CN106250999A - The methods, devices and systems of prediction turnover rate - Google Patents
The methods, devices and systems of prediction turnover rate Download PDFInfo
- Publication number
- CN106250999A CN106250999A CN201510301080.9A CN201510301080A CN106250999A CN 106250999 A CN106250999 A CN 106250999A CN 201510301080 A CN201510301080 A CN 201510301080A CN 106250999 A CN106250999 A CN 106250999A
- Authority
- CN
- China
- Prior art keywords
- turnover rate
- dimensional parameter
- moment
- log
- hour
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
This application discloses a kind of methods, devices and systems predicting turnover rate, in order to solve the problem that in prior art, the forecasting inaccuracy of turnover rate is true.This recording method includes: obtain the turnover rate data in preset duration before the first moment;Turnover rate data are decomposed into hour of log-on dimensional parameter, registration time length dimensional parameter and calendar time dimensional parameter, wherein, hour of log-on dimensional parameter represents the hour of log-on power of influence to turnover rate, registration time length dimensional parameter represents the registration time length power of influence to turnover rate, and calendar time dimensional parameter represents the calendar time power of influence to turnover rate;According to decomposing the hour of log-on dimensional parameter of gained, registration time length dimensional parameter and calendar time dimensional parameter, calculate the turnover rate in the second moment, wherein, after being engraved in for the first moment when second.
Description
Technical field
The application relates to Internet technical field, particularly relates to a kind of predict the method for turnover rate, device and be
System.
Background technology
For enterprise, especially Internet enterprises, client is a kind of important resource.To client more
In-depth study and understanding, hold with the loss law to client, even carries out the turnover rate of client
Prediction, it is possible to allow enterprise make coping strategy in advance, save more client, thus bring to enterprise more
Interests.
In prior art, the Forecasting Methodology of common churn rate is: stream to client on predetermined period
Mistake rate carries out adding up and record, to form the time series of one-dimensional.Utilize common statistical model, set up visitor
Relational expression between family turnover rate and some setting variablees, and be trained by the history turnover rate data of record
After be applied to the prediction of turnover rate.
But, the loss behavior of client is often a kind of general performance of inherent law and external environment condition, existing
Turnover rate Forecasting Methodology in limited degree, the relation of variable and turnover rate can only be described, it is impossible to
Excavate the immanent cause constituting customer churn, and then cause predicting the outcome of turnover rate the most accurate.
Summary of the invention
The embodiment of the present application provides a kind of methods, devices and systems predicting turnover rate, to realize flowing user
Mistake rate is predicted more accurately.
A kind of method predicting turnover rate that the embodiment of the present application provides, including:
Under the control of one or more calculating devices being configured with executable instruction,
The turnover rate data in preset duration before Operation Server obtained for the first moment;
When described turnover rate data are decomposed into hour of log-on dimensional parameter, registration time length dimensional parameter and calendar
Between dimensional parameter, wherein, described hour of log-on dimensional parameter represents the hour of log-on power of influence to turnover rate,
Described registration time length dimensional parameter represents the registration time length power of influence to turnover rate, and described calendar time dimension is joined
Number represents the calendar time power of influence to turnover rate;
According to decomposing the hour of log-on dimensional parameter of gained, registration time length dimensional parameter and calendar time dimension ginseng
Number, calculates the turnover rate in the second moment, wherein, after being engraved in described first moment when described second;
Feed back described second moment turnover rate data to described Operation Server.
A kind of method predicting turnover rate that the embodiment of the present application provides, including:
Under the control of one or more calculating devices being configured with executable instruction,
Obtain the turnover rate data in preset duration before the first moment;
Described turnover rate data are decomposed into hour of log-on dimensional parameter and registration time length dimensional parameter, wherein,
Described hour of log-on dimensional parameter represents the hour of log-on power of influence to turnover rate, and described registration time length dimension is joined
Number represents the registration time length power of influence to turnover rate;
Hour of log-on dimensional parameter according to decomposition gained and registration time length dimensional parameter, calculated for the second moment
Turnover rate, wherein, after being engraved in described first moment when described second;
Feed back described second moment turnover rate data to described Operation Server.
A kind of method predicting turnover rate that the embodiment of the present application provides, including:
Under the control of one or more calculating devices being configured with executable instruction,
Obtain the turnover rate data in preset duration before the first moment;
Described turnover rate data are decomposed into hour of log-on dimensional parameter and calendar time dimensional parameter, wherein,
Described hour of log-on dimensional parameter represents the hour of log-on power of influence to turnover rate, and described calendar time dimension is joined
Number represents the calendar time power of influence to turnover rate;
Hour of log-on dimensional parameter according to decomposition gained and calendar time dimensional parameter, calculated for the second moment
Turnover rate, wherein, after being engraved in described first moment when described second;
Feed back described second moment turnover rate data to described Operation Server.
A kind of method predicting turnover rate that the embodiment of the present application provides, including:
Under the control of one or more calculating devices being configured with executable instruction,
Obtain the turnover rate data in preset duration before the first moment;
Described turnover rate data are decomposed into registration time length dimensional parameter and calendar time dimensional parameter, wherein,
Described registration time length dimensional parameter represents the registration time length power of influence to turnover rate, and described calendar time dimension is joined
Number represents the calendar time power of influence to turnover rate;
Registration time length dimensional parameter according to decomposition gained and calendar time dimensional parameter, calculated for the second moment
Turnover rate, wherein, after being engraved in described first moment when described second;
Feed back described second moment turnover rate data to described Operation Server.
A kind of device predicting turnover rate that the embodiment of the present application provides, including:
Acquisition module, for the turnover rate data obtained before the first moment in preset duration;
Decomposing module, for being decomposed into hour of log-on dimensional parameter, registration time length dimension by described turnover rate data
Degree parameter and calendar time dimensional parameter, wherein, described hour of log-on dimensional parameter represents hour of log-on convection current
The power of influence of mistake rate, described registration time length dimensional parameter represents the registration time length power of influence to turnover rate, described
Calendar time dimensional parameter represents the calendar time power of influence to turnover rate;
Prediction module, for according to decompose the hour of log-on dimensional parameter of gained, registration time length dimensional parameter and
Calendar time dimensional parameter, calculates the turnover rate in the second moment, wherein, is engraved in described first when described second
After moment;
Feedback module, is used for feeding back described second moment turnover rate data to described Operation Server.
A kind of device predicting turnover rate that the embodiment of the present application provides, including:
Acquisition module, for the turnover rate data obtained before the first moment in preset duration;
Decomposing module, for being decomposed into hour of log-on dimensional parameter and registration time length dimension by described turnover rate data
Degree parameter, wherein, described hour of log-on dimensional parameter represents the hour of log-on power of influence to turnover rate, described
Registration time length dimensional parameter represents the registration time length power of influence to turnover rate;
Prediction module, for the hour of log-on dimensional parameter according to decomposition gained and registration time length dimensional parameter,
Calculate the turnover rate in the second moment, wherein, after being engraved in described first moment when described second.
A kind of device predicting turnover rate that the embodiment of the present application provides, including:
Acquisition module, for the turnover rate data obtained before the first moment in preset duration;
Decomposing module, for being decomposed into hour of log-on dimensional parameter and calendar time dimension by described turnover rate data
Degree parameter, wherein, described hour of log-on dimensional parameter represents the hour of log-on power of influence to turnover rate, described
Calendar time dimensional parameter represents the calendar time power of influence to turnover rate;
Prediction module, for the hour of log-on dimensional parameter according to decomposition gained and calendar time dimensional parameter,
Calculate the turnover rate in the second moment, wherein, after being engraved in described first moment when described second;
Feedback module, is used for feeding back described second moment turnover rate data to described Operation Server.
