CN105354725A

CN105354725A - Prediction method and system of promotion effect of application

Info

Publication number: CN105354725A
Application number: CN201510818267.6A
Authority: CN
Inventors: 余建兴
Original assignee: ZHUHAI DUOWAN INFORMATION TECHNOLOGY Ltd
Current assignee: ZHUHAI DUOWAN INFORMATION TECHNOLOGY Ltd
Priority date: 2015-11-20
Filing date: 2015-11-20
Publication date: 2016-02-24

Abstract

The invention discloses a prediction method and system of a promotion effect of an application. The prediction method comprises the following steps: according to the attribute data of a current application and the operational data of the current application in a promotion channel within preset time, determining the characteristics of the current application; and according to the characteristics of the current application and an established prediction model corresponding to the promotion channel, predicting the promotion effect of the current application. Through the above way, the use ratio of a channel promotion resource niche can be improved, and revenue benefits can be effectively improved.

Description

A kind of Forecasting Methodology of application effect and system

Technical field

The present invention relates to a kind of Forecasting Methodology and system of application effect.

Background technology

The fast development of science and technology, a large amount of short-terms that is applied in continues to bring out, and these application are promoted by each channel resource position, application of such as playing.

The quality widely applied is uneven, and the application of low value not only consumes a large amount of popularizations and operation resource, also reduces the utilization factor of channel resource position, and then has a strong impact on the business revenue of application platform.

Traditional application recruitment evaluation is generally the method adopting Manual definition's rule, after the application phase, add up business revenue, determines application inferior.For example game promotion adds up the indexs such as new registration user that this game obtained by channel and corresponding business revenue for 15 days afterwards, if business revenue higher position judges that game is adapted at this channel promotion, otherwise is judged to be not suitable for.This mode reaction time is slow, fails Timeliness coverage poor quality apply and change candidate popularization and apply to stop loss.

How to monitor the promotion effect of each channel, and realize stopping loss fast being one and having major issue to be solved for poor quality application.

Summary of the invention

The technical matters that the present invention mainly solves be how predicted application in the promotion effect of each channel.

First aspect, the embodiment of the present invention provides a kind of Forecasting Methodology of application effect, and described method comprises: the attribute data and the operation data of described current application in channels in the schedule time that obtain current application; According to described attribute data and described operation data, determine the feature of described current application; According to the forecast model of the described channels of correspondence that the characteristic sum of described current application has been set up, predict the promotion effect of described current application.

Wherein, the attribute data of described current application comprises at least one in application subject matter, style, image quality, type, operation platform and form, described operation data comprises the registration behavior of user application, login behavior, at least one of supplementing with money in behavior and consumer behavior, described according to described attribute data and described operation data, determine that the feature of described current application comprises: according to described attribute data and described operation data, determine the foundation characteristic of described current application and derivative feature, wherein, described foundation characteristic comprises the accumulative enrollment that in the described schedule time, described channels brings to described current application, login times, log in number of days, log in duration, login trend, supplement number of times with money, supplement number of days with money, recharge amount, supplement trend with money, consumption number of times, consumption number of days, spending amount, at least one in the propensity to consume, described derivative feature comprises the paying next day rate that described channels brings to described current application, next day supplements number with money per capita, the same day conversion ratio, next day conversion ratio and next day retention ratio at least one.

Wherein, the forecast model of the described described channels of correspondence set up according to the characteristic sum of described current application, predict that the promotion effect of described current application comprises: according to the feature of described current application, and predict the income in the following schedule time of described current application according to the forecast model of the described channels of the correspondence set up; Whether the income judging in described current application following schedule time of predicting is greater than the income median that all history in described channels is applied; When the income in the described current application the predicted following schedule time is greater than the income median that all history in described channels applies, determine that described current application is the application being applicable to continuing to promote at described channels, otherwise, determine that described current application is the application being not suitable for continuing to promote at described channels.

Wherein, according to the forecast model of the described channels of correspondence that the characteristic sum of described current application has been set up, before predicting the promotion effect of described current application, also comprise: by least one method in proper vector normalization and the screening of validity feature vector, pre-service is carried out to the feature of described current application.

Wherein, describedly by proper vector normalization, pre-service is carried out to the feature of described current application and comprise: by maximal value-Returning to one for minimum value, pre-service is carried out to the feature of described current application; Describedly pre-service is carried out to the feature of described current application comprise by the screening of validity feature vector: the key character being filtered out described current application by any one following mode: numeric type Feature change coefficient is less than predetermined threshold, determine that described numeric type feature is inessential; Or numeric type character difference is less than predetermined threshold, determine that described numeric type feature is inessential; Or the class label of classifying type feature is greater than predetermined threshold, determine that described classifying type feature is inessential; Or the quantity of the class label of classifying type feature is greater than predetermined threshold, determine that described classifying type feature is inessential.

Wherein, before the step of the attribute data of the described application of described acquisition and the operation data in the channels schedule time, also comprise: collect attribute data that the history on each channels applies and the operation data in the schedule time respectively; The attribute data apply described history and the operation data in the schedule time process, and obtain history described in characteristic sum that described history applies and apply the income of the schedule time; On described channels, all history is applied the income of the schedule time and is got median as threshold value, and the income that described in the characteristic sum apply described history, history is applied carries out the forecast model that learning training obtains corresponding described channels.

Wherein, on described channels, all history is applied the income of the schedule time and is got median as threshold value, the income that described in the characteristic sum apply described history, history is applied is carried out the forecast model that learning training obtains corresponding described channels and is comprised: on described channels, all history is applied the income of the schedule time and got median as threshold value, and the income that described in the characteristic sum adopting support vector machine classifier to apply described history, history is applied carries out the forecast model that learning training obtains corresponding described channels.

Wherein, described method also comprises: according to the promotion effect of described current application, returns corresponding Promotion Strategy.

Wherein, the described promotion effect according to described current application, the step returning corresponding Promotion Strategy comprises: according to the promotion effect of described current application, returns the candidate popularization list of application of corresponding described current channels.

Wherein, the candidate popularization list of application of the described current channels of described correspondence is to predict that operation benefits arranges from high to low.

Wherein, game application is applied as described in.

Second aspect, a kind of prognoses system of application effect is provided, described system comprises acquisition module, determination module and prediction module, wherein: described acquisition module is for obtaining attribute data and the operation data of described current application in channels in the schedule time of current application; Described determination module is used for according to described attribute data and described operation data, determines the feature of described current application; Described prediction module is used for the forecast model of the described channels of correspondence set up according to the characteristic sum of described current application, predicts the promotion effect of described current application.

