WO2019056502A1 - Variety game result prediction method and apparatus, and storage medium - Google Patents

Abstract

A variety game result prediction method and apparatus, and a storage medium, the method comprising: extracting historical cycle data of a variety game from a preset data source, and extracting features of game players from the historical cycle data (S10); performing pairing conversion on the features of the game players in a historical cycle to obtain a plurality of result tags generated by the pairing conversion of the features of the paired game players (S20); training a logic regression model by utilizing the generated result tags, and determining a model coefficient and a model prediction function (S30); and substituting a current cycle of the variety game and the features of the game players in the current cycle into the predetermined prediction function to calculate function values of the game players, and ranking the game results of the game players according to a function value sequence (S40). With the present invention, variety game results are predicted by carrying out calculations for the cycle of the variety game and the features of the game players, so that the accuracy of prediction results is increased.

Description

Variety game result prediction method, device and storage medium

This patent application is based on the Chinese invention patent application filed on September 25, 2017, with the application number of 201710876238.4, entitled "Variety Competition Results Prediction Method, Apparatus and Storage Medium", and requires its priority.

Technical field

The present application relates to the field of computer data mining, and in particular, to a method, device, and computer readable storage medium for predicting results of a variety game.

Background technique

With the increasing popularity of social networking sites in life, many variety games, such as: I am a singer or a good voice in China, will be promoted on social platforms in order to gain more attention from the audience. While attracting the attention of fans, there are also many viewers who will express their support and opinions on the social platform.

At present, in predicting the outcome of the competition, some organizations use the support rate of the audience on the social platform to predict the outcome of the contestants. However, these methods are heuristics, not learned from historical data. The accuracy is low and it is not possible to accurately predict the outcome of the game.

Summary of the invention

The present application provides a method, device and computer readable storage medium for predicting results of a variety game, which can accurately predict the results of the current variety game according to the historical data of the variety game.

To achieve the above objective, the present application provides a method for predicting results of a variety game, the method comprising:

Data extraction step: extracting historical period data of a variety game from a preset data source, and extracting characteristics of each contestant from the historical period data;

a conversion step: pairing and converting the characteristics of each contestant in the historical period to obtain a result label generated by the feature pairing conversion of the plurality of matched contestants;

Model training step: training the logistic regression model with the generated result tags to determine the model coefficients and the prediction function of the model;

The first predicting step: substituting the current period of the variety game and the characteristics of each contestant in the current period into the prediction function, determining the function value of each contestant, and ranking the contestants according to the magnitude of the function value. The results are ranked.

Preferably, the expression of the prediction function is:

S=f(x)=W ^T X

Where T represents the period of the variety game, W ^T represents the model coefficient of the prediction function, X represents the characteristics of the contestant, and S represents the function value of the contestant.

Preferably, the model training step comprises:

Data collection step: obtaining the historical period of the variety game and the feature matching conversion result label of the contestant;

Detection step: the contestant's feature pairing conversion result label is substituted into the type model for training, and the loss function is used to detect the pros and cons of the prediction model;

Judging step: using the empirical risk function to perform statistics on the value of the loss function, and determining whether the value of the empirical risk function is less than the first preset value;

If the value of the empirical risk function is greater than or equal to the first preset value, it indicates that the logistic regression model is not the optimal model, adjusting the model coefficients, and returning to the training step until the empirical risk function value is less than the first preset value;

If the empirical risk function value is less than the first preset value, it indicates that the logistic regression model is the optimal model.

Preferably, the method further comprises:

The second prediction step: substituting the current period of the variety game and the characteristics of the two predetermined contestants into the prediction function respectively, and predicting the result of the match between the two contestants.

Preferably, the second prediction step further comprises:

If the difference between the function values of the two competitors is less than the second preset value, it is predicted that the winning rates of the two participating players are the same;

If the difference between the function values of the two competitors is greater than or equal to the second preset value, the competitor corresponding to the larger function value is predicted to win.

In addition, the present application further provides an electronic device including: a memory, a processor, and a variety game result prediction system stored on the memory and operable on the processor, the variety game result prediction system Executed by the processor, the following steps can be implemented:

Data extraction step: extracting historical period data of a variety game from a preset data source, and extracting characteristics of each contestant from the historical period data;

a conversion step: pairing and converting the characteristics of each contestant in the historical period to obtain a result label generated by the feature pairing conversion of the plurality of matched contestants;

Model training step: training the logistic regression model with the generated result tags to determine the model coefficients and the prediction function of the model;

The first predicting step: substituting the current period of the variety game and the characteristics of each contestant in the current period into the prediction function, determining the function value of each contestant, and ranking the contestants according to the magnitude of the function value. The results are ranked.

