CN108009926A - Method, information processor and readable storage medium storing program for executing for user's classification - Google Patents
Method, information processor and readable storage medium storing program for executing for user's classification Download PDFInfo
- Publication number
- CN108009926A CN108009926A CN201711282768.2A CN201711282768A CN108009926A CN 108009926 A CN108009926 A CN 108009926A CN 201711282768 A CN201711282768 A CN 201711282768A CN 108009926 A CN108009926 A CN 108009926A
- Authority
- CN
- China
- Prior art keywords
- user
- family
- information
- head
- throwing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/06—Asset management; Financial planning or analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Finance (AREA)
- Data Mining & Analysis (AREA)
- Operations Research (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- Accounting & Taxation (AREA)
- Technology Law (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Embodiment of the invention discloses that method and information processor for user's classification.This method includes：Based on the context data associated with registered user, the characteristic information at the not first family of coming into operation in extraction registered user, the characteristic information at the not first family of coming into operation includes subscriber identity information and user behavior information；The characteristic information at the not first family of coming into operation extracted is input to prediction model, to predict the user that will be carried out first throwing within a predetermined period of time in not first family of coming into operation and thrown without head, wherein, the prediction model is generated according to the characteristic variable and target variable extracted based on the context data associated with history registry user, the characteristic variable includes historical user's identity information and historical user's behavioural information, and the target variable includes being used to describe the first information thrown of historical user.
Description
Technical field
Put it briefly, this disclosure relates to which data processing field, more specifically to the method classified for user, believes
Cease processing unit and computer-readable recording medium.
Background technology
It is increasingly perfect with network service, conventionally by cabinet face, scene etc. provide service on investment mode start it is convex
Limitation is showed, it cannot meet the needs of investor realizes investment activity whenever and wherever possible, and provide service on investment by network
Mode become to become more and more popular, its reason is largely that it solves conventional investment activity in the time, geographically
Limitation, realizes that whenever and wherever possible easily investment disclosure satisfy that the investment demand of user.Investee involved by investment activity is usual
Be have certain values can investment assets, such as it can include but is not limited to finance product, stock, fund, insurance, outer
Remittance, staple commodities, art collection product, real estate, patent right, music copyright etc..When by network implementation investments activity, investment
Person is needed in the mechanism with providing service on investment (for example, bank, internet financing corporation, securities broker company, insurance company etc. invest
Mechanism) registration on associated investment platform is (for example, directly registration or passing through third party and logging in access registration or other means and note
Volume), and investment activity is carried out with the identity of registered user after succeeding in registration.
Due to user's mobility enhancing that Netowrk tape comes, for investment platform, in each of subscriber lifecycle management
A stage (user's introducing, growth, maturation, decline and loss), which tries lifting user's retention ratio and increase user's value, to be become very
It is important.Traditionally, business and market analysis are generally basede in a manner of lifting user's retention ratio and conversion ratio, can be from product path
Set out with aspects such as user's customs, fractionation analysis such as is carried out to registration channel, cuts down the low channel of conversion ratio, or to investee
Analyzed using funnel retention ratio, transformed for the low investee's level of payment stage conversion rate, i.e. based on business and
An algorithm is done in plain data analysis, and is classified to user and carried out operation management.However, transported from algorithm to user
Battalion's system, generally manually carries out data transmission, without accomplishing to automate.Pass through the use of network access investment platform
Amount amount may be very huge, considers from business and the timeliness of operation, and the user of investment platform is lifted using traditional approach
No matter retention ratio and registration conversion ratio efficiency or effect are all not ideal enough.It is each that user invests (hereinafter referred to as " head is thrown ") first
For user, whether overlapping investment (hereinafter referred to as " multiple to throw ") generates very big influence for the experience of link, it is thus determined that first throw
Targeted customer and the first throwing targeted customers of non-then pointedly carry out operation management, conversion ratio are thrown so that lifting is first, for improving use
The viscosity at family has great importance so as to lift user's retention ratio.
Therefore, there is an urgent need for a kind of improved method to classify to user.
The content of the invention
The algorithm that is traditionally based on business and market analysis and establishes carries out user's classification and operation management, so
Do that speed is slow, efficiency is low and effect is undesirable, therefore such solution can not meet current network investment platform pair
The requirement of business and the timeliness of operation.The present invention is in view of the above problems, propose the method for user's classification, information processing
Device and computer-readable recording medium.
The first aspect of the disclosure provides a kind of method for user's classification, including：Based on related to registered user
The context data of connection, extracts the characteristic information at the not first family of coming into operation in the registered user, the feature at the not first family of coming into operation
Information includes subscriber identity information and user behavior information；And the characteristic information input at not first family of coming into operation described in being extracted
To prediction model, with what is thrown in not first family of coming into operation described in prediction by the user for carrying out first throwing within a predetermined period of time and without head
User.
The second aspect of the disclosure provides a kind of information processor for user's classification, including：Memory, it is used
In store instruction；And processor, it is coupled to the memory, and the processor is configured as performing based on described instruction
Operate below：Based on the context data associated with registered user, the spy for extracting the not first family of coming into operation in the registered user
Reference ceases, and the characteristic information at the not first family of coming into operation includes subscriber identity information and user behavior information；And it will be extracted
The characteristic information at the not first family of coming into operation is input to prediction model, will within a predetermined period of time in family to predict that the non-head comes into operation
The user for carrying out the user of first throwing and being thrown without head.