Wherein, the attribute data of described current application comprises at least one in application subject matter, style, image quality, type, operation platform and form, described operation data comprises the registration behavior of user application, login behavior, at least one of supplementing with money in behavior and consumer behavior, described determination module is used for according to described attribute data and described operation data, determine the foundation characteristic of described current application and derivative feature, wherein, described foundation characteristic comprises the accumulative enrollment that in the described schedule time, described channels brings to described current application, login times, log in number of days, log in duration, login trend, supplement number of times with money, supplement number of days with money, recharge amount, supplement trend with money, consumption number of times, consumption number of days, spending amount, at least one in the propensity to consume, described derivative feature comprises the paying next day rate that described channels brings to described current application, next day supplements number with money per capita, the same day conversion ratio, next day conversion ratio and next day retention ratio at least one.

Wherein, described prediction module comprises predicting unit, judging unit and determining unit, wherein: described predicting unit is used for the feature according to described current application, and predicts the income in the following schedule time of described current application according to the forecast model of the described channels of the correspondence set up; Whether described judging unit is greater than for the income judging in described current application following schedule time of predicting the income median that all history in described channels applies; When the described determining unit income be used within the described current application the predicted following schedule time is greater than the income median that all history in described channels applies, determine that described current application is the application being applicable to continuing to promote at described channels, otherwise, determine that described current application is the application being not suitable for continuing to promote at described channels.

Wherein, described prediction module also comprises pretreatment unit, and described pretreatment unit is used for carrying out pre-service by least one method in proper vector normalization and the screening of validity feature vector to the feature of described current application.

Wherein, described pretreatment unit carries out pre-service by maximal value-Returning to one for minimum value to the feature of described current application; Or described pretreatment unit is used for the key character being filtered out described current application by any one following mode: numeric type Feature change coefficient is less than predetermined threshold, determines that described numeric type feature is inessential; Or numeric type character difference is less than predetermined threshold, determine that described numeric type feature is inessential; Or the class label of classifying type feature is greater than predetermined threshold, determine that described classifying type feature is inessential; Or the quantity of the class label of classifying type feature is greater than predetermined threshold, determine that described classifying type feature is inessential.

Wherein, described prognoses system also comprises model training module, wherein, described model training module comprises data collection module, processing unit and training unit, wherein: the attribute data that described data collection module is applied for the history of collecting respectively on each channels and the operation data in the schedule time; Described processing unit is used for the attribute data applied described history and the operation data in the schedule time processes, and obtains history described in characteristic sum that described history applies and applies the income of the schedule time; Described training unit goes median as threshold value for applying the income of the schedule time using all history on described channels, and the income that described in the characteristic sum apply described history, history is applied carries out the forecast model that learning training obtains corresponding described channels.

Wherein, described training unit gets median as threshold value for applying the income of the schedule time using all history on described channels, and the income that described in the characteristic sum adopting support vector machine classifier to apply described history, history is applied carries out the forecast model that learning training obtains corresponding described channels.

Wherein, described prognoses system also comprises and returns module, described in return module for the promotion effect according to described current application, return corresponding Promotion Strategy.

Wherein, game application is applied as described in.

The invention has the beneficial effects as follows: the situation being different from prior art, the present invention is the operation data in channels in the schedule time according to the attribute data of current application and current application, determine the feature of current application, according to the forecast model of the corresponding channels that the characteristic sum of current application has been set up, the promotion effect of prediction current application.By such mode, poor quality can be stopped in time being applied in popularization in channel by prediction effect, increase substantially the utilization factor of channel promotion resource-niche, and effectively improve business revenue benefit.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of the Forecasting Methodology of a kind of application effect that the embodiment of the present invention provides;

Fig. 2 is the process flow diagram of the prediction current application promotion effect that the embodiment of the present invention provides;

Fig. 3 is the process flow diagram of the method for the forecast model of the corresponding channel of foundation that the embodiment of the present invention provides

Fig. 4 is that different pieces of information point is separated the schematic diagram realizing classification by the segmentation plane that the embodiment of the present invention provides

Fig. 5 is the structural representation of the prognoses system of a kind of application effect that the embodiment of the present invention provides;

Fig. 6 is the structural representation of the prediction module that the embodiment of the present invention provides;

Fig. 7 is the structural representation of the model training module that the embodiment of the present invention provides.

Embodiment

Below in conjunction with the accompanying drawing in the embodiment of the present invention, be clearly and completely described the technical scheme in the embodiment of the present invention, obviously, described embodiment is only the present invention's part embodiment, instead of whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art, not making the every other embodiment obtained under creative work prerequisite, belong to the scope of protection of the invention.

It should be noted that, the term used in embodiments of the present invention is only for the object describing specific embodiment, and not intended to be limiting the present invention." one ", " described " and " being somebody's turn to do " of the singulative used in the embodiment of the present invention and appended claims is also intended to comprise most form, unless context clearly represents other implications.It is also understood that term "and/or" used herein refer to and comprise one or more project of listing be associated any or all may combine.

Existing application recruitment evaluation, just can will carry out often after the application phase, and such mode reaction time is slow, and the residence time causing poor quality to be applied in channels is long, affects utilization factor and the income of channel promotion resource-niche.The embodiment of the present invention provides a kind of Forecasting Methodology and system of application effect, be intended to solve the above technical matters existed in prior art, poor quality can be stopped in time being applied in channel by promotion effect prediction to promote, increase substantially the utilization factor of channel promotion resource-niche, and effectively improve business revenue benefit.

The method that the following embodiment of the present invention provides, all application can be applicable to, in the following embodiment of the present invention, to play, application illustrates method of the present invention as an example, but method of the present invention has more than and is limited to game application, can be such as other application, such as microblogging application, space application etc.

Refer to Fig. 1, Fig. 1 is the process flow diagram of the Forecasting Methodology of a kind of application effect that the embodiment of the present invention provides, and as shown in the figure, the Forecasting Methodology of the application effect of the present embodiment comprises the following steps:

S101: the attribute data and the operation data of current application in channels in the schedule time that obtain current application.

The attribute data of current application and the operation data in channels in the schedule time are the basis and information source of predicting current application promotion effect.

The attribute data of current application refers to the attribute data of application itself, can obtain from application developers, and these attribute datas such as include but not limited to it is application subject matter, style, image quality, type, operation platform and form etc.Current application operation data of the schedule time in channels refers to that being applied in each channel resource position does in the process promoted, the operation data of record user application in each money application of acquisition.These data can portray the fancy grade of user to application, and the business revenue ability of application.These operation datas comprise user application in the schedule time registration behavior, login behavior, supplement behavior and consumer behavior etc. with money.

