CN107886369A

CN107886369A - A kind of method, apparatus and electronic equipment for marketing activity effect prediction

Info

Publication number: CN107886369A
Application number: CN201711222432.7A
Authority: CN
Inventors: 樊翀
Original assignee: Koubei Shanghai Information Technology Co Ltd
Current assignee: Koubei Shanghai Information Technology Co Ltd
Priority date: 2017-11-29
Filing date: 2017-11-29
Publication date: 2018-04-06
Also published as: WO2019105226A1

Abstract

This application discloses a kind of method for marketing activity effect prediction, the history marketing data by obtaining different shops is used as input sample, calculates the marketing activity effect data of different shops before and after marketing activity；The marketing activity effect data of target shops are chosen, the marketing activity effect for handling to obtain the target shops by data smoothing estimates initial value；Marketing activity effect data based on different shops, discount rate regulatory factor is built by linear fit, initial value is estimated using the discount rate regulatory factor to the marketing activity effect to calibrate, the calibration value of acquisition is the target shops marketing activity effect discreet value, and marketing activity effect can not be more accurately estimated in real time according to the change of discount rate so as to solve the problems, such as that existing marketing activity effect estimates scheme.

Description

A kind of method, apparatus and electronic equipment for marketing activity effect prediction

Technical field

The application is related to O2O catering industry marketing domains, and in particular to a kind of method for marketing activity effect prediction. The application is related to one kind and is used for marketing activity effect prediction meanss simultaneously, and a kind of pre- for marketing activity effect for realizing The electronic equipment of the method for survey.

Background technology

In increasingly keen competition environment, the marketing methods of O2O catering industry businessmans emerge in an endless stream, the rush of various item Pin activity is constantly released.Businessman is when a marketing activity is created, it is to be understood that the movable effect, i.e., its can give shops Any interests is brought to businessman, this can allow businessman to accomplish to know what's what, and specify the value of marketing activity, in a planned way follow the prescribed order Do marketing activity.

At present, estimating for marketing activity effect, it is common practice that with reference to the similar marketing activity of the history of the shops Effect data, i.e., former effect should be also similar with the animation effect created now, and discount rate is treated as marketing activity A dimension, the effect of history similar active is calculated by the account of the history of simple geo-statistic difference discount rate activity.

Existing marketing activity prediction scheme, discount rate change is turned to key factor and quantitatively estimates active belt Effect, thus the problem of marketing activity effect can not more accurately being estimated according to the change of discount rate in real time be present.It is true On, discount rate influences very big, example for the BI indexs (Business Indicators, or business index) of many animation effects Such as a certificate for having threshold, threshold typically can be higher than shops's visitor's unit price, if the certificate discount dynamics is enough big, it is sufficient to and it is attractive, that Many users are in order to reach will being consumed more using threshold for certificate, so as to improve shops's visitor's unit price, for another example for multiple purchase Rate, if the consumption sent send certificate attractive enough, the secondary of stretchy user is turned one's head, and purchase rate is naturally toward rising again.

The content of the invention

The application provides a kind of method for marketing activity effect prediction, is estimated with solving existing marketing activity effect The problem of scheme can not more accurately estimate marketing activity effect in real time according to the change of discount rate.

The application provides a kind of device for marketing activity effect prediction in addition.

The application also provides a kind of electronic equipment for realizing the method for marketing activity effect prediction.

The application provides a kind of method for marketing activity effect prediction, including：

The history marketing data of different shops is obtained as input sample, calculates the marketing of different shops before and after marketing activity Animation effect data；

The marketing activity effect data of target shops are chosen, the marketing for handling to obtain the target shops by data smoothing is lived Dynamic result estimate initial value；

Marketing activity effect data based on different shops, discount rate regulatory factor is built by linear fit, uses institute State discount rate regulatory factor initial value is estimated to the marketing activity effect and calibrate, the calibration value of acquisition is the target door Shop marketing activity effect discreet value.

Optionally, the marketing activity effect data, it is business index lifting values of the shops before and after marketing activity, at least Including following any indexs：Objective unit price lifting percentage point and again purchase rate lifting percentage point.

Optionally, the marketing activity effect data for choosing target shops, including following processing：

Similar shops's model of target shops is established, the similar door of target shops is determined based on similar shops's model Shop, the marketing activity effect of the target shops and its similar shops is chosen from the marketing activity effect data of the different shops Fruit data.

Optionally, the similar shops's model for establishing target shops, including：

Shops's characteristic value is obtained from one or more dimensions, the shops similarity row based on KNN algorithms or based on setting Sequence method, similar shops's model of the target shops is established according to the characteristic value；Wherein, one or more of dimensions, Including at least following any dimensions：Average daily stroke count, pen unit price, the affiliated Merchant ID of shops, position feature.

Optionally, shops's sequencing of similarity method of the setting, including：

The similar shops of candidate of the target shops is filtered out by preparatory condition；

According to each dimensional characteristics value of the similar shops of candidate, each dimension in similar shops sorts proportion, it is and similar The characteristic value degree of closeness of shops's dimension corresponding with the target shops, determines shops's sequencing of similarity.

Optionally, the preparatory condition, including：

Shops belongs to Tong Cheng with target；And/or

Belong to same classification with target shops.

Optionally, the marketing activity effect for handling to obtain the target shops by data smoothing estimates initial value, bag Include following smoothing processings：The average of visitor's unit price lifting percentage point of target shops and its similar shops is taken as the target shops Objective unit price estimate initial value, take the multiple purchase rate of target shops and its similar shops lifting percentage point average as the target door The multiple purchase rate in shop estimates initial value.

Optionally, the marketing activity effect data for choosing target shops, it is to choose granting amount and the amount of checking and writing off is more than and set The animation effect data of the history marketing activity of definite value.

Optionally, the marketing activity effect data based on different shops, discount rate is built by linear fit and adjusted The factor, including：Visitor's unit price based on history marketing activity lifts percentage point and again purchase rate lifting percentage point, passes through least square method Linear fit is carried out, draws the calculation formula that discount rate regulatory factor changes with discount rate.