The third aspect of the disclosure provides a kind of computer-readable recording medium for being stored thereon with instruction.Described instruction
Method as described above is realized when executed.
Different from the life cycle management of traditionally user (particularly network investment platform user), the present invention takes into full account
To profession and user's feature and the timeliness of operation, multifaceted targeted customer is predicted by prediction model, is contributed to
The layering operation of differentiation is taken different classes of targeted customer, conversion ratio is thrown effectively to lift the first of registered user, is not required to
Want manual intervention and time efficiency is higher.
Brief description of the drawings
With reference to attached drawing and with reference to described in detail below, feature, advantage and the other side of the presently disclosed embodiments will become
Must be more obvious, show some embodiments of the disclosure by way of example, and not by way of limitation herein, in the accompanying drawings：
Fig. 1 show user and investment platform it is exemplary interact 100 schematic diagram；
Fig. 2 is shown according to the flow chart of the illustrative methods 200 for being used for user's classification of the embodiment of the present invention；
Fig. 3 is shown according to the schematic diagram of the exemplary information processing unit 300 for being used for user's classification of the embodiment of the present invention；
And
Fig. 4 is shown according to a specific example 400 of the method for being used for user's classification of the embodiment of the present invention.
Embodiment
Each exemplary embodiment of the disclosure is described in detail below with reference to attached drawing.Although be described below illustrative methods,
Device is included in the software and/or firmware performed among other components on hardware, it should be noted that these examples are only illustrative
, it should not see and be restricted.For example, it is contemplated that within hardware exclusively, in software exclusively or any group in hardware and software
It can implement any or all hardware, software and fastener components in conjunction.Therefore, although illustrative methods and device are described below,
But those skilled in the art should be easily understood that, there is provided example not only for realizing these method and apparatus modes.
In addition, the flow chart and block diagram in attached drawing show method and system according to various embodiments of the present disclosure can
Architectural framework, function and the operation that can be realized.It should be noted that the function of being marked in square frame can also be according to different from attached drawing
The order marked occurs.For example, two square frames succeedingly represented can essentially perform substantially in parallel, or they have
When can also perform in a reverse order, this depends on involved function.It should also be noted that flow chart and/or
The combination of each square frame and flow chart in block diagram and/or the square frame in block diagram, can use function or behaviour as defined in performing
The dedicated hardware based system made is realized, or can be realized using the combination of specialized hardware and computer instruction.
Word " exemplary " expression " serving as example, example or illustration " is used herein.Described herein as " example
Any embodiment of property " is all not necessarily construed to for other embodiments be preferable or advantageous.
Fig. 1 show user and investment platform it is exemplary interact 100 schematic diagram.Fig. 1 includes multiple users 102, multiple
User equipment 104, network 108 and server 112.User 102 accesses network by user equipment 104 via data link 106
108, and the investment platform on server 112 is accessed to realize interaction via data link 110.In other examples, investment is flat
Platform may reside within the infrastructure similar with server 112.In other examples, user equipment 104 can also pass through number
(not shown) server 112 is directly accessed according to link 106.User equipment 104 can be cell phone, personal digital assistant
(PDA), radio modem, wireless telecom equipment, handheld device, tablet PC, laptop computer, desktop computer
Or other any equipment that can be achieved to interact with investment platform etc..Data link 106 and 110 can include wired, wireless or mixed
Close connection.For example, user 102 can use the application on user equipment 104 (for example, browser application or special for investment platform
Application etc. of door exploitation) interacted to realize with investment platform.User from introduce complete registration, using application, to it is following may
Capital participation etc. it is each during there may be data, these data can produce value to solving business demand, this paper
It is middle that the data for interacting and producing described between user and investment platform are referred to as context data.It is it is, for example, possible to use each
Kind big data framework (for example, Kafka, Flume or its any combination etc.) and/or other data structures use registration to realize
Collection, transmission and the storage of the context data at family, wherein, some can be online (real-time) to Data Collection, some can be with
It is offline.It should be appreciated that quantity, type and the framework of the user, user equipment and server in Fig. 1 are merely exemplary,
It is rather than restricted.
Fig. 2 is shown according to the flow chart of the illustrative methods 200 for being used for user's classification of the embodiment of the present invention.Such as flow chart
Shown, method 200 comprises the following steps：
Step S201：Based on the context data associated with registered user, extract the non-head in registered user and come into operation family
Characteristic information, the characteristic information at not first family of coming into operation includes subscriber identity information and user behavior information.In this step, with note
The context data that volume user is associated can be from using big data framework and in the database stored or other data structures
Obtain.Registered user is come into operation family and head has come into operation family including non-head, is handled herein for non-head families of coming into operation, non-head comes into operation
Family is potential investment colony, as described above, lifts first conversion ratio of throwing and has great importance for lifting user's retention ratio.
The characteristic information at not first family of coming into operation can be automatically extracted for example, by machine learning algorithm.The characteristic information extracted can be
Characteristic information in context data or the characteristic information obtained based on context data.Characteristic information can include user's body
Part information and user behavior information.For example, subscriber identity information can include but is not limited to the ID of user, gender, educational background, duty
Industry, age, annual income, hour of log-on and source, real name information, credit grade etc..For example, user behavior information can include but
Be not limited to participate in action message, supplement with money enchashment behavior, binding bank card, call-information, different time sections application behavior, using row
For statistical information etc..