Preferably, the expression of the prediction function is:

S=f(x)=W ^T X

Where T represents the period of the variety game, W ^T represents the model coefficient of the prediction function, X represents the characteristics of the contestant, and S represents the function value of the contestant.

Preferably, the model training step comprises:

Data collection step: obtaining the historical period of the variety game and the feature matching conversion result label of the contestant;

Detection step: the contestant's feature pairing conversion result label is substituted into the type model for training, and the loss function is used to detect the pros and cons of the prediction model;

Judgment step: use the empirical risk function to count the value of the loss function to determine the empirical risk Whether the function value is less than the first preset value;

If the value of the empirical risk function is greater than or equal to the first preset value, it indicates that the logistic regression model is not the optimal model, adjusting the model coefficients, and returning to the training step until the empirical risk function value is less than the first preset value;

If the empirical risk function value is less than the first preset value, it indicates that the logistic regression model is the optimal model.

Preferably, when the variety game result prediction system is executed by the processor, the following steps are also implemented:

a second prediction step: substituting the current period of the variety game and the characteristics of the two predetermined contestants into the prediction function, respectively, and predicting the result of the match between the two contestants;

If the difference between the function values of the two competitors is less than the second preset value, it is predicted that the winning rates of the two participating players are the same;

If the difference between the function values of the two competitors is greater than or equal to the second preset value, the competitor corresponding to the larger function value is predicted to win.

In addition, in order to achieve the above object, the present application further provides a computer readable storage medium, which includes a variety game result prediction system, where the variety game result prediction system is executed by a processor, Any step in the method of predicting the results of the variety game.

The result prediction method, the electronic device and the computer readable storage medium proposed by the present application generate a feature pairing conversion result label according to the historical period of the variety game and the characteristics of the contestant, and then use the feature pairing conversion result label to train the logistic regression model to obtain the prediction. The function, using the generated prediction function to predict the ranking of each contestant in the current variety game, can improve the accuracy of the game result prediction.

DRAWINGS

1 is a schematic diagram of a preferred embodiment of an electronic device of the present application;

2 is a functional block diagram of a preferred embodiment of the result prediction system of the variety game of FIG. 1;

3 is a flow chart of a first embodiment of a method for predicting a result of a variety game of the present application;

4 is a flow chart of a second embodiment of a method for predicting a result of a variety game of the present application.

Detailed ways

It is understood that the specific embodiments described herein are merely illustrative of the application and are not intended to be limiting.

FIG. 1 is a schematic diagram of a preferred embodiment of an electronic device 1 of the present application.

In this embodiment, the electronic device 1 may be a terminal device having a computing function, such as a server, a smart phone, a tablet computer, a portable computer, or a desktop computer.

The electronic device 1 includes a memory 11, a processor 12, a display 13, a communication bus 3, and a variety game result prediction system 10 stored in the memory 11. The electronic device 1 is connected to one or more servers 4 via a network 2, and the social platform provides social users on the network through the server 4. The corresponding service. The historical period data of the variety game is extracted by the server 4 in the social platform. The extracted data is transmitted to the processor 12 via the communication bus 3. The network 2 may include a network such as a local area network, a wide area network, or a metropolitan area network, and may be a wired network or a wireless network (such as WI-FI). The communication bus 3 is used to implement connection communication between these components.

The server 4 may be a file server, a database server, an application server, or other terminal device that can communicate with the electronic device 1 through the network 2.

The memory 11 includes at least one type of readable storage medium. The at least one type of readable storage medium may be a non-volatile storage medium such as a flash memory, a hard disk, a multimedia card, a card type memory 11, or the like. In some embodiments, the memory 11 may be an internal storage unit of the electronic device 1, such as a hard disk of the electronic device 1. In other embodiments, the memory 11 may also be an external storage unit of the electronic device 1, such as a plug-in hard disk equipped on the electronic device 1, a smart memory card (SMC), and security. Digital (Secure Digital, SD) card, flash card (Flash Card), etc.

In this embodiment, the memory 11 can be used not only for storing application software installed on the electronic device 1 and various types of data, such as training of the variety game result prediction system 10, preset type text information data, and logistic regression model. It can also be used to temporarily store data that has been output or will be output.

The processor 12, in some embodiments, may be a Central Processing Unit (CPU), microprocessor or other data processing chip for running program code or processing data stored in the memory 11, such as performing a variety game result. The function of the system 10 and the like are predicted.