Example is applied as with game, the attribute data of game application itself can comprise game subject matter, style, image quality, type (as action, risk, simulation, role playing etc.), Platform Type is (as page trip, end trip, hand trip etc.), and form is (as unit, many people, large-scale many people etc.).And the game application operation data of the schedule time in channels comprise game player registration, log in, supplement with money and consumer behavior.

Wherein, the schedule time can be the time pre-set.When based on when being to the prediction of the promotion effect of current application, this schedule time is preferably set to 2 days.Namely promote by current application being put into channels the operation data collected two days later and be applied in this channel 2 days.Certainly, can proper extension or shorten this schedule time according to actual needs.

S102: according to attribute data and operation data, determines the feature of current application.

Specifically, be according to attribute data and operation data, the foundation characteristic determining current application and the derivative feature built based on foundation characteristic.Wherein, foundation characteristic comprises accumulative enrollment that in the schedule time, channels brings to current application, login times, login number of days, logs in duration, login trend, supplements number of times with money, supplements number of days with money, recharge amount, supplement trend, consumption number of times, consumption number of days, spending amount, propensity to consume etc. with money, derivative feature comprises paying next day rate that channels brings to described current application, next day supplements number per capita with money, the same day conversion ratio, next day conversion ratio and next day retention ratio etc.

For application 2 days of playing, foundation characteristic comprises registrations in first 2 days of collection, logs in, supplements with money, and consumer behavior feature, as statistics plays first 2 days day by this channel to the accumulative enrollment brought of playing; Login times/number of days/duration/trend, supplements the number of times/number of days/amount of money/trend with money, the consumption number of times/number of days/amount of money/trend etc.Wherein the trend of each index does fitting a straight line by this index in statistics each value of 2 days a few days ago, portrays trend feature with the slope of matching.

Derivative feature comprises the paying next day rate (=next day the number of supplementing with money/same day enrollment) brought to playing by this channel, supplement ARPU value (recharge amount/next day=next day supplements number with money) next day per capita with money, conversion ratio on the same day (enrollment/same day on=same day clicks number), conversion ratio next day (enrollment/next day=next day clicks number), retention ratio next day (=next day logs in retention number/same day enrollment).These derivative features can weigh the characteristic of game all sidedly.

S103: according to the forecast model of the corresponding channels that the characteristic sum of current application has been set up, the promotion effect of prediction current application.

According to the forecast model of the corresponding channels that the characteristic sum of current application has been set up, the promotion effect of prediction current application.

Wherein, preferred as one, before the promotion effect of prediction current application, carry out pre-service to the feature of current application, these pretreatment operation can allow and predict the outcome more accurately, rationally.Wherein, pre-service can be carried out by least one method in proper vector normalization and the screening of validity feature vector to the feature of current application, thus determine that in the feature of current application, key character is for predicting the promotion effect of current application.

Particularly, proper vector normalization process is as follows: each foundation characteristic (i.e. apply property data, operation situation data) and derivative feature characteristic of correspondence value as an element of proper vector, the dimension of vector is then aforesaid foundation characteristic or derivative feature.The numerical range disunity of each vector element, such as be characterized as login duration, the scope of its eigenwert between 1 to 3600, and may be characterized as login times, and the scope of its eigenwert may in 1 to 100 times.Need to be normalized proper vector, in implementation process, adopt maximal value-Returning to one for minimum value, be (preset features value-this dimension element minimum value)/(this dimension element maximal value-this dimension element minimum value).

And the process of validity feature vector screening is as follows: from feature self information amount aspect, important feature carry information is more, and namely eigenwert differs greatly, and the embodiment of the present invention weighs the importance of feature by following 4 class indexs:

1) numeric type Feature change coefficient is less than reservation threshold, then this variable is considered as inessential variable;

2) numeric type character difference is less than reservation threshold, then this variable is inessential;

When aforesaid dimensional characteristics is numeric type, such as recharge amount, log in duration etc., calculate two groups of statistical values that this feature is corresponding, comprise the coefficient of variation (standard deviation/mean value of=normal distribution) and standard deviation.Compared by the coefficient of variation and standard deviation and predetermined threshold, determine whether feature belongs to key character.

3) if the number of certain class label of classifying type feature is greater than reservation threshold, then this variable is inessential;

4) if the quantity of classifying type feature class label is greater than reservation threshold, then this variable is inessential.

When above-mentioned dimensional characteristics is classification type, for example (,) applicating category, subject matter etc., calculate two groups of statistical values that this feature is corresponding, comprise classification frequency (such as subject matter, class label may be 5 classes, as role playing, shooting class, strategy manages class, intelligence development reasoning class and chess category, the frequency that these five kinds values occur), with the quantity of classification, compare based on the number of class label and the quantity of class label and predetermined threshold, determine whether feature belongs to key character.

By above-mentioned proper vector normalization and the screening of validity feature vector, export the feature that some discriminations are higher, namely important feature.The judgment rule of these key characters is as follows: be a) numeric type for dimensional characteristics, the general coefficient of variation of important feature is larger, and standard deviation is larger; For example standard deviation is the feature of 0, and it is all the same for representing all eigenwerts, and discrimination is 0, and this category feature is just inessential; On the contrary, if standard deviation is larger, so this characteristic area calibration is just large, is this kind ofly characterized as key character; If b) this dimensional characteristics is classification type, frequency and the quantity of important feature general category are all average; The for example sample of certain class, the application image quality of 99% is all 3D, and so this characteristic area calibration is very little, namely inessential, otherwise, be key character.

Based on key character and the forecast model of corresponding channels set up of the current application exported after pre-service, the promotion effect of prediction current application.

Wherein, please consult Fig. 2 further, Fig. 2 is the process flow diagram of the forecast model of the corresponding channels set up according to the characteristic sum of current application that the embodiment of the present invention provides, the promotion effect of prediction current application, as shown in the figure, comprises following sub-step:

S201: according to the feature of current application, and according to the income in the following schedule time of the forecast model prediction current application of the corresponding channels set up.

Income in the following schedule time of current application can be accumulated earnings, and also can be that registered user accumulates cost per capita etc., preferred as one, the embodiment of the present invention spends per capita using registered user's accumulation and illustrates as income.

Here the following schedule time can be the time pre-set, such as 10 days, 15 days etc.Be 15 days for the following schedule time, namely according to the feature of current application, by the forecast model of corresponding channels set up, the income that current application will be promoted in future 15 days can be predicted.

