CN106682053A - Optimization method and device for putting scheme of game top-up sales promotion - Google Patents
Optimization method and device for putting scheme of game top-up sales promotion Download PDFInfo
- Publication number
- CN106682053A CN106682053A CN201610096792.6A CN201610096792A CN106682053A CN 106682053 A CN106682053 A CN 106682053A CN 201610096792 A CN201610096792 A CN 201610096792A CN 106682053 A CN106682053 A CN 106682053A
- Authority
- CN
- China
- Prior art keywords
- money
- game
- advertising campaign
- supplemented
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0207—Discounts or incentives, e.g. coupons or rebates
- G06Q30/0209—Incentive being awarded or redeemed in connection with the playing of a video game
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- Finance (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Development Economics (AREA)
- Accounting & Taxation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Entrepreneurship & Innovation (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Probability & Statistics with Applications (AREA)
- Game Theory and Decision Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Economics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an optimization method and an optimization device for a putting scheme of a game top-up sales promotion. The method comprises the following steps of 1, clustering collected multiple sets of game top-up sales promotion data to respectively classify the sets of game top-up sales promotion data into N different top-up sales promotion types, wherein N is more than and equal to 2, each set of game top-up sales promotion data comprises a plurality of indexes of which each is used for indicating set top-up sales promotion items and activity time of the top-up sales promotion, and the data value of each index correspondingly represents whether the corresponding top-up sales promotion item occurs, or characters of the time period of the activity, or the time length of the activity; and S2, determining which indexes of the game top-up sales promotion data mainly influence the top-up limit of each top-up sales promotion type by using a multi-factor variance analysis method based on the game top-up sales promotion data under each top-up sales promotion type. According to the method and the device provided by the invention, design of the game top-up sales promotion can be effectively optimized.
Description
Technical field
The present invention relates to advertising campaign is supplemented in a kind of game with money throws in scheme optimization method and device.
Background technology
In the whole life process of game, supplement advertising campaign with money and suffer from very important effect.At the game initial stage, fill
Value advertising campaign is to increase the catalyst that game player retains wish, in gaming the phase, and it is to adjust game Jing to supplement advertising campaign with money
The effective means of Ji system, in the game later stage, supplements advertising campaign with money and is more to increase the requisite design of game income.
Current industry is supplemented the form of advertising campaign with money and is emerged in an endless stream, and Game Developer is constantly trying to various modes stimulates trip
Play player's supplements and consumes desire with money.In the face of current industry this it is omnifarious supplement advertising campaign form with money, game designer compels
Being essential will can provide the technical method for supplementing advertising campaign design guidance with money, help them when design games supplement advertising campaign with money
Can accomplish to shoot the arrow at the target, and the scheme that can solve the problems, such as to play when supplementing advertising campaign with money throws in project settings hardly possible.
Additionally, current industry is substantially carried out for the analysis that advertising campaign is supplemented in game with money after activity end, cause
Activities Design initial stage game designer cannot supplement advertising campaign with money to designed game and make desired value estimation, also just cannot be
Supplement with money and corresponding design logarithm value is made before advertising campaign is thrown in be adjusted, reach the optimum mesh of animation effect
's.If the Expected Results that advertising campaign is supplemented in a kind of designed game of technical method prediction with money can be provided, can preferably help
Help game designer that corresponding numerical value adjustment is done to done design.
The content of the invention
Present invention aims to the deficiencies in the prior art, there is provided it is excellent that advertising campaign input scheme is supplemented in one kind game with money
Change method, mainly solves the problems, such as to play and supplements the design hardly possible of advertising campaign presence with money.The present invention also further is solved for supplementing with money
The difficult problem of advertising campaign design initial stage prediction.
For achieving the above object, the present invention is employed the following technical solutions：
One kind game is supplemented advertising campaign with money and throws in scheme optimization method, is comprised the following steps：
S1. advertising campaign data are supplemented in the multigroup game to collecting with money, using clustering method, are supplemented each group game with money promotion and are lived
Dynamic data be included into respectively N kinds it is different supplement advertising campaign type, N >=2 with money；
Advertising campaign data are supplemented in wherein per group game with money includes multiple indexs, and each index is respectively used to represent filling for setting
Value advertising campaign item and supplement activity time of advertising campaign with money, the data value of each index accordingly represent whether occurs it is corresponding
The characteristics of time interval or the duration of activity of supplementing advertising campaign item or generation activity with money；
S2. advertising campaign data are supplemented with money based on every kind of game supplemented with money under advertising campaign type, using multifactor variance point
Analysis method, it is determined that affecting every kind of major influence factors for supplementing amount with money for supplementing advertising campaign type with money to be that advertising campaign is supplemented in game with money
Which index of data.
Further：
Step S1 is comprised the following steps：
A) supplement selection N group game in advertising campaign data with money from multigroup game using random algorithm and supplement advertising campaign with money
Data, as N number of initial cluster center；
B) calculate described multigroup other group game supplemented with money in advertising campaign data of playing one by one according to Euclidean distance to supplement with money
Advertising campaign data and the Euclidean distance of each cluster centre, supplement multigroup game with money advertising campaign data and are divided into N one by one
Class；
C) average of each index of advertising campaign data is supplemented in the game under calculating per class with money, and the average of each index is made
For new cluster centre；
D) respectively advertising campaign data are supplemented with money as one group with the game under all kinds of, calculates standard deviation, by itself and setting
The threshold comparison of group distance, group inner distance；
If standard deviation is less than the threshold value of setting, or the iterations for calculating is more than the maximum iteration time of setting
When, using this classification results as final classification result；Otherwise circulation step a)-d), until obtaining final classification result.