Optionally, the electronic device 1 may further include a user interface, and the user interface may include an input unit such as a keyboard, a voice output device such as an audio, a headphone, etc., optionally, the user interface may further include a standard wired interface and a wireless interface. .

Display 13 may be referred to as a display or display unit as appropriate. In some embodiments, the display 13 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch sensor, or the like. The display is used to display information processed in the electronic device 1 and a user interface for displaying visualizations, such as: displaying the rankings of the contestants of the predicted variety game.

In the apparatus embodiment shown in FIG. 1, the program code of the variety game result prediction system 10 is stored in the memory 11 as a computer storage medium, and when the processor 12 executes the program code of the variety game result prediction system 10, the following functions are realized. :

Extracting a historical period of a variety game from a preset data source, and extracting characteristics of each contestant from the historical period data;

Performing pairing conversion on the characteristics of each participating player in the historical period, and obtaining feature pair conversion conversion result tags of the plurality of matching competitors are stored in the memory 11;

The logistic regression model is trained by using the feature pairing conversion result tag of each paired contestant to determine the model coefficient and the prediction function of the model;

The current period of the variety game and the characteristics of each participant in the current cycle are respectively substituted into a predetermined prediction function, and the function values of the respective contestants are obtained, and the order of the function values is sequentially The results of each contestant's competition are ranked to predict the results of the variety game.

For details, please refer to the detailed description of the function module diagram of the variety game result prediction system 10 below.

As shown in FIG. 2, it is a functional block diagram of a preferred embodiment of the variety game result prediction system 10 of FIG. A module as referred to in this application refers to a series of computer program instructions that are capable of performing a particular function.

In this embodiment, the variety game result prediction system 10 includes: a data extraction module 110, a pairing conversion module 120, a model training module 130, and a prediction module 140.

The data extraction module 110 is configured to extract historical period data of a variety game from the preset data source, and extract features of each contestant from the historical period data. The preset data source includes social platforms such as Sina Weibo, Tencent Weibo, and Sohu Weibo. The variety game may be a variety show such as "I am a singer", "China's good voice" or "Chinese new song". For example, when the variety game result prediction system 10 needs to predict the results of the "China New Songs" variety game, the processor 12 extracts the "China New Songs" variety game from Sina, Tencent and Sohu. During the period, the historical data of the period T is also collected, and the Weibo materials of the contestants are also extracted from various Weibo platforms. Then, the contestant's feature X is extracted from the competitor's Weibo: X={x ₁ , x ₂ , . . . , x _n }, wherein the feature X includes information such as the number of microblogs, the number of followers, and the number of fans. Thereby obtaining a set of SINGER={singer ₁ ,singer ₂ ,...,singer _n } composed of the identification information (such as name or entry number) of the group of artists, and their corresponding features X={x ₁ , x ₂ , . x _n }. Assuming singer singer _i , after determining the singer singer _i feature, the singer singer _i is characterized by X _i = {xi ₁ , xi ₂ , ... xi _n }. Xi ₁ wherein the characteristic index singer singer _i Twitter, features xi ₂ Follow index singer singer _i, features xi ₃ refers to the number of fans and the like singer singer _i are the same as the case of other singers, not repeated here.

The conversion module 120 is configured to perform pairing conversion on the characteristics of each participating player in the historical period to obtain a result label generated by the feature pairing conversion of the plurality of matching contestants. For example, the singer singer _i and the singer singer _j are two contestants of the same competition, and the processor 12 pairs and converts the characteristics of the singer _i and the singer singer _{j to} obtain the singer pair (singer _i , singer _j ) a result tag, if the singer singer _i win singer singer _j, then (singer _i, singer _j) results label Y (X _i, X _j) is 1, whereas, if the singer singer _j win singer singer _i, the (singer _i, singer _j) results label Y (X _i, X _j) is 0. Suppose there are n contestants in the variety game, then get the n*(n-1) result label, there is no single singer's label, only the singer wins and loses the label. If the singer singer _i wins the singer singer _j , then two result tags are generated here: the first tag Y(X _i , X _j ) is 1, and the eigenvalue is equal to X _i -X _j . The other record label Y(X _j , X _i ) is 0, and the feature value is equal to X _j -X _i .

singer_j∈SINGER:

The model training module 130 uses the generated result tag to train the logistic regression model to determine the model coefficients and the prediction function of the model. After the pairing conversion module 120 calculates the generated result tag by using the feature of the competitor, the model training module 130 substitutes the result tag into the logistic regression model to obtain the model coefficient and the prediction function of the model. For example, after the conversion module 120 generates the result label, it is assumed that the prediction function: S=f(x)=W ^T X, so that S=f(x)=W ^T X is satisfied.