Step S202：The characteristic information at the not first family of coming into operation extracted is input to prediction model, to predict that non-head comes into operation
The user that will be carried out the user of first throwing within a predetermined period of time in family and be thrown without head, wherein, prediction model is that basis is based on
The context data associated with history registry user and the characteristic variable and target variable extracted generate, characteristic variable bag
Historical user's identity information and historical user's behavioural information are included, target variable includes being used to describe the first information thrown of historical user.
Historical user's identity information and historical user's behavioural information can be analogous respectively to subscriber identity information as described above and use
Family behavioural information, for describing the first information thrown of historical user can include but is not limited to whether user carried out first throwing, head is thrown
Amount of money etc..In this step, classified by prediction model come family of coming into operation to non-head, i.e. carry out within a predetermined period of time first
The user for throwing and being thrown without head.For example, the context data associated with history registry user can be from utilizing big data
Framework and the database stored obtain in other data structures.In this step, for example, by that will have known target change
The sample for measuring label generates prediction model to be trained according to characteristic variable.For example, the process of generation prediction model can be with
Including carrying out data cleansing and characteristic processing to context data, the historical data for obtaining having target variable label is as model
Training data and characteristic variable is obtained, and be based on model training data, using training algorithm (for example, machine learning algorithm
Deng) be trained and obtain prediction model.Data cleansing is data prediction step, such as data is filled, polishing, mark
Standardization, outlier processing etc., its purpose have two, and first is to allow data can use by cleaning, and second is to allow data to become more suitable
Close and carry out follow-up processing.Characteristic processing is to obtain and filter out significant characteristic variable and be inputted training algorithm progress
Model training.In one embodiment, prediction model can be realized using disaggregated model.It is for instance possible to use machine learning
In Random Forest model, which is made of more decision trees, and input independent variable includes the use extracted from context data
Family identity information and user behavior information, output dependent variable are classification results (for example, user tag), when sample to be sorted enters
During random forest, classified by more decision trees, finally chosen by the most classification of all decision trees selection number as most
Whole classification results.In one embodiment, prediction model can be realized using regression model.For example, it can equally use
Random Forest model, output dependent variable are class probability, when the one or more probability threshold value of selection, it is possible to achieve two classification and
More classification.In other examples, similar prediction effect can be realized using any other suitable model.
Alternatively, method 200 can also comprise the following steps：It is customized for the resource of predicted user to encourage use
Family.In this step, for the user predicted, relevant resource is customized for it, for example, pushing particular message to user, beating
Phone, send reward voucher etc., pointedly to encourage response of the potential user to resource.Alternatively, method 200 can also wrap
Include：Response based on the user predicted to customized resource, further to adjust the resource for user.For example, can be with
The transmission of reward voucher is adjusted based on responsiveness.
In this way, classified by prediction model to user so that follow-up operation is as far as possible to compared with there may be response
Targeted customer carry out, on the one hand save operation cost, on the other hand reduce excessive operation caused by non-targeted user bear
The harassing and wrecking of face effect.Meanwhile targeted customer responds the resource of such as least cost certificate etc, and can be thus to different use
Family carries out differentiation resource allocation (for example, cost certificate of different prices).
Come into operation family for the non-head in registered user, it is understood that there may be two types：First throw does not Add User, not first to throw storage
User.First throw can be distinguished by the length of hour of log-on to Add User and first throwing storage user.For example, by a days with
The user of lower registration is known as Adding User, and the user of registration in more than b days is known as storage user.
In one embodiment, non-head families of coming into operation can include not first throw and Add User, and prediction model can include
First disaggregated model.In this embodiment, the characteristic information that the not first throwing that step S202 can include being extracted Adds User
The first disaggregated model is input to, head throws middle the first mesh that first throwing will be carried out in the first predetermined amount of time that Adds User to predict
Mark user and the second targeted customer thrown without head.Here, the first disaggregated model is referred to as above in step S202
Classify for prediction model is described.
In this embodiment, alternatively, prediction model can include regression model, and method 200 can also include：Will
The characteristic information of the first object user extracted is input to regression model, to predict that the first of first object user throws amount of money model
Enclose.For example, regression model can include linearly or nonlinearly regression model, decision tree regression model etc..In one example, return
Model is returned to include GBDT (Gradient Boosting Decision Tree, gradient lifting decision tree) model, GBDT moulds
Type is lifted by more weak decision trees (regression tree) of iteration come Shared Decision Making, and the result of more decision trees is added up
Exported as final prediction, preferable effect can be attained by two aspects of training precision and generalization ability.Except passing
The GBDT model realization modes of system, GBDT models can also efficiently realize that xgboost is one big rule using xgboost
Mould, distributed general Gradient Boosting (GBDT) storehouse, it is realized under Gradient Boosting frames
The linear machine learning algorithm of GBDT and some broad sense, and with speed is fast, effect is good, can handle large-scale data, supports oneself
The advantages that defining loss function.Different from traditional GBDT implementations, the derivative information of single order is only utilized, xgboost is to damage
Lose function and done the Taylor expansion of second order, and add regular terms outside object function and integrally seek optimal solution, to weigh mesh
The decline of scalar functions and the complexity of model, avoid over-fitting.
