CN116531763A - Game data monitoring system and method - Google Patents

Abstract

The invention discloses a game data monitoring system and a game data monitoring method, wherein the game data monitoring system comprises a data acquisition module, a data correction module, a data analysis module, a central control module and a monitoring execution module, wherein the data acquisition module is placed into a game client database through a port, the data correction module builds a data correction model to correct abnormal data information to generate a correction result and a relative error, the data analysis module calculates the correction result to acquire reporting data needing to be monitored, and the central control module builds a data change rate prediction model to output a prediction result to generate corresponding control instructions.

Description

Game data monitoring system and method

Technical Field

The invention relates to the technical field of data analysis, in particular to a game data monitoring system and a game data monitoring method.

Background

With the development of the internet, the game becomes one of life entertainment projects necessary for current young people, the game contains multiple types of data, wherein operation data are closely related to the vitality of the game, player behavior data in the game, such as player login, chatting, operation and the like, are recorded in a log form, the operation data analyze and acquire the player behavior data, but the data volume of the player behavior data is huge, in the process of acquiring the operation data, the data of the player behavior data are abnormally increased gradually, the main reasons for generating the data abnormality are faults in the data acquisition and transmission processes, unreasonable data parameter setting caused by the cheating of the player, the factors such as interference of external conditions such as a network or equipment and the like on the data, the abnormal data influence the normal acquisition of the operation data, and wrong operation data can lead a game manager to not realize accurate monitoring of the game data, not manage the game in time, influence player experience and cause player loss.

Disclosure of Invention

The invention aims to provide a game data monitoring system and a game data monitoring method, which are used for solving the problems in the background technology.

In order to achieve the above purpose, the present invention provides the following technical solutions:

the game data monitoring system comprises a data acquisition module, a data correction module, a data analysis module, a central control module and a monitoring execution module, wherein the data acquisition module is placed in a game client database through a port, the data correction module is connected with the data acquisition module, the data analysis module is connected with the data correction module, and the central control module is connected with the data acquisition module, the data analysis module and the monitoring execution module;

the data acquisition module is used for acquiring unprocessed database information from the game client;

the data correction module is used for manufacturing a data change curve of the database information, generating a clustering result through processing the database information by secondary clustering, training and extracting a characteristic curve for the clustering result, identifying abnormal data information, constructing a data correction model for correcting the abnormal data information, and generating a correction result and a relative error;

the data analysis module calculates the correction result, acquires the report data to be monitored and reports the report data to the central control module;

the central control module makes a time trend graph, takes reported data of redundant variables as input variables, establishes a data change rate prediction model, outputs a prediction result, and generates a corresponding control instruction through the prediction result;

and the monitoring execution module sends an execution signal to the operation management end according to the control instruction and generates execution feedback.

Further, the data correction module mainly comprises the following five processing flows:

(1) clustering the data change curve once by using a fuzzy C-means algorithm to obtain a clustering result generated after database information processing;

(2) acquiring all data classification validity index V using fuzzy clustering validity index function _xb Obtaining the optimal clustering quantity according to the minimum data of the number of the clusters;

(3) taking the clustering result of the minimum intra-class distance and the optimal clustering center as a benchmark, using a radial basis neural network to realize secondary clustering of the data change curve, and obtaining various data characteristic curves;

(4) the base utilizes the smoothness and similarity characteristics and determines a bidirectional detection threshold value through a normal distribution theory, and identifies abnormal operation data of the power grid;

(5) and constructing a data correction model, generating a normal random number and normal random correction data on the basis of a model expression, obtaining a final correction result, and comparing the final correction result with abnormal data to obtain a relative error.