Optionally, it is described that linear fit is carried out by least square method, show that discount rate regulatory factor changes with discount rate Calculation formula, including：

Using discount rate and marketing activity effect, the intercept and slope of straight line are fitted, then discount rate regulatory factor is with folding Button rate change calculation formula be：Discount rate regulatory factor is multiplied by discount rate equal to slope adds intercept；

Wherein, intercept subtracts slope equal to animation effect average value and is multiplied by discount rate average value；By the folding of each similar shops Button rate deviation from average is multiplied by animation effect deviation from average, after summation again divided by each similar shops discount rate sum of sguares of deviation from mean, produce To slope；The deviation from average is the distance of actual value deviation average.

Optionally, the use discount rate regulatory factor is estimated initial value to the marketing activity effect and calibrated, Including using following calibration equations：

The marketing activity effect that the target shops marketing activity effect discreet value is equal to the target shops estimates initial value It is multiplied by discount rate regulatory factor.

Optionally, in addition to the target shops marketing activity effect discreet value is directed to, enters one using threshold regulatory factor Step amendment；Wherein, the threshold regulatory factor, it is that the order numbers of threshold are met according to history marketing activity and meet visitor's unit price The factor that the ratio calculation threshold of order numbers influences on marketing activity effect, is determined by following manner：

Above-mentioned ratio according to corresponding to the threshold that target shops marketing activity is set determines it；

Corresponding historical act effect, is obtained when the use of historical act effect corresponding to the ratio divided by ratio being 1 The numerical value, as threshold regulatory factor.

Optionally, it is described to be directed to the target shops marketing activity effect discreet value, it is further using threshold regulatory factor Amendment, including：The threshold regulatory factor is multiplied by the target shops marketing activity effect discreet value, obtains target shops By the marketing activity effect discreet value further corrected.

Optionally, the marketing activity, including at least following any Activity Types：Completely subtract, often completely subtract, consume and send.

Optionally, it is described calculate marketing activity before and after different shops marketing activity effect data, refer to be sought according to shops Pin activity starts visitor's unit price and purchase rate again in front and rear certain period of time, calculates visitor unit price of the shops before and after marketing activity respectively Lift percentage point and again purchase rate lifting percentage point.

The application also provides a kind of method for the prediction of businessman's animation effect, including：

The characteristic value of shops is obtained from following any dimensions：Average daily stroke count, pen unit price, the affiliated Merchant ID of shops, position are special Sign, according to the characteristic value, shops's sequencing of similarity method based on KNN algorithms or based on setting, foundation activity is applicable door Similar shops's model in shop；

The similar shops of shops is applicable based on similar shops's model determination activity, obtains similar shops's marketing activity effect Fruit data, handle to obtain the marketing activity effect predicted value that the activity is applicable shops by data smoothing；

The marketing activity effect value for being applicable activity shops polymerize, and obtains businessman's animation effect predicted value.

The similar shops of candidate that the activity is applicable shops is filtered out by preparatory condition；

According to each dimensional characteristics value of the similar shops of candidate, each dimension in similar shops sorts proportion, it is and similar Shops and the activity are applicable the characteristic value degree of closeness of the corresponding dimension of shops, determine shops's sequencing of similarity.

Optionally, the preparatory condition, including：

Applicable shops belongs to Tong Cheng with activity；And/or

Belong to same classification using shops with activity.

Optionally, each dimension proportion in similar shops sorts, is to determine each dimension according to following orders of priority Spend proportion：

The activity is using shops's sheet as most like shops；

Average daily stroke count similarity；The average daily stroke count similarity, equal to high average daily stroke count divided by low average daily stroke count；

The monovalent similarity of pen；The pen unit price similarity, monovalent equal to high pen divided by low pen unit price；

It is identical with the applicable affiliated Merchant ID of shops of the activity；

The distance that shops is applicable with activity is less than setting value；

Applicable shops belongs to an administrative area together with activity.

The application also provides a kind of device for marketing activity effect prediction, including：

Historical data acquiring unit, the history marketing data for obtaining different shops calculate marketing as input sample The marketing activity effect data of different shops before and after activity；

Unit tentatively is estimated, for choosing the marketing activity effect data of target shops, handles to obtain by data smoothing The marketing activity effect of the target shops estimates initial value；

Discreet value alignment unit, for the marketing activity effect data based on different shops, built and rolled over by linear fit Button rate regulatory factor, initial value is estimated to the marketing activity effect using the discount rate regulatory factor and calibrated, acquisition Calibration value is the target shops marketing activity effect discreet value.

The application also provides a kind of electronic equipment, including：

Memory, and processor；

The memory is used to store computer executable instructions, and the processor can perform for performing the computer Instruction：

Compared with prior art, the application has advantages below：

What the application provided is used for the method, apparatus and electronic equipment of marketing activity effect prediction, by obtaining not fellow disciple The history marketing data in shop is as input sample, the marketing activity effect data of different shops before and after calculating marketing activity；Choose The marketing activity effect data of target shops, the marketing activity effect for handling to obtain the target shops by data smoothing are estimated just Value；Marketing activity effect data based on different shops, discount rate regulatory factor is built by linear fit, uses the discount Rate regulatory factor is estimated initial value to the marketing activity effect and calibrated, and the calibration value of acquisition is the target shops marketing Animation effect discreet value, the marketing activity effect of target shops is obtained according to the history off-line data of the marketing activity effect of shops Initial value is estimated, and discount rate regulatory factor is constructed using these off-line datas, it is further to the marketing activity effect discreet value Regulation, so as to solve existing marketing activity effect estimate scheme can not according to the change of discount rate and in real time more accurately The problem of estimating marketing activity effect.

Brief description of the drawings

Fig. 1 is the process chart for the method for being used for marketing activity effect prediction that the embodiment of the present application provides；

Fig. 2 is the system of the actual deployment citing for the method for being used for marketing activity effect prediction that the embodiment of the present application provides Schematic diagram；

Fig. 3 is interaction effect figure after the embodiment of the present application actual deployment；

Fig. 4 is the process chart for the method for being used for the prediction of businessman's marketing activity effect that the embodiment of the present application provides；

Fig. 5 is the schematic diagram for the device for being used for marketing activity effect prediction that the embodiment of the present application provides；

Fig. 6 is that the electronics for being used to realize the method for marketing activity effect prediction that the embodiment of the present application provides is set Standby schematic diagram.

