CN109242522A - The foundation of target user's identification model, target user's recognition methods and device - Google Patents
The foundation of target user's identification model, target user's recognition methods and device Download PDFInfo
- Publication number
- CN109242522A CN109242522A CN201710561389.0A CN201710561389A CN109242522A CN 109242522 A CN109242522 A CN 109242522A CN 201710561389 A CN201710561389 A CN 201710561389A CN 109242522 A CN109242522 A CN 109242522A
- Authority
- CN
- China
- Prior art keywords
- user
- data information
- log
- activity
- history
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Marketing (AREA)
- Economics (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the invention discloses a kind of foundation of target user's identification model, target user's recognition methods and devices.Target user's identification model method for building up includes: the history drawing New activity information for obtaining at least two users；Extract the log-on data information that New activity information association is drawn with history；Using binary classifier algorithm, according to user type corresponding with history drawing New activity information and log-on data information, setting prediction model is trained, generate target user's identification model, the log-on data information of user to be identified is obtained later, by log-on data information input into target user's identification model trained in advance, judge whether user to be identified is target user, it can effectively solve that the technical issues of ulling up wool user can not be accurately identified in the prior art, make full use of whether true reflection user is the history drawing New activity information for ulling up wool user, optimize and existing ulls up wool user recognition technology, improve the accuracy and timeliness for ulling up wool user identification.
Description
Technical field
The present embodiments relate to the information processing technologies more particularly to a kind of target user's identification model to establish, target is used
Family recognition methods and device.
Background technique
With the fast development of internet industry, operator is in order to promote oneself website, it will usually which tissue is a large amount of to be drawn
New activity customizes preferential activity for new user, so that new user is become the user of website by register account number, allow a user to
Subsequent consumption is enough carried out, portfolio is expanded.
But currently, after thering is certain customers' register account number to obtain a large amount of activity reward by network platform activity, again not
The network platform is accessed or logs in, this certain customers are referred to as to ull up wool user.Ull up wool user is which network found
Platform has activity, just goes register account number, or even investment, to obtain the money of reward or the user of present.These ull up wool user's
In the presence of seriously destroying the network platform does movable purpose itself, a large amount of active resource is occupied, the network platform is unfavorable for
It develops in a healthy way.Accordingly, it is desirable to provide a kind of identify the method for ulling up wool user.
In the prior art, the method that wool user is ulled up in identification mainly includes ulling up wool subscriber blacklist mechanism and to ulling up wool
The simple restriction rule of user.Wherein, ulling up wool subscriber blacklist mechanism is ulled up what is determined according to historical user's log-on data
Wool user is added and ulls up wool subscriber blacklist.The defect of this scheme is: user can re-register a new account, pendulum
The de- limitation for ulling up wool subscriber blacklist；And wool subscriber blacklist mechanism is ulled up with hysteresis quality, it is easy to be ulled up wool user
Waste a large amount of resource.And it is mainly to the simple restrictive rule for ulling up wool user: is drawn in New activity in a large amount of network, to note
Volume account is defined, and such as same register account number, the same cell-phone number or the same shipping address can only be registered once.The party
The major defect of case are as follows: the subjective factor that restrictive condition is related to is relatively more, and restriction rule is simpler, can not limit a large amount of
Ull up wool user community.
Summary of the invention
The embodiment of the present invention provide a kind of target user's identification model establish, target user's recognition methods and device, with excellent
Change it is existing ull up wool user recognition technology, improve the accuracy and timeliness for ulling up wool user identification.
In a first aspect, the embodiment of the invention provides a kind of target user's identification model method for building up, this method comprises:
The history for obtaining at least two users draws New activity information；
Extract the log-on data information that New activity information association is drawn with the history；
Using binary classifier algorithm, the corresponding user type of New activity information and the registration are drawn according to the history
Data information is trained setting prediction model, generates target user's identification model, wherein the user type includes mesh
Mark user and non-targeted user.
Second aspect, the embodiment of the invention also provides a kind of target user's recognition methods, this method comprises:
Obtain the log-on data information of user to be identified；
By the log-on data information input into target user's identification model trained in advance, the use to be identified is obtained
The recognition result at family, wherein target user's identification model is by drawing New activity information association with the history of at least two users
Log-on data information and draw the corresponding user type training of New activity information to generate with the history；
Judge whether the user to be identified is target user according to the recognition result.