A kind of device predicting turnover rate that the embodiment of the present application provides, including:
Acquisition module, for the turnover rate data obtained before the first moment in preset duration;
Decomposing module, for being decomposed into registration time length dimensional parameter and calendar time dimension by described turnover rate data
Degree parameter, wherein, described registration time length dimensional parameter represents the registration time length power of influence to turnover rate, described
Calendar time dimensional parameter represents the calendar time power of influence to turnover rate;
Prediction module, for the registration time length dimensional parameter according to decomposition gained and calendar time dimensional parameter,
Calculate the turnover rate in the second moment, wherein, after being engraved in described first moment when described second;
Feedback module, is used for feeding back described second moment turnover rate data to described Operation Server.
A kind of system predicting turnover rate that the embodiment of the present application provides, including:
The device predicting turnover rate as above;And,
DBM, for storing the turnover rate data of client;
Data module, for obtaining the turnover rate data of client in DBM, and is processed into predetermined number
According to form;
Algoritic module, is used for providing decomposition algorithm and model algorithm;Wherein,
Described decomposing module is additionally operable to by obtaining the turnover rate data after data module processes, and calls calculation
Turnover rate data after data module processes are decomposed into hour of log-on dimension by the decomposition algorithm in method module
Parameter, registration time length dimensional parameter and calendar time dimensional parameter;
Hour of log-on dimension is joined by model algorithm that described prediction module is additionally operable to call in algoritic module respectively
Number, registration time length dimensional parameter and calendar time dimensional parameter are predicted, and according to decomposing gained and prediction
Hour of log-on dimensional parameter, registration time length dimensional parameter and calendar time dimensional parameter calculate predetermined instant
Turnover rate;
Processing module, for obtaining the second moment turnover rate data from the device of described prediction turnover rate, and
According to the second moment turnover rate data genaration for the feedback information of user.
The embodiment of the present application provides a kind of method and apparatus predicting turnover rate, and the method is by the stream that will obtain
Mistake rate data are decomposed in hour of log-on dimension, registration time length dimension and calendar time dimension, and according to dividing
Solve the parameter of gained, during to hour of log-on dimensional parameter, registration time length dimensional parameter and the calendar of future time instance
Between dimensional parameter be predicted, thus obtain the turnover rate data of future time instance accordingly;Owing to hour of log-on is tieed up
Degree parameter and registration time length dimensional parameter represent hour of log-on and the registration time length power of influence to turnover rate respectively,
Namely describing the endogenous cause of ill of customer churn, calendar time dimensional parameter represents the calendar time impact on turnover rate
Power, namely describe the exopathogenic factor of customer churn, and the turnover rate comprehensive consideration finally predicted customer churn
Endogenous cause of ill and exopathogenic factor, therefore predict the outcome accurately.
Further, according to accurately predicting the outcome, the feedback information for user can be generated.Enter
And, be sent to client on the opportunity that Systematic selection is suitable, even also include with the transmission of this feedback information be
The system adjustment to user's internal data.Such that it is able to avoid or reduce the loss of user to a certain extent.
Accompanying drawing explanation
Accompanying drawing described herein is used for providing further understanding of the present application, constitutes the part of the application,
The schematic description and description of the application is used for explaining the application, is not intended that the improper limit to the application
Fixed.In the accompanying drawings:
The system architecture diagram of the prediction turnover rate that Fig. 1 a provides for the embodiment of the present application;
The schematic flow sheet of the prediction turnover rate that Fig. 1 provides for the embodiment of the present application;
The hour of log-on dimensional parameter being decomposed gained by turnover rate data that Fig. 2 provides for the embodiment of the present application
Schematic diagram;
The registration time length dimensional parameter being decomposed gained by turnover rate data that Fig. 3 provides for the embodiment of the present application
Schematic diagram;
The calendar time dimensional parameter being decomposed gained by turnover rate data that Fig. 4 provides for the embodiment of the present application
Schematic diagram;
Fig. 5 for the embodiment of the present application provide according to decompose the hour of log-on dimensional parameter of gained, registration time length
Dimensional parameter and calendar time dimensional parameter, calculate the idiographic flow schematic diagram of the turnover rate in the second moment;
The schematic flow sheet of the prediction turnover rate that Fig. 6 provides for the another embodiment of the application;
The schematic flow sheet of the prediction turnover rate that Fig. 7 provides for the another embodiment of the application;
The schematic flow sheet of the prediction turnover rate that Fig. 8 provides for the another embodiment of the application;
The module diagram of the device of the prediction turnover rate that Fig. 9 provides for the embodiment of the present application;
The module diagram of the system of the prediction turnover rate that Figure 10 provides for the embodiment of the present application.
Detailed description of the invention
For making the purpose of the application, technical scheme and advantage clearer, specifically real below in conjunction with the application
Execute example and technical scheme is clearly and completely described by corresponding accompanying drawing.Obviously, described
Embodiment is only some embodiments of the present application rather than whole embodiments.Based on the enforcement in the application
Example, the every other enforcement that those of ordinary skill in the art are obtained under not making creative work premise
Example, broadly falls into the scope of the application protection.
With reference to Fig. 1 a, shown is the networked environment 100 according to various embodiments.Networked environment
100 include the Operation Server carrying out data communication via network 112 and one or more clients 106
105, and Operation Server 105 or the computing environment 103 independent of Operation Server 105 can be integrated in.
Network 112 can include such as the Internet, in-house network, extranets, wide area network (WAN), LAN (LAN),
Wired network, wireless network or other suitable network etc., or any combination of two or more this kind of networks.
Run through the discussion of exemplary, it should be understood that term " data " and " information " can be the most mutual
Change for the text referring to may be present in computer based environment, image, audio frequency, video or any
The information of other form.
Client 106 may refer to be provided with the network equipment of application.Such network equipment can from hardware
To include server, desktop PC, laptop computer, tablet PC, smart phone, hand-held
Type computer, personal digital assistant (" PDA "), or the dress that other wired or wireless processor any drives
Put.From systems soft ware, can be the operating system being integrated with web browser, or be provided with special
The operating system of application;Such operating system can be operating system or the Linux behaviour of windows series
Make system etc., it is also possible to be Android, the IOS etc. in mobile platform.
Commercially available HTTP (HTTP) server can be included on Operation Server 105
Application, such as http server, Internet Information Service (IIS) and/or other server.
Once user uses client 106 and initiates registration request by network 112 to Operation Server 105,
Operation Server 105 can record the log-on message of this user, and according to registration request in Operation Server
Offer the account of preset authority.Follow-up each user accesses website or community by client 106, thus right
Time Operation Server 105 occurs to access, Operation Server 105 is based on carrying in information when accessing every time
The mark showing user identity of ID etc, can know and record such access.Particularly,
The access time of this user can be recorded, selectable, it is also possible to IP address during record access, or
The type of hardware of the network equipment used, or the operating system version that client release/client is based on
In the category information of basis one or more.Particularly, for there is the access of specific operation, such as, message is delivered
(as delivering message in social network sites), adds good friend, when buying virtual objects etc., and Operation Server
105 can record this operational access, and special by giving user after performing corresponding process interiorly or exteriorly
Fixed response.Usually, all operations request and general access, all can stay in Operation Server
Respective record.Operation Server can sort out all operations of different user respectively according to ID.Right
In the operation of interbehavior, such as mutual between different registration users, it is also possible to according to above-mentioned ID
Classify and constitute set.