Further, the specific generation steps of the optimal cluster number are as follows:

s1 sets x= { X ₁ ,x ₂ ,…,x _n The data sample is represented, and the clustering center and the fuzzy classification matrix respectively adopt C= [ C ] ₁ ,c ₂ ,…,c _c ]、A＝[a _ij ] _c×n Description;

s2, determining a value c, setting a weighted index and iteration times as m and l respectively, and initializing the A;

s3, calculating a clustering center C of all samples by using a formula, and updating a fuzzy classification matrix A;

s4, designating judgment accuracy E and enabling the judgment accuracy E>0, set up ||A ^l+1 -A ^l If the I is smaller than the judgment precision E, the iteration termination condition is not met, and if the I=l+1 is not met, the clustering centers C of all samples are continuously calculated;

s5, judging the values of the weighted indexes m and c in advance, and introducing Xie-Beni effectiveness index V _xb And calculate the effectiveness index V _xb The smaller the value of (c) is, the more excellent the clustering result is, and the optimal number of clusters is generated.

Further, the expression of the fuzzy C-means algorithm is as follows:

wherein, the number of samples and clusters is described by n and c respectively; the weighting index is described by m; d for Euclidean distance _ij Description; the membership of data point j to cluster center i is a _ij Description.

Further, in the fuzzy C-means algorithm, λ= [ λ ] is used ₁ ，…，λ _n ]The extremum of the operator implementation limiting condition is changed into an unlimited condition, and the specific description is as follows:

and (3) calculating the derivative of all the input parameters, and when the minimum value can be obtained, calculating the clustering center C and updating the fuzzy classification matrix A by using the following specific formulas:

and the result is a minimized objective function.

Further, the effectiveness index V introduced in the fuzzy clustering effectiveness index function _xb The specific expression of (2) is as follows:

wherein, the blurring factor is described by m; the cluster centers and the matrix of membership are described by V, U, respectively.

Further, the formula for outputting the secondary clustering of the data change curve by the radial basis function neural network is as follows:

wherein: w for connecting weight vectors _ik Description; the number of output nodes is described by s; r for radial basis function _i (x) Describing, a gaussian function is typically chosen.

Further, the data correction model uses the abnormal data obtained by comprehensive cloud correction identification to search for an abnormal rule through a historical data set and obtain a time a _t Corresponding abnormal data b _t { A for correction rule set ₁ →B ₁ ，A ₂ →B ₂ ，…，A _l →B _l By parameter A ₁ (E _x1 ，E _n1 ，E _e1 )，…A _l (E _xl ，E _nl ，E _el ) Creating a cloud generator, and a _t As input to the cloud generator, the maximum value of the search output results,obtaining a _t Membership A _i Obtain the best correction rule A _i →B _i History correction cloud B as correction information at this time _i (E _xb ，E _nb ，E _eb ) At the same time a by the current time dataset _t Setting corresponding abnormal data L= { L (i, a) |i=1, 2, …, n }, taking the abnormal data L needing to be corrected as a current time data set, and acquiring a current correction cloud L by using a reverse operation algorithm _i Finally, comprehensive history correction cloud B _i And the current correction cloud L _i Generating a data correction model S (E _x ，E _n ，H _e ) The specific expression is:

further, the data change rate prediction model is constructed by using a support vector regression machine, the operation data change rate model adopts a principal component analysis method, the dimension of the acquired report data, which is affected by interference, is reduced, redundant variables are removed, normalization processing is needed before the report data is used as input variables, adverse effects of singular sample data are eliminated, super parameters in the data change rate prediction model constructed by using the support vector regression machine are selected, an artificial bee colony algorithm is adopted to optimize and select the super parameters, a training sample set and a test sample set are divided by randomly grouping the report data which is used as the input variables, the training data change rate prediction model is realized by inputting the training sample set, the test sample set is applied to evaluate the model performance, the average absolute error, the mean square error and the decision coefficient are used for judging the measurement model prediction precision, and the data change rate prediction model construction is completed.

Further, the mean absolute error MAE, the mean square error RMSE and the determination coefficient R ² Is defined as:

where n represents the number of test samples,and->The actual, predicted and average values of the corrosion rate are shown, respectively.