Singer _j ∈SINGER:

singer_j∈SINGER,当singer_i<R singer_j时，S_i>S_j。Correct

Singer _j ∈SINGER, when singer _i <R singer _j , S _i >S _j .

Substituting S _i , S _j into the prediction function S=f(x)=W ^T X yields:

S _i -S _j =W ^T X _i -W ^T X _j =W ^T (X _i -X _j )>0

Thereby getting:

Convert to a logistic regression model:

This logistic regression model is thus obtained as a logistic regression model. Then the n*(n-1) result labels are trained on a pair of logistic regression models to obtain an optimal logistic regression model, and the model coefficients W ^T and the model's prediction function S=f(x)=W ^T X are determined.

The training steps of the logistic regression model are as follows:

Data collection step: Get the historical period of the variety game and the feature pair conversion result label of the contestant. Assuming that there are n contestants in the variety game, the historical period T and n*(n-1) result labels of the variety game are obtained.

Training steps: The result labels of the contestants are trained into the prediction type model one by one, and the loss function is used to judge the pros and cons of the prediction model. For example, the n*(n-1) feature pairing conversion result tags of n competitors are substituted into the prediction results in the logistic regression model. Suppose the loss function is a 0-1 loss function:

Where Y is the result tag value of singer singer _i and singer singer _j , and y(X) is the predicted value of singer singer _i and singer singer _j in the logistic regression model. If the result label value Y is equal to the predicted value y(X), L(Y, y(X)) is 0, indicating that the logistic regression model is accurate; if the result label value Y is not equal to the predicted value y(X), L (Y, y(X)) is 1, indicating that the logistic regression model prediction is not accurate enough. Among them, the loss function can be a square loss function, an absolute loss function, and a logarithmic loss function.

Judgment step: using the empirical risk function to calculate the value of the loss function, and determine whether the value of the empirical risk function is less than the first preset value. Assume the empirical risk function:

Where E _emp [L(Y, y(X))] is the average loss of the model with respect to the training samples, and N is the number of result labels. In the present embodiment, the number of result labels: N=n*(n-1). If the empirical risk function value is greater than the first preset value, the logistic regression model is not the optimal model, and the model coefficient W ^{T is} adjusted and returned to the training step until the empirical risk function value is less than the first preset value; if the empirical risk function value If the value is smaller than the first preset value, the logistic regression model is the optimal model.

The prediction module 140 is configured to substitute the current period of the variety game and the characteristics of each contestant into a predetermined prediction function to predict the result of the match between the contestants.

Preferably, the prediction module 140 includes a first prediction unit 141 and a second prediction unit 142.

The first prediction unit 141 is configured to substitute the characteristics of each contestant in the current period and the current period of the variety game into a predetermined prediction function, and obtain function values of each contestant, and participate in each contest according to the order of the function values. The players are ranked. For example, in the "China New Songs" variety game, the processor 12 extracts the current period T of the game and the feature X corresponding to all the competitors in the current cycle, and substitutes it into the prediction function S=f(x)=W ^T X Find the function value W ^T X of all contestants. Finally, the function values W ^T X of all the contestants are sorted, and the contestants corresponding to the maximum of the three function values are selected as the champion candidates and the results are displayed on the display 13.

The second prediction unit 142 is configured to substitute the current period of the variety game and the characteristics of the two predetermined contestants into the prediction function, respectively, and predict the result of the match between the two contestants. For example, the "China New Songs" variety game is conducted to the singer singer ₁ and the singer singer ₂ to decide the stage of the runner-up, and the second prediction unit 142 needs to predict the result of the match between the singer singer ₁ and the singer singer ₂ . The processor 12 extracts "New Chinese song" current period T, singer singer ₁ wherein X ₁ and feature information of a singer ₂ singer information X _2, and substituted into the prediction function S = f (x) = W T X in We were determined singer singer function value W ^T X _{1 1} and ₂ singer singer function value W ^T X _2. Finally singer singer function value W ^T X _{1 1} and ₂ singer singer function value W ^T X ₂ as difference. If the difference between the function values of the two competitors is less than the second preset value, the probability of winning the two participating players is the same on the display 13; if the difference between the function values of the two participating players is greater than or equal to the first The second preset value indicates that the contestant with a larger function value wins.

As shown in FIG. 3, it is a flowchart of the first embodiment of the method for predicting the results of the variety game of the present application.

In this embodiment, the variety game result prediction method includes: step S10 - step S40.