Alternatively, in this embodiment, method 200 can also include：First throwing based on the first object user predicted
Amount of money scope is customized for the resource of first object user to encourage first object user.For example, the amount of money can be thrown for head
In the user of different interval ranges, different frequency and dynamics is taken to carry out PUSH message, make a phone call, send reward voucher etc. to encourage use
Family is converted on faster or more the first direction for throwing number, and first conversion ratio is thrown so as to be lifted.
In another embodiment, non-head come into operation family can include it is first throw storage user, and prediction model can wrap
Include the second disaggregated model.In this embodiment, the not first feature for throwing storage user that step S202 can include being extracted is believed
Breath is input to the second disaggregated model, to predict that not first throw in storage user will carry out the 3rd of first throwing in the second predetermined amount of time
Targeted customer.Here, the second disaggregated model be referred to as above in step S202 for prediction model it is described come into
Row classification.
Alternatively, method 200 can also include：The context data associated with history registry user periodically carries out
Renewal.For example, the context data that registered user produces within a period can be collected, transmits and store then be added
To the context data associated with history registry user so that the context data associated with history registry user can week
Phase property it is updated, is updated so as to fulfill the closed loop of data, while the updated context data can be used for training in advance
Model is surveyed, so as to fulfill automatically updating for training pattern.
Fig. 3 is shown according to the schematic diagram of the exemplary information processing unit 300 for being used for user's classification of the embodiment of the present invention.
Device 300 can include：Memory 301 and the processor 302 for being coupled to memory 301.Memory 301 for storing instruction,
Processor 302 is configured as realizing such as the described method of any embodiment in Fig. 2 based on the instruction of the storage of memory 301.
As shown in figure 3, device 300 can also include communication interface 303, for carrying out information exchange with miscellaneous equipment.This
Outside, device 300 can also include bus 304, and memory 301, processor 302 and communication interface 303 are by bus 304 come each other
Communicate.
Memory 301 can include volatile memory, can also include nonvolatile memory.Processor 302 can be with
It is central processing unit (CPU), microcontroller, application-specific integrated circuit (ASIC), digital signal processor (DSP), field-programmable
Gate array (FPGA) or other programmable logic device or the one or more collection for being configured as realizing the embodiment of the present invention
Into circuit.
In order to preferably express the design of the present invention, illustrated with reference to a specific example.
Fig. 4 is shown according to a specific example 400 of the method for being used for user's classification of the embodiment of the present invention.
In example 400, some users 402 are registered as financing platform user by the APP that manages money matters, with financing platform
In interaction, user from introduce complete registration, using APP, to it is following may capital participation etc. it is each during there may be
Context data 404.Context data 404 can be collected, transmit and store big data equipment 406.For example, it is based on
Kafka and Flume frameworks carry out context data and collect transmission, and necessary parsing storage is finally carried out to context data and is arrived
SparkHive.In some scenes, part storage can also be carried out in oracle, mysql and mongodb.
User this period before first throw is registered to is relatively short and without investment behavior, and the data that user produces are few, because
This characteristic information has been taken into the key issue that must be faced.For example, what can be stored from big data equipment 406 uses with registration
The characteristic information 408 at not first family of coming into operation is extracted in the context data that family is associated, characteristic information 408 can include user identity
Information and user behavior information.For example, subscriber identity information can the ID including user, gender, educational background, occupation, age, year receipts
Enter, hour of log-on (for example, registration number of days etc.), registration source (for example, clicked on from the web advertisement redirect, registered users are recommended
Link registration, third party software redirect registration or directly registration etc.), real name information, credit grade etc..For example, user behavior is believed
Breath can include binding identity card and bank card, the respondent behavior to history operation activity, call-information (for example, financing platform
Customer service and user's communication etc.), supplement enchashment behavior, click behavior etc. on financing APP with money.Based on to internet financial business
Consider, these data contribute to the prediction for being to the not first first investment bank at family of coming into operation.It is also possible to stored from big data equipment 406
The context data associated with history registry user in extract characteristic variable 410 and target variable 412 (for example, via such as
Data cleansing described above and characteristic processing).Characteristic variable 410 can include historical user's identity information and historical user's row
For information.Target variable 412 can include being used to describe the first information thrown of historical user.According to training algorithm (for example, GBDT is calculated
Method), feature based variable 410 and target variable 412 train the historical data with user tag to obtain prediction model 414.
Then, the characteristic information 408 at the not first family of coming into operation extracted is input to the prediction model 414 generated, so as to export user
Whether label 416, such as non-head families of coming into operation carry out first throwing, first throw amount of money scope etc..User 402 can produce over a period
Context data, the context data are collected, transmit and store big data equipment 406, then pair with history registry user's phase
Associated context data is updated, and is updated so as to fulfill the closed loop of data, while updated data can be used for training
Prediction model 414, so as to fulfill automatically updating for prediction model 414.It is alternatively possible to send user tag 416 to operation
Management end (not shown), to carry out specific aim operation.
According to registration number of days, non-head comes into operation family can be including head throwings Add User and/or head throws storage user.Prediction
Model 414 can include classification and/or regression model.