Above-mentioned user access information can be stored with data base 115.Additionally, it is all right in data base 115
The data of storage include such as applying and require data, business rules, client end capacity data, application market number
According to, customer data etc..Application in data base 115 can with provided to be included in application by developer
Those application correspondences in market.Application can include such as Mobile solution, HTML 5 (HTML5)
Application, desktop application and/or other application.
Computing environment 103 can include that any other of such as server computer or offer computing capability is
System.Alternatively, computing environment 103 can use one or more calculating equipment, the plurality of calculating
Equipment can be arranged to the most one or more server group or calculate unit or other device.Such as, many
Individual calculating equipment can collectively form cloud computing resources, grid computing resource and/or other distributed meter any
Calculate device.The each method embodiment of following the application, can be configured with the one or more of executable instruction
Calculate and perform under the control of device.
According to various embodiments, various application and/or other function can be performed in computing environment 103
Property.The assembly performed in computing environment 103 such as includes that accessing data introduces service, accesses data analysis
Service, application data introducing service, application data analysis service and other application not detailed herein,
Service, process, system, engine or functional.
It addition, various data can be obtained by computing environment 103, such as, obtain from above-mentioned data base 115.
As scrutable, data base 115 can represent multiple data base 115.It is stored in data base 115
Data such as operation with various embodiments described below is associated.
The method of the prediction turnover rate that Fig. 1 provides for the embodiment of the present application, specifically includes following steps:
S11: the turnover rate data in preset duration before Operation Server obtained for the first moment.
First moment can be any one historical juncture including current time, the first moment of acquisition
Turnover rate data before can be the history of any preset time period including current turnover rate data
Turnover rate data.Exemplarily, if needing to predict the turnover rate of the subsequent time of current time, can be right
The turnover rate data in preset duration before current time and current time should be obtained.
In concrete acquisition process, can be that the main body predicting turnover rate as required is (such as operation website
Enterprise) different characteristics, add up the pass of turnover rate and calendar time date according to suitable measurement period
System.Exemplarily, here to illustrate by the time point at the end of month of every month statistics, within i.e. 1 year, have 12
Individual churn rate data.Using month as abscissa, turnover rate, as vertical coordinate, can obtain client's stream
Mistake rate and the corresponding relation of calendar time sequence.Meanwhile, also need to carry out noting by client for turnover rate data
Volume time vintage and the statistics of registration time length age.Unit is done equally, such as client 2014 with the moon
Registration in October in year, then, during calendar time date=201411, the registration time length age=1 of client, during registration
Between vintage=201410.Hour of log-on vintage can be that user passes through client 106 and accesses first
This time recorded by Operation Server 105 during registration Operation Server 105.Based on needs, Ke Yishe
Put the storage form of this hour of log-on vintage, such as when being converted into binary representation, occupy the size of byte,
Typically can set 2~4 bytes with storage.Calendar time date can be Operation Server 105 be
The system time.Clock generator in Operation Server 105 can cause and affect system time.During this system
Between most operation or services requiring time for benchmark and then can be affected in system.Registration time length age can be
Lead time between this time point and the hour of log-on vintage that obtain on sometime.Such as root
According to calendar time date obtained by the system time of Operation Server 105 and hour of log-on vintage phase
Every natural law, moon number and year number (can be as accurate as hourage, the number of minutes etc. as required).To run off
Rate data according to calendar time date, registration time length age and hour of log-on vintage carry out statistics be for
The turnover rate facilitating following step is decomposed and prediction process.
It should be noted that the turnover rate data that the turnover rate data of preset duration may refer to during this period can
To meet posterior data process needs, this can be an empirical value being preset, it is also possible to be root
Constantly adjust according to the turnover rate data obtained.For example, it is assumed that the first moment was 201410, and
Find in above-mentioned turnover rate data press the statistic processes of hour of log-on vintage: turnover rate data are first
It is continuous print on the time point that each measurement period before moment is corresponding, and the turnover rate number before the first moment
It is vintage=201410 according to corresponding hour of log-on at the latest, then shows that the turnover rate data now obtained are
Suitably;If do not connected time on the time point that each measurement period that turnover rate data are before the first moment is corresponding
Continue, or hour of log-on at the latest corresponding to turnover rate data before the first moment be not vintage=201410,
Then can increase preset duration further to obtain more turnover rate data.
S12: turnover rate data are decomposed into hour of log-on dimensional parameter, registration time length dimensional parameter and calendar
Time dimension parameter.
Ginseng Fig. 2 to Fig. 4, hour of log-on dimensional parameter represents the hour of log-on power of influence to turnover rate, registration
Duration dimensional parameter represents the registration time length power of influence to turnover rate, when calendar time dimensional parameter represents calendar
Between power of influence to turnover rate.And in concrete catabolic process, due to obtained with
Hour of log-on, registration time length and the calendar time data that turnover rate is corresponding, therefore can be respectively with hour of log-on
Vintage, registration time length age and calendar time date are coordinate, set up with hour of log-on dimensional parameter,
Registration time length dimensional parameter and the corresponding relation of calendar time dimensional parameter.
Decompose the hour of log-on dimensional parameter that obtains and registration time length dimensional parameter describe the inherence of client because of
The element impact on turnover rate, calendar time dimensional parameter then describes the external environmental factor shadow to turnover rate
Ring.Ginseng Fig. 2, for hour of log-on dimensional parameter, the such as visitor of hour of log-on vintage=201410
Family is mainly student group, and the client of hour of log-on vintage=201411 is mainly on-line shop storekeeper
Colony, it is assumed that storekeeper colony of on-line shop is less susceptible to run off relative to student group, shows hour of log-on dimensional parameter
On be then that hour of log-on dimensional parameter during hour of log-on vintage=201411 is relative to hour of log-on
Hour of log-on dimensional parameter during vintage=201410 is less, namely storekeeper colony of on-line shop is relative to student group
Body is less on the impact of turnover rate.Ginseng Fig. 3, for registration time length parameter, general registration time length is the longest
Client is more likely run off, and shows the increase being then as hour of log-on age on registration time length dimensional parameter,
Corresponding registration time length parameter is substantially in the trend increased.Ginseng Fig. 4, for calendar time parameter, such as
When calendar time date=201308, owing to some policy behavior of enterprise result in the note of substantial amounts of client
Pin, shows on calendar time dimensional parameter calendar time parameter when being then calendar time age=201308
Other calendar time relatively is significantly higher.By turnover rate data hour of log-on dimension, registration time length dimension and
The decomposition of calendar time dimension, describes the inside and outside factor causing customer churn, it is simple to following step
Turnover rate prediction steps.
In this step, decompose and obtain hour of log-on dimensional parameter, registration time length dimensional parameter and calendar time dimension
The process of degree parameter specifically can be such that
First, such as when setting turnover rate with hour of log-on dimensional parameter, registration time length dimensional parameter, calendar
Between there is corresponding relation: y=f (vintage)+f (age)+f (date) between dimensional parameter;Wherein y is for running off
Rate, f (vintage) be hour of log-on dimensional parameter, f (age) be registration time length dimensional parameter, f (date)
For calendar time dimensional parameter.Then these parameters are carried out following iterative process:
Iteration 1: f (vintage) and f (age) carries out initializing (giving one the most respectively to preset
Initial value), and then calculate f (date)=y-f (vintage)-f (age), and in step S11
Count the relation of turnover rate data and calendar time, therefore can be in the hope of f (date) value now;
Iteration 2: carry out initializing (the f that will try to achieve in iteration 1 by f (date) and f (vintage)
(date) value still gives above-mentioned default as the initial value of now f (date), f (vintage)
Initial value), and then calculate f (age)=y-f (vintage)-f (date), and step S11 has been united
Count out the relation of turnover rate data and registration time length, therefore can be in the hope of f (age) value now;
Iteration 3: carry out initializing (the f (date) that will try to achieve in iteration 1 by f (age) and f (date)
Being worth the initial value as now f (date), the f tried to achieve in iteration 2 (age) value is as f (age) now
Initial value), and then calculate f (vintage)=y-f (age)-f (date), and in step S11
Count the relation of turnover rate data and hour of log-on parameter, therefore can be in the hope of f (vintage) now
Value;
…
Iteration N: carry out initializing (the f (date) that will try to achieve in iteration N-2 by f (age) and f (date)
Being worth the initial value as now f (date), the f tried to achieve in iteration N-1 (age) value is as f now
(age) initial value), and then calculate f (vintage)=y-f (age)-f (date).