Step S2 is comprised the following steps：
Game under respectively to supplement advertising campaign type with money supplements advertising campaign data with money as group, quadratic sum, interaction in calculating group
Effect quadratic sum, sum of squares between groups, the group internal degree of freedom, the free degree between the interaction free degree and group；
Based on the quadratic sum and the free degree that are calculated, intra-class variance, interaction variance and between-group variance are calculated；
Based on the variance for being calculated, calculating is used as between the conspicuousness and each index of each index of influence factor
The conspicuousness of interaction factor；
Significance P value according to corresponding to calculated conspicuousness F value determines it, by significance P value and
The confidence level of setting is compared, and judges whether corresponding index is major influence factors according to comparative result；Or, based on institute
The free degree of calculating obtains the relevant parameter of the F values for needing inquiry, the F values for obtaining of being tabled look-up according to the confidence level of setting, then
By the conspicuousness of the interaction factor between the conspicuousness of each index for calculating and each index, with the F values ratio that obtains of tabling look-up
Compared with size, judge whether corresponding index is major influence factors according to comparative result.
It is further comprising the steps of：
S3. advertising campaign data set is supplemented with money using existing game, using the main affecting factors of step S2 determination as certainly
Variable, amount is supplemented with money as dependent variable using actual play, obtains supplementing active prediction with money by the method fitting of multiple linear regression
Model.
It is further comprising the steps of：
S4. supplement the game collected with money activity data input fitting prediction module, obtain supplementing predicted value with money.
It is further comprising the steps of：
S5. after active prediction model is supplemented in acquisition with money, according to supplementing predicted value and the corresponding phase for actually supplementing amount with money with money
Prediction effect is judged to error, if relative error is more than the relative error threshold value of setting, return to step S3, and it is pre- to adjust fitting
Regain after the relevant parameter of survey it is new supplement active prediction model with money, otherwise show that forecast model can use.
One kind game is supplemented advertising campaign with money and throws in scheme optimization device, including：
Cluster module, it supplements advertising campaign data with money to the multigroup game collected, and using clustering method, each group game is filled
Value advertising campaign data be included into respectively N kinds it is different supplement advertising campaign type, N >=2 with money；
Advertising campaign data are supplemented in wherein per group game with money includes multiple indexs, and each index is respectively used to represent filling for setting
Value advertising campaign item and supplement activity time of advertising campaign with money, the data value of each index accordingly represent whether occurs it is corresponding
The characteristics of time interval or the duration of activity of supplementing advertising campaign item or generation activity with money；
Influence factor determining module, it is based on every kind of game supplemented with money under advertising campaign type and supplements advertising campaign data with money,
Using multifactor analysis of variance method, it is determined that affect it is every kind of supplement advertising campaign type with money the major influence factors for supplementing amount with money be
Which index of advertising campaign data is supplemented in game with money.
The cluster module includes：
Initial cluster center selecting module, it is supplemented with money in advertising campaign data from multigroup game using random algorithm and is selected
Select N group game and supplement advertising campaign data with money, as N number of initial cluster center；
Advertising campaign data categorization module is supplemented in game with money, and it calculates one by one multigroup game and supplements rush with money according to Euclidean distance
Advertising campaign data and the Euclidean distance of each cluster centre are supplemented in other group game in pin activity data with money, by multigroup game
Supplement advertising campaign data with money and be divided into N classes one by one；
The equal of each index of advertising campaign data is supplemented in new cluster centre computing module, its game under calculating per class with money
Value, using the average of each index as new cluster centre；
Classification results determining module, it supplements advertising campaign data with money as one group, calculates standard with the game under all kinds of respectively
Deviation, by itself and the group distance of setting, the threshold comparison of group inner distance；
If standard deviation is less than the threshold value of setting, or the iterations for calculating is more than the maximum iteration time of setting
When, using this classification results as final classification result；Otherwise turn to enter by from the initial cluster center selecting module
Row circular treatment, until obtaining final classification result.
Also include：
Supplement active prediction model generation module with money, it supplements advertising campaign data set with money using existing game, with the shadow
The main affecting factors that the factor of sound determining module determines supplement amount with money as dependent variable as independent variable using actual play, pass through
The method fitting of multiple linear regression obtains supplementing active prediction model with money.
Also include：
Prediction effect detection module, its after active prediction model is supplemented in acquisition with money, according to supplementing predicted value and corresponding with money
The relative error for actually supplementing amount with money judges prediction effect, if relative error is more than the relative error threshold value of setting, return is filled
Value active prediction model generation module, regains after the relevant parameter for adjusting fitting prediction and new supplements active prediction mould with money
Type, otherwise shows that forecast model can use.
