges whether empirical risk function value is small
In the first preset value;
If empirical risk function value is more than or equal to the first preset value, then it represents that the Logic Regression Models are not optimal modelses, are adjusted
Integral mould coefficient, and training step is returned to until empirical risk function value is less than the first preset value;
If empirical risk function value is less than the first preset value, then it represents that the Logic Regression Models are optimal models.
4. variety result of the match Forecasting Methodology according to claim 1, it is characterised in that this method also includes:
Second prediction steps:The feature of the current period of variety match and predetermined two contestants is substituted into institute respectively
Anticipation function is stated, predicts the result of the match between two contestants.
5. variety result of the match Forecasting Methodology according to claim 4, it is characterised in that second prediction steps are also wrapped
Include:
If the difference of the functional value of two contestants is less than the second preset value, the average of wins phase of two contestants is predicted
Together;
If the difference of the functional value of two contestants is more than or equal to the second preset value, predict corresponding to larger functional value
Contestant wins.
6. a kind of electronic installation, it is characterised in that described device includes:Memory, processor and it is stored on the memory
And the variety result of the match forecasting system that can be run on the processor, the variety result of the match forecasting system is by the place
Manage device to perform, following steps can be achieved:
Data pick-up step:From preset data source, the history cycle data of variety match are extracted, and from the history cycle
The feature of each contestant is extracted in data;
Switch process:Pairing conversion is carried out to the feature of each contestant in the history cycle, obtains multiple pairing ginsengs
Match the result label of the feature pairing conversion generation of player;
Model training step:Using the result label training Logic Regression Models of generation, the prediction of model coefficient and model is determined
Function;
First prediction steps:The feature for each contestant of current period and current period that the variety is competed substitutes into described respectively
In anticipation function, the functional value of each contestant is obtained, and according to the match of the size order of functional value to each contestant
As a result ranking is carried out.
7. electronic installation according to claim 6, it is characterised in that the expression formula of the anticipation function is:
S=f (x)=WTX
Wherein T represents the cycle of variety match, WTThe model coefficient of the anticipation function is represented, X represents the feature of contestant, S generations
The functional value of the table contestant.
8. electronic installation according to claim 6, it is characterised in that the model training step includes:
Data collection steps:Obtain the history cycle of variety match and the feature pairing transformation result label of contestant;
Detecting step:The feature pairing transformation result label of contestant is substituted into the type model one by one and trained, utilizes damage
Lose the quality of the function check forecast model;
Judgment step:The value of loss function is counted using empirical risk function, judges whether empirical risk function value is small
In the first preset value;
If empirical risk function value is more than or equal to the first preset value, then it represents that the Logic Regression Models are not optimal modelses, are adjusted
Integral mould coefficient, and training step is returned to until empirical risk function value is less than the first preset value;
If empirical risk function value is less than the first preset value, then it represents that the Logic Regression Models are optimal models.
9. electronic installation according to claim 6, it is characterised in that the variety result of the match forecasting system is by the place
Manage device to perform, also realize following steps:
Second prediction steps:The feature of the current period of variety match and predetermined two contestants is substituted into institute respectively
Anticipation function is stated, predicts the result of the match between two contestants;
If the difference of the functional value of two contestants is less than the second preset value, the average of wins phase of two contestants is predicted
Together;
If the difference of the functional value of two contestants is more than or equal to the second preset value, predict corresponding to larger functional value
Contestant wins.