Further, in the specific process of constructing the data change rate prediction model by the support vector regression mechanism, the following steps are: setting x _i To input variable, y _i R is the output variable ⁿ For an n-dimensional input space, R is the output space, m is the number of samples, then assume the sample set is:

D＝{(x _i ，y _i )|x _i ∈R ⁿ ，y _i ∈R，i＝1，2，…，m}

the regression prediction function of the vector regression machine is as follows:

wherein ω is a weight vector;is a nonlinear mapping function; b is a bias vector;

introducing a relaxation variable constraint θ, θ ^* And Lagrangian function constraint, obtaining a regression prediction function of the final vector regression machine:

wherein alpha is _i ，Is Lagrangian multiplier, K (x _i ，y _i ) Is a kernel function of the vector regression machine.

Further, the monitoring method of the data monitoring system comprises the following steps:

step one, accessing a data acquisition module into a game client, copying a player behavior record log in a database in real time, and storing the record log as database information;

step two, transmitting the database information to a data correction module to perform primary clustering on the data change curve by using a fuzzy C-means algorithm, obtaining a clustering result generated after the database information is processed, and obtaining all data classification effectiveness indexes V by using a fuzzy clustering effectiveness index function _xb Obtaining the optimal clustering quantity, taking the clustering result of the minimum intra-class distance and the optimal clustering center as a benchmark, using a radial basis neural network to realize secondary clustering of the data change curve, obtaining various data characteristic curves, determining a two-way detection threshold value by using smoothness and similarity characteristics and through a normal distribution theory, identifying abnormal power grid operation data, constructing a data correction model, generating normal random numbers and normal random correction data on the basis of a model expression, obtaining a final correction result, and comparing the final correction result with the abnormal data to obtain relative errors.

Step three, the data analysis module obtains the final correction result and the relative error, calculates the reported data to be monitored based on the final correction result and the relative error, and sends the reported data to the central control module after the calculation is completed;

step four, the central control module receives the reported data and makes a time trend graph, then adopts a principal component analysis method to reduce the dimension of the interference influence factor of the obtained reported data, eliminates redundant variables, normalizes the reported data to eliminate the adverse effect of singular sample data, selects super parameters in a data change rate prediction model built by a support vector regression mechanism, adopts an artificial bee colony algorithm to optimize and select the super parameters, and realizes the construction of a data change rate prediction model frame;

fifthly, randomly sequencing reported data serving as input variables, taking the first 80% of sample data as a training sample set and the remaining 20% of sample data as a test sample set, realizing a training data change rate prediction model by inputting the training sample set, evaluating model performance by applying the test sample set, judging and measuring model prediction precision by using an average absolute error, a mean square error and a decision coefficient, completing data change rate prediction model construction, and outputting a prediction result;

step six, obtaining the value of the predicted result P, and establishing a sample set, and recording as { P } ₁ ，P ₂ ，···，P _m M is the total number of net load values of the power grid with negative values, and m is a natural number less than or equal to n;

step seven, obtaining the mean value and standard deviation in the sample set, normalizing the data by using the mean value and standard deviation, and using the dataAdjusting the standard parameters to [0,1 ]]Setting the extremum of the control command transmission interval as f (x) ₁ Namely, the sending mechanism of the control instruction is as follows:

when f (x) min is less than or equal to f (x)<f(x) ₁ Generating and sending a control instruction;

when f (x) ₁ ≤f(x)<f (x) max, no control instruction is generated;

and step eight, the monitoring execution module informs game management personnel according to the control instruction, reminds the management personnel to timely manage the game, sends execution feedback information to the central control module, and the central control module reconstructs a data change rate prediction model to perform secondary prediction by taking the execution feedback information receiving time point as the start point so as to realize recheck monitoring.