Step S10: Extract historical period data of a variety game from the preset data source, and extract features of each contestant from the historical period data. The preset data source may be a social platform such as Sina Weibo, Tencent Weibo, and Sohu Weibo. The variety game may be a variety show such as "I am a singer", "China's good voice" or "Chinese new song". For example, when the variety game result prediction system 10 needs to predict the results of the "China New Songs" variety game, the processor 12 extracts the "China New Songs" variety game from Sina, Tencent and Sohu. The historical data of the period T is also selected, and the Weibo materials of the contestants are also selected from various Weibo platforms. Then, the feature X of the competitor is extracted from the preset type text information of the microblog: X={x ₁ , x ₂ , . . . , x _n }, wherein the feature X includes information such as the number of microblogs, the number of followers, and the number of fans. . Thereby obtaining a set of SINGER={singer ₁ ,singer ₂ ,...,singer _n } composed of the identification information (such as name or entry number) of the group of artists, and their corresponding features X={x ₁ , x ₂ , . x _n }. Suppose you need to extract the characteristics of singer singer _i , you can find the singer _i 's microblog in Sina Weibo, and extract the singer _i 's microblog information, and then determine the singer _i 's feature X _i , the singer _i 's feature representation For X _i ={xi ₁ ,xi ₂ ,...xi _n }, other singers also extract feature information by a similar method.

In step S20, pairing conversion is performed on the characteristics of each participating player in the historical period to obtain a feature pairing conversion result label of the plurality of matching competitors. For example, the singer singer _i and the singer singer _j are two contestants of the same competition, and the processor 12 pairs and converts the characteristics of the singer _i and the singer singer _{j to} obtain the singer pair (singer _i , singer _j ) a result tag, if the singer singer _i win singer singer _j, then (singer _i, singer _j) results label Y (X _i, X _j) is 1, whereas, if the singer singer _j win singer singer _i, the (singer _i, singer _j) results label Y (X _i, X _j) is 0. Assuming that there are n contestants in the variety game, n*(n-1) feature pairing conversion result label records are obtained. If the singer singer _i wins the singer singer _j , then two result tags are generated here: the first tag Y(X _i , X _j ) is 1, and the eigenvalue is equal to X _i -X _j . The other record label Y(X _j , X _i ) is 0, and the feature value is equal to X _j -X _i . There is no label for a single singer, only the label for winning or losing between singers. Finally, the generated feature pair conversion result tag is stored in the memory 11.

In step S30, the logistic regression model is trained by using the result tags of the respective matched competitors to determine the model coefficients and the prediction function of the model. For example, suppose the prediction function: S = f(x) = W ^T X, and makes the prediction function S = f(x) = W ^T X satisfy: Singer _j ∈SINGER:

singer_j∈SINGER,当singer_i<R singer_j时，S_i>S_j。Correct

Singer _j ∈SINGER, when singer _i <R singer _j , S _i >S _j .

Substituting S _i , S _j into the prediction function S=f(x)=W ^T X yields:

S _i -S _j =W ^T X _i -W ^T X _j =W ^T (X _i -X _j )>0

Thereby getting:

Convert to a logistic regression model:

This logistic regression model is thus obtained as a logistic regression model. Then n*(n-1) feature pairing conversion result labels are trained one by one by logistic regression model to obtain the optimal logistic regression model, and the model coefficient W ^T and the model's prediction function S=f(x)=W are determined. ^T X.

Step S40, for extracting the current period of the variety game and the characteristics of each contestant in the current period are respectively substituted into a predetermined prediction function, determining the function values of each contestant, and performing the contestants according to the order of the function values. Ranking. For example, in the Chinese New Song Variety Competition, the processor 12 extracts the current period T of the game and the feature X corresponding to all the contestants, and substitutes it into the prediction function S=f(x)=W ^T X to find all The function value of the contestant is W ^T X. Finally, the function values W ^T X of all the contestants are sorted from large to small, and the contestants corresponding to the largest three function values are selected as the champion candidates and the results are displayed on the display 13. For example, the singer singer _i , the singer singer _j , and the singer singer _k respectively have the largest function values W ^T X _i , W ^T X _j , W ^T X _k , and the display shows: singer singer _i , singer singer _j , singer singer _k is Champion candidate.

The method for predicting the result of the variety game proposed in this embodiment, by using the historical period of the variety game and the characteristics of the contestant to generate the feature pairing conversion result label, training the logistic regression model to obtain the prediction function, and finally substituting the current period of the variety game and each The characteristics of the contestants are predicted to function, determine the ranking of the contestants, and improve the accuracy of the predicted game results.