Wherein, according to the feature of current application, and according to the income in the following schedule time of the forecast model prediction current application of the corresponding channels set up, a kind of possible implementation can be the feature based on current application, from forecast model, find one or more history identical or close with current application feature to apply, apply income in the schedule time as a reference with history, estimate the income obtained in the current application schedule time.The feature of such as current application is A, B, C, D, find the close history of feature with it to apply and comprise 2, that history is applied 1 and comprised feature A, B, C respectively, history is applied 2 and is comprised feature B, D, and the income that current application income in the given time can apply 1 and 2 based on history is averaged as with reference to income.Certainly, when specific implementation, the weight size that can further contemplate each feature accurately estimates the income in the current application schedule time further.

S202: whether the income judging in current application following schedule time of predicting is greater than the income median that all history in channels is applied.

Whether the income judging in current application following schedule time of predicting is greater than the income median that all history in channels is applied, if be greater than, then performs S203, otherwise, perform S204.

As the preferred scheme of one, the embodiment of the present invention with registered user's accumulation per capita cost weigh application and whether be applicable to continuing to promote in channel.In other words, for given channel, if certain money is applied in this channel promote the schedule time, the accumulative registered user obtained is more, and particularly these registered users cost is higher, and so this promotion effect being applied in channel is better.

But on the other hand, the flow of each channel has difference, and user crowd is also different, cause the threshold values weighing channel effect (namely registered user spends per capita) variant.Consider the income threshold values being difficult to reasonably define each channel by artificial method, the present invention adopts following strategy to define the revenue threshold of channel: for given channel, collect channel each money in half a year registered user be applied in continuous 15 days to add up to spend per capita, get the threshold values of its median as this channel.That is, example is applied as with game, channel to be reached the standard grade a lot of game in statistics the first half day, produce different incomes (namely the registered user of 15 days is accumulative spends per capita), income belongs to medium (namely income is greater than the median of history) on the upper side and has belonged to game, otherwise belongs to bad game; If namely given game prediction is greater than the median of history in the statistics income of 15 days in the future, be then considered to high-quality game, determine the game being applicable to continuing to promote in channel, otherwise be considered to low-quality game, be not suitable for continuing to promote in channel.

S203: determine that current application is the application being applicable to continuing to promote at channels.

Being greater than when predicting the income in the current application following schedule time income median that all history in channels applies, determining that current application is the application being applicable to continuing to promote at channels.

S204: determine that current application is the application being not suitable for continuing to promote at channels.

Being not more than when predicting the income in the current application following schedule time income median that all history in channels applies, determining that current application is the application being not suitable for continuing to promote at channels.

Wherein, the promotion effect for current application predicts the outcome, and directly can return and predict the outcome accordingly, such as directly return described current application and be applicable to continuing to promote at channels, or described current application is not suitable for continuing to promote at channels.

Also can to return the corresponding mark that predicts the outcome.Such as return 1 and be designated potentiality application, be namely applicable to the application continuing to promote at channels, return 0 and be designated the application of non-potentiality, be namely not suitable for the application continuing to promote at channels.

That is, by said method of the present invention, can predict the promotion effect of application, predict that it promotes the income that can bring within the following schedule time.Be applied as example with game, for each channel, be collected in the game the promoted operation feature of nearest 2 days above.Particularly, collect attribute data and the operation data of game, determine the foundation characteristic of game and derivative feature.According to these features, by the forecast model of corresponding channel set up and measurable and monitor this game and whether be applicable to continuing at this channel promotion, whether extra earning is good such as to identify this game popularization of following 15 days.For example test day is September 1, collects August 30 to this operation data of 2 days on August 31, within 15 days after September 1, whether has good business revenue potentiality by this game of model prediction.If predict the income median that income that this is played is greater than history in this channel and applies, show that these game business revenue potentiality are better, otherwise, show that these game business revenue potentiality are bad.Return and predict the outcome, such as return 1, represent following 15 days and be applicable to continuing to promote in this channel, return 0 representative and within 15 days, be not suitable for continuing to promote in this channel future.

As the preferred implementation of one, wherein, after the promotion effect predicting current application, can also generate and return corresponding Promotion Strategy.

Wherein, Promotion Strategy is promoted for applicable continuation respectively and is not suitable for continuing to promote and returning different Promotion Strategies, specifically, being the candidate popularization list of application returning corresponding current channel: continue the application at channel promotion for being not suitable for, finding out alternative candidate's application; For the game of applicable continuation at channel promotion, identify whether to there is more excellent candidate's application.

Particularly, for applicable continuation in the application of specifying channel promotion, consider because channel resource position is limited, there are some not at the new opplication that this channel was recommended, and these new opplication may produce higher income, by identifying that the candidate that these may produce more high yield applies and returns, selecting more excellent to be applied in this channel promotion for user as Promotion Strategy, thus producing better income.

Be applied as example with game, concrete strategy is as follows:

1) candidate's new game this channel of reaching the standard grade respectively is allowed to test 2 days;

2) collect the operation data of 2 days, utilize model prediction whether to be adapted at this channel promotion;

3) if candidate's new game is adapted at channel promotion, the game that in 2 days, new registration user effort is maximum is therefrom selected to promote.

Continuing specifying the application of channel promotion for being not suitable for, selecting most suitable replacement candidates and doing to promote the application library promoted from waiting to reach the standard grade and changing.Wherein, as an example, the method for a kind of selection of the embodiment of the present invention finds to differ larger application candidate as an alternative with the apply property being not suitable for promoting from application library.The another kind of method selected determines that the history of this channel applies the good application of middle income, finds with the identical or close application candidate's application as an alternative of the apply property that income is good from application library.Certainly, can be other way selection replacement candidates application, the embodiment of the present invention be not exhaustive one by one yet.

Wherein, as the preferred implementation of one, return again after prediction running income (popularization index) sequence is carried out to replacement candidates.The sequence logic of such as replacement candidates, according to the descending sequence of following popularization index, preferentially promotes the high game of desired value.Promoting index is add up pre-capita consumption the new registration user of each channel in nearest 15 days, namely applies the business revenue ability of the overall situation, preferentially promotes in the good application of each platform business revenue ability.

Situation is promoted in the operation adopting method of the present invention can monitor the application in each channel, is promoted the operation benefits of data and the measurable following schedule time (such as following 15 days) by the operation of the application schedule time (such as 2 days).Further, the present invention can generate corresponding Promotion Strategy automatically based on the promotion effect of prediction, reduces the hold-up time be applied in channel inferior, promotes the popularization efficiency of channel.