Compared with the prior art, the method has the advantages that the data correction module corrects the player behavior data by identifying the abnormal data of the player behavior data, establishes a data correction model, calculates the operation data according to the corrected player behavior data, reduces the abnormal data generated by the influence of the abnormal data of the player behavior data, ensures the accuracy of the operation data during monitoring, avoids meaningless monitoring behaviors of the abnormal data monitoring, and predicts the change rate of the operation data by training and simulating the operation data calculated after correction, builds a data change rate prediction model, can judge the descending trend of the operation data in advance by predicting the operation data, and sends a control instruction, so that a manager can conveniently manage games in time, and the player experience is improved.

Drawings

In order to more clearly illustrate the technical solutions of the present invention, the drawings that are needed in the technical description of the present invention will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 is a schematic diagram of the overall structure of the present invention.

Detailed Description

The present invention will be further described with reference to the following embodiments, and all other embodiments obtained by those skilled in the art without making any inventive effort are intended to fall within the scope of the present invention.

Example 1

Referring to fig. 1, the invention provides a game data monitoring system, which comprises a data acquisition module, a data correction module, a data analysis module, a central control module and a monitoring execution module, wherein the data acquisition module is placed into a game client database through a port, the data correction module is connected with the data acquisition module, the data analysis module is connected with the data correction module, and the central control module is connected with the data acquisition module, the data analysis module and the monitoring execution module;

the data acquisition module is used for acquiring unprocessed database information from the game client;

the data correction module is used for manufacturing a data change curve of the database information, generating a clustering result through processing the database information by secondary clustering, training and extracting a characteristic curve for the clustering result, identifying abnormal data information, constructing a data correction model for correcting the abnormal data information, and generating a correction result and a relative error;

the data analysis module calculates the correction result, acquires the report data to be monitored and reports the report data to the central control module;

the central control module makes a time trend graph, takes reported data of redundant variables as input variables, establishes a data change rate prediction model, outputs a prediction result, and generates a corresponding control instruction through the prediction result;

and the monitoring execution module sends an execution signal to the operation management end according to the control instruction and generates execution feedback.

By adopting the technical scheme, the game data monitoring method comprises the following steps:

step one, accessing a data acquisition module into a game client, copying a player behavior record log in a database in real time, and storing player behavior data as database information;

step two, transmitting player behavior data to a data correction module to perform primary clustering on the data change curve by using a fuzzy C-means algorithm, obtaining a clustering result generated after database information processing, and obtaining all data classification effectiveness indexes V by using a fuzzy clustering effectiveness index function _xb Obtaining optimal clustering quantity, taking the clustering result of the minimum intra-class distance and the optimal clustering center as a benchmark, using a radial basis neural network to realize secondary clustering of data change curves, obtaining various data characteristic curves, determining a bidirectional detection threshold value by using smoothness and similarity characteristics and through a normal distribution theory, identifying abnormal power grid operation data, constructing a data correction model, generating normal random numbers and normal random correction data on the basis of a model expression to obtain a final correction result, andand comparing the obtained error with the abnormal data to obtain a relative error.

Step three, the data analysis module obtains the final correction result and the relative error, calculates operation data as reporting data needing to be monitored based on the final correction result and the relative error, and sends the reporting data to the central control module after the calculation is completed;

step four, the central control module receives the reported data and makes a time trend graph, then adopts a principal component analysis method to reduce the dimension of the interference influence factor of the acquired operation data, eliminates redundant variables, normalizes the operation data to eliminate the adverse effect of singular sample data, selects super parameters in a data change rate prediction model constructed by a support vector regression mechanism, and adopts an artificial bee colony algorithm to optimize and select the super parameters so as to realize the construction of a data change rate prediction model frame;

fifthly, randomly sequencing operation data serving as input variables, taking the first 80% of sample data as a training sample set and the remaining 20% of sample data as a test sample set, realizing a training data change rate prediction model by inputting the training sample set, evaluating model performance by applying the test sample set, judging and measuring model prediction precision by using an average absolute error, a mean square error and a decision coefficient, completing data change rate prediction model construction, and outputting a prediction result of the operation data;