Embodiment

Many details are elaborated in the following description in order to fully understand the application.But the application can be with Much it is different from other manner described here to implement, those skilled in the art can be in the situation without prejudice to the application intension Under do similar popularization, therefore the application is not limited by following public specific implementation.

The application provides a kind of method for marketing activity effect prediction.The application is related to a kind of for work of marketing simultaneously The device of dynamic effect prediction and a kind of electronic equipment for being used for realization and being used for marketing activity effect prediction.In following embodiment In be described in detail one by one.

The application one embodiment provides a kind of method for marketing activity effect prediction.

A kind of side for marketing activity effect prediction provided below in conjunction with Fig. 1 to Fig. 3 the application one embodiment The embodiment of method illustrates.Wherein Fig. 1 is the method for being used for marketing activity effect prediction that the application one embodiment provides Process chart；Fig. 2 is the practical application for the method for being used for marketing activity effect prediction that the application one embodiment provides The system schematic of citing；Fig. 3 is interaction effect figure after the application one embodiment actual deployment.

Method shown in Fig. 1 for marketing activity effect prediction, including：

Step S101, the history marketing data of different shops is obtained as input sample, calculates difference before and after marketing activity The marketing activity effect data of shops；

Step S102, the marketing activity effect data of target shops are chosen, handle to obtain the target door by data smoothing The marketing activity effect in shop estimates initial value；

Step S103, the marketing activity effect data based on different shops, by linear fit build discount rate regulation because Son, initial value is estimated to the marketing activity effect using the discount rate regulatory factor and calibrated, the calibration value of acquisition is The target shops marketing activity effect discreet value.

At present, O2O catering industries businessman marketing plan and marketing program are constantly weeded out the old and bring forth the new, and marketing activity effect is carried out More accurately predict, then allow businessman to specify the value of marketing activity, in a planned way follow the prescribed order and do marketing activity, and to battalion The effect of pin activity carries out real-time estimate, and businessman can be helped more reasonably to configure marketing activity resource, expense and suitable Alternative, so as to set the activity for more meeting market expectations and marketing purpose.Existing general predictive method is to refer to the shops History similar marketing activity effect data, i.e., before effect should be also similar with the animation effect created now.It is true On, only these can not carry out accurately prediction in real time to the effect for the marketing activity that will be disposed.Wherein, discount rate pair Influenceed in the BI indexs (Business Indicators, or business index) of many animation effects very big, such as one have threshold Certificate, threshold typically can be higher than shops's visitor's unit price, if the certificate discount dynamics is enough big, it is sufficient to attractive, then many users are Reach will being consumed more using threshold for certificate, so as to improve shops's visitor's unit price, thus discount rate is to influence prediction marketing to live One of key factor of dynamic effect, and can not be only that the account of the history of simple geo-statistic difference discount rate activity calculates history The effect of similar active；In addition, existing predictor method does not account for the fluctuation situation of shops's management state yet.

The embodiment of the present application is used to market so that businessman's marketing activity effect of O2O catering industries is estimated as an example to described The method of animation effect prediction is described in detail.In the embodiment of the present application, model explicitly is built for discount rate, with folding The change of button rate and more accurately estimate animation effect in real time, rather than simply discount rate as activity a dimension The account of the history of different discount rate activities is only counted, the similar of shops periphery is considered when calculating the effect of history similar active The history marketing activity effect of shops, and build model together with the data of these similar shops and give a forecast, with every activity every The granularity in shop estimates animation effect, and so as to carry out more accurately prediction in real time to marketing activity effect, businessman can also For the prediction result for estimating effect and carrying out polymerization and obtaining entirety of the applicable shops for the activity of reaching the standard grade.So-called similar shops is to seek During selling active prediction, for the shops close with the feature of target shops of target shops selection.Wherein, so-called target door Shop, refer to the applicable movable shops of the selected activity of reaching the standard grade when businessman releases marketing activity, i.e., reached the standard grade battalion by shops's granularity Pin activity, such as public praise net businessman can be selected with freedom and flexibility under which shops which advertising campaign be applicable.

The step S101, the history marketing datas of different shops is obtained as input sample, before and after calculating marketing activity The marketing activity effect data of different shops.

Using the history marketing activity data of shops, the effect of the marketing activity to that will reach the standard grade carries out ex ante forecasting, obtained To reliable animation effect prediction result can be as the contrast object pre-established, the contrast object is used for the aprowl phase Or marketing activity is assessed after activity end, by will really market, situation carries out check analysis with the contrast object, so as to Instruct shops or the operation of businessman.

The purpose of this step is the history marketing initial data based on different shops, to calculate different shops's history marketing Animation effect data.

So-called marketing activity effect data, it is business index lifting values of the shops before and after marketing activity, comprises at least down State any index：Objective unit price lifting percentage point and again purchase rate lifting percentage point.Different shops before and after the calculating marketing activity Marketing activity effect data, refer to start the visitor's unit price and purchase rate again in front and rear certain period of time according to shops's marketing activity, point Not Ji Suan shops before and after marketing activity visitor unit price lifting percentage point and again purchase rate lifting percentage point, for example, pull marketing work It is dynamic start before initial data in the latter moon calculated, different shops can be same businessman each shops or The shops of different businessmans.Wherein, visitor's unit price, refers to that each customer averagely buys goods amount；The rate of purchase again refers to Under certain special time window, the frequent customer of secondary consumption accounts for total transaction UV ratio, so-called UV, is independent visitor (Unique Visitor), a computer client for accessing website is a visitor, 00:00-24:Identical client can only be counted in 00 Calculate once, it is clear that advertising campaign can lift UV amounts.

O2O catering industry businessman's marketing activity a multitude of names, including at least following any Activity Types：Completely subtract, often completely subtract, Consumption is sent.Wherein, it is described completely to subtract, it is a kind of Activity Type, refers to and meet that threshold can is preferential corresponding to enjoying, such as completely 100 subtract 10 yuan；It is described often completely to subtract, it is completely to subtract different another Activity Types from described, refers to and every meet that how many times of cans of threshold are enjoyed Corresponding preferential multiple, such as often expire 100 and subtract 10 yuan, 300 yuan of consumer spending can be preferential 30 yuan；The consumption is sent, and is another Kind Activity Type, a reward voucher can be got automatically after referring to successfully consumption, its preferential dynamics is typically more every than completely subtracting corresponding to shops Full deactivation is dynamic big, can be used when user's secondary consumption.

