CN105956873A - Game data intelligent analysis method - Google Patents

Abstract

The invention discloses a game data intelligent analysis method. The game data intelligent analysis method comprises the following steps of: receiving, in real time, data fed back from a client and caching the data in a Redis first-in first-out queue; extracting data from the Redis first-in first-out queue; validating the extracted data; according to a preset statistical target, deduplicating the data; counting each user in the data as a vector, and carrying out difference statistical processing of all users' individual information; extracting user behavior characteristics, carrying out difference detection and calculating by using a mutual information entropy; calculating a joint entropy of two events; calculating mutual information MI; dividing a behavior difference value by the MI to obtain a new difference value; calculating to obtain an entropy HIi of the i-th behavior and an entropy of the i-1th behavior, solving a joint distribution rate of the two entropys, and solving an MIi Value to obtain a difference value Di; analyzing the difference value Di and including the analysis results into a statistical result database. The game data intelligent analysis method of the invention makes it possible to conduct fast analysis and processing of data and achieves high practicability.

Description

A kind of intelligent analysis method of game data

Technical field

The present invention relates to data analysis field, a kind of intelligent analysis method of game data.

Background technology

Data Analysis refers to be analyzed collecting the mass data come with suitable statistical analysis technique, carries Take useful information and form conclusion and to data in addition research and the process of summary in detail.This process is also quality management The support process of system.In practicality, data analysis can help people to judge, in order to takes appropriate action.Data analysis Fundamentals of Mathematics the most establish in early days in 20th century, but until the appearance of computer just makes practical operation be possibly realized, and make Obtain data analysis to be promoted.Data analysis is the product that mathematical and computer sciences combines.

Data analysis is widely used in computer system, such as in game operation field, by carrying out game data Data analysis, it is possible to grasp the status information of whole game player, as logged in, online amount and wastage etc., thus according to object for appreciation The status information of family, it is possible to planning and management to game provide advisory opinion.The analytical cycle of existing data analysing method Longer, poor in timeliness, gradually it is difficult to meet the turn of the market maked rapid progress.

Summary of the invention

It is an object of the invention to provide a kind of intelligent analysis method of game data, to solve in above-mentioned background technology The problem proposed.

For achieving the above object, the present invention provides following technical scheme:

A kind of intelligent analysis method of game data, step is as follows:

1) data of real-time reception client feedback, and by data buffer storage in Redis fifo queue, wherein client The data of end feedback have pre-defined data form；

2) from Redis fifo queue, data are extracted；

3) data extracted being carried out legitimate verification, if verifying legal, then carrying out step 4), otherwise, abandon this data, Return step 2)；

4) according to the statistics target preset, data are carried out duplicate removal；

5) being added up by each user's individuality in data is a vector, and the information that all users are individual is carried out Variant statistical Process:

$Fd=\underset{x=1}{\overset{H}{\Σ}}\underset{y=1}{\overset{W}{\Σ}}|f_1(x,y)-f_2(x,y)|/HW$

Wherein H is game behavior quantity, and W is user's sample size, and (x is y) user's behavior every day vector, with this, counts f Calculate the mean difference numerical value of user's behavior on the two, in like manner individual consumer is individually added up, calculate user and playing Game behavior fluctuation situation in journey, to set up the user behavior curve of cyclical fluctuations；

6) extract user behavior feature and Difference test, use Mutual information entropy to calculate:

$H\left(x\right)=-\underset{x\∈A\left(X\right)}{\Σ}{P}_X\left(x\right){\mathrm{logP}}_X\left(x\right)$

Wherein, now the unit of entropy is behavior number；

7) combination entropy of two events of calculating:

$H(x,y)=-\underset{x\∈A\left(X\right),y\∈A\left(y\right)}{\Σ}{P}_{xy}(x,y){\mathrm{logP}}_{xy}(x,y);$

8) mutual information MI is calculated:

MI (x, y)=H (x)+H (y)-H (x, y)；

9) with MI, behavior difference being business, obtaining new difference is:

$D_i=\underset{j=0}{\overset{N}{\Σ}}\left(H_i\right(j)-H_i-(j-1\left)\right)/MI$

Wherein, N is the number of the time interval divided, and Hi (j) is the probability in i-th behavior generation jth day；

10) the entropy HIi calculating i-th behavior is:

${\mathrm{HI}}_i=\underset{j=1}{\overset{N}{\Σ}}-H_i\left(j\right){\mathrm{logH}}_i\left(j\right)$

In like manner, calculate the entropy of the i-th-1 behavior, and seek the two Joint Distribution rate, obtain the value of MIi, with poor Value Di；

11) difference Di is resolved, and analysis result is counted in statistical result data base.

As the further scheme of the present invention: described data form is JSON data form, the game that described packet contains Behavioural information is logon data, login IP, hour of log-on, data generation time, data content.

As the present invention further scheme: step 2) in legitimate verification for check data whether meet pre-fixing Formula, and data are the most complete.

As the present invention further scheme: step 4) according to preset statistics target judge data the need of Duplicate removal, if desired, then call Redis interface and carry out duplicate removal, if need not, then enters next step.

Compared with prior art, the invention has the beneficial effects as follows:

Data can quickly be analyzed and processed by the present invention, to obtain the Behavioral change data of user, contributes to game Operator makes adjustment in time, and to cater to the hobby of user, beneficially gaming operators holds market.

Detailed description of the invention

Below in conjunction with detailed description of the invention, technical scheme is described in more detail.