Add User if there is not first throw, then the characteristic information 408 of the user extracted is input to prediction model 414
The first disaggregated model, with predicting from the same day to following distance registration x1 days can first throwing first object user (for example, with
Family 1,2 and will not first the second targeted customer (for example, user 4) thrown 3) and within x1 days.For the first first object thrown of meeting
User, operation management end can take the measures such as phone, APP pushed informations or hair certificate to be stimulated, it is desirable to which user can throw head
Time further shifts to an earlier date, so as to lift overall registration head throwing rates；For will not first the second targeted customer thrown, operation management end
Corresponding certificate stimulation of increasing the interest can be given, it is desirable to which user throws direction transformation towards head.It is possible to further the first mesh that will be extracted
The characteristic information 408 of mark user is input to the regression model of prediction model 414, to predict that the first of first object user throws amount of money model
Enclose.The amount of money is thrown in the first object user of different interval ranges for head, and operation management end can take different frequency and dynamics
APP pushed informations, phone, the measure such as certificate of increasing the interest stimulated, it is desirable to user is on faster or more the first direction for throwing number
Converted.
For example, user 1 is had found with good personal credit by Data Collection and frequentlys click on finance product, user
2 annual income is higher and is bundled with bank card and checks high value finance product, and the annual income of user 3 is medium and has supplemented with money certain
The amount of money, user 4 is professional and occasionally browses financing message.The feature letter extracted based on these from context data
Breath, is predicted by prediction model 414, and user 1,2 and 3, which is predicted to be, first within x1=30 days to throw, and user
1st, 2 and 3 estimated first amount of money scope of throwing is respectively 50000~100000,200000~500000 and 2000~5000 yuan, accordingly
Ground, operation management end can send 1.5%, 2.5% and 0.5% certificate of increasing the interest respectively to user 1,2 and 3 and be stimulated；And user 4
It is predicted to be and first within x1=30 days will not throws, correspondingly, operation management end can sends 0.25% certificate of increasing the interest to user 4 and give
To stimulate.Further pay close attention to, collect, seeking advice from the behaviors such as certain class finance product for example, user 1 after certificate of increasing the interest is received, generates,
Operation management end can be preferably that user 1 reserves the finance product of certain amount or adjusts other resources for user 1 etc..Class
As, operation management end can be according to the response of user 2,3 and 4, further to adjust the resource for these users.
Storage user is thrown if there is not first, then the characteristic information 408 of the user extracted is input to prediction model 414
The second disaggregated model, with predicting from the same day to following distance registration x2~x3 days can first throwing targeted customer (for example, with
Family 6 and 7).For the first targeted customer thrown of meeting, operation management end can take the measures such as phone, APP pushed informations or hair certificate to give
To stimulate, it is desirable to which user can further shift to an earlier date the head throwing times, so as to lift overall registration head throwing rates.For example, to 6 He of user
7 prediction and Resources Customization can be similar to described above for user 1,2 and 3.
In the example 400, using the user tag 416 of the output of prediction model 414, differentiation fortune is done at operation management end
Battalion, to lift business objective.
Whether the present invention produces expected behavior predicting user on the scheduled time by prediction model to excavate mesh
User is marked, contributes to user operating side to make targeted customer towards the conversion of first throwing direction or further shift to an earlier date first throwing time or the first throwing of raising
The amount of money, so that under conditions of cost control, relatively precisely and significantly lifts user's head throwing rates, and then realize user's retention ratio
Lifting.Meanwhile compared with traditional user's classification and operation based on business rule, the present invention fully takes into account business and user
Feature and the timeliness of operation, all processing can all be completed by computer disposal, it is not necessary to which manual intervention, time efficiency are higher.
In addition, alternatively, the above-mentioned method for being used for user's classification can pass through computer program product, i.e., tangible meter
Calculation machine readable storage medium storing program for executing is realized.Computer program product can include computer-readable recording medium, containing for
Perform the computer-readable program instructions of various aspects of the disclosure.Computer-readable recording medium can be kept and deposit
Storage is performed the tangible device for the instruction that equipment uses by instruction.Computer-readable recording medium can for example be but not limited to electricity and deposit
Store up equipment, magnetic storage apparatus, light storage device, electromagnetism storage device, semiconductor memory apparatus or above-mentioned any appropriate
Combination.The more specifically example (non exhaustive list) of computer-readable recording medium includes：Portable computer diskette, hard disk,
Random access memory (RAM), read-only storage (ROM), erasable programmable read only memory (EPROM or flash memory), static state
Random access memory (SRAM), Portable compressed disk read-only storage (CD-ROM), digital versatile disc (DVD), memory stick,
Floppy disk, mechanical coding equipment, the punch card for being for example stored thereon with instruction or groove internal projection structure and above-mentioned any conjunction
Suitable combination.Computer-readable recording medium used herein above is not interpreted instantaneous signal in itself, such as radio wave or
The electromagnetic wave of the other Free propagations of person, the electromagnetic wave propagated by waveguide or other transmission mediums are (for example, pass through fiber optic cables
Light pulse) or by electric wire transmit electric signal.
In addition, computer-readable program instructions or computer program product for performing various aspects of the disclosure
It can store beyond the clouds, when needing to call, user can be stored in by mobile Internet, fixed network or other network access
The computer-readable program instructions for being used to perform various aspects of the disclosure on high in the clouds, so that the implementation basis disclosure is each
Technical solution disclosed in aspect.
The foregoing is merely the alternative embodiment of the disclosure, embodiment of the disclosure is not limited to, for this area
Technical staff for, embodiment of the disclosure can have various modifications and variations.It is all embodiment of the disclosure spirit and
Within principle, any modification for being made, equivalence replacement, improvement etc., should be included in embodiment of the disclosure protection domain it
It is interior.