As shown in FIG. 4, it is a flowchart of the second embodiment of the method for predicting the result of the variety game of the present application.

In this embodiment, the variety game result prediction method includes: step S10 - step S70. The steps S10 to S40 are substantially the same as those in the first embodiment, and are not described herein again.

In step S50, it is determined whether the difference between the prediction function values of the two participating players of the current variety game is less than the second preset value. For example, the Chinese New Songs Variety Competition went to the stage where the singer Singer ₁ and the singer Singer ₂ decided to win the championship, and the second prediction unit 142 needed to predict the result of the match between the singer singer ₁ and the singer singer ₂ . The processor 12 extracts the current song held Chinese New period T, singer singer ₁ wherein X ₁ and feature information of a singer ₂ singer information X _2, and substituted into the prediction function S = f (x) = W T X in We were determined singer singer function value W ^T X _{1 1} and ₂ singer singer function value W ^T X _2. Finally singer singer function value W ^T X _{1 1} and ₂ singer singer function value W ^T X ₂ as a difference, the difference is determined and the prediction function values of the two singers is less than a second predetermined value.

In step S60, if the difference between the function values of the two participating players is less than the second preset value, the winning rate of both participating players is 50%. For example, when the value of _a singer singer function W ^T X ₁ and function ₂ singer Singer W ^T X ₂ as a difference value, the difference is smaller than a second predetermined value, the display 13 shows "singer singer singer singer _{1 2} win rate the same".

Step S70, if the difference between the function values of the two participating players is greater than or equal to the second preset value, the contestant corresponding to the larger function value wins. For example, when the value of _a singer singer function W ^T X ₁ and function ₂ singer Singer W ^T X ₂ as a difference value, the difference is greater than a second predetermined value, assuming the singer singer function value W ^T X _{1 1} is greater than Singer The function value of singer ₂ is W ^T X ₂ , and the display 13 displays "Singer Singer ₁ wins".

Compared with the first embodiment, the prediction method of the variety game result proposed by the embodiment generates a feature pairing conversion result label according to the historical period of the variety game and the characteristics of the contestant, and then uses the feature pairing conversion result label to train the logistic regression model to obtain the prediction. The function can finally use the generated prediction function to predict the ranking of each contestant in the current variety game. You can also use the prediction function to predict the results of the two singers in the same stage in the process of the current variety game, and improve the accuracy of the game result prediction. And comprehensive.

In addition, the embodiment of the present application further provides a computer readable storage medium, which includes a variety game result prediction system 10, and the variety game result prediction system 10 is executed by a processor to:

Data extraction step: extracting historical period data of a variety game from a preset data source, and extracting characteristics of each contestant from the historical period data;

a conversion step: performing pairing conversion on characteristics of each contestant in the historical period to obtain a feature pairing conversion result label of a plurality of matched contestants;

Model training step: training the logistic regression model by using the feature pairing conversion result label of each paired contestant to determine the model coefficient and the prediction function of the model;

a first prediction step: substituting the current period of the variety game and the characteristics of each contestant in the current period into the prediction function, and determining the function value of each contestant according to the function value The order of the competition is to rank the results of each contestant.

Preferably, the expression of the prediction function is:

S=f(x)=W ^T X

Where T represents the period of the variety game, W ^T represents the model coefficient of the prediction function, X represents the characteristics of the contestant, and S represents the function value of the contestant.

Preferably, when the variety game result prediction system is executed by the processor, the following steps are also implemented:

a second prediction step: substituting the current period of the variety game and the characteristics of the two predetermined contestants into the prediction function, respectively, and predicting the result of the match between the two contestants;

Preferably, the second prediction step further comprises:

If the difference between the function values of the two competitors is less than the second preset value, it is predicted that the winning rates of the two participating players are the same;

If the difference between the function values of the two competitors is greater than or equal to the second preset value, the competitor corresponding to the larger function value is predicted to win.

The specific implementation manner of the computer readable storage medium of the present application is substantially the same as the specific implementation manner of the above-mentioned variety game result prediction method, and details are not described herein again.

The serial numbers of the embodiments of the present application are merely for the description, and do not represent the advantages and disadvantages of the embodiments.

Through the description of the above embodiments, those skilled in the art can clearly understand that the foregoing embodiment method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be through hardware, but in many cases, the former is better. Implementation. Based on such understanding, the technical solution of the present application, which is essential or contributes to the prior art, may be embodied in the form of a software product stored in a storage medium (such as ROM/RAM as described above). , a disk, an optical disk, including a number of instructions for causing a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) to perform the methods described in the various embodiments of the present application.