Present invention beneficial effect compared with prior art：
Game in the past is supplemented advertising campaign designer with money and often there are problems that two kinds when advertising campaign is supplemented in design with money：1) activity
Form is same in the whole text.This form for supplementing advertising campaign with money frequently can lead to the feeling of freshness of game player to be reduced, and is supplemented desire with money and is held
It is continuous to decline, supplement with money and be accustomed to the problems such as gradually fixing, there is a certain degree of harm for game income is pulled；2) activity form is total
In change.Although this Activities Design form can constantly stimulate the feeling of freshness of player, specific aim is not strong, while being also required to
The human resources of more game programmers are spent, cost performance is not high.The present invention supplements activity data with money according to existing game, leads to
Cross clustering method to classify the activity of supplementing with money of playing, for different supplement advertising campaign type with money, with variance analysis method,
It is determined that affecting every kind of major influence factors for supplementing amount with money for supplementing advertising campaign type with money to be that advertising campaign data are supplemented in game with money
Which index, so that game is supplemented advertising campaign designer with money and can preferentially be set with reference to major influence factors according to purpose of design
Meter game supplement advertising campaign with money, efficiently solve going game supplement with money advertising campaign throw in without purpose, supplement with money advertising campaign throw
Put the difficult problem of project settings.Therefore, to make to supplement with money advertising campaign design more targeted for the present invention so that game designer can root
According to the starting point of design movable input scheme reasonable in design, the purpose for pulling game income is reached with reaching the best price/performance ratio.
In addition, in the past game designer supplements the effect that advertising campaign scheme brings with money and cannot make conjunction for self-designed
The desired value of the science of reason, needs the game after activity to supplement quota data with money and does Late Stage Verification when more.This thing
There is serious delay drawback in the method verified afterwards, it is impossible to which reaching " has before input and pull trip so as to meet according to ground Adjusted Option
Play income reach desired value " purpose.Present invention further propose that the method predicted using fitting for the activity of supplementing with money, according to not
The effect of advertising campaign input is supplemented in type together and different types of design, prediction with money, is provided more to supplement advertising campaign with money
Rational expected reference value, and portable propelling designer's adjusted design supplements advertising campaign with money so that designer is for supplementing rush with money
The relevant parameter adjustment of pin activity more targetedly can reach and throw in the optimum purpose of effect.
Description of the drawings
Fig. 1 is that the schematic flow sheet that scheme optimization embodiment of the method is thrown in advertising campaign is supplemented in present invention game with money；
Fig. 2 is that advertising campaign cluster process schematic diagram is supplemented in game with money in the embodiment of the present invention；
Fig. 3 is that advertising campaign multifactor analysis of variance process schematic is supplemented in game with money in the embodiment of the present invention.
Specific embodiment
Embodiments of the present invention are elaborated below.It is emphasized that what the description below was merely exemplary,
Rather than in order to limit the scope of the present invention and its application.
Refering to Fig. 1, in one embodiment, one kind game is supplemented advertising campaign with money and throws in scheme optimization method, including following
Step：
S1. advertising campaign data are supplemented in the multigroup game to collecting with money, using clustering method, are supplemented each group game with money promotion and are lived
Dynamic data be included into respectively N kinds it is different supplement advertising campaign type, N >=2 with money；
Advertising campaign data are supplemented in wherein per group game with money includes multiple indexs, and each index is respectively used to represent filling for setting
Value advertising campaign item and supplement activity time of advertising campaign with money, whether the data value of each index is accordingly represented supplements with money
The duration of the characteristics of time interval or activity of advertising campaign item or generation activity；
S2. advertising campaign data are supplemented with money based on every kind of game supplemented with money under advertising campaign type, using multifactor variance point
Analysis method, it is determined that affecting every kind of major influence factors for supplementing amount with money for supplementing advertising campaign type with money to be that advertising campaign is supplemented in game with money
Which index of data.
For example, " whether advertising campaign data are supplemented in per group of game with money can include " whether throw in VIP privilege activity ",
Throw in single and supplement class activity with money ", " whether throw in reward double class activity ", " whether throw in store prescribe a time limit discounting class activity ", " be
No is the traditional festival period ", multiple indexs such as " input number of days ", the wherein data value of index can be logical value " 0 " and " 1 ",
Whether representative there occurs the activity item representated by the index, and the value of such as index " whether throwing in VIP privilege activities " is 1, generation
Table has thrown in VIP privilege activities, is worth for 0, and VIP privilege activities are not thrown in representative；In addition, the data value of index can also be generation
There is the characteristics of time interval of activity time in table, such as index " whether being the traditional festival period " is worth for 0, and representative is not traditional festival
Period, it is worth for 1, representative is the traditional festival period；The data value of index can also be the duration of deputy activity, and for example index " is thrown
Put number of days ", its value number of days that directly expression activity is persistently thrown in.
As shown in figure 1, in a preferred embodiment, the method is further comprising the steps of：
S3. advertising campaign data set is supplemented with money using existing game, using the main affecting factors of step S2 determination as certainly
Variable, amount is supplemented with money as dependent variable using actual play, obtains supplementing active prediction with money by the method fitting of multiple linear regression
Model.
In a further embodiment, the method can also be comprised the following steps：
Supplement the game collected with money activity data input fitting prediction module, obtain supplementing predicted value with money.
In a further preferred embodiment, the method is further comprising the steps of：
After active prediction model is supplemented in acquisition with money, according to supplement with money predicted value and it is corresponding actually supplement with money amount it is relative by mistake
Difference judges prediction effect, if relative error is more than the relative error threshold value of setting, return to step S3, and adjusts fitting prediction
Regain after relevant parameter it is new supplement active prediction model with money, otherwise show that forecast model can use.