step six, obtaining the value of the predicted result P, and establishing a sample set { P } ₁ ，P ₂ ，···，P _m M is the total number of net load values of the power grid with negative values, and m is a natural number less than or equal to n;

step seven, obtaining the mean value and standard deviation in the sample set, normalizing the data by using the mean value and standard deviation, and using the dataAdjusting the standard parameters to [0,1 ]]Setting the extremum of the control command transmission interval as f (x) ₁ Namely, the sending mechanism of the control instruction is as follows:

when f (x) min is less than or equal to f (x)<f(x) ₁ Generating and sending a control instruction;

when f (x) ₁ ≤f(x)<f (x) max, no control instruction is generated;

and step eight, the monitoring execution module informs game management personnel according to the control instruction, reminds the management personnel to timely manage the game, sends execution feedback information to the central control module, and the central control module reconstructs a data change rate prediction model to perform secondary prediction by taking the execution feedback information receiving time point as the start point so as to realize recheck monitoring.

Example 2

Referring to fig. 1, the same parts as those in embodiment 1 in this embodiment are not repeated, and it should be noted that the specific process of the data correction module for correcting data is:

step one: clustering the data change curves once by using a fuzzy C-means algorithm, and setting X= { X ₁ ,x ₂ ,…,x _n The data sample is represented, and the clustering center and the fuzzy classification matrix respectively adopt C= [ C ] ₁ ,c ₂ ,…,c _c ]、A＝[a _ij ] _c×n Description; c, determining a value, setting a weighted index and iteration times as m and l respectively, and initializing the A; calculating a clustering center C of all samples by using a formula, and updating the fuzzy classification matrix A; designating judgment accuracy epsilon and making judgment accuracy epsilon>0, set up ||A ^l+1 -A ^l If the I is smaller than the judgment precision E, the iteration termination condition is not met, and if the I=l+1 is not met, the clustering centers C of all samples are continuously calculated; judging the values of the weighted indexes m and c in advance, and introducing Xie-Beni effectiveness index V _xb And calculate the effectiveness index V _xb The smaller the value of (2) is, the more excellent the clustering result is, and the optimal clustering quantity is generated;

based on the clustering results of the minimum intra-class distance and the optimal clustering center, the radial basis function neural network is used for realizing secondary clustering of the data change curves, various data characteristic curves are obtained, and the output formula for realizing the secondary clustering of the data change curves by the radial basis function neural network is as follows:

wherein: w for connecting weight vectors _ik Description; the number of output nodes is described by s; r for radial basis function _i (x) Describing, a gaussian function is typically chosen.

Step three, the base utilizes the smoothness and similarity characteristics and determines a bidirectional detection threshold value through a normal distribution theory, and identifies abnormal operation data of the power grid;

searching for abnormal rules through a historical data set, and obtaining a moment a _t Corresponding abnormal data b _t { A for correction rule set ₁ →B ₁ ，A ₂ →B ₂ ，…，A _l →B _l By parameter A ₁ (E _x1 ，E _n1 ，E _e1 )，…A _l (E _xl ，E _nl ，E _el ) Creating a cloud generator, and a _t As input to the cloud generator, searching for the maximum of the output results to obtain a _t Membership A _i Obtain the best correction rule A _i →B _i History correction cloud B as correction information at this time _i (E _xb ，E _nb ，E _eb ) At the same time a by the current time dataset _t Setting corresponding abnormal data L= { L (i, a) |i=1, 2, …, n }, taking the abnormal data L needing to be corrected as a current time data set, and acquiring a current correction cloud L by using a reverse operation algorithm _i Finally, comprehensive history correction cloud B _i And the current correction cloud L _i Generating a data correction model S (E _x ，E _n ，H _e ) The specific expression is:

generating normal random numbers and normal random correction data on the basis of the model expression to obtain a final correction result, and comparing the final correction result with abnormal data to obtain a relative error.