10. a kind of computer-readable recording medium, it is characterised in that the computer-readable recording medium includes variety match
Prediction of result system, when the system variety result of the match forecasting system is executed by processor, it can be achieved as in claim 1 to 5
The step of any one variety result of the match Forecasting Methodology.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710876238.4A CN107807908A (en) | 2017-09-25 | 2017-09-25 | Variety result of the match Forecasting Methodology, device and storage medium |
PCT/CN2017/108805 WO2019056502A1 (en) | 2017-09-25 | 2017-10-31 | Variety game result prediction method and apparatus, and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710876238.4A CN107807908A (en) | 2017-09-25 | 2017-09-25 | Variety result of the match Forecasting Methodology, device and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107807908A true CN107807908A (en) | 2018-03-16 |
Family
ID=61583927
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710876238.4A Pending CN107807908A (en) | 2017-09-25 | 2017-09-25 | Variety result of the match Forecasting Methodology, device and storage medium |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN107807908A (en) |
WO (1) | WO2019056502A1 (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109091868A (en) * | 2018-08-14 | 2018-12-28 | 腾讯科技(深圳)有限公司 | Method, apparatus, computer equipment and the storage medium that battle behavior determines |
CN110458328A (en) * | 2019-07-11 | 2019-11-15 | 清华大学 | Black Swan event decision method and device based on subjective and objective associated prediction |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102385719A (en) * | 2011-11-01 | 2012-03-21 | 中国科学院计算技术研究所 | Regression prediction method and device |
CN105989441A (en) * | 2015-02-11 | 2016-10-05 | 阿里巴巴集团控股有限公司 | Model parameter adjustment method and device |
CN107103419A (en) * | 2017-04-20 | 2017-08-29 | 北京航空航天大学 | One kind of groups software development process analogue system and method |
-
2017
- 2017-09-25 CN CN201710876238.4A patent/CN107807908A/en active Pending
- 2017-10-31 WO PCT/CN2017/108805 patent/WO2019056502A1/en active Application Filing
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109091868A (en) * | 2018-08-14 | 2018-12-28 | 腾讯科技(深圳)有限公司 | Method, apparatus, computer equipment and the storage medium that battle behavior determines |
CN109091868B (en) * | 2018-08-14 | 2019-11-22 | 腾讯科技(深圳)有限公司 | Method, apparatus, computer equipment and the storage medium that battle behavior determines |
CN110458328A (en) * | 2019-07-11 | 2019-11-15 | 清华大学 | Black Swan event decision method and device based on subjective and objective associated prediction |
Also Published As
Publication number | Publication date |
---|---|
WO2019056502A1 (en) | 2019-03-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110909176B (en) | Data recommendation method and device, computer equipment and storage medium | |
CN104239535B (en) | A kind of method, server, terminal and system for word figure | |
CN110009401A (en) | Advertisement placement method, device and storage medium based on user's portrait | |
CN108629043A (en) | Extracting method, device and the storage medium of webpage target information | |
CN108280115A (en) | Identify the method and device of customer relationship | |
CN106688215A (en) | Automated click type selection for content performance optimization | |
CN115002200B (en) | Message pushing method, device, equipment and storage medium based on user portrait | |
CN105469262B (en) | Commodity information acquisition of information and integration method and device | |
CN112528616B (en) | Service form generation method and device, electronic equipment and computer storage medium | |
CN114722281B (en) | Training course configuration method and device based on user portrait and user course selection behavior | |
CN112559923A (en) | Website resource recommendation method and device, electronic equipment and computer storage medium | |
CN109194689A (en) | Abnormal behaviour recognition methods, device, server and storage medium | |
CN112733023A (en) | Information pushing method and device, electronic equipment and computer readable storage medium | |
CN108228720A (en) | Identify method, system, device, terminal and the storage medium of target text content and artwork correlation | |
CN107239577A (en) | A kind of information cuing method, terminal and computer-readable recording medium | |
CN107704486A (en) | The device that a kind of registrable property of the figurative mark based on artificial intelligence judges | |
CN114663198A (en) | Product recommendation method, device and equipment based on user portrait and storage medium | |
CN110083774A (en) | Using determination method, apparatus, computer equipment and the storage medium of recommendation list | |
CN111858834A (en) | Case dispute focus determining method, case dispute focus determining device, case dispute focus determining equipment and case dispute focus determining medium based on AI | |
CN107807908A (en) | Variety result of the match Forecasting Methodology, device and storage medium | |
CN113360768A (en) | Product recommendation method, device and equipment based on user portrait and storage medium | |
CN115423535A (en) | Product purchasing method, device, equipment and medium based on market prior big data | |
CN116405712A (en) | OTT video-based multimedia pushing method, device, equipment and storage medium | |
CN109934631A (en) | Question and answer information processing method, device and computer equipment | |
CN113627160A (en) | Text error correction method and device, electronic equipment and storage medium |
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 |
Application publication date: 20180316 |
|
RJ01 | Rejection of invention patent application after publication |