The third aspect, the embodiment of the invention also provides a kind of target user's identification models to establish device, which includes:
Action message obtains module, and the history for obtaining at least two users draws New activity information；
Log-on data information extraction modules draw the log-on data of New activity information association to believe for extracting with the history
Breath；
Target user's identification model generation module, it is new living according to being drawn with the history for using binary classifier algorithm
The dynamic corresponding user type of information and the log-on data information are trained setting prediction model, generate target user and know
Other model, wherein the user type includes target user and non-targeted user.
Fourth aspect, the embodiment of the invention also provides a kind of target user's identification device, which includes:
Log-on data data obtaining module, for obtaining the log-on data information of user to be identified；
Recognition result obtains module, for the log-on data information input to target user trained in advance to be identified mould
In type, the recognition result of the user to be identified is obtained, wherein target user's identification model is by at least two users'
History draws the log-on data information of New activity information association and user type corresponding with history drawing New activity information to instruct
Practice and generates；
Target user's judgment module, for judging whether the user to be identified is that target is used according to the recognition result
Family.
The embodiment of the present invention draws New activity information by obtaining the history of at least two users；It extracts and draws New activity with history
The log-on data information of information association；Using binary classifier algorithm, according to user class corresponding with history drawing New activity information
Type and log-on data information are trained setting prediction model, generate target user's identification model, wherein user type packet
Target user and non-targeted user are included, later when whether judge user is target user, obtains the registration number of user to be identified
It is believed that breath, by log-on data information input into target user's identification model trained in advance, judge user to be identified whether be
The technological means of target user can be solved effectively that the technical issues of ulling up wool user can not be accurately identified in the prior art, be filled
Divide using can really reflect whether user is the history drawing New activity information for ulling up wool user, optimization is existing to ull up wool user
Identification technology improves the accuracy and timeliness for ulling up wool user identification.
Detailed description of the invention
Fig. 1 is a kind of flow chart for target user's identification model method for building up that the embodiment of the present invention one provides；
Fig. 2 is a kind of flow chart of target user's identification model method for building up provided by Embodiment 2 of the present invention；
Fig. 3 is a kind of flow chart for target user's recognition methods that the embodiment of the present invention three provides；
Fig. 4 is the structural schematic diagram that a kind of target user's identification model that the embodiment of the present invention four provides establishes device；
Fig. 5 is a kind of structural schematic diagram for target user's identification device that the embodiment of the present invention five provides.
Specific embodiment
The present invention is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining the present invention rather than limiting the invention.It also should be noted that in order to just
Only the parts related to the present invention are shown in description, attached drawing rather than entire infrastructure.
Embodiment one
Fig. 1 is a kind of flow chart for target user's identification model method for building up that the embodiment of the present invention one provides, this implementation
The method of example can be established device by target user's identification model to execute, which can be by way of hardware and/or software
It realizes, and can generally be integrated in completion target user's identification model and establish in the Modeling Server of function, participated in storage user
The data server of New activity data is drawn to be used cooperatively, wherein Modeling Server and data server can be same server
Or belong to same server cluster, or different servers, the present embodiment is to this and is not limited.The present embodiment
Method specifically include:
S101, the history for obtaining at least two users draw New activity information.
In the present embodiment, the history of user draws New activity information to refer to that user participates in the network platform and draws in New activity, uses
Family register account number or user's logging in network platform browsing or the historical data information of purchase.Wherein, history draws New activity can be
Each network platform formulates various preferential activities for new user to promote the website of oneself, be such as hungry, outside Meituan
The preferential Securities that makes a reservation for feeding back to new user is sold, after for another example new user downloads and registers certain software APP, number that APP software is given
Not equal cash red packet.User is during participating in various drawing New activities, in register account number, can bind various personal letters
Breath, such as user account title, user account password, user contact details, user's shipping address, these information are stored in net
In network Platform Server, the history for constituting user draws New activity information.
S102, extraction and history draw the log-on data information of New activity information association.
In the present embodiment, it is drawn in New activity information from the history of user and extracts the log-on data information being associated.Its
In, log-on data information may include user account similarity, user password number of repetition, user contact details number of repetition,
At least one of consignee's number of repetition, user's shipping address number of repetition, user's registration time and activity on-line time difference.