In further embodiments, a kind of game is supplemented advertising campaign with money and throws in scheme optimization device, including：
Cluster module, it supplements advertising campaign data with money to the multigroup game collected, and using clustering method, each group game is filled
Value advertising campaign data be included into respectively N kinds it is different supplement advertising campaign type, N >=2 with money；
Advertising campaign data are supplemented in wherein per group game with money includes multiple indexs, and each index is respectively used to represent filling for setting
Value advertising campaign item and supplement activity time of advertising campaign with money, the data value of each index accordingly represent whether occurs it is corresponding
The characteristics of time interval or the duration of activity of supplementing advertising campaign item or generation activity with money；
Influence factor determining module, it is based on every kind of game supplemented with money under advertising campaign type and supplements advertising campaign data with money,
Using multifactor analysis of variance method, it is determined that affect it is every kind of supplement advertising campaign type with money the major influence factors for supplementing amount with money be
Which index of advertising campaign data is supplemented in game with money.
In a preferred embodiment, the cluster module includes：
Initial cluster center selecting module, it is supplemented with money in advertising campaign data from multigroup game using random algorithm and is selected
Select N group game and supplement advertising campaign data with money, as N number of initial cluster center；
Advertising campaign data categorization module is supplemented in game with money, and it calculates one by one multigroup game and supplements rush with money according to Euclidean distance
Advertising campaign data and the Euclidean distance of each cluster centre are supplemented in other group game in pin activity data with money, by multigroup game
Supplement advertising campaign data with money and be divided into N classes one by one；
The equal of each index of advertising campaign data is supplemented in new cluster centre computing module, its game under calculating per class with money
Value, using the average of each index as new cluster centre；
Classification results determining module, it supplements advertising campaign data with money as one group, calculates standard with the game under all kinds of respectively
Deviation, by itself and the group distance of setting, the threshold comparison of group inner distance；
If standard deviation is less than the threshold value of setting, or the iterations for calculating is more than the maximum iteration time of setting
When, using this classification results as final classification result；Otherwise turn to enter by from the initial cluster center selecting module
Row circular treatment, until obtaining final classification result.
In a preferred embodiment, the device also includes：
Supplement active prediction model generation module with money, it supplements advertising campaign data set with money using existing game, with the shadow
The main affecting factors that the factor of sound determining module determines supplement amount with money as dependent variable as independent variable using actual play, pass through
The method fitting of multiple linear regression obtains supplementing active prediction model with money.
In a preferred embodiment, the device also includes：
Prediction effect detection module, its after active prediction model is supplemented in acquisition with money, according to supplementing predicted value and corresponding with money
The relative error for actually supplementing amount with money judges prediction effect, if relative error is more than the relative error threshold value of setting, return is filled
Value active prediction model generation module, regains after the relevant parameter for adjusting fitting prediction and new supplements active prediction mould with money
Type, otherwise shows that forecast model can use.
Embodiment is described in detail below in conjunction with instantiation.
Advertising campaign Type division is supplemented in game with money
That collects game supplements advertising campaign data with money, forms raw data table.The field that tables of data is included is as shown in the table：
Table 1 is supplemented activity data with money and collects list of fields
Field " whether throwing in VIP privilege activities " can be extracted, whether " whether throw in single and supplement class activity with money " " throws in prize
Encourage double class activity ", " whether throw in accumulation and supplement class activity with money ", " whether throw in store prescribe a time limit discounting class activity ", " time throws in
Class activity is purchased by group in limited time ", " whether game New function being released during activity ", " whether be the traditional festival period ", " throw in day
Input of the data in number " as cluster module.Concrete cluster process is as shown in Figure 2.
A) three initial cluster centers are determined using random algorithm：
CENTER1(A1,B1…1),ENTER2(A2,B2…2),ENTER3(A3,B3…3)
Wherein (Ai,Bi…Ii) represent whether field throws in VIP privilege activities respectively ", " whether throw in single and supplement class work with money
It is dynamic ", " whether throwing in the double class activity of reward ", whether " whether throw in accumulation and supplement class activity with money " " throws in store to prescribe a time limit discounting class
Activity ", " time throws in and purchases by group class activity in limited time ", " whether game New function being released during activity ", " whether it is traditional festival
Period ", " input number of days " this 9 desired values.
B) other are calculated one by one according to Euclidean distance and supplements activity data (A with moneyi,Bi…Ii) with the Euclidean distance of cluster centre,
Supplement other with money activity data and be divided into three classes one by one, it is specific as follows：
For certain supplements activity data (A with moneyi,Bi…Ii), its distance with three cluster centres is calculated according to following formula：
Euclidean distance：
Chosen distance it is minimum certain class is included into the activity of supplementing with money.For remaining activity data of supplementing with money was computed repeatedly
Journey.Until all activities of supplementing with money are all included into certain classification.
C) by above-mentioned steps, can obtain each class supplements activity data with money.Calculate the average of 9 indexs in each class.
Using average as new cluster centre:
D) by standard deviation and threshold value, the threshold comparison of group inner distance of the group distance of setting.When the threshold less than setting
Value, or the number of times of iteration more than setting maximum iteration time when, using this classification results as each supplement with money activity
Classification results, output result；Otherwise circulation step (a)-(d).
Standard deviation refers to each data point in group to the average distance of mean value.
Specific formula for calculation is：
The value of each data point wherein inside X expressions group,For the mean value in group.