In the embodiment of the present application, which kind of advertising campaign businessman can release according to shops's granularity selection, and can be lived to releasing Dynamic target shops predicts its effect before activity, comprehensive during prediction to use the shops number similar to the target shops According to so as to balance the influence of the target shops management state fluctuation to prediction effect.Specifically, the data pulled over 1 year are come Visitor's unit price of statistics every shops per activity and again purchase rate lifting percentage point, every activity to past 1 year is per shops, statistical activity The respective visitor's unit price and purchase rate again of the latter moon, then calculates corresponding lifting percentage point, as the type activity before beginning Visitor's unit price of the shops and again the historical act effect data of purchase rate lifting percentage point.If a shops has multiple same in history The activity of type, then visitor's unit price lifting percentage point and again purchase rate lifting percentage point take it is multiple it is movable averagely.

It should be noted that historical act data are being analyzed according to every activity per shops's granularity in actual treatment When, Activity Type can be segmented, or different analysis purposes is divided activity by major class, for example, activity is divided into Two major classes, including completely subtract/often completely subtract class activity and class activity is sent in consumption, the initial transaction behavior for user and later hand over respectively It is easy to be, i.e., the initial transaction behavior for user, a major class is incorporated into by completely subtracting/often completely subtracting；Can also be in order to more segment analysis number According to as two independent Activity Types, the activity cost profit margin per disaggregated classification is such as analyzed.In addition, when bottom data is endless When standby, such as the similar shops in periphery did not did activity in history, then was rule of thumb adjusted, and abandoned not doing the shop of activity, or Person is revealed all the details using the empirical value of the global performance of business.

The step S102, the marketing activity effect data of target shops are chosen, handle to obtain the mesh by data smoothing The marketing activity effect of mark shops estimates initial value.

The purpose of this step is that the animation effect that the target shops is tentatively estimated out based on shops's historical act data is pre- Estimate initial value.

In the history marketing activity effect data of the different shops obtained from step S101, the marketing of target shops is extracted Animation effect data, according to these marketing activity historical datas, the marketing activity effect that target shops releases is carried out in advance pre- Survey.It is predicted it should be noted that the target shops historical data of itself can be chosen, its similar door can also be chosen simultaneously The marketing activity historical data in shop, the data basis being predicted collectively as the marketing activity effect to the target shops.Together When consider the marketing activity historical datas of these similar shops, the effect for the marketing activity that can be released to target shops be made more Reliable prediction.

In the embodiment of the present application, so-called animation effect refers to objective unit price and again purchase rate lifting degree, for animation effect Estimate visitor's unit price lifting percentage point including prediction marketing activity and purchase rate lifts percentage point again, carry out marketing activity prediction When, choose the marketing activity effect number of target shops and its similar shops by establishing similar shops's model of target shops According to.

Specifically, the marketing activity effect data for choosing target shops, including following processing：Establish target shops Similar shops's model, the similar shops of target shops is determined based on similar shops's model, from the marketing of the different shops The marketing activity effect data of the target shops and its similar shops are chosen in animation effect data.

The similar shops's model for establishing target shops, including：Shops's characteristic value, base are obtained from one or more dimensions In KNN algorithms or shops's sequencing of similarity method based on setting, the phase of the target shops is established according to the characteristic value Like shops's model；Wherein, one or more of dimensions, including at least following any dimensions：Average daily stroke count, pen unit price, shops Affiliated Merchant ID, position feature.

In the embodiment of the present application, for the various dimensions characteristic value of different shops, the sequencing of similarity side of shops based on setting Method establishes similar shops's model of the target shops, specifically, shops's sequencing of similarity method of the setting, including：

According to each dimensional characteristics value of the similar shops of candidate, each dimension in similar shops sorts proportion, it is and similar The characteristic value degree of closeness of shops's dimension corresponding with the target shops, determines shops's sequencing of similarity.Wherein, the default bar Part, including：Shops belongs to Tong Cheng with target；And/or belong to same classification with target shops.For example, it will belong to same with target shops City, and then symbol screens therefrom as similar shops's Candidate Set according still further to other dimensions in the shops for belonging to same three-level classification The similar shops of conjunction condition.

In the embodiment of the present application, shops's characteristic value is obtained from multiple dimensions, including：Average daily stroke count, pen are monovalent, belonging to shops Merchant ID, position feature.

So-called average daily stroke count, is averagely daily transaction stroke count, a water Dan Weiyi transaction.

The pen unit price, refer to average dealing money corresponding to each transaction record (a water list), it is typically total to consume The amount of money divided by consumption stroke count measure.

The affiliated Merchant ID of shops, i.e. shops pid (or partner ID), such as KFC shops pid is Bai Sheng collection Group, the pid that O2O caterers can be registered when promoting website and registering, each same shops of its subordinate typically share the pid, Same pid shops's similarity is very high.

The position feature, be with target shops residing for geographical position distance and present position whether in same administration Region, in O2O catering industries, apart near shops because customer group's scope is similar thus generally as choosing similar shops A comparison dimension, for example, belonging to an administrative area together with the target shops present position.

In addition, in the embodiment of the present application, to similar shops's model of target shops foundation, include to similar shops's stepping It is secondary, namely it is ranked up according to similarity, so when selecting similar shops to recall marketing activity, can preferentially it call together More like shops is gone back to, due to paying the utmost attention to the historical data of more like shops, can help to improve the prediction essence of animation effect Degree；Wherein, it is so-called to recall marketing activity, refer to activity data and other relevant informations for obtaining the marketing activity.And Establish when being sorted in similar shops's model process to similar shops, consider each dimension proportion in similar shops sorts, its In, each dimension proportion in similar shops sorts, is to determine each dimension proportion according to following orders of priority：

The target shops is most like shops in itself；

It is identical with the affiliated Merchant ID of target shops；

Setting value is less than with the distance of target shops；

An administrative area is belonged to together with target shops.