Wherein, the reflection of user account similarity is user during participating in drawing New activity, when carrying out account registration,
The account title of user and the similitude of other users account title.Wherein it is possible to calculate current use by cosine similarity
The account title at family and a large amount of history draw the similitude of other each user account titles in New activity, and by the account of active user
Number title and history draw account of the maximum value of the cosine similarity in New activity in other each account titles as active user
Number similarity.User password number of repetition refers to user when participating in drawing the registration of New activity account, and the account for account setting is close
Code draws the number of repetition of other users account number cipher in New activity with a large amount of history.User contact details number of repetition refers to user
When participating in drawing the registration of New activity account, reserved contact method is contacted with what other users in a large amount of history drawing New activity left
The number of repetition of mode.Consignee's number of repetition refers to that consignee's name that user reserves and a large amount of history draw other in New activity
The duplicate number of consignee's name that user leaves.User's shipping address number of repetition refers to shipping address that user reserves and big
Amount history draws the number of repetition for the shipping address that other users leave in New activity.User's registration time and activity on-line time are poor
It the time difference for referring to the time of user's registration account and drawing New activity online, can be embodied in the form of timestamp.
S103, using binary classifier algorithm, draw the corresponding user type of New activity information and registration number according to history
It is believed that breath, is trained setting prediction model, generates target user's identification model.
Wherein, the user type includes target user and non-targeted user.
In the present embodiment, user type includes target user and two kinds of non-targeted user, can only for a certain user
It selects one and belongs to target user or non-targeted user.Firstly, according to user's history draw New activity information to corresponding user type into
Row definition determines that history draws the corresponding user type of New activity information.Illustratively, user is after drawing New activity register account number
In preset time, when not browsing drawing New activity web page platform behavior, defining the user is target user, otherwise defines the user
For non-targeted user.For example, if user A is not browsed within the 28 day time for having registered by stages happy drawing New activity account
Happy webpage by stages when being more not logged on buying behavior, judges that user A for target user, otherwise, judges user A for non-targeted use
Family.
The identification of target user belongs to binary classification forecasting problem, therefore, using binary classifier algorithm, according to history
The corresponding user type of New activity information and log-on data information are drawn, setting prediction model is trained, to generate target use
Family identification model.Wherein, setting prediction model belong to binary classification model, can for decision tree, neural network, logistic regression,
Any one in the binary classifiers such as discriminatory analysis.Target user's identification model is for judging whether user to be identified is that target is used
Family.
Technical solution provided in an embodiment of the present invention, the history by obtaining at least two users draw New activity information；It mentions
Take the log-on data information that New activity information association is drawn with history；Using binary classifier algorithm, New activity is drawn according to history
The corresponding user type of information and log-on data information are trained setting prediction model, generate target user's identification model
Technological means, can effectively solve that the technical issues of ulling up wool user can not be accurately identified in the prior art, make full use of energy
Whether enough true reflection user is the history drawing New activity information for ulling up wool user, and optimization is existing to ull up wool user identification skill
Art improves the accuracy and timeliness for ulling up wool user identification.
Embodiment two
Fig. 2 is a kind of flow chart of target user's identification model method for building up provided by Embodiment 2 of the present invention.This implementation
Example is optimized based on above-described embodiment, and in the present embodiment, history draws New activity information further include: user's participation activity
Data information；And binary classifier algorithm is used by described, according to user type corresponding with history drawing New activity information
And the log-on data information, setting prediction model is trained, generates target user's identification model, optimization are as follows: use two
Meta classifier algorithm draws the corresponding user type of New activity information, the log-on data information and described according to the history
User participates in activity data information, is trained to setting prediction model, generates target user's identification model.Correspondingly, this reality
The method for applying example specifically includes:
S201, the history for obtaining at least two users draw New activity information.
S202, extraction and history draw the log-on data information of New activity information association and user to participate in activity data information.
In the present embodiment, it can also include that user participates in activity data information, institute that the history of user, which draws New activity information,
Stating user and participating in activity data information includes: that Webpage total amount, user are browsed before user's participation activity total duration, user place an order
At least one of total duration is stopped in loose-leaf.Wherein, user's participation activity total duration refers to that user completes account registration
Afterwards, it places an order from login account arrival loose-leaf to the loose-leaf in the network platform and buys total duration used.Preferably, also
Other users participation activity total duration in New activity can be drawn to carry out greatly active user's participation activity total duration and a large amount of history
Small comparison, and being ranked up by sequence from small to large, using the locating sorting position of active user's participation activity total duration as
The parametric variable when training of target user's identification model.Preferably, it by active user's participation activity total duration and can also preset
Duration threshold value is compared, wherein preset duration threshold value is to be drawn other users participation activity in New activity total according to a large amount of history
Duration is for statistical analysis to be obtained.Using the comparison result of active user's participation activity total duration and preset duration threshold value as mesh
Mark the parametric variable when training of user's identification model.User place an order before browsing Webpage total amount refer to user from login account to
Place an order up to loose-leaf to the loose-leaf in the network platform Webpage total amount when buying, browsed in total.User is in activity
The page stop total duration refer to user from login account reach loose-leaf after, the network platform loose-leaf stop it is total when
It is long.It extracts and draws the log-on data information of New activity information association and user to participate in activity data information with history.