Group distance is calculated：
Group inner distance is calculated：
The cluster result of similar table 2 below can finally be obtained：
Advertising campaign cluster result is supplemented in the game of table 2 with money
After above-mentioned cluster process, throw in each time supplement with money advertising campaign data be all assigned to red-letter day type,
In filling some classification of the type that disappears, franchise type in these three.
Difference supplements advertising campaign analysis of Influential Factors with money
According to cluster result, advertising campaign tables of data 1 is supplemented with money based on original game, obtain new game and supplement promotion work with money
Dynamic tables of data, convenient here for description, the table for still newly obtaining is called that advertising campaign tables of data is supplemented in game with money.As described above,
When the advertising campaign of supplementing with money after above-mentioned cluster process, thrown in each time has all been assigned to red-letter day type, has filled the type that disappears, privilege
In a certain class of the type in these three.On the basis of table 1, increased a line " Activity Type " for record cluster result, most
The structure of the data set for obtaining afterwards is as shown in table 3.
Activity data literary name section list is supplemented in the game of table 3 with money
By in field data input multiplicity module in above-mentioned table, the detailed process of multiplicity is as shown in Figure 3.
The circular of each step in Fig. 3 is illustrated below by way of example.
It is A to assume that the influence factor that amount may be subject to is supplemented in certain game with money, and two kinds of B, the wherein level of A have three kinds, point
Not Yong A1, A2, A3 represents that the level of B has two kinds, and respectively with B1, B2 is representing.Game under different A, B levels is supplemented with money
Amount is as shown in the table.
The multifactor analysis of variance case data collection of table 4
Calculate quadratic sum
Total sum of squares：
Quadratic sum in group：
Interaction quadratic sum：
Interaction refers to that is supplemented with money in hypothesis and be not only subject to A, and the impact of B, the various combination of A, B also can produce impact to supplementing with money
Sum of squares between groups：
Calculate the free degree
Total free degree：
dfT=abr-1=2 × 3 × 5-1=29
The group internal degree of freedom：
dfA=a-1=2-1=1 dfB=b-1=3-1=2
The interaction free degree：
dfAXB=(a-1) (b-1) -1=2-1=1
The free degree between group：
dfe=ab (r-1)=3 × 2 × (5-1) -1=24
Calculate variance
Intra-class variance：
Interaction variance：
Between-group variance：
Calculate conspicuousness
A factor conspicuousnesses：
B factor conspicuousnesses：
A, B interaction factor conspicuousness：
As a result understand
According to calculated conspicuousness F value, according to the known conversion in statistics between F values and significance P value
Relation, it may be determined that the level of signifiance P value corresponding to it.Significance P (is generally selected with the confidence level of setting
0.05th, 0.01) it is compared, can determine whether whether corresponding index is major influence factors according to comparative result.
In addition, the Table of the corresponding F (r-1, nr-r) of inquiry can be passed through, corresponding F is obtained0.05And F0.01Value, root
According to F values and the relation of significance P value, you can to obtain significance P value and 0.05 and 0.01 magnitude relationship.Work as meter
The F of certain factor for obtaining is more than F0.05When, i.e., when significance P value is less than 0.05, show that the factor affects notable, when
The F values of calculated certain factor are more than F0.01When, i.e., when significance P value is less than 0.01, show that the factor affects pole
Its is notable.In actual use, by comparing magnitude relationship of the significance P value of variance analysis with 0.05 and 0.01 just
Can determine whether the influence degree of factor.Here 0.05,0.01 two values it is selected be to learn according to the common standards in statistics
The more commonly used boundary value is selected in section.Under normal circumstances, experimental result reaches 0.05 confidence level, it may be said that between data
Significant difference, reaches 0.01 confidence level, it may be said that difference is extremely notable between data.
According to above calculated group of internal degree of freedom, the free degree is obtained between the interaction free degree and group needs inquiry
F values relevant parameter.After it have selected confidence level, it is possible to according to the corresponding row of these parameter queries F Distribution value tables and
Row, and for example (2,24).
For above-mentioned calculated result, understand respectively：
FA>F(2,24)0.01=5.61, i.e. significance P value are less than 0.01, show A factors for the impact for supplementing amount with money
It is extremely notable；
FB>F(1,24)0.01=7.82, i.e. significance P value are less than 0.01, show B factors for the impact for supplementing amount with money
It is extremely notable；
FAB<F(2,24)0.05=3.42, i.e. significance P value are more than 0.05, show A, and B interactions are for supplementing amount with money
Impact it is not notable.
According to the Computing Principle of the above-mentioned multifactor analysis of variance, the multifactor side of simulated data sets in the present invention can be obtained
The result of difference analysis, it is as shown in the table：
Activity multifactor analysis of variance result is supplemented in the game of table 5 with money
From the result of the multifactor analysis of variance, it may be seen that affecting the principal element of game income.According to notable
The numerical value of property can be determined which influence factor is.It is general when choosing significance P value less than or equal to 0.05, it is corresponding because
Element is influence factor, and significance is less, and impact property is bigger.
The multifactor analysis of variance result of the simulated data sets used in the present invention shows to supplement with money in advertising campaign design
Supplement amount of each factor to playing has an impact.Difference game supplements the results of analysis of variance that promotion data collection obtains with money not
One and it is same, needs understood according to specific the results of analysis of variance.