Specifically, the expression of the fuzzy C-means algorithm is:

wherein, the number of samples and clusters is described by n and c respectively; the weighting index is described by m; d for Euclidean distance _ij Description; the membership of data point j to cluster center i is a _ij Description.

Specifically, in the fuzzy C-means algorithm, λ= [ λ ] is used ₁ ，…，λ _n ]The extremum of the operator implementation limiting condition is changed into an unlimited condition, and the specific description is as follows:

and (3) calculating the derivative of all the input parameters, and when the minimum value can be obtained, calculating the clustering center C and updating the fuzzy classification matrix A by using the following specific formulas:

and the result is a minimized objective function.

Specifically, the effectiveness index V introduced in the fuzzy clustering effectiveness index function _xb The specific expression of (2) is as follows:

wherein, the blurring factor is described by m; the cluster centers and the matrix of membership are described by V, U, respectively.

Example 3

Referring to fig. 1, the same parts as those in embodiment 1 are not repeated in this embodiment, and it should be noted that the data change rate prediction model is constructed by using a support vector regression machine, and x is set _i To input variable, y _i R is the output variable ⁿ For an n-dimensional input space, R is the output space, m is the number of samples, then assume the sample set is:

D＝{(x _i ，y _i )|x _i ∈R ⁿ ，y _i ∈R，i＝1，2，…，m}

the regression prediction function of the vector regression machine is as follows:

wherein ω is a weight vector;is a nonlinear mapping function; b is a bias vector;

introducing a relaxation variable constraint θ, θ ^* And Lagrangian function constraint, obtaining a regression prediction function of the final vector regression machine:

wherein alpha is _i ，Is Lagrangian multiplier, K (x _i ，y _i ) Is a kernel function of the vector regression machine.

The operation data change rate model adopts a principal component analysis method, the dimension of the acquired report data, which is affected by interference, is reduced, redundant variables are removed, normalization processing is needed before the report data is used as an input variable, the adverse effect of singular sample data is eliminated, super parameters in a data change rate prediction model constructed by a support vector regression mechanism are selected, an artificial bee colony algorithm is adopted to optimize and select the super parameters, a training sample set and a test sample set are divided, the training data change rate prediction model is realized by inputting the training sample set, the test sample set is applied to evaluate the performance of the model, and the model prediction accuracy is judged by using average absolute error, mean square error and decision coefficient to complete the construction of the data change rate prediction model.

Specifically, the mean absolute error MAE, the mean square error RMSE and the determination coefficient R ² Is defined as:

wherein n represents the number of test samples, y ⁱ ，And->The actual, predicted and average values of the corrosion rate are shown, respectively.

The foregoing description is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and it should be understood that the technical scheme and the inventive concept according to the present invention are equivalent or changed within the scope of the present invention disclosed by the present invention by those skilled in the art.

Claims (8)

1. The utility model provides a monitored control system of recreation data, includes data acquisition module, data correction module, data analysis module, central control module, control execution module, in the data acquisition module put into the game client database through the port, data correction module with data acquisition module connects, data analysis module is connected with data correction module, central control module with data acquisition module data analysis module and control execution module all connects its characterized in that:

the data acquisition module is used for acquiring unprocessed database information from the game client;

the data correction module is used for manufacturing a data change curve of the database information, generating a clustering result through processing the database information by secondary clustering, training and extracting a characteristic curve for the clustering result, identifying abnormal data information, constructing a data correction model for correcting the abnormal data information, and generating a correction result and a relative error;

the data analysis module calculates the correction result, acquires the report data to be monitored and reports the report data to the central control module;

the central control module makes a time trend graph, takes reported data of redundant variables as input variables, establishes a data change rate prediction model, outputs a prediction result, and generates a corresponding control instruction through the prediction result;

and the monitoring execution module sends an execution signal to the operation management end according to the control instruction and generates execution feedback.