Here the purpose carrying out n times iteration is so that f (vintage), f (age), f (date)
Gradually tend to convergence, namely the value of f (vintage), f (age), f (date) is through n times iteration
After the most no longer change.And by the f (vintage) now obtained, f (age), f (date) really
It is set to and decomposes the hour of log-on dimensional parameter of gained, registration time length dimensional parameter, calendar according to turnover rate data
Time dimension parameter.
It should be noted that merely just with turnover rate and hour of log-on dimensional parameter, registration time length dimension ginseng
As a example by number, calendar time dimensional parameter have the corresponding relation of y=f (vintage)+f (age)+f (date), show
The concrete decomposition method of meaning turnover rate data.In other embodiments, turnover rate is joined with hour of log-on dimension
Number, registration time length dimensional parameter, calendar time dimensional parameter are it is also possible that have corresponding relation such as:
Y=α * f (vintage)+β * f (age)+γ * f (date), α, β, γ are weight coefficient;Or
Y=f (vintage) * f (age)+f (age) * f (date)+f (date) * f (vintage) etc..
Further, in other embodiments, implementing of decomposition can be to utilize other engineering configured
Practise algorithm, such as: EM algorithm (Expectation Maximiztion, EM) etc., the most not
Do more concrete example again.
S13: according to decomposing the hour of log-on dimensional parameter of gained, registration time length dimensional parameter and calendar time
Dimensional parameter, calculates the turnover rate in the second moment.
After being engraved in for the first moment when second, after the turnover rate in the second moment of calculating can be current time
The turnover rate of future time, namely turnover rate is predicted.
Ginseng Fig. 5, step S13 specifically includes:
S131: according to the hour of log-on dimensional parameter decomposing gained, it was predicted that before the first moment and the second moment
The respectively corresponding hour of log-on at the latest of turnover rate data between hour of log-on dimensional parameter.
Owing to the most the turnover rate data before the first moment being pressed by the client enrollment time
Vintage is added up, therefore the registration at the latest that turnover rate data before may determine that for the first moment are corresponding
Time.And for hour of log-on at the latest corresponding to the turnover rate data before the second moment, can be to adopt here
The first moment and second is increased on the hour of log-on at the latest that turnover rate data before being used in for the first moment are corresponding
The mode of the time difference between moment, it is also possible to be that the second moment was defaulted as the turnover rate data before the second moment
Corresponding hour of log-on at the latest.
Similarly, the hour of log-on dimensional parameter predicted here can be with the moon be prediction cycle unit.Example
Hour of log-on at the latest as corresponding in, it is assumed that turnover rate data before the first moment 201403 is 201403,
Second moment was 201411, then the hour of log-on dimensional parameter doped is hour of log-on
The hour of log-on dimensional parameter in nine months between vintage=201403~201411.
S132: according to the registration time length dimensional parameter decomposing gained, it was predicted that the second moment increased relative to the first moment
Add the registration time length dimensional parameter of duration part.
Similarly, according in step S11 to turnover rate data according to the statistics of client enrollment duration age, can
The maximum registration time length that turnover rate data before determining for the first moment are corresponding, and relative according to the second moment
The time difference in the first moment, it may be determined that need the registration time length of prediction to increase the registration time length dimensional parameter of part.
Similarly, the registration time length dimensional parameter predicted here can also be with the moon be prediction cycle unit.
When assuming for the first moment, the maximum registration time length age=6 that the registration time length dimensional parameter that counts is corresponding, and
Second moment was 6 with the time difference in the first moment, then the registration time length dimensional parameter doped is registration time length
Six registration time length dimensional parameter during age=7~12.
S133: according to the calendar time dimensional parameter decomposing gained, it was predicted that calendar time is in the first moment and the
Calendar time dimensional parameter between two moment;
Similarly, the registration time length dimensional parameter predicted here can also be with the moon be prediction cycle unit.
Assuming that the first moment was 201403, the second moment was 201411, then the calendar time dimensional parameter doped
For the calendar time dimensional parameter in nine months between calendar time date=201403~201411.
Above-mentioned, hour of log-on dimensional parameter, registration time length dimensional parameter and calendar time dimensional parameter are carried out
During prediction, owing to three groups of data are all seasonal effect in time series variablees, thus can use regression model and/
Or time series models are carried out above-mentioned three groups of data modeling and forecasting respectively.
Such as, with based on time series models to hour of log-on dimensional parameter, registration time length dimensional parameter and day
Last as a example by a dimensional parameter is predicted and be specifically described:
Time series { xtPrediction refer to: if xtAlready known, it is desirable to t+l (the i.e. second moment)
Future values xi+lIt is predicted, this predictive value xtL () represents, and become at initial time t forward
The predictive value of step l.The purpose of prediction is to make the forecast error of time series future value the least.Due to
Forecast error is a stochastic variable, it should make this variance of a random variable minimum.It is desirable to so select
Select predictive value xi(l) so that E [xt(l)]2=E [xt+l-xt(l)]2Becoming minimum, such prediction becomes
At moment t to future time instance t+l least standard criterion, and least standard criterion is by xt+lConditional expectation
Draw, it may be assumed that
xt(l)=E [xt+l|xt, xt-1..., x1]。
Forecast model according to ARMA
The foundation of ARMA forecast model: arma modeling is write as:
Use Green's function GjBy xt+lBeing divided into two parts, i.e. front l item is with the α of future time instancet+jIt is worth relevant, by
GlαtThe Part II started, cumulative and information including moment t.To this end, predictive value is write as
The error term that can estimate, i.e. α1, αt-l... weighted sum.Thus desired predictive value is:
Wherein, weighter factor G 't+jTo carry out preferably, so that the mean-square value of forecast error minimizes.And it is pre-
The expression formula surveying error is:
According to it is assumed that E [α during i ≠ jiαj]=0, the mean square expected value of forecast error is:
It is obvious that as optimum G 't+jEqual to G during actual valuel+j, E [et(l)]2For minimum.Therefore can obtain
Go out best predictor:
xt(l)=Et(xt+l)。
Here, can for hour of log-on dimensional parameter, registration time length dimensional parameter and calendar time dimensional parameter
To build above-mentioned time series { x respectivelyt, when can obtain each comfortable t+l (the i.e. second moment) accordingly
Predictive value.
S134: according to decomposing gained and the hour of log-on dimensional parameter of prediction, decomposing gained and the registration of prediction
Duration dimensional parameter, decomposition gained and the calendar time dimensional parameter of prediction, calculate the turnover rate in the second moment.
Step S134 can be equivalent to " inverse step " of step S12, by decomposing gained and the note of prediction
Volume time dimension parameter, decomposition gained and the registration time length dimensional parameter of prediction, decomposition gained and the day of prediction
Last a dimensional parameter to merge and collect, can again obtain turnover rate corresponding with calendar time date
Relation, namely turnover rate data.Correspondingly, turnover rate data now include the loss in the second moment
Rate, namely achieve the prediction to the second moment turnover rate.