S203, using binary classifier algorithm, draw the corresponding user type of New activity information, log-on data according to history
Information and user participate in activity data information, are trained to setting prediction model, generate target user's identification model.
In the present embodiment, according to user type corresponding with history drawing New activity information, log-on data information and user
Activity data information is participated in, setting prediction model is trained, target user's identification model is generated.Preferably, log-on data
Information include user account similarity, user password number of repetition, user contact details number of repetition, consignee's number of repetition,
All the information in user's shipping address number of repetition, user's registration time and activity on-line time difference, activity data information
Webpage total amount is browsed before placing an order including user's participation activity total duration, user, user stops in total duration in loose-leaf
All the information.Using log-on data information and activity data information as the independent variable of setting prediction model, user type is made
Model training is carried out to set the dependent variable of prediction model, to generate target user's identification model.Preferably, using logistic regression
Model is trained, wherein logistic regression is a kind of probabilistic classification models, be using the probability of target user as dependent variable,
The recurrence mould established using user's characteristic information (log-on data information and/or user participate in activity data information) as independent variable
Type.
Technical solution provided in an embodiment of the present invention, the history by obtaining at least two users draw New activity information；It mentions
It takes and draws the log-on data information of New activity information association and user to participate in activity data information with the history；Using binary classification
Device algorithm is joined according to user type corresponding with history drawing New activity information, the log-on data information and the user
With activity data information, setting prediction model is trained, target user's identification model is generated, optimizes target user's identification
Model further improves the accuracy and timeliness for ulling up wool user identification.
Embodiment three
Fig. 3 is a kind of flow chart for target user's recognition methods that the embodiment of the present invention three provides, the method for the present embodiment
It can be executed by target user's identification device, which can be realized by way of hardware and/or software, and can generally be integrated
It is identified in server in the target user for completing target user's identification function, establishes function with for completing target user's identification model
The Modeling Server of energy is used cooperatively, wherein target user identifies server device and Modeling Server can be same server
Or belong to same server cluster, or different servers, the present embodiment is to this and is not limited.The present embodiment
Method specifically include:
S301, the log-on data information for obtaining user to be identified.
In the present embodiment, the log-on data information of user to be identified is obtained, wherein the log-on data information includes using
Family account similarity, user password number of repetition, user contact details number of repetition, consignee's number of repetition, user's place of acceptance
At least one of location number of repetition, user's registration time and activity on-line time difference.The log-on data information conduct that will acquire
The characteristic information of user to be identified.Preferably, template is extracted to the log-on data of user to be identified according to preset user characteristics
Information carries out feature extraction, the characteristic information as user to be identified.Illustratively, with { user account similarity, user password
Number of repetition, user contact details number of repetition, consignee's number of repetition, user's shipping address number of repetition, when user's registration
Between it is poor with activity on-line time as preset user characteristics extract template, extract the characteristic information of user to be identified.
S302, by log-on data information input into target user's identification model trained in advance, obtain user to be identified
Recognition result.
Wherein, target user's identification model is by the registration with the history of at least two users drawing New activity information association
Data information and user type training corresponding with history drawing New activity information generate.
In the present embodiment, log-on data information target user trained in advance is input to as user's characteristic information to know
In other model, the recognition result of user to be identified is obtained.Preferably, by log-on data information according to data vector { user account
Similarity, user password number of repetition, user contact details number of repetition, consignee's number of repetition, user's shipping address repeat
Number, user's registration time and activity on-line time are poor } form, be input in advance trained target user's identification model,
Obtain the recognition result of user to be identified.