2. A game data monitoring system according to claim 1, wherein: the data correction module mainly comprises the following five processing flows:

(1) clustering the data change curve once by using a fuzzy C-means algorithm to obtain a clustering result generated after database information processing;

(2) acquiring all data classification validity index V using fuzzy clustering validity index function _xb Obtaining the optimal clustering number by the minimum data of the (4)An amount of;

(3) taking the clustering result of the minimum intra-class distance and the optimal clustering center as a benchmark, using a radial basis neural network to realize secondary clustering of the data change curve, and obtaining various data characteristic curves;

(4) the base utilizes the smoothness and similarity characteristics and determines a bidirectional detection threshold value through a normal distribution theory, and identifies abnormal operation data of the power grid;

(5) and constructing a data correction model, generating a normal random number and normal random correction data on the basis of a model expression, obtaining a final correction result, and comparing the final correction result with abnormal data to obtain a relative error.

3. A game data monitoring system according to claim 2, wherein: the specific generation steps of the optimal cluster number are as follows:

s1 sets x= { X ₁ ,x ₂ ,…,x _n The data sample is represented, and the clustering center and the fuzzy classification matrix respectively adopt C= [ C ] ₁ ,c ₂ ,…,c _c ]、A＝[α _ij ] _c×n Description;

s2, determining a value c, setting a weighted index and iteration times as m and l respectively, and initializing the A;

s3, calculating a clustering center C of all samples by using a formula, and updating a fuzzy classification matrix A;

s4, designating judgment accuracy E and enabling the judgment accuracy E>0, set up ||A ^l+1 -A ^l If the I is smaller than the judgment precision E, the iteration termination condition is not met, and if the I=l+1 is not met, the clustering centers C of all samples are continuously calculated;

s5, judging the values of the weighted indexes m and c in advance, and introducing Xie-Beni effectiveness index V _xb And calculate the effectiveness index V _xb The smaller the value of (c) is, the more excellent the clustering result is, and the optimal number of clusters is generated.

4. A game data monitoring system according to claim 2, wherein: the effectiveness index V introduced in the fuzzy clustering effectiveness index function _xb The specific expression of (2) is as follows:

wherein, the blurring factor is described by m; the cluster centers and the matrix of membership are described by V, U, respectively.

5. A game data monitoring system according to claim 2, wherein: the data correction model uses the abnormal data obtained by comprehensive cloud correction identification to search for an abnormal rule through a historical data set and obtain a moment a _t Corresponding abnormal data b _t { A for correction rule set ₁ →B ₁ ，A ₂ →B ₂ ，…，A _l →B _l By parameter A ₁ (E _x1 ，E _n1 ，E _e1 )，…A _l (E _xl ，E _nl ，E _el ) Creating a cloud generator, and a _t As input to the cloud generator, searching for the maximum of the output results to obtain a _t Membership A _i Obtain the best correction rule A _i →B _i History correction cloud B as correction information at this time _i (E _xb ，E _nb ，E _eb ) At the same time a by the current time dataset _t Setting corresponding abnormal data L= { L (i, a) |i=1, 2, …, n }, taking the abnormal data L needing to be corrected as a current time data set, and acquiring a current correction cloud L by using a reverse operation algorithm _i Finally, comprehensive history correction cloud B _i And the current correction cloud L _i A data correction model is generated.

6. A game data monitoring system according to claim 1, wherein: the data change rate prediction model is constructed by using a support vector regression machine, the operation data change rate model adopts a principal component analysis method, the dimension of the interference influence factor of the acquired report data is reduced, redundant variables are removed, normalization processing is needed before the report data is used as an input variable, adverse effects of singular sample data are eliminated, super parameters in the data change rate prediction model constructed by using the support vector regression machine are selected, an artificial bee colony algorithm is adopted to optimize and select the super parameters, the report data used as the input variable is randomly grouped, a training sample set and a test sample set are divided, the training data change rate prediction model is realized by inputting the training sample set, the test sample set is applied to evaluate the model performance, the average absolute error, the mean square error and the decision coefficient are used for judging the measurement model prediction precision, and the data change rate prediction model construction is completed.