S14: feed back described second moment turnover rate data to described Operation Server.
It should be noted that the second moment here can also be a historical juncture, for by the of prediction
The value of the value of the turnover rate in two moment and the turnover rate in the second actual moment is compared, or DIGEN further
According to comparison result, the model of prediction is carried out domestication etc..Further, according to accurately predicting the outcome,
The feedback information for user can be generated.And then, it is sent to client on the opportunity that Systematic selection is suitable, very
To the adjustment also including system of users internal data with the transmission of this feedback information.Such that it is able to certain journey
Avoid or reduce the loss of user on degree.
The method of the prediction turnover rate that Fig. 6 provides for the embodiment of the present application, specifically includes following steps:
S21: obtain the turnover rate data in preset duration before the first moment.
S22: turnover rate data are decomposed into hour of log-on dimensional parameter and registration time length dimensional parameter.
S23: according to hour of log-on dimensional parameter and the registration time length dimensional parameter of decomposition gained, calculate second
The turnover rate in moment;
S24: feed back described second moment turnover rate data to described Operation Server.
In the present embodiment, by turnover rate data being carried out in hour of log-on dimension and registration time length dimension point
Solve, and the parameter obtained according to decomposition, these two dimensions carry out hour of log-on dimensional parameter and note respectively
The calculating of volume duration dimensional parameter, can excavate the relation of turnover rate and client's endogenous cause of ill, improves the accurate of prediction
Degree.Further, according to accurately predicting the outcome, the feedback information for user can be generated.Enter
And, be sent to client on the opportunity that Systematic selection is suitable, even also include with the transmission of this feedback information be
The system adjustment to user's internal data.Such that it is able to avoid or reduce the loss of user to a certain extent.
The method of the prediction turnover rate that Fig. 7 provides for the embodiment of the present application, specifically includes following steps:
S31: obtain the turnover rate data in preset duration before the first moment.
S32: described turnover rate data are decomposed into hour of log-on dimensional parameter and calendar time dimensional parameter.
S33: according to hour of log-on dimensional parameter and the calendar time dimensional parameter of decomposition gained, calculate second
The turnover rate in moment.
S34: feed back described second moment turnover rate data to described Operation Server.
In the present embodiment, by turnover rate data being carried out in hour of log-on dimension and calendar time dimension point
Solve, and the parameter obtained according to decomposition, these two dimensions are carried out hour of log-on dimensional parameter and day respectively
Last the calculating of a dimensional parameter, turnover rate and the endo relation of client can be excavated, improve the standard of prediction
Exactness.Further, according to accurately predicting the outcome, the feedback information for user can be generated.
And then, it is sent to client on the opportunity that Systematic selection is suitable, even also includes with the transmission of this feedback information
The adjustment of system of users internal data.Such that it is able to avoid or reduce the loss of user to a certain extent.
The method of the prediction turnover rate that Fig. 8 provides for the embodiment of the present application, specifically includes following steps:
S41: obtain the turnover rate data in preset duration before the first moment.
S42: described turnover rate data are decomposed into registration time length dimensional parameter and calendar time dimensional parameter.
S43: according to registration time length dimensional parameter and the calendar time dimensional parameter of decomposition gained, calculate second
The turnover rate in moment.
S44: feed back described second moment turnover rate data to described Operation Server.
In the present embodiment, by turnover rate data being carried out in registration time length dimension and calendar time dimension point
Solve, and the parameter obtained according to decomposition, these two dimensions are carried out registration time length dimensional parameter and day respectively
Last the calculating of a dimensional parameter, turnover rate and the endo relation of client can be excavated, improve the standard of prediction
Exactness.Further, according to accurately predicting the outcome, the feedback information for user can be generated.
And then, it is sent to client on the opportunity that Systematic selection is suitable, even also includes with the transmission of this feedback information
The adjustment of system of users internal data.Such that it is able to avoid or reduce the loss of user to a certain extent.
The device of the prediction turnover rate that Fig. 9 provides for the embodiment of the present application, including:
Acquisition module 51, for the turnover rate data obtained before the first moment in preset duration;
Decomposing module 52, during for being decomposed into hour of log-on dimensional parameter, registration by described turnover rate data
Long dimensional parameter and calendar time dimensional parameter, wherein, described hour of log-on dimensional parameter represents hour of log-on
Power of influence to turnover rate, described registration time length dimensional parameter represents the registration time length power of influence to turnover rate,
Described calendar time dimensional parameter represents the calendar time power of influence to turnover rate;
Prediction module 53, for according to decomposing the hour of log-on dimensional parameter of gained, registration time length dimension ginseng
Number and calendar time dimensional parameter, calculate the turnover rate in the second moment, wherein, be engraved in described when described second
After first moment;
Feedback module 54, is used for feeding back described second moment turnover rate data to described Operation Server.
In the present embodiment, described prediction module 53 specifically for:
According to the hour of log-on dimensional parameter decomposing gained, it was predicted that the loss before the first moment and the second moment
Hour of log-on dimensional parameter between the hour of log-on at the latest that rate data are the most corresponding;
According to the registration time length dimensional parameter decomposing gained, it was predicted that the second moment increased duration relative to the first moment
The registration time length dimensional parameter of part;
According to the calendar time dimensional parameter decomposing gained, it was predicted that calendar time is in the first moment and the second moment
Between calendar time dimensional parameter;
According to the registration time length dimension decomposing gained and the hour of log-on dimensional parameter of prediction, decomposition gained and prediction
Degree parameter, decomposition gained and the calendar time dimensional parameter of prediction, calculate the turnover rate in the second moment.
In the present embodiment, described decomposing module 52 uses machine learning algorithm to carry out described turnover rate data
Decompose, described prediction module use regression model and/or time series models to hour of log-on dimensional parameter,
Registration time length dimensional parameter and calendar time dimensional parameter are predicted.
Continue ginseng Fig. 9, introduce the device of the prediction turnover rate that the embodiment of the present application provides, including:
Acquisition module 51, for the turnover rate data obtained before the first moment in preset duration;
Decomposing module 52, during for being decomposed into hour of log-on dimensional parameter and registration by described turnover rate data
Long dimensional parameter, wherein, described hour of log-on dimensional parameter represents the hour of log-on power of influence to turnover rate,
Described registration time length dimensional parameter represents the registration time length power of influence to turnover rate;
Prediction module 53, for according to hour of log-on dimensional parameter and the registration time length dimension ginseng decomposing gained
Number, calculates the turnover rate in the second moment, wherein, after being engraved in described first moment when described second;
Feedback module 54, is used for feeding back described second moment turnover rate data to described Operation Server.
Continue ginseng Fig. 9, introduce the device of the prediction turnover rate that the embodiment of the present application provides, including:
Acquisition module 51, for the turnover rate data obtained before the first moment in preset duration;
Decomposing module 52, in time being decomposed into hour of log-on dimensional parameter and calendar by described turnover rate data
Between dimensional parameter, wherein, described hour of log-on dimensional parameter represents the hour of log-on power of influence to turnover rate,
Described calendar time dimensional parameter represents the calendar time power of influence to turnover rate;
Prediction module 53, for according to hour of log-on dimensional parameter and the calendar time dimension ginseng decomposing gained
Number, calculates the turnover rate in the second moment, wherein, after being engraved in described first moment when described second.