Preferably, the input parameter of target user's identification model further include: user participates in activity data information；Its
In, the user participate in activity data information include user's participation activity total duration, user browsed before placing an order Webpage total amount,
User stops at least one of total duration in loose-leaf.It illustratively, will { user account similarity, user password repetition
Number, user contact details number of repetition, consignee's number of repetition, user's shipping address number of repetition, the user's registration time with
Activity on-line time is poor, user's participation activity total duration, and user browses Webpage total amount before placing an order, and user stops in loose-leaf
Stay total duration } characteristic information as user to be identified, it is input in target user's identification model trained in advance, and will output
As a result the recognition result as user to be identified.Wherein, the recognition result of user to be identified can be a probability numbers, may be used also
To be the Boolean variable of one 0 or 1, the present embodiment is not construed as limiting the form of expression of the recognition result of user to be identified.
S303, judge whether user to be identified is target user according to recognition result.
Judge identify whether user is target user according to recognition result.Illustratively, as target user trained in advance
Identification model is when being obtained by Logic Regression Models training, and the recognition result of user to be identified is a probability numbers.Judgement should
Whether probability value is greater than predetermined probabilities value, when the probability value of recognition result is greater than predetermined probabilities value, determines that user to be identified is
Target user, otherwise, user to be identified are non-targeted user.The output result of definition target user's identification model trained in advance
For 0 or 1 Boolean variable, when exporting result is 1, definition user is target user；When output result is 0, defining user is
Non-targeted user.So judging that user to be identified for target user, otherwise, defines when the recognition result of user to be identified is 1
User to be identified is non-targeted user.
Technical solution provided in an embodiment of the present invention, by the log-on data information for obtaining user to be identified；Number will be registered
It is believed that breath is input in target user's identification model trained in advance, the recognition result of user to be identified is obtained, wherein target is used
Family identification model is by drawing the log-on data information of New activity information association with the history of at least two users and drawing with history new
The corresponding user type training of action message generates；Judge whether user to be identified is target user according to recognition result, it can be with
It effectively solves that the technical issues of ulling up wool user can not be accurately identified in the prior art, making full use of true reflection user is
No is the history drawing New activity information for ulling up wool user, and optimization is existing to ull up wool user recognition technology, and wool user is ulled up in raising
The accuracy and timeliness of identification.
Example IV
Fig. 4 is the structure chart that a kind of target user's identification model that the embodiment of the present invention four provides establishes device.Such as Fig. 4 institute
Show, described device includes: that action message obtains module 401, log-on data information extraction modules 402 and target user's identification
Model generation module 403, in which:
Action message obtains module 401, and the history for obtaining at least two users draws New activity information；Log-on data letter
Extraction module 402 is ceased, for extracting the log-on data information for drawing New activity information association with the history；Target user identifies mould
Type generation module 403, for using binary classifier algorithm, according to user type corresponding with history drawing New activity information
And the log-on data information, setting prediction model is trained, generates target user's identification model, wherein the user
Type includes target user and non-targeted user.
Technical solution provided in an embodiment of the present invention, the history by obtaining at least two users draw New activity information；It mentions
Take the log-on data information that New activity information association is drawn with history；Using binary classifier algorithm, New activity is drawn according to history
The corresponding user type of information and log-on data information are trained setting prediction model, generate target user's identification model
Technological means, can effectively solve that the technical issues of ulling up wool user can not be accurately identified in the prior art, make full use of energy
Whether enough true reflection user is the history drawing New activity information for ulling up wool user, and optimization is existing to ull up wool user identification skill
Art improves the accuracy and timeliness for ulling up wool user identification.
On the basis of the various embodiments described above, the log-on data information includes user account similarity, user password weight
Again number, user contact details number of repetition, consignee's number of repetition, user's shipping address number of repetition, user's registration time
At least one of with activity on-line time difference.
On the basis of the various embodiments described above, the history draws New activity information further include: user participates in activity data letter
Breath；
Target user's identification model generation module, is used for:
Using binary classifier algorithm, according to user type corresponding with history drawing New activity information, the registration
Data information and the user participate in activity data information, are trained to setting prediction model, generate target user and identify mould
Type.
On the basis of the various embodiments described above, when user's participation activity data information includes: that user's participation activity is total
Long, user browses Webpage total amount, user and stops at least one of total duration in loose-leaf before placing an order.
On the basis of the various embodiments described above, the prediction model that sets is Logic Regression Models.
Target user's identification model provided by the embodiment of the present invention, which establishes device, can be used for executing any implementation of the invention
Target user's identification model method for building up that example provides, has corresponding functional module, realizes identical beneficial effect.