7. The system for monitoring game data according to claim 6, wherein: in the specific process of constructing the data change rate prediction model by the support vector regression mechanism, the support vector regression mechanism comprises the following steps of: setting x _i To input variable, y _i R is the output variable ⁿ For an n-dimensional input space, R is the output space, m is the number of samples, then assume the sample set is:

D＝{(x _i ，y _i )|x _i ∈R ⁿ ，y _i ∈R，i＝1，2，…，m}

the regression prediction function of the vector regression machine is as follows:

wherein ω is a weight vector;is a nonlinear mapping function; b is a bias vector;

introducing a relaxation variable constraint θ, θ ^* And Lagrangian function constraint, obtaining a regression prediction function of the final vector regression machine:

wherein alpha is _i ，Is Lagrangian multiplier, K (x _i ，y _i ) Is a kernel function of the vector regression machine.

8. A method of monitoring game data according to any one of claims 1 to 7, wherein: the monitoring method of the data monitoring system comprises the following steps:

step one, accessing a data acquisition module into a game client, copying a player behavior record log in a database in real time, and storing the record log as database information;

step two, transmitting the database information to a data correction module to perform primary clustering on the data change curve by using a fuzzy C-means algorithm, obtaining a clustering result generated after the database information is processed, and obtaining all data classification effectiveness indexes V by using a fuzzy clustering effectiveness index function _xb Obtaining the optimal clustering quantity, taking the clustering result of the minimum intra-class distance and the optimal clustering center as a benchmark, using a radial basis neural network to realize secondary clustering of the data change curve, obtaining various data characteristic curves, determining a two-way detection threshold value by using smoothness and similarity characteristics and through a normal distribution theory, identifying abnormal power grid operation data, constructing a data correction model, generating normal random numbers and normal random correction data on the basis of a model expression, obtaining a final correction result, and comparing the final correction result with the abnormal data to obtain relative errors.

Step three, the data analysis module obtains the final correction result and the relative error, calculates the reported data to be monitored based on the final correction result and the relative error, and sends the reported data to the central control module after the calculation is completed;

step four, the central control module receives the reported data and makes a time trend graph, then adopts a principal component analysis method to reduce the dimension of the interference influence factor of the obtained reported data, eliminates redundant variables, normalizes the reported data to eliminate the adverse effect of singular sample data, selects super parameters in a data change rate prediction model built by a support vector regression mechanism, adopts an artificial bee colony algorithm to optimize and select the super parameters, and realizes the construction of a data change rate prediction model frame;

fifthly, randomly sequencing reported data serving as input variables, taking the first 80% of sample data as a training sample set and the remaining 20% of sample data as a test sample set, realizing a training data change rate prediction model by inputting the training sample set, evaluating model performance by applying the test sample set, judging and measuring model prediction precision by using an average absolute error, a mean square error and a decision coefficient, completing data change rate prediction model construction, and outputting a prediction result;

step six, obtaining the value of the predicted result P, and establishing a sample set, and recording as { P } ₁ ，P ₂ ，···，P _m M is the total number of net load values of the power grid with negative values, and m is a natural number less than or equal to n;

step seven, obtaining the mean value and standard deviation in the sample set, normalizing the data by using the mean value and standard deviation, and using the dataAdjusting the standard parameters to [0,1 ]]Setting the extremum of the control command transmission interval as f (x) ₁ Namely, the sending mechanism of the control instruction is as follows:

when f (x) min is less than or equal to f (x)<f(x) ₁ Generating and sending a control instruction;

when f (x) ₁ ≤f(x)<f (x) max, no control instruction is generated;

and step eight, the monitoring execution module informs game management personnel according to the control instruction, reminds the management personnel to timely manage the game, sends execution feedback information to the central control module, and the central control module reconstructs a data change rate prediction model to perform secondary prediction by taking the execution feedback information receiving time point as the start point so as to realize recheck monitoring.