Continue ginseng Fig. 9, introduce the device of the prediction turnover rate that the embodiment of the present application provides, including:
Acquisition module 51, for the turnover rate data obtained before the first moment in preset duration;
Decomposing module 52, in time being decomposed into registration time length dimensional parameter and calendar by described turnover rate data
Between dimensional parameter, wherein, described registration time length dimensional parameter represents the registration time length power of influence to turnover rate,
Described calendar time dimensional parameter represents the calendar time power of influence to turnover rate;
Prediction module 53, for according to registration time length dimensional parameter and the calendar time dimension ginseng decomposing gained
Number, calculates the turnover rate in the second moment, wherein, after being engraved in described first moment when described second;
Feedback module 54, is used for feeding back described second moment turnover rate data to described Operation Server.
The system of the prediction turnover rate that Figure 10 provides for the embodiment of the present application, including: predict as above
The device of turnover rate and DBM 54, data module 55, algoritic module 56 and processing module
57。
DBM 54 is for storing the turnover rate data of client, in addition, in DBM 54
Also manage the Back ground Information of client, log in, browse, the basis behavioral data such as purchase.
Data module 55 is used for the turnover rate data utilizing acquisition module 51 to obtain client in DBM,
And it is processed into predetermined data form.Predetermined data form mentioned here refers to that being available for decomposing module is carried out
The data form decomposed, meanwhile, data module 55 is additionally operable to guarantee and obtains enough from DBM
Data volume, if data volume is undesirable, points out the most accordingly.
Algoritic module 56 is used for providing decomposition algorithm, optimized algorithm and model algorithm.Model algorithm includes patrolling
Collect recurrence, neutral net, time series models etc..Algoritic module 56 is open module, can basis
Need freely to increase more different types of algorithm, and corresponding calling interface is provided.
Decomposing module 52 is for obtaining the turnover rate data after data module processes, and calls algoritic module
Turnover rate data after data module 55 processes are decomposed into hour of log-on dimension ginseng by the decomposition algorithm in 56
Number, registration time length dimensional parameter and calendar time dimensional parameter.
When prediction module 53 decomposes the hour of log-on dimensional parameter of gained, registration for obtaining decomposing module 52
Long dimensional parameter and calendar time dimensional parameter, call the model algorithm in algoritic module 56 respectively to registration
Time dimension parameter, registration time length dimensional parameter and calendar time dimensional parameter are predicted, and according to decomposition
The hour of log-on dimensional parameter of gained and prediction, registration time length dimensional parameter and calendar time dimensional parameter calculate
The turnover rate of predetermined instant.
Processing module 57 is for from the prediction turnover rate described in any one of described claim 7 to 12
Device obtains the second moment turnover rate data, and according to the second moment turnover rate data genaration for user's
Feedback information.
The methods, devices and systems of the prediction turnover rate that the embodiment of the present application provides, the method will be by obtaining
Turnover rate data decompose in hour of log-on dimension, registration time length dimension and calendar time dimension, and root
According to the parameter of decomposition gained, hour of log-on dimensional parameter, registration time length dimensional parameter and the day to future time instance
Last a dimensional parameter to be predicted, thus obtain the turnover rate data of future time instance accordingly;During due to registration
Between dimensional parameter and registration time length dimensional parameter represent hour of log-on and the registration time length shadow to turnover rate respectively
The power of sound, namely describe the endogenous cause of ill of customer churn, calendar time dimensional parameter represents that calendar time is to turnover rate
Power of influence, namely describe the exopathogenic factor of customer churn, and the turnover rate comprehensive consideration finally predicted client
The endogenous cause of ill run off and exopathogenic factor, therefore predict the outcome accurately.
Further, according to accurately predicting the outcome, the feedback information for user can be generated.Enter
And, be sent to client on the opportunity that Systematic selection is suitable, even also include with the transmission of this feedback information be
The system adjustment to user's internal data.Such that it is able to avoid or reduce the loss of user to a certain extent.
Those skilled in the art are it should be appreciated that embodiments of the invention can be provided as method, system or meter
Calculation machine program product.Therefore, the present invention can use complete hardware embodiment, complete software implementation or knot
The form of the embodiment in terms of conjunction software and hardware.And, the present invention can use and wherein wrap one or more
Computer-usable storage medium containing computer usable program code (include but not limited to disk memory,
CD-ROM, optical memory etc.) form of the upper computer program implemented.
The present invention is with reference to method, equipment (system) and computer program product according to embodiments of the present invention
The flow chart of product and/or block diagram describe.It should be understood that can by computer program instructions flowchart and
/ or block diagram in each flow process and/or flow process in square frame and flow chart and/or block diagram and/
Or the combination of square frame.These computer program instructions can be provided to general purpose computer, special-purpose computer, embedding
The processor of formula datatron or other programmable data processing device is to produce a machine so that by calculating
The instruction that the processor of machine or other programmable data processing device performs produces for realizing at flow chart one
The device of the function specified in individual flow process or multiple flow process and/or one square frame of block diagram or multiple square frame.
These computer program instructions may be alternatively stored in and computer or the process of other programmable datas can be guided to set
In the standby computer-readable memory worked in a specific way so that be stored in this computer-readable memory
Instruction produce and include the manufacture of command device, this command device realizes in one flow process or multiple of flow chart
The function specified in flow process and/or one square frame of block diagram or multiple square frame.
These computer program instructions also can be loaded in computer or other programmable data processing device, makes
Sequence of operations step must be performed to produce computer implemented place on computer or other programmable devices
Reason, thus the instruction performed on computer or other programmable devices provides for realizing flow chart one
The step of the function specified in flow process or multiple flow process and/or one square frame of block diagram or multiple square frame.
In a typical configuration, calculating equipment includes one or more processor (CPU), input/defeated
Outgoing interface, network interface and internal memory.
Internal memory potentially includes the volatile memory in computer-readable medium, random access memory
(RAM) and/or the form such as Nonvolatile memory, such as read only memory (ROM) or flash memory (flash RAM).
Internal memory is the example of computer-readable medium.
Computer-readable medium includes that removable media permanent and non-permanent, removable and non-can be by appointing
Where method or technology realize information storage.Information can be computer-readable instruction, data structure, program
Module or other data.The example of the storage medium of computer include, but are not limited to phase transition internal memory (PRAM),
Static RAM (SRAM), dynamic random access memory (DRAM), other kinds of at random
Access memorizer (RAM), read only memory (ROM), Electrically Erasable Read Only Memory (EEPROM),
Fast flash memory bank or other memory techniques, read-only optical disc read only memory (CD-ROM), digital multi light
Dish (DVD) or other optical storage, magnetic cassette tape, the storage of tape magnetic rigid disk or other magnetic storage set
Standby or any other non-transmission medium, can be used for the information that storage can be accessed by a computing device.According to herein
In define, computer-readable medium does not include temporary computer readable media (transitory media),
Data signal and carrier wave such as modulation.
Also, it should be noted term " includes ", " comprising " or its any other variant are intended to non-
Comprising of exclusiveness, so that include that the process of a series of key element, method, commodity or equipment not only wrap
Include those key elements, but also include other key elements being not expressly set out, or also include for this process,
The key element that method, commodity or equipment are intrinsic.In the case of there is no more restriction, statement " include
One ... " key element that limits, it is not excluded that including the process of described key element, method, commodity or setting
Other identical element is there is also in Bei.
It will be understood by those skilled in the art that embodiments herein can be provided as method, system or computer journey
Sequence product.Therefore, the application can use complete hardware embodiment, complete software implementation or combine software and
The form of the embodiment of hardware aspect.And, the application can use and wherein include calculating one or more
The computer-usable storage medium of machine usable program code (include but not limited to disk memory, CD-ROM,
Optical memory etc.) form of the upper computer program implemented.