Difference is supplemented the expection of advertising campaign design with money and supplements prediction with money
According to the influence factor obtained in above-mentioned, using existing game promotion data collection is supplemented with money, returned by multiple linear
The method returned can obtain supplementing active prediction model with money, so as to amount desired value is supplemented in game when obtaining throwing in different schemes with money.
It is specific as follows：
It is y to assume that amount is supplemented in game with money, and each influence factor is A, B, C ...., then the purpose of multiple regression is to find conjunction
Suitable parameter b1,b2,b3... .. causes the calculated match value y of following equalitiespThe minimum condition of sum of square of deviations can be met.
yp=b0+b1×A+b2×B+b3×C+…+bm×X
Sum of square of deviations：
Wherein, yp_iI-th numerical value in be fitted the sequence for obtaining, yiI-th numerical value in supplement amount sequence with money.b1,
b2,b3... the parameter that .. takes for model needs
The process for solving parameter is as follows：
The minimum condition of sum of square of deviations is converted
Then problem is converted into the extreme-value problem for seeking above-mentioned function expression, according to the original of " it is extreme point that derivative is 0 point "
Reason, to Q derivation is carried out, and can ask for parameter b1,b2,b3…..
……
Above-mentioned equation can be converted to matrix form：
Finally can be in the hope of parameter matrixFor：
Last digital simulation goodness
Wherein R2For the goodness of fit, whether can use for decision model equation.SST is total sum of squares, and measurement is actual
The squared difference and size of supplementing amount and average with money, SSR is regression sum of square, and measurement is match value and actual value average
Squared difference and size, SSE is residual sum of squares (RSS), and measurement is squared difference and size between match value and actual value.He
Be specifically calculated as follows shown in：
Wherein, yiAmount is supplemented with money for actual,Actually to supplement the average of amount, y with moneyp_iVolume is supplemented with money for what fitting was obtained
Degree.
The calculated goodness of fit shows that more greatly models fitting effect is better, and the accuracy for prediction will be higher.When
When the goodness of fit is satisfied by the threshold value for setting, show that the model equation is available.
Now we have been obtained for the model equation for prediction, it is assumed that have another set to supplement the data set of promotion with money,
The data integration is herein test set as described in hereinbefore table 1 by the form of data set.The model for obtaining is applied into the test
What concentration can be calculated prediction supplements amount with money.Concrete mode is exactly by the factor of influence in the following model expression tried to achieve
The value of A～X replaces with the occurrence in test set.
yp=b0+b1×A+b2×B+b3×C+…+bm×X
Then the amount of supplementing with money that prediction is obtained is asked for into relative error average with the amount of actually supplementing with money of test set.It is relative to miss
Differ from being calculated as follows shown in formula for average：
Wherein, N is the line number of the test set, yiAmount, y are supplemented with money for i-th reality in test setp_iObtain for prediction
Supplement amount with money i-th.
When the relative error average asked for meets the threshold value standard of our settings, the accuracy for being considered as the model is can
With what is received.So far, we have just obtained the forecast model that can be used and accurately sexual satisfaction is required.
Additionally, according to the parameter matrix being the previously calculatedValue positive and negative a certain factor of influence can be learnt for filling
The pulling of value amount is positive or negative sense, it is possible to which the Activities Design person of auxiliary game is when advertising campaign scheme is supplemented in design with money
Corresponding setting value is carried out targeted specifically.
Above content is with reference to concrete/preferred embodiment further description made for the present invention, it is impossible to recognized
Being embodied as of the fixed present invention is confined to these explanations.For general technical staff of the technical field of the invention,
Without departing from the inventive concept of the premise, it can also make some replacements or modification to the embodiment that these have been described,
And these are substituted or variant should all be considered as belonging to protection scope of the present invention.
Claims (10)
1. a kind of game is supplemented advertising campaign with money and throws in scheme optimization method, it is characterised in that comprised the following steps：
S1. advertising campaign data are supplemented in the multigroup game to collecting with money, using clustering method, supplement each group game with money advertising campaign number
According to be included into respectively N kinds it is different supplement advertising campaign type, N >=2 with money；
Advertising campaign data are supplemented in wherein per group game with money includes multiple indexs, and each index is respectively used to represent that what is set supplements rush with money
Pin activity item and supplement activity time of advertising campaign with money, whether the data value of each index is accordingly represented occurs to fill accordingly
Value advertising campaign item or the characteristics of time interval or the duration of activity of generation activity；
S2. advertising campaign data are supplemented with money based on every kind of game supplemented with money under advertising campaign type, using multifactor analysis of variance side
Method, it is determined that affecting every kind of major influence factors for supplementing amount with money for supplementing advertising campaign type with money to be that advertising campaign data are supplemented in game with money
Which index.
2. game as claimed in claim 1 is supplemented advertising campaign with money and throws in scheme optimization method, it is characterised in that step S1 includes
Following steps：
A) supplement selection N group game in advertising campaign data with money from multigroup game using random algorithm and supplement advertising campaign number with money
According to as N number of initial cluster center；
B) described multigroup other group game supplemented with money in advertising campaign data of playing are calculated one by one according to Euclidean distance and supplements promotion with money
Activity data and the Euclidean distance of each cluster centre, supplement multigroup game with money advertising campaign data and are divided into N classes one by one；
C) average of each index for supplementing advertising campaign data with money per the game under class is calculated, using the average of each index as new
Cluster centre；
D) respectively advertising campaign data are supplemented with money as one group with the game under all kinds of, standard deviation is calculated, by between its group with setting
Distance, the threshold comparison of group inner distance；
If standard deviation is less than the threshold value of setting, or the iterations for calculating more than the maximum iteration time for setting, by
This classification results are used as final classification result；Otherwise circulation step a)-d), until obtaining final classification result.