Embodiment five
Fig. 5 is a kind of structural schematic diagram for target user's identification device that the embodiment of the present invention five provides.As shown in figure 5,
Described device includes: log-on data data obtaining module 501, recognition result acquisition module 502 and target user's judgment module
503, in which:
Log-on data data obtaining module 501, for obtaining the log-on data information of user to be identified；Recognition result obtains
Module 502, it is described wait know for into target user's identification model trained in advance, obtaining the log-on data information input
The recognition result of other user, wherein target user's identification model is by drawing New activity information with the history of at least two users
Associated log-on data information and user type training corresponding with history drawing New activity information generate；Target user sentences
Disconnected module 503, for judging whether the user to be identified is target user according to the recognition result.
Technical solution provided in an embodiment of the present invention, by the log-on data information for obtaining user to be identified；Number will be registered
It is believed that breath is input in target user's identification model trained in advance, the recognition result of user to be identified is obtained, wherein target is used
Family identification model is by drawing the log-on data information of New activity information association with the history of at least two users and drawing with history new
The corresponding user type training of action message generates；Judge whether user to be identified is target user according to recognition result, it can be with
It effectively solves that the technical issues of ulling up wool user can not be accurately identified in the prior art, making full use of true reflection user is
No is the history drawing New activity information for ulling up wool user, and optimization is existing to ull up wool user recognition technology, and wool user is ulled up in raising
The accuracy and timeliness of identification.
On the basis of the various embodiments described above, the log-on data information includes user account similarity, user password weight
Again number, user contact details number of repetition, consignee's number of repetition, user's shipping address number of repetition, user's registration time
At least one of with activity on-line time difference.
On the basis of the various embodiments described above, the input parameter of target user's identification model further include: user participates in
Activity data information；
Wherein, it includes that user's participation activity total duration, user browse net before placing an order that the user, which participates in activity data information,
Page page total amount, user stop at least one of total duration in loose-leaf.
Target user's identification device provided by the embodiment of the present invention can be used for executing any embodiment of that present invention offer
Target user's recognition methods has corresponding functional module, realizes identical beneficial effect.
Note that the above is only a better embodiment of the present invention and the applied technical principle.It will be appreciated by those skilled in the art that
The invention is not limited to the specific embodiments described herein, be able to carry out for a person skilled in the art it is various it is apparent variation,
It readjusts and substitutes without departing from protection scope of the present invention.Therefore, although being carried out by above embodiments to the present invention
It is described in further detail, but the present invention is not limited to the above embodiments only, without departing from the inventive concept, also
It may include more other equivalent embodiments, and the scope of the invention is determined by the scope of the appended claims.
Claims (16)
1. a kind of target user's identification model method for building up characterized by comprising
The history for obtaining at least two users draws New activity information；
Extract the log-on data information that New activity information association is drawn with the history；
Using binary classifier algorithm, the corresponding user type of New activity information and the log-on data are drawn according to the history
Information is trained setting prediction model, generates target user's identification model, wherein the user type includes that target is used
Family and non-targeted user.
2. the method according to claim 1, wherein the log-on data information include user account similarity,
User password number of repetition, consignee's number of repetition, user's shipping address number of repetition, is used user contact details number of repetition
At least one of family registion time and activity on-line time difference.
3. the method according to claim 1, wherein the history draws New activity information further include: user participates in
Activity data information；
It is described to use binary classifier algorithm, the corresponding user type of New activity information and the registration are drawn according to the history
Data information is trained setting prediction model, generates target user's identification model, comprising:
Using binary classifier algorithm, according to user type corresponding with history drawing New activity information, the log-on data
Information and the user participate in activity data information, are trained to setting prediction model, generate target user's identification model.
4. according to the method described in claim 3, it is characterized in that, it includes: user's ginseng that the user, which participates in activity data information,
Webpage total amount, user are browsed before placing an order with activity total duration, user stops at least one of total duration in loose-leaf.
5. method according to claim 1 to 4, which is characterized in that the prediction model that sets is logistic regression mould
Type.
6. a kind of target user's recognition methods characterized by comprising
Obtain the log-on data information of user to be identified；
By the log-on data information input into target user's identification model trained in advance, obtain the user's to be identified
Recognition result, wherein target user's identification model is by the note with the history of at least two users drawing New activity information association
Volume data information and user type training corresponding with history drawing New activity information generate；
Judge whether the user to be identified is target user according to the recognition result.