The above is only a preferred embodiment of the present application, and is not intended to limit the scope of the patent application, and the equivalent structure or equivalent process transformations made by the specification and the drawings of the present application, or directly or indirectly applied to other related technical fields. The same is included in the scope of patent protection of this application.

Claims (20)

A method for predicting results of a variety game, characterized in that the method comprises: Data extraction step: extracting historical period data of a variety game from a preset data source, and extracting characteristics of each contestant from the historical period data; a conversion step: pairing and converting the characteristics of each contestant in the historical period to obtain a result label generated by the feature pairing conversion of the plurality of matched contestants; Model training step: training the logistic regression model with the generated result tags to determine the model coefficients and the prediction function of the model; The first predicting step: substituting the current period of the variety game and the characteristics of each contestant in the current period into the prediction function, determining the function value of each contestant, and ranking the contestants according to the magnitude of the function value. The results are ranked. The variety game result prediction method according to claim 1, wherein the expression of the prediction function of the model is: S=f(x)=W ^T X Where T represents the period of the variety game, W ^T represents the model coefficient of the prediction function, X represents the characteristics of the contestant, and S represents the function value of the contestant. The method for predicting results of a variety game according to claim 1, wherein the model training step comprises: Data collection step: obtaining the historical period of the variety game and the feature matching conversion result label of the contestant; Detection step: the contestant's feature pairing conversion result label is substituted into the type model for training, and the loss function is used to detect the pros and cons of the prediction model; Judging step: using the empirical risk function to perform statistics on the value of the loss function, and determining whether the value of the empirical risk function is less than the first preset value; If the value of the empirical risk function is greater than or equal to the first preset value, it indicates that the logistic regression model is not the optimal model, adjusting the model coefficients, and returning to the training step until the empirical risk function value is less than the first preset value; If the empirical risk function value is less than the first preset value, it indicates that the logistic regression model is the optimal model. The method for predicting results of a variety game according to claim 1, wherein the method further comprises: The second prediction step: substituting the current period of the variety game and the characteristics of the two predetermined contestants into the prediction function respectively, and predicting the result of the match between the two contestants. The method of predicting a variety game result according to claim 4, wherein the second predicting step further comprises: If the difference between the function values of the two competitors is less than the second preset value, it is predicted that the winning rates of the two participating players are the same; If the difference between the function values of the two competitors is greater than or equal to the second preset value, the competitor corresponding to the larger function value is predicted to win. An electronic device, comprising: a memory, a processor, and a variety game result prediction system stored on the memory and operable on the processor, wherein the variety game result prediction system is The processor executes, and the following steps can be implemented: Data extraction step: extracting historical period data of a variety game from a preset data source, and extracting characteristics of each contestant from the historical period data; a conversion step: pairing and converting the characteristics of each contestant in the historical period to obtain a result label generated by the feature pairing conversion of the plurality of matched contestants; Model training step: training the logistic regression model with the generated result tags to determine the model coefficients and the prediction function of the model; The first predicting step: substituting the current period of the variety game and the characteristics of each contestant in the current period into the prediction function, determining the function value of each contestant, and ranking the contestants according to the magnitude of the function value. The results are ranked. The electronic device according to claim 6, wherein the expression of the prediction function is: S=f(x)=W ^T X Where T represents the period of the variety game, W ^T represents the model coefficient of the prediction function, X represents the characteristics of the contestant, and S represents the function value of the contestant. The electronic device according to claim 6, wherein the model training step comprises: Data collection step: obtaining the historical period of the variety game and the feature matching conversion result label of the contestant; Detection step: the contestant's feature pairing conversion result label is substituted into the type model for training, and the loss function is used to detect the pros and cons of the prediction model; Judging step: using the empirical risk function to perform statistics on the value of the loss function, and determining whether the value of the empirical risk function is less than the first preset value; If the value of the empirical risk function is greater than or equal to the first preset value, it indicates that the logistic regression model is not the optimal model, adjusting the model coefficients, and returning to the training step until the empirical risk function value is less than the first preset value; If the empirical risk function value is less than the first preset value, it indicates that the logistic regression model is the optimal model. The electronic device according to claim 6, wherein the variety game result prediction system is executed by the processor, and the following steps are further implemented: The second prediction step: substituting the current period of the variety game and the characteristics of the two predetermined contestants into the prediction function respectively, and predicting the result of the match between the two contestants. The electronic device according to claim 9, wherein the second predicting step further comprises: if the difference between the function values of the two participating players is less than the second preset value, predicting the winning of the two participating players The same rate; If the difference between