3. game as claimed in claim 1 is supplemented advertising campaign with money and throws in scheme optimization method, it is characterised in that step S2 includes
Following steps：
Game under respectively to supplement advertising campaign type with money supplements advertising campaign data with money as group, quadratic sum, interaction in calculating group
Quadratic sum, sum of squares between groups, the group internal degree of freedom, the free degree between the interaction free degree and group；
Based on the quadratic sum and the free degree that are calculated, intra-class variance, interaction variance and between-group variance are calculated；
Based on the variance for being calculated, calculating is used as the friendship between the conspicuousness F value and each index of each index of influence factor
The conspicuousness F value of mutual factor；
Significance P value according to corresponding to calculated conspicuousness F value determines it, by significance P value and setting
Confidence level be compared, judge whether corresponding index is major influence factors according to comparative result；Or, based on being calculated
The free degree obtain the relevant parameter of the F values for needing inquiry, tabled look-up the F values that obtain according to the confidence level of setting, then will count
The conspicuousness of the interaction factor between the conspicuousness of each index for calculating and each index, with the F values for obtaining of tabling look-up than larger
It is little, judge whether corresponding index is major influence factors according to comparative result.
4. the game as described in any one of claims 1 to 3 is supplemented advertising campaign with money and throws in scheme optimization method, it is characterised in that
It is further comprising the steps of：
S3. using it is existing game supplement advertising campaign data set with money, using step S2 determine main affecting factors as independent variable,
Amount is supplemented with money as dependent variable using actual play, obtains supplementing active prediction model with money by the method fitting of multiple linear regression.
5. the game as described in any one of claims 1 to 3 is supplemented advertising campaign with money and throws in scheme optimization method, it is characterised in that
It is further comprising the steps of：
S4. supplement the game collected with money activity data input fitting prediction module, obtain supplementing predicted value with money.
6. the game as described in any one of claims 1 to 3 is supplemented advertising campaign with money and throws in scheme optimization method, it is characterised in that
It is further comprising the steps of：
S5. after active prediction model is supplemented in acquisition with money, according to supplement with money predicted value and it is corresponding actually supplement with money amount it is relative by mistake
Difference judges prediction effect, if relative error is more than the relative error threshold value of setting, return to step S3, and adjusts fitting prediction
Regain after relevant parameter it is new supplement active prediction model with money, otherwise show that forecast model can use.
7. a kind of game is supplemented advertising campaign with money and throws in scheme optimization device, it is characterised in that included：
Cluster module, it supplements advertising campaign data with money to the multigroup game collected, and using clustering method, supplements each group game with money rush
Pin activity data be included into respectively N kinds it is different supplement advertising campaign type, N >=2 with money；
Advertising campaign data are supplemented in wherein per group game with money includes multiple indexs, and each index is respectively used to represent that what is set supplements rush with money
Pin activity item and supplement activity time of advertising campaign with money, whether the data value of each index is accordingly represented occurs to fill accordingly
Value advertising campaign item or the characteristics of time interval or the duration of activity of generation activity；
Influence factor determining module, it is based on every kind of game supplemented with money under advertising campaign type and supplements advertising campaign data with money, uses
Multifactor analysis of variance method, it is determined that affecting every kind of major influence factors for supplementing amount with money for supplementing advertising campaign type with money to be game
Which index of advertising campaign data supplemented with money.
8. game as claimed in claim 7 is supplemented advertising campaign with money and throws in scheme optimization device, it is characterised in that the cluster mould
Block includes：
Initial cluster center selecting module, it is supplemented with money in advertising campaign data from multigroup game using random algorithm and selects N
Advertising campaign data are supplemented in group game with money, used as N number of initial cluster center；
Advertising campaign data categorization module is supplemented in game with money, and it calculates one by one multigroup game and supplements promotion work with money according to Euclidean distance
Advertising campaign data and the Euclidean distance of each cluster centre are supplemented in other group game in dynamic data with money, and multigroup game is supplemented with money
Advertising campaign data are divided into one by one N classes；
The average of each index of advertising campaign data is supplemented in new cluster centre computing module, its game under calculating per class with money, will
The average of each index is used as new cluster centre；
Classification results determining module, it supplements advertising campaign data with money as one group, calculates standard deviation with the game under all kinds of respectively,
By itself and the group distance of setting, the threshold comparison of group inner distance；
Wherein, if standard deviation is less than the threshold value for setting, or the iterations for calculating is more than the maximum iteration time of setting
When, using this classification results as final classification result；Otherwise turn to enter by from the initial cluster center selecting module
Row circular treatment, until obtaining final classification result.
9. as claimed in claim 7 or 8 game is supplemented advertising campaign with money and throws in scheme optimization device, it is characterised in that also included：
Supplement active prediction model generation module with money, its using it is existing game supplement advertising campaign data set with money, with the impact because
The main affecting factors that plain determining module determines supplement amount with money as dependent variable, by polynary as independent variable using actual play
The method fitting of linear regression obtains supplementing active prediction model with money.