Although describe embodiment of the disclosure by reference to some specific embodiments, it should be appreciated that, the disclosure
Embodiment is not limited to disclosed specific embodiment.Embodiment of the disclosure be intended to appended claims spirit and
In the range of included various modifications and equivalent arrangements.Scope of the following claims meets broadest explanation, so that comprising
All such modifications and equivalent structure and function.
Claims (13)
- A kind of 1. method for user's classification, it is characterised in that including：Based on the context data associated with registered user, the feature for extracting the not first family of coming into operation in the registered user is believed Breath, wherein, the characteristic information at the not first family of coming into operation includes subscriber identity information and user behavior information；AndThe characteristic information at not first family of coming into operation is input to prediction model described in being extracted, will in family to predict that the non-head comes into operation The user for carrying out the user of first throwing within a predetermined period of time and being thrown without head, wherein, the prediction model be according to be based on Context data that history registry user is associated and the characteristic variable and target variable extracted generate, wherein, the spy Sign variable includes historical user's identity information and historical user's behavioural information, and the target variable includes being used to describe historical user The information that head is thrown.
- 2. according to the method described in claim 1, it is characterized in that, the non-head comes into operation, family includes not first throw and Adds User, and And the prediction model includes the first disaggregated model,Wherein, the characteristic information at not first family of coming into operation is input to prediction model described in being extracted, to predict that the non-head comes into operation Include in family by the user for carrying out first throwing within a predetermined period of time and without the head users thrown：It is not first described in being extracted to throw The characteristic information to Add User is input to first disaggregated model, will be pre- first in being Added User with first throwing described in prediction The second targeted customer that the first object user of first throwing is carried out in section of fixing time and is thrown without head.
- 3. according to the method described in claim 2, it is characterized in that, the prediction model includes regression model, the method is also Including：The characteristic information of the first object user extracted is input to the regression model, to predict the first object The first of user throws amount of money scope.
- 4. according to the method described in claim 3, it is characterized in that, the regression model includes GBDT models.
- 5. according to the method described in claim 1, it is characterized in that, the non-head comes into operation family include it is first throw storage user, and And the prediction model includes the second disaggregated model,Wherein, the characteristic information at not first family of coming into operation is input to prediction model described in being extracted, to predict that the non-head comes into operation Include in family by the user for carrying out first throwing within a predetermined period of time and without the head users thrown：It is not first described in being extracted to throw The characteristic information of storage user is input to second disaggregated model, with first throw will be not pre- second in storage user described in prediction The 3rd targeted customer of first throwing is carried out in section of fixing time.
- 6. the according to the method described in claim 1, it is characterized in that, context data associated with the history registry user Periodically it is updated.
- A kind of 7. information processor for user's classification, it is characterised in that including：Memory, it is for storing instruction；AndProcessor, it is coupled to the memory, and the processor is configured as performing following operation based on described instruction：Based on the context data associated with registered user, the feature for extracting the not first family of coming into operation in the registered user is believed Breath, the characteristic information at the not first family of coming into operation include subscriber identity information and user behavior information；The characteristic information at not first family of coming into operation is input to prediction model described in being extracted, will in family to predict that the non-head comes into operation The user for carrying out the user of first throwing within a predetermined period of time and being thrown without head, wherein, the prediction model be according to be based on Context data that history registry user is associated and the characteristic variable and target variable extracted generate, the characteristic variable Including historical user's identity information and historical user's behavioural information, the target variable includes being used to describe the first throwing of historical user Information.
- 8. information processor according to claim 7, it is characterised in that the non-head come into operation family include it is first throw it is newly-increased User, and the prediction model includes the first disaggregated model,Wherein, the characteristic information at not first family of coming into operation is input to prediction model described in being extracted, to predict that the non-head comes into operation Include in family by the user for carrying out first throwing within a predetermined period of time and without the head users thrown：It is not first described in being extracted to throw The characteristic information to Add User is input to first disaggregated model, will be pre- first in being Added User with first throwing described in prediction The second targeted customer that the first object user of first throwing is carried out in section of fixing time and is thrown without head.
- 9. information processor according to claim 8, it is characterised in that the prediction model includes regression model,Wherein, the processor is additionally configured to perform following operation based on described instruction：First mesh that will be extracted The characteristic information of mark user is input to the regression model, to predict that the first of the first object user throws amount of money scope.
- 10. information processor according to claim 9, it is characterised in that the regression model includes GBDT models.
- 11. information processor according to claim 7, it is characterised in that non-head families of coming into operation include not first throw and deposit User is measured, and the prediction model includes the second disaggregated model,Wherein, the characteristic information at not first family of coming into operation is input to prediction model described in being extracted, to predict that the non-head comes into operation Include in family by the user for carrying out first throwing within a predetermined period of time and without the head users thrown：It is not first described in being extracted to throw The characteristic information of storage user is input to second disaggregated model, with first throw will be not pre- second in storage user described in prediction The 3rd targeted customer of first throwing is carried out in section of fixing time.
- 12. information processor according to claim 7, it is characterised in that associated with the history registry user Context data is periodically updated.