Wherein, in order to prove the effect that the present invention brings further, respectively the same application promoted being done to contemporaneity by adopting the method for method of the present invention and artificial rule respectively and doing topsis, business revenue income is compared.Be applied as example with game below and carry out effectiveness comparison:

AB is adopted to test the evaluating system performance of multiple months, wherein A group is the Promotion Strategy adopting forecast model, B group is the Promotion Strategy of artificial rule (according to the business revenue of game after 15 days), and two groups of tactics of the game quantity promoted are consistent, and contemporaneity is promoted.For example certain channel given, has two to promote resource-niche, and wherein the Promotion Strategy of first resource-niche adopts the Promotion Strategy that A group forecast model provides, and the Promotion Strategy of second resource-niche adopts the Promotion Strategy of artificial B group.These two groups of resource-niches have different renewal frequencies and strategy; By statistics by the income of this channel resource position correspondence, for weighing two kinds of tactful qualities.Evaluation metrics is the cost per capita that each channel new registration user in 15 days adds up; Wherein adopt the Promotion Strategy that provides of forecast model, spend per capita provide 85 yuan of strategy by artificial rule and rise to 98 yuan in August, 2015, September rises to 101 yuan by artificial rule 90 yuan of providing strategy.

Visible, method of the present invention, can stop poor quality being applied in popularization in channel in time by prediction effect, increase substantially the utilization factor of channel resource position and improve business revenue benefit.

The method of the above embodiment of the present invention, that the forecast model of the corresponding channel set up based on the embodiment of the present invention realizes, in order to further distinct method of the present invention, the method further embodiment of the present invention being set up to the forecast model of corresponding channel below in conjunction with accompanying drawing is described in detail.

Refer to Fig. 3, Fig. 3 is the process flow diagram of the method for the forecast model of the corresponding channel of foundation that the embodiment of the present invention provides, and as shown in the figure, the method for the forecast model of the corresponding channel of foundation of the present embodiment comprises the following steps:

S301: collect attribute data that the history on each channels applies and the operation data in the schedule time respectively.

For each channels, collect the attribute data applied of history and the operation data in the schedule time respectively, these data set up basis and the information source of forecast model.

The attribute data that history is applied refers to the attribute data of application itself.Such as include but not limited to it is application subject matter, style, image quality, type, operation platform and form etc.History is applied the operation data of the schedule time in channels and is referred to that being applied in each channel resource position does in the process promoted, the operation data of record user application in each money application of acquisition.These data can portray the fancy grade of user to application, and the business revenue ability of application.These operation datas comprise user application in the schedule time registration behavior, login behavior, supplement behavior and consumer behavior etc. with money.

Wherein, the schedule time can be the time pre-set.When based on Modling model, for be all that history is applied, therefore, the data that can collect can be long period scopes so that model set up more accurate.The such as schedule time can be set to 15 days.Namely put into channels and promote to collect afterwards for 15 days by history is applied and be applied in the operation data of in this channel 15 days.Certainly, can proper extension or shorten this schedule time according to actual needs.

S302: the attribute data apply history and the operation data in the schedule time process, obtains the income that characteristic sum history that history applies applies the schedule time.

Processed by the attribute data applied history and the operation data in the schedule time, thus the history that can determine the feature and correspondence setting up forecast model applies the income (also can be defined as target labels) of the schedule time, as the training data of Modling model.

Particularly, being applied as example to play, a given channel, collecting the operation data of playing and adding up 2 days a few days ago in this channel, for building characteristic of correspondence vector; On the other hand, collect game at this channel statistics operation situation of 15 days in the future, as business revenue, user's registration amount etc., for building the target labels of training.For example statistics day is July 1, collects this game in the statistics characteristic of 2 days a few days ago (attributive character that namely operation data of this game of period on June 29 to June 30 in certain channel is relevant with game).On the other hand, collect in adding up 15 days in the future the operation situation data obtained at this channel of playing, and judge the target labels (such as potentiality game or the game of non-potentiality) of this game according to service logic.If game operation data meets rule, be noted as 1 (potentiality game), otherwise be noted as 0 (non-potentiality game).Based on the historical data of different statistics days, can obtain the characteristic of game and the target labels of this game correspondence, these data are used for training pattern.

Wherein, the feature that above-described history is applied, comprises based on attribute data and the determined foundation characteristic of operation data and the derivative feature that built by foundation characteristic.Foundation characteristic comprises accumulative enrollment that in the schedule time, channels brings to current application, login times, login number of days, logs in duration, login trend, supplements number of times with money, supplements number of days with money, recharge amount, supplement trend, consumption number of times, consumption number of days, spending amount, propensity to consume etc. with money, derivative feature comprises paying next day rate that channels brings to described current application, next day supplements number per capita with money, the same day conversion ratio, next day conversion ratio and next day retention ratio etc.These derivative features can weigh the characteristic of application comprehensively.

The income applying the schedule time can be accumulated earnings, and also can be that registered user accumulates cost per capita etc., preferred as one, the embodiment of the present invention spends per capita using registered user's accumulation and illustrates as income.

Here the schedule time can be the time pre-set, such as 10 days, 15 days etc.Be 15 days for the schedule time, namely according to the operation data that history is applied, can determine that history applies the risk return profile in 15 days, registered user's accumulation that such as history was applied in 15 days spends per capita.

S303: all history is applied the income of the schedule time and got median as threshold value on channels, the income that the characteristic sum history applied history is applied carries out the forecast model that learning training obtains corresponding channels.

A given channel, based on all historical usage adding up the business revenue data in the schedule time in the future, thus can determine that history applies the income of the schedule time.With the training objective of this income definition channel, namely quantize the game at this channel with high value.According to service logic, usually spend per capita with registered user and weigh application and whether be adapted at channel promotion.In other words, for given channel, promote 15 days if certain money is applied in this channel, the accumulative registered user obtained is more, and particularly these registered users cost is higher, and so this promotion effect being applied in channel is better.On the other hand, the flow of each channel has difference, and user crowd is also different, causes the threshold values weighing channel effect (namely registered user spends per capita) variant.