10. game as claimed in claim 9 is supplemented advertising campaign with money and throws in scheme optimization device, it is characterised in that also included：
Prediction effect detection module, its after active prediction model is supplemented in acquisition with money, according to supplementing predicted value and corresponding reality with money
The relative error for supplementing amount with money judges prediction effect, if relative error is more than the relative error threshold value of setting, work is supplemented in return with money
Dynamic forecast model generation module, regain after the relevant parameter for adjusting fitting prediction it is new supplement active prediction model with money, it is no
Then show that forecast model can use.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2015107624848 | 2015-11-10 | ||
| CN201510762484 | 2015-11-10 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106682053A true CN106682053A (en) | 2017-05-17 |
Family
ID=58839161
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610096792.6A Pending CN106682053A (en) | 2015-11-10 | 2016-02-22 | Optimization method and device for putting scheme of game top-up sales promotion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106682053A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108646688A (en) * | 2018-05-31 | 2018-10-12 | 成都天衡智造科技有限公司 | A kind of process parameter optimizing analysis method based on recurrence learning |
| CN109460778A (en) * | 2018-10-12 | 2019-03-12 | 中国平安人寿保险股份有限公司 | Active evaluation method, apparatus, electronic equipment and storage medium |
| WO2019200600A1 (en) * | 2018-04-20 | 2019-10-24 | 上海荟萃网络科技有限公司 | Sampling simulation-based quick a/b testing method |
| CN112686543A (en) * | 2020-12-31 | 2021-04-20 | 上海掌门科技有限公司 | Service index processing method, electronic equipment and computer readable storage medium |
-
2016
- 2016-02-22 CN CN201610096792.6A patent/CN106682053A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019200600A1 (en) * | 2018-04-20 | 2019-10-24 | 上海荟萃网络科技有限公司 | Sampling simulation-based quick a/b testing method |
| CN108646688A (en) * | 2018-05-31 | 2018-10-12 | 成都天衡智造科技有限公司 | A kind of process parameter optimizing analysis method based on recurrence learning |
| CN108646688B (en) * | 2018-05-31 | 2019-05-07 | 成都天衡智造科技有限公司 | A kind of process parameter optimizing analysis method based on recurrence learning |
| CN109460778A (en) * | 2018-10-12 | 2019-03-12 | 中国平安人寿保险股份有限公司 | Active evaluation method, apparatus, electronic equipment and storage medium |
| CN109460778B (en) * | 2018-10-12 | 2024-04-09 | 中国平安人寿保险股份有限公司 | Activity evaluation method, activity evaluation device, electronic equipment and storage medium |
| CN112686543A (en) * | 2020-12-31 | 2021-04-20 | 上海掌门科技有限公司 | Service index processing method, electronic equipment and computer readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105260474B (en) | A kind of microblog users influence power computational methods based on information exchange network | |
| CN104330721B (en) | IC Hardware Trojan detecting method and system | |
| CN106682053A (en) | Optimization method and device for putting scheme of game top-up sales promotion | |
| CN108694390A (en) | A kind of cuckoo search improves the modulated signal sorting technique of grey wolf Support Vector Machines Optimized | |
| CN103309857B (en) | A kind of taxonomy determines method and apparatus | |
| CN107862457B (en) | Method for extracting stage scheduling rules of reservoir | |
| CN109214444A (en) | Game Anti-addiction decision-making system and method based on twin neural network and GMM | |
| Puram et al. | Understanding the effect of contextual factors and decision making on team performance in Twenty20 cricket: an interpretable machine learning approach | |
| McHale | Assessing the fairness of the golf handicapping system in the UK | |
| Munir et al. | Predicting a T20 cricket match result while the match is in progress | |
| Kutek et al. | Development and application of model characteristics for optimizing the educational and training process of qualified athletes | |
| Zhu | Classification and optimization of basketball players’ training effect based on particle swarm optimization | |
| CN110909890B (en) | Game artificial intelligence training method and device, server and storage medium | |
| Chen et al. | A normalized score-based weighted PageRank algorithm on ranking prediction of basketball games | |
| An et al. | Research on the Application of Apriori Algorithm in the Teaching of Ball Sports Techniques and Tactics | |
| Azcárate et al. | Variability of professional soccer players’ perceived match load after successive matches | |
| Stevenson et al. | Modelling career trajectories of cricket players using Gaussian processes | |
| Dey et al. | Modified group decision algorithm for performance appraisal of Indian premier league cricketers | |
| Li | Appliance of apriori algorithm on technical-tactics analysis of volleyball | |
| CN109543843A (en) | A kind of game user selection method based on machine learning | |
| Sadurska et al. | Fusion of a fuzzy rule-based method and other decision-making models in injury prediction problem in football | |
| Qiu et al. | Prediction of Wordle Puzzles Based on Neural Networks and K-Means Improved Models | |
| Zhao et al. | Analysis of the Running Ability Mining Model of Football Trainers Based on Dynamic Incremental Clustering Algorithm | |
| Vohra et al. | Markov Cricket: Using Forward and Inverse Reinforcement Learning to Model, Predict And Optimize Batting Performance in One-Day International Cricket | |
| Zhang et al. | A Methodological Research on Evaluating Competitive Capability and Growth Rate of Basketball Players |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170517 |