- 13. a kind of computer-readable recording medium, it is characterised in that the computer-readable recording medium storage has instruction, institute State the method for instructing and being used for realization when executed as any one of claim 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711282768.2A CN108009926A (en) | 2017-12-07 | 2017-12-07 | Method, information processor and readable storage medium storing program for executing for user's classification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711282768.2A CN108009926A (en) | 2017-12-07 | 2017-12-07 | Method, information processor and readable storage medium storing program for executing for user's classification |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108009926A true CN108009926A (en) | 2018-05-08 |
Family
ID=62057316
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711282768.2A Pending CN108009926A (en) | 2017-12-07 | 2017-12-07 | Method, information processor and readable storage medium storing program for executing for user's classification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108009926A (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108764994A (en) * | 2018-05-24 | 2018-11-06 | 深圳前海桔子信息技术有限公司 | A kind of user behavior guidance method, device, server and storage medium |
| CN109102176A (en) * | 2018-07-25 | 2018-12-28 | 珠海宏桥高科技有限公司 | A kind of trade finance appraisal procedure and device based on block chain |
| CN109684549A (en) * | 2018-12-24 | 2019-04-26 | 拉扎斯网络科技（上海）有限公司 | Target data prediction technique, device, electronic equipment and computer storage medium |
| CN109886349A (en) * | 2019-02-28 | 2019-06-14 | 成都新希望金融信息有限公司 | A kind of user classification method based on multi-model fusion |
| CN110555714A (en) * | 2018-06-04 | 2019-12-10 | 百度在线网络技术（北京）有限公司 | method and apparatus for outputting information |
| CN110717537A (en) * | 2019-10-08 | 2020-01-21 | 支付宝(杭州)信息技术有限公司 | Method and device for training user classification model and executing user classification prediction |
| CN111260416A (en) * | 2020-02-13 | 2020-06-09 | 支付宝(杭州)信息技术有限公司 | Method and device for determining associated user of object |
| CN111311338A (en) * | 2020-03-30 | 2020-06-19 | 网易（杭州）网络有限公司 | User value prediction method and user value prediction model training method |
| CN111680382A (en) * | 2019-02-25 | 2020-09-18 | 北京嘀嘀无限科技发展有限公司 | Grade prediction model training method, grade prediction device and electronic equipment |
| CN111737575A (en) * | 2020-06-19 | 2020-10-02 | 北京字节跳动网络技术有限公司 | Content distribution method and device, readable medium and electronic equipment |
| CN112070533A (en) * | 2020-08-28 | 2020-12-11 | 上海连尚网络科技有限公司 | Method and equipment for predicting user retention |
| CN112308638A (en) * | 2020-10-29 | 2021-02-02 | 北京中科闻歌科技股份有限公司 | False invoice behavior detection method and device, electronic equipment and storage medium |
| CN112529236A (en) * | 2019-09-18 | 2021-03-19 | 泰康保险集团股份有限公司 | Target object identification method and device, electronic equipment and storage medium |
| CN113297478A (en) * | 2021-04-25 | 2021-08-24 | 上海淇玥信息技术有限公司 | Information pushing method and device based on user life cycle and electronic equipment |
| CN113538070A (en) * | 2020-10-30 | 2021-10-22 | 深圳市九九互动科技有限公司 | User life value cycle detection method and device and computer equipment |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040117239A1 (en) * | 2002-12-17 | 2004-06-17 | Mittal Parul A. | Method and system for conducting online marketing research in a controlled manner |
| US20130151383A1 (en) * | 2011-12-13 | 2013-06-13 | Opera Solutions, Llc | Recommender engine for collections treatment selection |
| CN106021376A (en) * | 2016-05-11 | 2016-10-12 | 上海点荣金融信息服务有限责任公司 | Method and device for processing user information |
| CN106204063A (en) * | 2016-06-30 | 2016-12-07 | 北京奇艺世纪科技有限公司 | A kind of paying customer's method for digging and device |
| CN106447384A (en) * | 2016-08-31 | 2017-02-22 | 五八同城信息技术有限公司 | Method and apparatus for determining object user |
| CN106897282A (en) * | 2015-12-17 | 2017-06-27 | 阿里巴巴集团控股有限公司 | The sorting technique and equipment of a kind of customer group |
-
2017
- 2017-12-07 CN CN201711282768.2A patent/CN108009926A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040117239A1 (en) * | 2002-12-17 | 2004-06-17 | Mittal Parul A. | Method and system for conducting online marketing research in a controlled manner |
| US20130151383A1 (en) * | 2011-12-13 | 2013-06-13 | Opera Solutions, Llc | Recommender engine for collections treatment selection |
| CN106897282A (en) * | 2015-12-17 | 2017-06-27 | 阿里巴巴集团控股有限公司 | The sorting technique and equipment of a kind of customer group |
| CN106021376A (en) * | 2016-05-11 | 2016-10-12 | 上海点荣金融信息服务有限责任公司 | Method and device for processing user information |
| CN106204063A (en) * | 2016-06-30 | 2016-12-07 | 北京奇艺世纪科技有限公司 | A kind of paying customer's method for digging and device |
| CN106447384A (en) * | 2016-08-31 | 2017-02-22 | 五八同城信息技术有限公司 | Method and apparatus for determining object user |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108764994A (en) * | 2018-05-24 | 2018-11-06 | 深圳前海桔子信息技术有限公司 | A kind of user behavior guidance method, device, server and storage medium |
| CN110555714A (en) * | 2018-06-04 | 2019-12-10 | 百度在线网络技术（北京）有限公司 | method and apparatus for outputting information |