Be difficult to based on by artificial method the threshold values reasonably defining each channel, the embodiment of the present invention adopts following strategy to define the training objective of channel.For given channel, collect its each money in half a year registered user be applied in continuous 15 days and add up to spend per capita, get the threshold values of its median as this channel.That is, channel to be reached the standard grade a lot of application in statistics the first half day, produce different incomes (namely the registered user of 15 days is accumulative spends per capita), apply middle income belong to medium (namely income is greater than the median of history) on the upper side belong to potentiality application, otherwise belong to non-potentiality application.If namely given the be applied in statistics income of 15 days is in the future greater than the median of history, is then considered to high-quality application (such as can be labeled as 1), otherwise is considered to low-quality application (such as can be labeled as 0).

The forecast model obtained to make training more accurately, rationally.The embodiment of the present invention is before the income that the characteristic sum history applied history is applied carries out learning training, pre-service is carried out to the feature that history is applied, with in the feature applied from history, get rid of insignificant feature, filter out key character for model training.

Wherein, can carry out by any one in proper vector normalization and the screening of validity feature vector or two kinds of modes the feature combined history is applied and carry out pre-service.

Aforesaid dimensional characteristics is numeric type, such as recharge amount, logs in duration etc., calculates two groups of statistical values that this feature is corresponding, comprise the coefficient of variation (standard deviation/mean value of=normal distribution) and standard deviation.Compared by the coefficient of variation and standard deviation and predetermined threshold, determine whether feature belongs to key character.

Above-mentioned dimensional characteristics is classification type, such as applicating category, subject matter etc., calculate two groups of statistical values that this feature is corresponding, comprise classification frequency (such as subject matter, class label may be 5 classes, as role playing, shooting class, strategy manages class, intelligence development reasoning class and chess category, the frequency that these five kinds values occur), with the quantity of classification, compare based on the number of class label and the quantity of class label and predetermined threshold, determine whether feature belongs to key character.

By above-mentioned proper vector normalization and the screening of validity feature vector, export the feature that some discriminations are higher, namely important feature.The judgment rule of these key characters is as follows: be a) numeric type for dimensional characteristics, the general coefficient of variation of important feature is larger, and standard deviation is larger; For example standard deviation is the feature of 0, and it is all the same for representing all eigenwerts, and discrimination is 0, and this category feature is just inessential; On the contrary, if standard deviation is larger, so this characteristic area calibration is just large; If b) this dimensional characteristics is classification type, frequency and the quantity of important feature general category are all average; The for example sample of certain class, the application image quality of 99% is all 3D, and so this characteristic area calibration is very little, namely inessential.

The key character applied based on the history exported after pre-service carries out model training.

As the preferred implementation of one, the embodiment of the present invention carries out to the income that the characteristic sum history that history is applied is applied the forecast model that learning training obtains corresponding channels with support vector machine classifier (SVM).This sorter can calculate the remote sensing (schematic diagram with reference to figure 4) of positive and negative training sample (i.e. suitable and improper sample) in vector space.That is, by sorter, being adapted at the application of channel promotion and being not suitable for having carried out cutting in the application of channel promotion.Apply based on history in this channel in forecast model, feature based set history of forming is applicable in applying promoting and be not suitable for the large class of popularization two.

And the embodiment of the present invention trains the forecast model obtained also to be equivalent to a support vector machine classifier, identify the classification of test sample book according to the division of the plane of forecast model.Briefly, lineoid is suitable and improper application cutting; The feature that a given test sample book is current, if this feature in vector space in the training sample suitable applications side, namely the situation of this sample follows the state of suitable applications in history very similar, and so this application is predicted to be and is adapted at promoting in this channel, and the value namely continuing to promote is larger.Accordingly, if given sample in vector space in the training sample improper application side, be predicted to be improper continuation and promote in channel.

The Forecasting Methodology of the application effect that the above embodiment of the present invention provides, according to attribute data and the operation data of current application in channels in the schedule time of current application, determine the feature of current application, according to the forecast model of the corresponding channels that the characteristic sum of current application has been set up, the promotion effect of prediction current application.By such mode, poor quality can be stopped in time being applied in popularization in channel by prediction effect, increase substantially the utilization factor of channel promotion resource-niche, and effectively improve business revenue benefit.

Refer to Fig. 5, Fig. 5 is the structural representation of the prognoses system of a kind of application effect that the embodiment of the present invention provides, as shown in the figure, the prognoses system 100 of the application effect of the present embodiment comprises acquisition module 11, determination module 12 and prediction module 13, wherein:

Acquisition module 11 is for obtaining attribute data and the operation data of current application in channels in the schedule time of current application.

Acquisition module 11 obtains the attribute data of current application and the operation data in channels in the schedule time, and these data are to the application promotion effect basis of predicting and information source.

Determination module 12, for according to attribute data and operation data, determines the feature of current application.

Specifically, determination module 12 is according to attribute data and operation data, the foundation characteristic determining current application and the derivative feature built based on foundation characteristic.Wherein, foundation characteristic comprises accumulative enrollment that in the schedule time, channels brings to current application, login times, login number of days, logs in duration, login trend, supplements number of times with money, supplements number of days with money, recharge amount, supplement trend, consumption number of times, consumption number of days, spending amount, propensity to consume etc. with money, derivative feature comprises paying next day rate that channels brings to described current application, next day supplements number per capita with money, the same day conversion ratio, next day conversion ratio and next day retention ratio etc.

The forecast model of corresponding channels of prediction module 13 for having set up according to the characteristic sum of current application, the promotion effect of prediction current application.

The forecast model of the corresponding channels that prediction module 13 has been set up according to the characteristic sum of current application, the promotion effect of prediction current application.

Wherein, please consult Fig. 6 further, Fig. 6 is the structural representation of the prediction module 13 of the present embodiment, and as shown in the figure, the prediction module 13 of the present embodiment specifically can comprise predicting unit 131, judging unit 132 and determining unit 133, wherein:

Predicting unit 131 for the feature according to current application, and predicts the income in the following schedule time of described current application according to the forecast model of the corresponding channels set up.

Whether judging unit 132 is greater than for the income judging in current application following schedule time of predicting the income median that all history in channels applies.

Whether the income that judging unit 132 judges in current application following schedule time of predicting is greater than the income median that all history in channels is applied, and determines whether current application is applicable to continuing to promote in this channel with this.

But on the other hand, the flow of each channel has difference, and user crowd is also different, cause the threshold values weighing channel effect (namely registered user spends per capita) variant.Consider the income threshold values being difficult to reasonably define each channel by artificial method, the present invention adopts following strategy to define the revenue threshold of channel: for given channel, collect channel each money in half a year registered user be applied in continuous 15 days to add up to spend per capita, get the threshold values of its median as this channel.That is, example is applied as with game, channel to be reached the standard grade a lot of game in statistics the first half day, produce different incomes (namely the registered user of 15 days is accumulative spends per capita), income belongs to medium (namely income is greater than the median of history) on the upper side and has belonged to game, otherwise belongs to bad game; If namely given game prediction is greater than the median of history in the statistics income of 15 days in the future, be then considered to high-quality game, determine the application being applicable to continuing to promote in channel, otherwise be considered to low-quality game, be not suitable for continuing to promote in channel.