Specifically, the embodiment of the present application, A to H totally 8 Sort Priorities are defined as to similar shops, by target shops certainly The history marketing activity data of body examine the data of rate, priority A as override during prediction；Secondly, average daily stroke count phase It is also very big like degree, Sort Priority B, its marketing effectiveness difference of the very big Liang Jia shops of average daily stroke count difference, therefore the dimension Characteristic value priority is forward in the ranking, average daily stroke count similarity is the phase calculated with high average daily stroke count divided by low average daily stroke count To value, such as using shops of the average daily stroke count similarity less than 1.5 as similar shops；Sort Priority C is pen unit price similarity, Because animation effect is related to sales volume, therefore the similarity degree of pen unit price dimension is important, in addition, the value is taken from high pen Monovalent divided by low pen unit price, such as small 1.5 shops of being equal to of pen unit price similarity is similar shops；Sort Priority D is, with institute It is identical to state the affiliated Merchant ID of target shops；Sort Priority E to H is position feature, including：Distance is less than 300 meters, and its sequence is excellent First level E；Distance is less than 500 meters, Sort Priority F；Distance is less than 1000 meters, and Sort Priority G is；It is same with the shops Area, Sort Priority H.

It should be noted that during the similar shops of selection target shops, KNN sorting algorithms can also be used, for difference The various dimensions characteristic value of shops, the output each shops similar with target shops, so-called KNN algorithms, i.e. K arest neighbors (kNN, or k- NearestNeighbor) sorting algorithm, it is one of Data Mining Classification technical method.So-called K arest neighbors, refers to each sample It can be represented with its immediate k neighbour, its core concept is if the k in feature space most phases of a sample Most of in adjacent sample belong to some classification, then the sample falls within this classification, and with sample in this classification Characteristic.This method on categorised decision it is determined that only determine according to the classification of one or several closest limited samples Classification belonging to sample to be divided.

In addition, the marketing activity effect data for choosing target shops and its similar shops, are to choose granting amount and core Sales volume is more than the animation effect data of the history marketing activity of setting value.It is right i.e. after the similar shops of target shops is selected Marketing activity carries out following Screening Treatments：Choose granting amount and the amount of checking and writing off is more than the marketing activity of setting value, such as choose 10 The activity of similar shops.I.e. during the activity of similar shops is selected, it is also necessary to which the activity to history is done a certain degree of Screening, because can just have the high confidence level of comparison, it is necessary to which the historical act has sufficient statistical sample for historical act, Sample could be used as to build model.Specifically, the embodiment of the present application is only to choose those granting amounts more than 50 amounts of checking and writing off to be more than 10 activity.It is described to check and write off, refer to that the certificate that marketing activity is sent out is used in off-line transaction.The setting of certificate or card granting amount threshold value Purpose, on the one hand it is to filter out rubbish activity, some marketing platforms are implicitly present in many invalid activations, and this active user is basic It can not perceive, so need not also be used as the reference of future anticipation, second, granting amount is more than 50, then it is it is considered that this movable Every BI indexs such as check and write off rate, visitor's unit price lifting, and purchase rate lifting again just has statistical significance.Blocking threshold value of the amount of checking and writing off more than 10 is One soft-condition (so-called soft-condition, not being mandatory condition), i.e., prioritizing selection meets the activity of this condition, when meeting condition Amount of activity does not reach the limitation that setting value just loosens this condition, to choose enough activities as far as possible.Block the amount of checking and writing off Threshold value be in order to select the higher activity of quality as far as possible because with the development of platform, movable mass more and more higher, rubbish Activity is fewer and fewer, and the measured movement reference meaning of matter is naturally bigger.

The activity of 10 similar shops is selected in the embodiment of the present application, obtains the animation effect number of these shops's marketing activities According to, including before and after each comfortable marketing activity of these shops visitor unit price lifting percentage point and again purchase rate lifted percentage point, by right The marketing activity effect that the animation effect data progress data smoothing handles to obtain the target shops estimates initial value.The warp Cross data smoothing and handle to obtain the marketing activity effect of the target shops and estimate initial value, including following smoothing processings：Take target The average of visitor's unit price lifting percentage point of shops and its similar shops estimates initial value as the objective unit price of the target shops, takes mesh The average of the multiple purchase rate of mark shops and its similar shops lifting percentage point estimates initial value as the multiple purchase rate of the target shops.

In addition, if there is the activity of multiple same types in a shops in history, then visitor is monovalent, again the lifting percentage point of purchase rate Take multiple movable be averaged.When bottom data is incomplete, such as the similar shops in periphery did not did activity in history, then according to warp Test and be adjusted, abandon not doing the shop of activity, or revealed all the details using the empirical value of the global performance of business.For above-mentioned processing In get per activity per shops animation effect data take result estimate of its average value as the type marketing activity at the beginning of Value.

The step S103, the marketing activity effect data based on different shops, discount rate is built by linear fit and adjusted The factor is saved, initial value is estimated to the marketing activity effect using the discount rate regulatory factor and calibrated, the calibration value of acquisition As described target shops marketing activity effect discreet value.

Due to discount rate change, to objective unit price and again, the BI such as purchase rate Index Influences very greatly, therefore are carried out to marketing activity effect More accurate prediction in real time then needs to take into account the influence of discount rate.

The purpose of this step, be according to different shops's historical act data, by linear fit build discount rate regulation because Son, estimate initial value for the target shops animation effect obtained in step S102 and calibrate.

The marketing activity effect data based on different shops, discount rate regulatory factor, bag are built by linear fit Include：Visitor's unit price based on history marketing activity lifts percentage point and again purchase rate lifting percentage point, enters line by least square method Property fitting, draw the calculation formula that discount rate regulatory factor changes with discount rate.It should be noted that structure discount rate regulation because Sample data used in son, can be historical act data or statistics the owning with city same type of similar shops Animation effect numerical value average of the activity under specific discount rate, carries out linear fit to discount rate with these sample datas, obtains The slope and intercept of the line correspondences fitted.Specific to the embodiment of the present application, with visitor's unit price lifting of same city same type activity Percentage point and again purchase rate lift percentage point as sample, and select only to consider the activity more than or equal to 5 foldings during sample, because small In 5 foldings amount of activity very little, animation effect fluctuation is obvious.