| CN109102176A (en) * | 2018-07-25 | 2018-12-28 | 珠海宏桥高科技有限公司 | A kind of trade finance appraisal procedure and device based on block chain |
| CN109684549A (en) * | 2018-12-24 | 2019-04-26 | 拉扎斯网络科技（上海）有限公司 | Target data prediction technique, device, electronic equipment and computer storage medium |
| CN111680382A (en) * | 2019-02-25 | 2020-09-18 | 北京嘀嘀无限科技发展有限公司 | Grade prediction model training method, grade prediction device and electronic equipment |
| CN109886349A (en) * | 2019-02-28 | 2019-06-14 | 成都新希望金融信息有限公司 | A kind of user classification method based on multi-model fusion |
| CN109886349B (en) * | 2019-02-28 | 2019-11-19 | 成都新希望金融信息有限公司 | A kind of user classification method based on multi-model fusion |
| CN112529236A (en) * | 2019-09-18 | 2021-03-19 | 泰康保险集团股份有限公司 | Target object identification method and device, electronic equipment and storage medium |
| CN110717537A (en) * | 2019-10-08 | 2020-01-21 | 支付宝(杭州)信息技术有限公司 | Method and device for training user classification model and executing user classification prediction |
| CN110717537B (en) * | 2019-10-08 | 2022-04-12 | 支付宝(杭州)信息技术有限公司 | Method and device for training user classification model and executing user classification prediction |
| CN111260416A (en) * | 2020-02-13 | 2020-06-09 | 支付宝(杭州)信息技术有限公司 | Method and device for determining associated user of object |
| CN111311338A (en) * | 2020-03-30 | 2020-06-19 | 网易（杭州）网络有限公司 | User value prediction method and user value prediction model training method |
| CN111737575A (en) * | 2020-06-19 | 2020-10-02 | 北京字节跳动网络技术有限公司 | Content distribution method and device, readable medium and electronic equipment |
| CN111737575B (en) * | 2020-06-19 | 2023-11-14 | 北京字节跳动网络技术有限公司 | Content distribution method, content distribution device, readable medium and electronic equipment |
| CN112070533A (en) * | 2020-08-28 | 2020-12-11 | 上海连尚网络科技有限公司 | Method and equipment for predicting user retention |
| CN112308638A (en) * | 2020-10-29 | 2021-02-02 | 北京中科闻歌科技股份有限公司 | False invoice behavior detection method and device, electronic equipment and storage medium |
| CN113538070A (en) * | 2020-10-30 | 2021-10-22 | 深圳市九九互动科技有限公司 | User life value cycle detection method and device and computer equipment |
| CN113538070B (en) * | 2020-10-30 | 2023-12-26 | 深圳市九九互动科技有限公司 | User life value cycle detection method and device and computer equipment |
| CN113297478A (en) * | 2021-04-25 | 2021-08-24 | 上海淇玥信息技术有限公司 | Information pushing method and device based on user life cycle and electronic equipment |
| CN113297478B (en) * | 2021-04-25 | 2022-06-21 | 上海淇玥信息技术有限公司 | Information pushing method and device based on user life cycle and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108009926A (en) | Method, information processor and readable storage medium storing program for executing for user's classification | |
| Han et al. | Segmentation of telecom customers based on customer value by decision tree model | |
| Reeves et al. | The determinants of tendering periods for PPP procurement in the UK: an empirical analysis | |
| CN108446972A (en) | Bank's Supervision of credit method, apparatus and fund position manage system | |
| Ning et al. | How does the digital economy promote green technology innovation by manufacturing enterprises? Evidence from China | |
| CN108109063A (en) | For the method, apparatus and computer readable storage medium of prediction label predicted value | |
| CN104115178A (en) | Methods and systems for predicting market behavior based on news and sentiment analysis | |
| Wissner | ICT, growth and productivity in the German energy sector–On the way to a smart grid? | |
| Feng et al. | Design and simulation of human resource allocation model based on double-cycle neural network | |
| Liu et al. | A dynamic model for managing cultural tourism | |
| Peng et al. | What determines the subsidy decision bias of local governments? An enterprise heterogeneity perspective | |
| Wang et al. | Industrial policy and the rise of China’s strategic emerging industries | |
| CN113592287A (en) | Risk assessment method and device | |
| CN109741172B (en) | Credit early warning method, device, system and storage medium | |
| Zhai et al. | Big data analysis of accounting forecasting based on machine learning | |
| Nalluri et al. | Modelling the FinTech adoption barriers in the context of emerging economies—An integrated Fuzzy hybrid approach | |
| Xiang et al. | Operational risk allocation in urban rail transit public–private partnership projects | |
| Bian et al. | Construction of Application Model of Accounting Framework Platform for Industry‐Finance Integration Management under the Background of Multimedia Technology | |
| Fernandes et al. | A model for valuing new technologies under a pull incentives environment | |
| CN114926204A (en) | Data processing device and method based on data value | |
| Nan et al. | [Retracted] Corporate Marketing Strategy Analysis with Machine Learning Algorithms | |
| Mani et al. | Diffusion of digital payments in India, 2011–2012 through 2020–2021: role of its sectoral system of innovation | |
| Zhai et al. | Research on Application of Meticulous Nursing Scheduling Management Based on Data‐Driven Intelligent Optimization Technology | |
| CN112116253A (en) | Method, device and system for selecting central mesh point | |
| Dwyer | Future economic development in tourism |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180508 |