the function values of the two competitors is greater than or equal to the second preset value, the competitor corresponding to the larger function value is predicted to win. A computer readable storage medium, comprising: a variety game result prediction system, wherein when the system game result prediction system is executed by a processor, the following steps can be implemented: Data extraction step: extracting historical period data of a variety game from a preset data source, and extracting characteristics of each contestant from the historical period data; a conversion step: pairing and converting the characteristics of each contestant in the historical period to obtain a result label generated by the feature pairing conversion of the plurality of matched contestants; Model training step: training the logistic regression model with the generated result tags to determine the model coefficients and the prediction function of the model; The first predicting step: substituting the current period of the variety game and the characteristics of each contestant in the current period into the prediction function, determining the function value of each contestant, and ranking the contestants according to the magnitude of the function value. The results are ranked. The computer readable storage medium of claim 11 wherein the expression of the predictive function of the model is: S=f(x)=W ^T X Where T represents the period of the variety game, W ^T represents the model coefficient of the prediction function, X represents the characteristics of the contestant, and S represents the function value of the contestant. The computer readable storage medium of claim 11, wherein the model training step comprises: Data collection step: obtaining the historical period of the variety game and the feature matching conversion result label of the contestant; Detection step: the contestant's feature pairing conversion result label is substituted into the type model for training, and the loss function is used to detect the pros and cons of the prediction model; Judging step: using the empirical risk function to perform statistics on the value of the loss function, and determining whether the value of the empirical risk function is less than the first preset value; If the value of the empirical risk function is greater than or equal to the first preset value, it indicates that the logistic regression model is not the optimal model, adjusting the model coefficients, and returning to the training step until the empirical risk function value is less than the first preset value; If the empirical risk function value is less than the first preset value, it indicates that the logistic regression model is the optimal model. The computer readable storage medium according to claim 11, wherein when the unified variety game result prediction system is executed by the processor, the method further implements: The second prediction step: substituting the current period of the variety game and the characteristics of the two predetermined contestants into the prediction function respectively, and predicting the result of the match between the two contestants. The computer readable storage medium of claim 14, wherein the second predicting step further comprises: If the difference between the function values of the two competitors is less than the second preset value, it is predicted that the winning rates of the two participating players are the same; If the difference between the function values of the two competitors is greater than or equal to the second preset value, the competitor corresponding to the larger function value is predicted to win. A variety game result prediction system, characterized in that the system result prediction system can be executed by a processor to implement the following steps: Data extraction step: extracting historical period data of a variety game from a preset data source, and extracting characteristics of each contestant from the historical period data; a conversion step: pairing and converting the characteristics of each contestant in the historical period to obtain a result label generated by the feature pairing conversion of the plurality of matched contestants; Model training step: training the logistic regression model with the generated result tags to determine the model coefficients and the prediction function of the model; The first predicting step: substituting the current period of the variety game and the characteristics of each contestant in the current period into the prediction function, determining the function value of each contestant, and ranking the contestants according to the magnitude of the function value. The results are ranked. The variety game result prediction system according to claim 16, wherein the expression of the prediction function of the model is: S=f(x)=W ^T X Where T represents the period of the variety game, W ^T represents the model coefficient of the prediction function, X represents the characteristics of the contestant, and S represents the function value of the contestant. The variety game result prediction system according to claim 16, wherein the model training step comprises: Data collection step: obtaining the historical period of the variety game and the feature matching conversion result label of the contestant; Detection step: the contestant's feature pairing conversion result label is substituted into the type model for training, and the loss function is used to detect the pros and cons of the prediction model; Judgment step: use the empirical risk function to count the value of the loss function to determine the empirical risk Whether the function value is less than the first preset value; If the value of the empirical risk function is greater than or equal to the first preset value, it indicates that the logistic regression model is not the optimal model, adjusting the model coefficients, and returning to the training step until the empirical risk function value is less than the first preset value; If the empirical risk function value is less than the first preset value, it indicates that the logistic regression model is the optimal model. The variety game result prediction system according to claim 16, wherein when the system result prediction system is executed by the processor, it is also possible to: The second prediction step: substituting the current period of the variety game and the characteristics of the two predetermined contestants into the prediction function respectively, and predicting the result of the match between the two contestants. The variety game result prediction system according to claim 19, wherein the second prediction step further comprises: If the difference between the function values of the two competitors is less than the second preset value, it is predicted that the winning rates of the two participating players are the same; If the difference between the function values of the two competitors is greater than or equal to the second preset value, the competitor corresponding to the larger function value is predicted to win.

Images