The foregoing is only embodiments herein, be not limited to the application.For this area skill
For art personnel, the application can have various modifications and variations.All institutes within spirit herein and principle
Any modification, equivalent substitution and improvement etc. made, within the scope of should be included in claims hereof.
Claims (13)
1. the method predicting turnover rate, it is characterised in that including:
Under the control of one or more calculating devices being configured with executable instruction,
The turnover rate data in preset duration before Operation Server obtained for the first moment;
When described turnover rate data are decomposed into hour of log-on dimensional parameter, registration time length dimensional parameter and calendar
Between dimensional parameter, wherein, described hour of log-on dimensional parameter represents the hour of log-on power of influence to turnover rate,
Described registration time length dimensional parameter represents the registration time length power of influence to turnover rate, and described calendar time dimension is joined
Number represents the calendar time power of influence to turnover rate;
According to decomposing the hour of log-on dimensional parameter of gained, registration time length dimensional parameter and calendar time dimension ginseng
Number, calculates the turnover rate in the second moment, wherein, after being engraved in described first moment when described second;
Feed back described second moment turnover rate data to described Operation Server.
2. the method for claim 1, it is characterised in that according to the hour of log-on dimension decomposing gained
Degree parameter, registration time length dimensional parameter and calendar time dimensional parameter, calculate the turnover rate in the second moment, tool
Body includes:
According to the hour of log-on dimensional parameter decomposing gained, it was predicted that the loss before the first moment and the second moment
Hour of log-on dimensional parameter between the hour of log-on at the latest that rate data are the most corresponding;
According to the registration time length dimensional parameter decomposing gained, it was predicted that the second moment increased duration relative to the first moment
The registration time length dimensional parameter of part;
According to the calendar time dimensional parameter decomposing gained, it was predicted that calendar time is in the first moment and the second moment
Between calendar time dimensional parameter;
According to the registration time length dimension decomposing gained and the hour of log-on dimensional parameter of prediction, decomposition gained and prediction
Degree parameter, decomposition gained and the calendar time dimensional parameter of prediction, calculate the turnover rate in the second moment.
3. method as claimed in claim 2, it is characterised in that use machine learning algorithm to described stream
Mistake rate data are decomposed, use regression model and/or time series models to hour of log-on dimensional parameter,
Registration time length dimensional parameter and calendar time dimensional parameter are predicted.
4. the method predicting turnover rate, it is characterised in that including:
Under the control of one or more calculating devices being configured with executable instruction,
Obtain the turnover rate data in preset duration before the first moment;
Described turnover rate data are decomposed into hour of log-on dimensional parameter and registration time length dimensional parameter, wherein,
Described hour of log-on dimensional parameter represents the hour of log-on power of influence to turnover rate, and described registration time length dimension is joined
Number represents the registration time length power of influence to turnover rate;
Hour of log-on dimensional parameter according to decomposition gained and registration time length dimensional parameter, calculated for the second moment
Turnover rate, wherein, after being engraved in described first moment when described second;
Feed back described second moment turnover rate data to described Operation Server.
5. the method predicting turnover rate, it is characterised in that including:
Under the control of one or more calculating devices being configured with executable instruction,
Obtain the turnover rate data in preset duration before the first moment;
Described turnover rate data are decomposed into hour of log-on dimensional parameter and calendar time dimensional parameter, wherein,
Described hour of log-on dimensional parameter represents the hour of log-on power of influence to turnover rate, and described calendar time dimension is joined
Number represents the calendar time power of influence to turnover rate;
Hour of log-on dimensional parameter according to decomposition gained and calendar time dimensional parameter, calculated for the second moment
Turnover rate, wherein, after being engraved in described first moment when described second;
Feed back described second moment turnover rate data to described Operation Server.
6. the method predicting turnover rate, it is characterised in that including:
Under the control of one or more calculating devices being configured with executable instruction,
Obtain the turnover rate data in preset duration before the first moment;
Described turnover rate data are decomposed into registration time length dimensional parameter and calendar time dimensional parameter, wherein,
Described registration time length dimensional parameter represents the registration time length power of influence to turnover rate, and described calendar time dimension is joined
Number represents the calendar time power of influence to turnover rate;
Registration time length dimensional parameter according to decomposition gained and calendar time dimensional parameter, calculated for the second moment
Turnover rate, wherein, after being engraved in described first moment when described second;
Feed back described second moment turnover rate data to described Operation Server.
7. the device predicting turnover rate, it is characterised in that including:
Acquisition module, for the turnover rate data obtained before the first moment in preset duration;
Decomposing module, for being decomposed into hour of log-on dimensional parameter, registration time length dimension by described turnover rate data
Degree parameter and calendar time dimensional parameter, wherein, described hour of log-on dimensional parameter represents hour of log-on convection current
The power of influence of mistake rate, described registration time length dimensional parameter represents the registration time length power of influence to turnover rate, described
Calendar time dimensional parameter represents the calendar time power of influence to turnover rate;
Prediction module, for according to decompose the hour of log-on dimensional parameter of gained, registration time length dimensional parameter and
Calendar time dimensional parameter, calculates the turnover rate in the second moment, wherein, is engraved in described first when described second
After moment;
Feedback module, is used for feeding back described second moment turnover rate data to described Operation Server.
8. device as claimed in claim 7, it is characterised in that described prediction module specifically for:
According to the hour of log-on dimensional parameter decomposing gained, it was predicted that the loss before the first moment and the second moment
Hour of log-on dimensional parameter between the hour of log-on at the latest that rate data are the most corresponding;
According to the registration time length dimensional parameter decomposing gained, it was predicted that the second moment increased duration relative to the first moment
The registration time length dimensional parameter of part;
According to the calendar time dimensional parameter decomposing gained, it was predicted that calendar time is in the first moment and the second moment
Between calendar time dimensional parameter;
According to the registration time length dimension decomposing gained and the hour of log-on dimensional parameter of prediction, decomposition gained and prediction
Degree parameter, decomposition gained and the calendar time dimensional parameter of prediction, calculate the turnover rate in the second moment.
9. device as claimed in claim 8, it is characterised in that described decomposing module uses machine learning
Described turnover rate data are decomposed by algorithm, and described prediction module uses regression model and/or time series
Hour of log-on dimensional parameter, registration time length dimensional parameter and calendar time dimensional parameter are predicted by model.
10. the device predicting turnover rate, it is characterised in that including:
Acquisition module, for the turnover rate data obtained before the first moment in preset duration;
Decomposing module, for being decomposed into hour of log-on dimensional parameter and registration time length dimension by described turnover rate data
Degree parameter, wherein, described hour of log-on dimensional parameter represents the hour of log-on power of influence to turnover rate, described
Registration time length dimensional parameter represents the registration time length power of influence to turnover rate;
Prediction module, for the hour of log-on dimensional parameter according to decomposition gained and registration time length dimensional parameter,
Calculate the turnover rate in the second moment, wherein, after being engraved in described first moment when described second;
Feedback module, is used for feeding back described second moment turnover rate data to described Operation Server.
11. 1 kinds of devices predicting turnover rate, it is characterised in that including:
Acquisition module, for the turnover rate data obtained before the first moment in preset duration;
Decomposing module, for being decomposed into hour of log-on dimensional parameter and calendar time dimension by described turnover rate data
Degree parameter, wherein, described hour of log-on dimensional parameter represents the hour of log-on power of influence to turnover rate, described
Calendar time dimensional parameter represents the calendar time power of influence to turnover rate;
Prediction module, for the hour of log-on dimensional parameter according to decomposition gained and calendar time dimensional parameter,
Calculate the turnover rate in the second moment, wherein, after being engraved in described first moment when described second;
Feedback module, is used for feeding back described second moment turnover rate data to described Operation Server.