When determining unit 133 is greater than for the income within the current application the predicted following schedule time income median that all history in channels applies, determine that current application is the application being applicable to continuing to promote at channels, otherwise, determine that current application is the application being not suitable for continuing to promote at channels.

Be greater than when predicting the income in the current application following schedule time income median that all history in channels applies, determining unit 133 determines that current application is the application being applicable to continuing to promote at channels.Be not more than when predicting the income in the current application following schedule time income median that all history in channels applies, determining unit 133 determines that current application is the application being not suitable for continuing to promote at channels.

Preferably, described prediction module 13 can further include pretreatment unit 134, and pretreatment unit 134 is for carrying out pre-service by least one method in proper vector normalization and the screening of validity feature vector to the feature of current application.

Before the promotion effect of prediction current application, the feature of pretreatment unit 134 pairs of current application carries out pre-service, and these pretreatment operation can allow and predict the outcome more accurately, rationally.Wherein, pre-service can be carried out by least one method in proper vector normalization and the screening of validity feature vector to the feature of current application, thus determine that in the feature of current application, key character is for predicting the promotion effect of current application.

Wherein, please continue to refer to Fig. 5, as the preferred implementation of one, the prognoses system of the present embodiment can further include and returns module 14, returns module 14 for returning predicting the outcome of application effect.

Wherein, promotion effect for current application predicts the outcome, return module 14 directly to return and predict the outcome accordingly, such as directly return described current application and be applicable to continuing to promote at channels, or described current application is not suitable for continuing to promote at channels.

Also can to return the corresponding mark that predicts the outcome.Such as return 1 and be designated potentiality application, be namely applicable to the application continuing to promote at channels, return mark 0 and be designated the application of non-potentiality, be namely not suitable for the application continuing to promote at channels.

As the preferred implementation of one, return module 14 also for after the promotion effect predicting current application, while returning the promotion effect of prediction or afterwards in other words, can also generate and return corresponding Promotion Strategy.

Be applied as example with game, concrete strategy is as follows:

Wherein, as the preferred implementation of one, return after module 14 pairs of replacement candidates carry out prediction running income (popularization index) sequence and return again.The sequence logic of such as replacement candidates, according to the descending sequence of following popularization index, preferentially promotes the high game of desired value.Promoting index is add up pre-capita consumption the new registration user of each channel in nearest 15 days, namely applies the business revenue ability of the overall situation, preferentially promotes in the good application of each platform business revenue ability.

Wherein, the realization of method of the present invention, that the forecast model of the corresponding channel set up based on the embodiment of the present invention realizes, therefore, please continue to refer to Fig. 5, the prognoses system of the application effect of the present embodiment also comprises model training module 15 further, wherein, please further combined with consulting Fig. 7, Fig. 7 is the structural representation of the model training module 15 that the embodiment of the present invention provides, as shown in the figure, the model training module 15 of the present embodiment comprises data collection module 151, processing unit 152 and training unit 153, wherein:

The attribute data that data collection module 151 is applied for the history of collecting respectively on each channels and the operation data in the schedule time.

For each channels, data collection module 151 collects attribute data that history applies and the operation data in the schedule time respectively, and these data set up basis and the information source of forecast model.

Processing unit 152 processes for the attribute data applied history and the operation data in the schedule time, obtains the income that characteristic sum history that history applies applies the schedule time.

Processing unit 152 is processed by the attribute data applied history and the operation data in the schedule time, thus the history that can determine the feature and correspondence setting up forecast model applies the income (also can be defined as target labels) of the schedule time, as the training data of Modling model.

The income that training unit 153 applies the schedule time for history all on channels goes median as threshold value, and the income that the characteristic sum history applied history is applied carries out the forecast model that learning training obtains corresponding channels.

The forecast model obtained to make training more accurately, rationally.Embodiment of the present invention processing unit 152 at training unit 153 before the income that the characteristic sum history applied history is applied carries out learning training, pre-service is carried out to the feature that history is applied, with in the feature applied from history, get rid of insignificant feature, filter out key character for model training.

Wherein, processing unit 152 can carry out by any one in proper vector normalization and the screening of validity feature vector or two kinds of modes the feature combined history is applied and carry out pre-service.

The key character that training unit 153 is applied based on the history exported after pre-service carries out model training.

As the preferred implementation of one, embodiment of the present invention training unit 153 carries out to the income that the characteristic sum history that history is applied is applied the forecast model that learning training obtains corresponding channels with support vector machine classifier (SVM).This sorter can calculate the remote sensing (schematic diagram with reference to figure 4) of positive and negative training sample (i.e. suitable and improper sample) in vector space.That is, by sorter, being adapted at the application of channel promotion and being not suitable for having carried out cutting in the application of channel promotion.Apply based on history in this channel in forecast model, feature based set history of forming is applicable in applying promoting and be not suitable for the large class of popularization two.

More than the Forecasting Methodology of application effect that provides of the embodiment of the present invention and the detailed description of system, be appreciated that, the embodiment of the present invention is by according to the attribute data of current application and the operation data of current application in channels in the schedule time, determine the feature of current application, according to the forecast model of the corresponding channels that the characteristic sum of current application has been set up, the promotion effect of prediction current application.By such mode, poor quality can be stopped in time being applied in popularization in channel by prediction effect, increase substantially the utilization factor of channel promotion resource-niche, and effectively improve business revenue benefit.

In several embodiment provided by the present invention, should be understood that, disclosed system, apparatus and method, can realize by another way.Such as, device embodiment described above is only schematic, such as, the division of described module or unit, be only a kind of logic function to divide, actual can have other dividing mode when realizing, such as multiple unit or assembly can in conjunction with or another system can be integrated into, or some features can be ignored, or do not perform.Another point, shown or discussed coupling each other or direct-coupling or communication connection can be by some interfaces, and the indirect coupling of device or unit or communication connection can be electrical, machinery or other form.

The described unit illustrated as separating component or can may not be and physically separates, and the parts as unit display can be or may not be physical location, namely can be positioned at a place, or also can be distributed in multiple network element.Some or all of unit wherein can be selected according to the actual needs to realize the object of the present embodiment scheme.