Specifically, described carry out linear fit by least square method, show that discount rate regulatory factor changes with discount rate Calculation formula, including：Using discount rate and marketing activity effect, the intercept and slope of straight line are fitted, then discount rate is adjusted The calculation formula that the factor changes with discount rate is：Discount rate regulatory factor is multiplied by discount rate equal to slope adds intercept；

The slope b and intercept a formula that the embodiment of the present application is gone out by least square fitting are as follows：

In formula,

Then discount rate regulatory factor computational methods：Discount rate regulatory factor=b* discount rates+a.Using discount rate regulation because Son is estimated initial value to the marketing activity effect and calibrated, including uses following calibration equations：

The marketing activity effect that the target shops marketing activity effect discreet value is equal to the target shops estimates initial value Discount rate regulatory factor is multiplied by, i.e.,：

Prediction visitor's unit price lifting percentage point=visitor's unit price lifting percentage point estimates initial value * discount rate regulatory factors；

Purchase rate lifting percentage point=again purchase rate lifting percentage point estimates initial value * discount rate regulatory factors again for prediction.

Further, since activity threshold reflects crowd characteristic and the crowd size of participation activity, for business index of marketing Improvement and activity landing have direct influence, therefore, provided in the embodiment of the present application be used for marketing activity effect predict Method, in addition to be directed to the target shops marketing activity effect discreet value, further corrected, carried using threshold regulatory factor High forecasting accuracy.Wherein, the threshold regulatory factor, it is that the order numbers of threshold are met according to history marketing activity and meet visitor The factor that the ratio calculation threshold of the order numbers of unit price influences on marketing activity effect, is determined by following manner：

Wherein, it is described to be directed to the target shops marketing activity effect discreet value, further repaiied using threshold regulatory factor Just, including：

The threshold regulatory factor is multiplied by the target shops marketing activity effect discreet value, obtains the warp of target shops Cross the marketing activity effect discreet value further corrected.

Specific to the embodiment of the present application, initial value is estimated to target shops marketing activity effect, adjusted by discount rate The factor and threshold regulatory factor two-stage amendment, obtain final animation effect predicted value, and the priority that this two-stage is adjusted when realizing is suitable Sequence does not influence prediction result.After the marketing activity effect prediction scheme actual deployment, user inputs the phase of marketing activity scheme Information is closed, such as：Promotion plan classification, maximum stand subtract amount of money etc., by the processing of the embodiment of the present application, can obtain this rush Effect is estimated in pin activity, and the system schematic of actual deployment is as shown in Fig. 2 user and the interaction effect of actual deployment system As shown in Figure 3.

Based on the embodiment of the method for being used for marketing activity effect prediction provided by the application, present invention also provides A kind of method for the prediction of businessman's animation effect.

Reference picture 4, it illustrates the method process chart for being used for businessman's animation effect and predicting provided according to the application. Because the present embodiment is based on the embodiment of the above-mentioned method for marketing activity effect prediction, so describing simpler Single, related part refers to the corresponding explanation of above-described embodiment.Described below is used for what businessman's animation effect was predicted The embodiment of method is only schematical.

The application provides a kind of method for the prediction of businessman's marketing activity effect, including：

Step S401, the characteristic value of shops is obtained from following any dimensions：Average daily stroke count, pen unit price, the affiliated businessman of shops ID, position feature, according to the characteristic value, shops's sequencing of similarity method based on KNN algorithms or based on setting, establish and live The dynamic similar shops's model for being applicable shops；

Step S402, the similar shops of shops is applicable based on similar shops's model determination activity, obtains similar shops Marketing activity effect data, handle to obtain the marketing activity effect discreet value that the activity is applicable shops by data smoothing；

Step S403, the marketing activity effect value for being applicable activity shops polymerize, and obtain the prediction of businessman's animation effect Value.

The embodiment of the present application is said by taking O2O catering industries as an example to the method for the prediction of businessman's animation effect It is bright.

In practical application, businessman reaches the standard grade the applicable shops of advertising campaign when promoting according to shops's granularity unrestricted choice, such as Public praise net, when being estimated to animation effect, the animation effect of shops, then the institute to the activity of reaching the standard grade are applicable first against activity There is the animation effect predicted value for being applicable shops to carry out polymerization and obtain the overall prediction result of businessman.The so-called applicable shops of activity, it is Refer to the applicable shops that businessman releases the activity of reaching the standard grade selected during marketing activity.

In the embodiment of the present application, when the promotion effect of the shops for being applicable the activity of reaching the standard grade is predicted, by shops's week The history marketing activity effect data of the similar shops on side are taken into account to give a forecast together, rather than simply just counts this and be applicable The account of the history of movable shops, it is thus possible to balance shops and manage fluctuation, more accurately predicted marketing activity effect.

The step S401, the characteristic value of shops is obtained from following any dimensions：Average daily stroke count, pen are monovalent, belonging to shops Merchant ID, position feature, according to the characteristic value, shops's sequencing of similarity method based on KNN algorithms or based on setting, build Vertical activity is applicable similar shops's model of shops.

The purpose of this step is similar shops's model that foundation activity is applicable shops, so as to pay the utmost attention to most like shops Marketing activity historical data, the marketing activity effect reached the standard grade to the shops make high-precision forecast.

In the embodiment of the present application, different shops's characteristic values are obtained from multiple dimensions, including：Average daily stroke count, pen unit price, shops Affiliated Merchant ID, position feature, and for the various dimensions characteristic value of different shops, shops's sequencing of similarity method based on setting Similar shops's model of the target shops is established, specifically, shops's sequencing of similarity method of the setting, including：

In addition, in the embodiment of the present application, to similar shops's model of target shops foundation, include to similar shops's stepping It is secondary, namely it is ranked up according to similarity, so when marketing activity is recalled to similar shops, can preferentially it recall More like shops, due to paying the utmost attention to the historical data of more like shops, it can help to improve the precision of prediction of animation effect. And when being sorted in establishing similar shops's model process to similar shops, consider each dimension institute's accounting in similar shops sorts Weight, wherein, each dimension proportion in similar shops sorts, is to determine each dimension proportion according to following orders of priority：

The activity is using shops's sheet as most like shops；

Applicable shops belongs to an administrative area together with activity.