12. 1 kinds of devices predicting turnover rate, it is characterised in that including:
Acquisition module, for the turnover rate data obtained before the first moment in preset duration;
Decomposing module, for being decomposed into registration time length dimensional parameter and calendar time dimension by described turnover rate data
Degree parameter, wherein, described registration time length dimensional parameter represents the registration time length power of influence to turnover rate, described
Calendar time dimensional parameter represents the calendar time power of influence to turnover rate;
Prediction module, for the registration time length dimensional parameter according to decomposition gained and calendar time dimensional parameter,
Calculate the turnover rate in the second moment, wherein, after being engraved in described first moment when described second;
Feedback module, is used for feeding back described second moment turnover rate data to described Operation Server.
13. 1 kinds of systems predicting turnover rate, it is characterised in that including:
The device of the prediction turnover rate as described in any one of claim 7 to 12;And,
DBM, for storing the turnover rate data of client;
Data module, for obtaining the turnover rate data of client in DBM, and is processed into predetermined number
According to form;
Algoritic module, is used for providing decomposition algorithm and model algorithm;Wherein,
Described decomposing module is additionally operable to by obtaining the turnover rate data after data module processes, and calls calculation
Turnover rate data after data module processes are decomposed into hour of log-on dimension by the decomposition algorithm in method module
Parameter, registration time length dimensional parameter and calendar time dimensional parameter;
Hour of log-on dimension is joined by model algorithm that described prediction module is additionally operable to call in algoritic module respectively
Number, registration time length dimensional parameter and calendar time dimensional parameter are predicted, and according to decomposing gained and prediction
Hour of log-on dimensional parameter, registration time length dimensional parameter and calendar time dimensional parameter calculate predetermined instant
Turnover rate;
Processing module, for from the prediction turnover rate described in any one of described claim 7 to 12
Device obtains the second moment turnover rate data, and according to the second moment turnover rate data genaration for user's
Feedback information.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510301080.9A CN106250999A (en) | 2015-06-03 | 2015-06-03 | The methods, devices and systems of prediction turnover rate |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510301080.9A CN106250999A (en) | 2015-06-03 | 2015-06-03 | The methods, devices and systems of prediction turnover rate |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106250999A true CN106250999A (en) | 2016-12-21 |
Family
ID=57626389
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510301080.9A Pending CN106250999A (en) | 2015-06-03 | 2015-06-03 | The methods, devices and systems of prediction turnover rate |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106250999A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106845722A (en) * | 2017-02-06 | 2017-06-13 | 腾讯科技(深圳)有限公司 | A kind of method and apparatus for predicting customer volume |
CN109840790A (en) * | 2017-11-28 | 2019-06-04 | 腾讯科技(深圳)有限公司 | Prediction technique, device and the computer equipment of customer churn |
CN111160651A (en) * | 2019-12-31 | 2020-05-15 | 福州大学 | STL-LSTM-based subway passenger flow prediction method |
CN116051154A (en) * | 2023-03-06 | 2023-05-02 | 吉林省国迅广告有限公司 | Media terminal data analysis management system |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20040236648A1 (en) * | 2003-05-22 | 2004-11-25 | Pershing Investments, Llc | Activity-driven, customer profitability calculation system |
CN101620692A (en) * | 2008-06-30 | 2010-01-06 | 上海全成通信技术有限公司 | Method for analyzing customer churn of mobile communication service |
CN103854065A (en) * | 2012-11-30 | 2014-06-11 | 西门子公司 | Customer loss prediction method and device |
CN104021433A (en) * | 2014-06-18 | 2014-09-03 | 上海美琦浦悦通讯科技有限公司 | System and method for achieving user loss prediction in data information communication network |
CN104504460A (en) * | 2014-12-09 | 2015-04-08 | 北京嘀嘀无限科技发展有限公司 | Method and device for predicating user loss of car calling platform |
CN104616173A (en) * | 2015-02-11 | 2015-05-13 | 北京嘀嘀无限科技发展有限公司 | Method and device for forecasting user loss |
-
2015
- 2015-06-03 CN CN201510301080.9A patent/CN106250999A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20040236648A1 (en) * | 2003-05-22 | 2004-11-25 | Pershing Investments, Llc | Activity-driven, customer profitability calculation system |
CN101620692A (en) * | 2008-06-30 | 2010-01-06 | 上海全成通信技术有限公司 | Method for analyzing customer churn of mobile communication service |
CN103854065A (en) * | 2012-11-30 | 2014-06-11 | 西门子公司 | Customer loss prediction method and device |
CN104021433A (en) * | 2014-06-18 | 2014-09-03 | 上海美琦浦悦通讯科技有限公司 | System and method for achieving user loss prediction in data information communication network |
CN104504460A (en) * | 2014-12-09 | 2015-04-08 | 北京嘀嘀无限科技发展有限公司 | Method and device for predicating user loss of car calling platform |
CN104616173A (en) * | 2015-02-11 | 2015-05-13 | 北京嘀嘀无限科技发展有限公司 | Method and device for forecasting user loss |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106845722A (en) * | 2017-02-06 | 2017-06-13 | 腾讯科技(深圳)有限公司 | A kind of method and apparatus for predicting customer volume |
CN109840790A (en) * | 2017-11-28 | 2019-06-04 | 腾讯科技(深圳)有限公司 | Prediction technique, device and the computer equipment of customer churn |
CN109840790B (en) * | 2017-11-28 | 2023-04-28 | 腾讯科技(深圳)有限公司 | User loss prediction method and device and computer equipment |
CN111160651A (en) * | 2019-12-31 | 2020-05-15 | 福州大学 | STL-LSTM-based subway passenger flow prediction method |
CN116051154A (en) * | 2023-03-06 | 2023-05-02 | 吉林省国迅广告有限公司 | Media terminal data analysis management system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10135836B2 (en) | Managing data privacy and information safety | |
CN110266510B (en) | Network control strategy generation method and device, network control method and storage medium | |
CN109389412B (en) | Method and device for training model, service equipment and user equipment | |
CN106250999A (en) | The methods, devices and systems of prediction turnover rate | |
US9742684B1 (en) | Adaptive service scaling | |
CN105608065A (en) | Chart generation method and apparatus | |
CN108108490B (en) | Hive table scan method, apparatus, computer equipment and storage medium | |
US9184994B2 (en) | Downtime calculator | |
US20180060887A1 (en) | Brand equity prediction | |
US10726367B2 (en) | Resource allocation forecasting | |
US11221938B2 (en) | Real-time collaboration dynamic logging level control | |
CN116263899A (en) | Dynamic enhancement supply chain strategy based on carbon emission targets | |
CN109710410A (en) | A kind of internet information resource distribution method and device | |
US10277607B2 (en) | Login performance | |
US20230025848A1 (en) | Simulating weather scenarios and predictions of extreme weather | |
Tamura et al. | Practical reliability and maintainability analysis tool for an open source cloud computing | |
CN117616436A (en) | Joint training of machine learning models | |
US20210132912A1 (en) | Software detection and modification | |
US20230177383A1 (en) | Adjusting machine learning models based on simulated fairness impact | |
CN112988455A (en) | Method, apparatus and computer program product for data backup | |
US20200012987A1 (en) | Advising audit ratings in a multiple-auditor environment | |
CN114629675B (en) | Method, system and storage medium for making security recommendations | |
US11683317B2 (en) | Behavior management | |
Aksentijević et al. | Application of social network analysis to port community systems | |
US20230027115A1 (en) | Event-based record matching |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20161221 |