In addition, each functional unit in each embodiment of the present invention can be integrated in a processing unit, also can be that the independent physics of unit exists, also can two or more unit in a unit integrated.Above-mentioned integrated unit both can adopt the form of hardware to realize, and the form of SFU software functional unit also can be adopted to realize.

If described integrated unit using the form of SFU software functional unit realize and as independently production marketing or use time, can be stored in a computer read/write memory medium.Based on such understanding, the part that technical scheme of the present invention contributes to prior art in essence in other words or all or part of of this technical scheme can embody with the form of software product, this computer software product is stored in a storage medium, comprising some instructions in order to make a computer equipment (can be personal computer, server, or the network equipment etc.) or processor (processor) perform all or part of step of method described in each embodiment of the present invention.And aforesaid storage medium comprises: USB flash disk, portable hard drive, ROM (read-only memory) (ROM, Read-OnlyMemory), random access memory (RAM, RandomAccessMemory), magnetic disc or CD etc. various can be program code stored medium.

The foregoing is only embodiments of the invention; not thereby the scope of the claims of the present invention is limited; every utilize instructions of the present invention and accompanying drawing content to do equivalent structure or equivalent flow process conversion; or be directly or indirectly used in other relevant technical fields, be all in like manner included in scope of patent protection of the present invention.

Claims

1. a Forecasting Methodology for application effect, is characterized in that, described method comprises:

Obtain attribute data and the operation data of described current application in channels in the schedule time of current application;

According to described attribute data and described operation data, determine the feature of described current application;

According to the forecast model of the described channels of correspondence that the characteristic sum of described current application has been set up, predict the promotion effect of described current application.

2. the Forecasting Methodology of application effect according to claim 1, is characterized in that, the attribute data of described current application comprises at least one in application subject matter, style, image quality, type, operation platform and form; Described operation data comprises the registration behavior of user application, login behavior, at least one of supplementing with money in behavior and consumer behavior;

Described according to described attribute data and described operation data, determine that the feature of described current application comprises:

According to described attribute data and described operation data, determine the foundation characteristic of described current application and derivative feature, wherein, described foundation characteristic comprises the accumulative enrollment that in the described schedule time, described channels brings to described current application, login times, log in number of days, log in duration, login trend, supplement number of times with money, supplement number of days with money, recharge amount, supplement trend with money, consumption number of times, consumption number of days, spending amount, at least one in the propensity to consume, described derivative feature comprises the paying next day rate that described channels brings to described current application, next day supplements number with money per capita, the same day conversion ratio, next day conversion ratio and next day retention ratio at least one.

3. Forecasting Methodology according to claim 1, is characterized in that, the forecast model of the described described channels of correspondence set up according to the characteristic sum of described current application, predicts that the promotion effect of described current application comprises:

According to the feature of described current application, and predict the income in the following schedule time of described current application according to the forecast model of the described channels of the correspondence set up;

Whether the income judging in described current application following schedule time of predicting is greater than the income median that all history in described channels is applied;

When the income in the described current application the predicted following schedule time is greater than the income median that all history in described channels applies, determine that described current application is the application being applicable to continuing to promote at described channels, otherwise, determine that described current application is the application being not suitable for continuing to promote at described channels.

4. Forecasting Methodology according to claim 3, is characterized in that, according to the forecast model of the described channels of correspondence that the characteristic sum of described current application has been set up, before predicting the promotion effect of described current application, also comprises:

By at least one method in proper vector normalization and the screening of validity feature vector, pre-service is carried out to the feature of described current application.

5. Forecasting Methodology according to claim 4, is characterized in that, describedly carries out pre-service by proper vector normalization to the feature of described current application and comprises:

By maximal value-Returning to one for minimum value, pre-service is carried out to the feature of described current application;

Describedly pre-service carried out to the feature of described current application comprise by the screening of validity feature vector: the key character being filtered out described current application by any one following mode:

Numeric type Feature change coefficient is less than predetermined threshold, determines that described numeric type feature is inessential; Or

Numeric type character difference is less than predetermined threshold, determines that described numeric type feature is inessential; Or

The class label of classifying type feature is greater than predetermined threshold, determines that described classifying type feature is inessential; Or

The quantity of the class label of classifying type feature is greater than predetermined threshold, determines that described classifying type feature is inessential.

6. Forecasting Methodology according to claim 1, is characterized in that, before the step of the attribute data of the described application of described acquisition and the operation data in the channels schedule time, also comprises:

Collect attribute data that the history on each channels applies and the operation data in the schedule time respectively;

The attribute data apply described history and the operation data in the schedule time process, and obtain history described in characteristic sum that described history applies and apply the income of the schedule time;

On described channels, all history is applied the income of the schedule time and is got median as threshold value, and the income that described in the characteristic sum apply described history, history is applied carries out the forecast model that learning training obtains corresponding described channels.

7. Forecasting Methodology according to claim 6, it is characterized in that, on described channels, all history is applied the income of the schedule time and is got median as threshold value, and the income that described in the characteristic sum apply described history, history is applied is carried out the forecast model that learning training obtains corresponding described channels and comprised:

On described channels, all history is applied the income of the schedule time and is got median as threshold value, and the income that described in the characteristic sum adopting support vector machine classifier to apply described history, history is applied carries out the forecast model that learning training obtains corresponding described channels.

8. Forecasting Methodology according to claim 1, is characterized in that, described method also comprises:

According to the promotion effect of described current application, return corresponding Promotion Strategy.

9. Forecasting Methodology according to claim 8, is characterized in that, the described promotion effect according to described current application, and the step returning corresponding Promotion Strategy comprises:

According to the promotion effect of described current application, return the candidate popularization list of application of corresponding described current channels.

10. Forecasting Methodology according to claim 9, is characterized in that, the candidate popularization list of application of the described current channels of described correspondence is to predict that operation benefits arranges from high to low.

The prognoses system of 11. 1 kinds of application effects, is characterized in that, described system comprises acquisition module, determination module and prediction module, wherein:

Described acquisition module is for obtaining attribute data and the operation data of described current application in channels in the schedule time of current application;

Described determination module is used for according to described attribute data and described operation data, determines the feature of described current application;

Described prediction module is used for the forecast model of the described channels of correspondence set up according to the characteristic sum of described current application, predicts the promotion effect of described current application.

12. prognoses systems according to claim 11, is characterized in that, described prognoses system also comprises and returns module, described in return module for the promotion effect according to described current application, return corresponding Promotion Strategy.