In addition, when selection activity is applicable the similar shops of shops, KNN sorting algorithms can also be used, for different shops Various dimensions characteristic value, the output each shops similar with target shops.

The step S402, the similar shops of shops is applicable based on similar shops's model determination activity, obtained similar Shops's marketing activity effect data, handle to obtain the marketing activity effect discreet value that the activity is applicable shops by data smoothing.

The purpose of this step is to obtain the animation effect prediction data that the activity is applicable shops.

When selection activity is applicable the marketing activity effect data of shops and its similar shops, it is necessary to choose granting amount and core Sales volume is more than the animation effect data of the history marketing activity of setting value.Marketing activity i.e. to demand is carried out at following screenings Reason：Choose granting amount and the amount of checking and writing off is more than the marketing activity of setting value, such as choose the activity of 10 similar shops.So choose Historical act have sufficient statistical sample, can just have higher confidence level.Specifically, the embodiment of the present application is only to choose those Granting amount is more than the activity that 50 amounts of checking and writing off are more than 10, so as to filter out rubbish activity.

The activity of 10 similar shops is selected in the embodiment of the present application, obtains the animation effect number of these shops's marketing activities According to, including before and after each comfortable marketing activity of these shops visitor unit price lifting percentage point and again purchase rate lifted percentage point, by right The animation effect data carry out data smoothing and handle to obtain the marketing activity effect discreet value of the target shops, specifically, Including following smoothing processings：The average of visitor's unit price lifting percentage point of target shops and its similar shops is taken as the target door The objective unit price in shop estimates initial value, takes the average of the multiple purchase rate of target shops and its similar shops lifting percentage point as the target The multiple purchase rate discreet value of shops.

The step S403, the marketing activity effect value for being applicable activity shops polymerize, and obtain businessman's animation effect Predicted value.

It is whole that the purpose of this step is that advertising campaign of the advertising campaign predicted value based on businessman's difference shops to businessman is made The prediction of body.

Specific to the embodiment of the present application, the different applicable shops for the activity of reaching the standard grade are made a prediction according to Activity Type, Each predicted value for being applicable shops is summed up to obtain overall the predicting of such activity of businessman.

Corresponding with a kind of embodiment for method for marketing activity effect prediction that the application provides, the application also carries A kind of device for marketing activity effect prediction is supplied.

Reference picture 5, it illustrates a kind of schematic device predicted for marketing activity effect provided according to the application. Because device embodiment is substantially similar to embodiment of the method, so describing fairly simple, related part refers to method reality Apply the corresponding explanation of example.Device embodiment described below is only schematical.

The application provides a kind of device for marketing activity effect prediction, including：

Historical data acquiring unit 501, the history marketing data for obtaining different shops calculate battalion as input sample The marketing activity effect data of different shops before and after pin activity；

Unit 502 tentatively is estimated, for choosing the marketing activity effect data of target shops, is handled by data smoothing Marketing activity effect to the target shops estimates initial value；

Discreet value alignment unit 503, for the marketing activity effect data based on different shops, built by linear fit Discount rate regulatory factor, initial value is estimated to the marketing activity effect using the discount rate regulatory factor and calibrated, obtain Calibration value be the target shops marketing activity effect discreet value.

Optionally, the marketing activity effect data for choosing target shops, including following processing：Establish target shops Similar shops's model, the similar shops of target shops is determined based on similar shops's model, from the marketing of the different shops The marketing activity effect data of the target shops and its similar shops are chosen in animation effect data.

Optionally, unit 502, including modeling subelement are tentatively estimated, for establishing similar shops's model of target shops, Including：

Optionally, the preparatory condition, including：

Shops belongs to Tong Cheng with target；And/or

Belong to same classification with target shops.

Optionally, it is described tentatively to estimate unit 502, including activity screening subelement, for choosing granting amount and the amount of checking and writing off More than the animation effect data of the history marketing activity of setting value.

Optionally, the described method for being used for marketing activity effect prediction, in addition to threshold adjustment unit, for for institute The marketing activity effect discreet value of target shops is stated, is further corrected using threshold regulatory factor；Wherein, threshold regulation because Son, it is that the order numbers of threshold are met according to history marketing activity and meet the ratio calculation threshold of the order numbers of visitor's unit price to marketing The factor that animation effect influences, is determined by following manner：

Present invention also provides a kind of electronic equipment for being used to realize the method for marketing activity effect prediction, ginseng According to Fig. 6, the schematic diagram of a kind of electronic equipment provided it illustrates the present embodiment.

The electronic equipment embodiment that the application provides describes fairly simple, and related part refers to above-mentioned offer The method for marketing activity effect prediction embodiment corresponding explanation.Embodiment described below is only Schematically.

The application provides a kind of electronic equipment, including：

Memory 601, and processor 602；

The memory 601 is used to store computer executable instructions, and the processor 602 is used to perform the computer Executable instruction：

Optionally, the processor 602 is additionally operable to perform following computer executable instructions：

Optionally, the preparatory condition, including：

Shops belongs to Tong Cheng with target；And/or

Belong to same classification with target shops.

Optionally, the processor 602 is additionally operable to perform following computer executable instructions：Choose granting amount and the amount of checking and writing off More than the animation effect data of the history marketing activity of setting value.

Optionally, the processor 602 is additionally operable to perform following computer executable instructions：Sought for the target shops Animation effect discreet value is sold, is further corrected using threshold regulatory factor；Wherein, the threshold regulatory factor, is according to history Marketing activity meets the order numbers of threshold and meets that the ratio calculation threshold of the order numbers of visitor's unit price influences on marketing activity effect The factor, determined by following manner：

Optionally, the processor 602 is additionally operable to perform following computer executable instructions：Opened according to shops's marketing activity Visitor's unit price and purchase rate again before and after beginning in certain period of time, visitor unit price lifting percentage of the shops before and after marketing activity is calculated respectively Point and again purchase rate lifting percentage point.

In a typical configuration, computing device includes one or more processors (CPU), input/output interface, net Network interface and internal memory.

Internal memory may include computer-readable medium in volatile memory, random access memory (RAM) and/or The forms such as Nonvolatile memory, such as read-only storage (ROM) or flash memory (flash RAM).Internal memory is computer-readable medium Example.