A kind of intelligent analysis method of game data, step is as follows:

1) data of real-time reception client feedback, and by data buffer storage in Redis fifo queue, wherein client The data of end feedback have pre-defined data form, in the present embodiment, it is preferred that described data form is JSON data lattice Formula, described packet produces the game behavioural information such as time, data content containing logon data, login IP, hour of log-on, data；

2) from Redis fifo queue, data are extracted；

3) data extracted being carried out legitimate verification, if verifying legal, then carrying out step 4), otherwise, abandon this data, Return step 2), wherein legitimate verification is for checking whether data meet predetermined format, and data are the most complete, for not being inconsistent Close predetermined format and incomplete data the most do not possess researching value, therefore should give up；

4) according to the statistics target preset, data are carried out duplicate removal, whether judges data according to default statistics target Need duplicate removal, if desired, then call Redis interface and carry out duplicate removal, if need not, then enter next step；

5) being added up by each user's individuality in data is a vector, and the information that all users are individual is carried out Variant statistical Process；

Wherein H is game behavior quantity, and W is user's sample Quantity, (x is y) user's behavior every day vector, with this, calculates the mean difference numerical value of user's behavior on the two, in like manner to individual f Body user individually add up, and calculates user's game behavior fluctuation situation in game process, to set up user behavior The curve of cyclical fluctuations；

6) extract user behavior feature and Difference test, use Mutual information entropy to calculate:

Wherein, now the unit of entropy is behavior number；

7) combination entropy of two events of calculating:

$H(x,y)=-\underset{x\∈A\left(X\right),y\∈A\left(y\right)}{\Σ}{P}_{xy}(x,y)\log P_{xy}(x,y);$

8) mutual information MI is calculated:

MI (x, y)=H (x)+H (y)-H (x, y)；

9) with MI, behavior difference being business, obtaining new difference is:

Wherein, N is the number of the time interval divided, Hi (j) For the probability in i-th behavior generation jth day；

10) the entropy HIi calculating i-th behavior is:

In like manner, calculate the entropy of the i-th-1 behavior, and ask the two to combine Distributive law, obtains the value of MIi, thus obtains difference Di, and the reaction of described difference Di is that user is for behavior between two events Diversity between feature and different behavior, thus two things when the change of monitoring user behavior and user play intuitively Correlation degree between part；

11) difference Di is resolved, and analysis result is counted in statistical result data base.

Data can quickly be analyzed and processed by the present invention, to obtain the Behavioral change data of user, contributes to game Operator makes adjustment in time, to cater to the hobby of user.

Above the better embodiment of the present invention is explained in detail, but the present invention is not limited to above-mentioned embodiment party Formula, in the ken that one skilled in the relevant art is possessed, it is also possible on the premise of without departing from present inventive concept Various changes can be made.

Claims (4)

1. the intelligent analysis method of a game data, it is characterised in that step is as follows:

1) data of real-time reception client feedback, and by data buffer storage in Redis fifo queue, wherein client is returned The data of feedback have pre-defined data form；

2) from Redis fifo queue, data are extracted；

3) data extracted being carried out legitimate verification, if verifying legal, then carrying out step 4), otherwise, abandon this data, return Step 2)；

4) according to the statistics target preset, data are carried out duplicate removal；

5) being added up by each user's individuality in data is a vector, carries out the information that all users are individual at Variant statistical Reason:

$Fd=\underset{x=1}{\overset{H}{\Σ}}\underset{y=1}{\overset{W}{\Σ}}|f_1(x,y)-f_2(x,y)|/HW$

Wherein H is game behavior quantity, and W is user's sample size, and (x is y) user's behavior every day vector, with this, calculates f The mean difference numerical value of user's behavior on the two, in like manner individually adds up individual consumer, calculates user in game process Game behavior fluctuation situation, to set up the user behavior curve of cyclical fluctuations；

6) extract user behavior feature and Difference test, use Mutual information entropy to calculate:

$H\left(x\right)=-\underset{x\∈A\left(X\right)}{\Σ}{P}_X\left(x\right){\mathrm{logP}}_X\left(x\right)$

Wherein, now the unit of entropy is behavior number；

7) combination entropy of two events of calculating:

$H(x,y)=-\underset{x\∈A\left(X\right),y\∈A\left(y\right)}{\Σ}{P}_{xy}(x,y){\mathrm{logP}}_{xy}(x,y);$

8) mutual information MI is calculated:

MI (x, y)=H (x)+H (y)-H (x, y)；

9) with MI, behavior difference being business, obtaining new difference is:

$D_i=\underset{j=0}{\overset{N}{\Σ}}\left(H_i\right(j)-H_i-(j-1\left)\right)/MI$

Wherein, N is the number of the time interval divided, and Hi (j) is the probability in i-th behavior generation jth day；

10) the entropy HIi calculating i-th behavior is:

${\mathrm{HI}}_i=\underset{j=1}{\overset{N}{\Σ}}-H_i\left(j\right){\mathrm{logH}}_i\left(j\right)$

In like manner, calculate the entropy of the i-th-1 behavior, and seek the two Joint Distribution rate, obtain the value of MIi, to obtain difference Di；

11) difference Di is resolved, and analysis result is counted in statistical result data base.

The intelligent analysis method of game data the most according to claim 1, it is characterised in that described data form is JSON data form, when the game behavioural information that described packet contains is logon data, login IP, hour of log-on, data generation Between, data content.

The intelligent analysis method of game data the most according to claim 1 and 2, it is characterised in that step 2) in legal Property be verified as checking whether data meet predetermined format, and data are the most complete.

The intelligent analysis method of game data the most according to claim 3, it is characterised in that step 4) middle according to presetting Statistics target judge that data are the need of duplicate removal, if desired, then call Redis interface and carry out duplicate removal, if need not, then enter Enter next step.