7. according to the method described in claim 6, it is characterized in that, the log-on data information include user account similarity,
User password number of repetition, consignee's number of repetition, user's shipping address number of repetition, is used user contact details number of repetition
At least one of family registion time and activity on-line time difference.
8. method according to claim 6 or 7, which is characterized in that the input parameter of target user's identification model is also
It include: that user participates in activity data information；
Wherein, it includes that user's participation activity total duration, user browse webpage page before placing an order that the user, which participates in activity data information,
Face total amount, user stop at least one of total duration in loose-leaf.
9. a kind of target user's identification model establishes device characterized by comprising
Action message obtains module, and the history for obtaining at least two users draws New activity information；
Log-on data information extraction modules, for extracting the log-on data information for drawing New activity information association with the history；
Target user's identification model generation module draws New activity letter according to the history for using binary classifier algorithm
Corresponding user type and the log-on data information are ceased, setting prediction model is trained, target user is generated and identifies mould
Type, wherein the user type includes target user and non-targeted user.
10. device according to claim 9, which is characterized in that the log-on data information include user account similarity,
User password number of repetition, consignee's number of repetition, user's shipping address number of repetition, is used user contact details number of repetition
At least one of family registion time and activity on-line time difference.
11. device according to claim 9, which is characterized in that the history draws New activity information further include: user participates in
Activity data information；
Target user's identification model generation module, is used for:
Using binary classifier algorithm, according to user type corresponding with history drawing New activity information, the log-on data
Information and the user participate in activity data information, are trained to setting prediction model, generate target user's identification model.
12. device according to claim 11, which is characterized in that it includes: user that the user, which participates in activity data information,
Participation activity total duration, user browse Webpage total amount, user and stop at least one in total duration in loose-leaf before placing an order
It is a.
13. device according to claim 1 to 4, which is characterized in that the prediction model that sets is logistic regression mould
Type.
14. a kind of target user's identification device characterized by comprising
Log-on data data obtaining module, for obtaining the log-on data information of user to be identified；
Recognition result obtains module, for by the log-on data information input to target user's identification model trained in advance
In, obtain the recognition result of the user to be identified, wherein target user's identification model at least two users by going through
History draws the log-on data information and user type training corresponding with history drawing New activity information of New activity information association
It generates；
Target user's judgment module, for judging whether the user to be identified is target user according to the recognition result.
15. device according to claim 14, which is characterized in that the log-on data information includes that user account is similar
Degree, user password number of repetition, user contact details number of repetition, consignee's number of repetition, user's shipping address number of repetition,
At least one of user's registration time and activity on-line time difference.
16. device according to claim 14 or 15, which is characterized in that the input parameter of target user's identification model
Further include: user participates in activity data information；
Wherein, it includes that user's participation activity total duration, user browse webpage page before placing an order that the user, which participates in activity data information,
Face total amount, user stop at least one of total duration in loose-leaf.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710561389.0A CN109242522A (en) | 2017-07-11 | 2017-07-11 | The foundation of target user's identification model, target user's recognition methods and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710561389.0A CN109242522A (en) | 2017-07-11 | 2017-07-11 | The foundation of target user's identification model, target user's recognition methods and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109242522A true CN109242522A (en) | 2019-01-18 |
Family
ID=65083941
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710561389.0A Pending CN109242522A (en) | 2017-07-11 | 2017-07-11 | The foundation of target user's identification model, target user's recognition methods and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109242522A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110838024A (en) * | 2019-10-16 | 2020-02-25 | 支付宝(杭州)信息技术有限公司 | Information pushing method, device and equipment based on deep reinforcement learning |