1st, computer-readable medium can be by any side including permanent and non-permanent, removable and non-removable media Method or technology realize that information stores.Information can be computer-readable instruction, data structure, the module of program or other numbers According to.The example of the storage medium of computer includes, but are not limited to phase transition internal memory (PRAM), static RAM (SRAM), dynamic random access memory (DRAM), other kinds of random access memory (RAM), read-only storage (ROM), Electrically Erasable Read Only Memory (EEPROM), fast flash memory bank or other memory techniques, read-only optical disc are read-only Memory (CD-ROM), digital versatile disc (DVD) or other optical storages, magnetic cassette tape, tape magnetic rigid disk storage or Other magnetic storage apparatus or any other non-transmission medium, the information that can be accessed by a computing device available for storage.According to Herein defines, and computer-readable medium does not include non-temporary computer readable media (transitory media), such as modulates Data-signal and carrier wave.

2nd, it will be understood by those skilled in the art that embodiments herein can be provided as method, system or computer program production Product.Therefore, the application can use the embodiment in terms of complete hardware embodiment, complete software embodiment or combination software and hardware Form.Moreover, the application can use the computer for wherein including computer usable program code in one or more can use The computer program product that storage medium is implemented on (including but is not limited to magnetic disk storage, CD-ROM, optical memory etc.) Form.

Although the application is disclosed as above with preferred embodiment, it is not for limiting the application, any this area skill Art personnel are not being departed from spirit and scope, can make possible variation and modification, therefore the guarantor of the application Shield scope should be defined by the scope that the application claim is defined.

Claims

A kind of 1. method for marketing activity effect prediction, it is characterised in that including：

The history marketing data of different shops is obtained as input sample, calculates the marketing activity of different shops before and after marketing activity Effect data；

The marketing activity effect data of target shops are chosen, the marketing activity for handling to obtain the target shops by data smoothing is imitated Fruit estimates initial value；

Marketing activity effect data based on different shops, discount rate regulatory factor is built by linear fit, uses the folding Button rate regulatory factor is estimated initial value to the marketing activity effect and calibrated, and the calibration value of acquisition is the battalion of target shops Sell animation effect discreet value.
2. the method according to claim 1 for marketing activity effect prediction, it is characterised in that the marketing activity effect Fruit data, it is business index lifting values of the shops before and after marketing activity, including at least following any indexs：Objective unit price lifting hundred Branch and again purchase rate lifting percentage point.
3. the method according to claim 1 or 2 for marketing activity effect prediction, it is characterised in that the selection mesh Mark the marketing activity effect data of shops, including following processing：

Similar shops's model of target shops is established, the similar shops of target shops is determined based on similar shops's model, from The marketing activity effect number of the target shops and its similar shops is chosen in the marketing activity effect data of the different shops According to.
4. the method according to claim 3 for marketing activity effect prediction, it is characterised in that described to establish target door Similar shops's model in shop, including：

Shops's characteristic value, the sequencing of similarity side of shops based on KNN algorithms or based on setting are obtained from one or more dimensions Method, similar shops's model of the target shops is established according to the characteristic value；Wherein, one or more of dimensions, at least Including following any dimensions：Average daily stroke count, pen unit price, the affiliated Merchant ID of shops, position feature.
5. the method according to claim 4 for marketing activity effect prediction, it is characterised in that the shops of the setting Sequencing of similarity method, including：

The similar shops of candidate of the target shops is filtered out by preparatory condition；

According to each dimensional characteristics value of the similar shops of candidate, each dimension in similar shops sorts proportion, and similar shops The characteristic value degree of closeness of dimension corresponding with the target shops, determines shops's sequencing of similarity.
6. the method according to claim 3 for marketing activity effect prediction, it is characterised in that described to be put down by data The marketing activity effect that sliding processing obtains the target shops estimates initial value, including following smoothing processings：Take target shops and its Similar shops visitor unit price lifting percentage point average estimate initial value as the objective unit price of the target shops, take target shops and The average of the multiple purchase rate lifting percentage point of its similar shops estimates initial value as the multiple purchase rate of the target shops.
7. the method according to claim 1 for marketing activity effect prediction, it is characterised in that the selection target door The marketing activity effect data in shop, it is to choose the animation effect number of granting amount and the amount of checking and writing off more than the history marketing activity of setting value According to.
8. the method according to claim 2 for marketing activity effect prediction, it is characterised in that described based on not fellow disciple The marketing activity effect data in shop, discount rate regulatory factor is built by linear fit, including：Visitor based on history marketing activity Unit price lifting percentage point and again purchase rate lifting percentage point, by least square method carry out linear fit, draw discount rate adjust because The calculation formula that son changes with discount rate.
9. the method according to claim 8 for marketing activity effect prediction, it is characterised in that described to pass through a most young waiter in a wineshop or an inn Multiplication carries out linear fit, draws the calculation formula that discount rate regulatory factor changes with discount rate, including：

Using discount rate and marketing activity effect, the intercept and slope of straight line are fitted, then discount rate regulatory factor is with discount rate The calculation formula of change is：Discount rate regulatory factor is multiplied by discount rate equal to slope adds intercept；

Wherein, intercept subtracts slope equal to animation effect average value and is multiplied by discount rate average value；By the discount rate of each similar shops Deviation from average is multiplied by animation effect deviation from average, after summation again divided by each similar shops discount rate sum of sguares of deviation from mean, that is, obtain tiltedly Rate；The deviation from average is the distance of actual value deviation average.
A kind of 10. method for the prediction of businessman's animation effect, it is characterised in that including：

The characteristic value of shops is obtained from following any dimensions：Average daily stroke count, pen unit price, the affiliated Merchant ID of shops, position feature, root According to the characteristic value, shops's sequencing of similarity method based on KNN algorithms or based on setting, foundation activity is applicable the phase of shops Like shops's model；

The similar shops of shops is applicable based on similar shops's model determination activity, obtains similar shops's marketing activity effect number According to handling to obtain the marketing activity effect predicted value that the activity is applicable shops by data smoothing；

The marketing activity effect value for being applicable activity shops polymerize, and obtains businessman's animation effect predicted value.