| CN111080305A (en) * | 2019-12-16 | 2020-04-28 | 中国建设银行股份有限公司 | Risk identification method and device and electronic equipment |
| CN111091408A (en) * | 2019-10-30 | 2020-05-01 | 北京天元创新科技有限公司 | User identification model creating method and device and identification method and device |
| CN111310863A (en) * | 2020-03-27 | 2020-06-19 | 北京奇艺世纪科技有限公司 | User detection method and device and electronic equipment |
| CN111782735A (en) * | 2020-07-01 | 2020-10-16 | 北京深演智能科技股份有限公司 | Wool party flow identification method and device |
| CN114119164A (en) * | 2021-11-30 | 2022-03-01 | 必要鸿源(北京)科技有限公司 | Commodity group purchase method, device, system and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090265639A1 (en) * | 2008-04-17 | 2009-10-22 | Gary Stephen Shuster | Evaluation of remote user attributes in a social networking environment |
| CN103853948A (en) * | 2012-11-28 | 2014-06-11 | 阿里巴巴集团控股有限公司 | User identity recognizing and information filtering and searching method and server |
| CN106022826A (en) * | 2016-05-18 | 2016-10-12 | 武汉斗鱼网络科技有限公司 | Cheating user recognition method and system in webcast platform |
| CN106777024A (en) * | 2016-12-08 | 2017-05-31 | 北京小米移动软件有限公司 | Recognize the method and device of malicious user |
| CN106897919A (en) * | 2017-02-28 | 2017-06-27 | 百度在线网络技术（北京）有限公司 | With the foundation of car type prediction model, information providing method and device |
-
2017
- 2017-07-11 CN CN201710561389.0A patent/CN109242522A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090265639A1 (en) * | 2008-04-17 | 2009-10-22 | Gary Stephen Shuster | Evaluation of remote user attributes in a social networking environment |
| CN103853948A (en) * | 2012-11-28 | 2014-06-11 | 阿里巴巴集团控股有限公司 | User identity recognizing and information filtering and searching method and server |
| CN106022826A (en) * | 2016-05-18 | 2016-10-12 | 武汉斗鱼网络科技有限公司 | Cheating user recognition method and system in webcast platform |
| CN106777024A (en) * | 2016-12-08 | 2017-05-31 | 北京小米移动软件有限公司 | Recognize the method and device of malicious user |
| CN106897919A (en) * | 2017-02-28 | 2017-06-27 | 百度在线网络技术（北京）有限公司 | With the foundation of car type prediction model, information providing method and device |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110838024A (en) * | 2019-10-16 | 2020-02-25 | 支付宝(杭州)信息技术有限公司 | Information pushing method, device and equipment based on deep reinforcement learning |
| CN111091408A (en) * | 2019-10-30 | 2020-05-01 | 北京天元创新科技有限公司 | User identification model creating method and device and identification method and device |
| CN111080305A (en) * | 2019-12-16 | 2020-04-28 | 中国建设银行股份有限公司 | Risk identification method and device and electronic equipment |
| CN111310863A (en) * | 2020-03-27 | 2020-06-19 | 北京奇艺世纪科技有限公司 | User detection method and device and electronic equipment |
| CN111310863B (en) * | 2020-03-27 | 2023-09-08 | 北京奇艺世纪科技有限公司 | User detection method and device and electronic equipment |
| CN111782735A (en) * | 2020-07-01 | 2020-10-16 | 北京深演智能科技股份有限公司 | Wool party flow identification method and device |
| CN114119164A (en) * | 2021-11-30 | 2022-03-01 | 必要鸿源(北京)科技有限公司 | Commodity group purchase method, device, system and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109242522A (en) | The foundation of target user's identification model, target user's recognition methods and device | |
| CN110569377B (en) | Media file processing method and device | |
| US20190104197A1 (en) | Discovering signature of electronic social networks | |
| CN104615608B (en) | A kind of data mining processing system and method | |
| CN107958317A (en) | A kind of method and apparatus that crowdsourcing participant is chosen in crowdsourcing project | |
| CN110298547A (en) | Methods of risk assessment, device, computer installation and storage medium | |
| CN102708153B (en) | Self-adaption finding and predicting method and system for hot topics of online social network | |
| CN106940732A (en) | A kind of doubtful waterborne troops towards microblogging finds method | |
| CN108764994A (en) | A kind of user behavior guidance method, device, server and storage medium | |
| CN110598070B (en) | Application type identification method and device, server and storage medium | |
| CN108280104A (en) | The characteristics information extraction method and device of target object | |
| US20210256423A1 (en) | Methods, apparatuses, and computing devices for trainings of learning models | |
| CN107071193A (en) | The method and apparatus of interactive answering system accessing user | |
| US11423307B2 (en) | Taxonomy construction via graph-based cross-domain knowledge transfer | |
| CN108960574A (en) | Quality determination method, device, server and the storage medium of question and answer | |
| CN104750674A (en) | Man-machine conversation satisfaction degree prediction method and system | |
| CN108564393A (en) | Potential customers' methods of marking, device and system | |
| US20200401663A1 (en) | Techniques for Generating a Topic Model | |
| Zhang et al. | Advertisement click-through rate prediction based on the weighted-ELM and adaboost algorithm | |
| US11120115B2 (en) | Identification method and apparatus | |
| CN110069686A (en) | User behavior analysis method, apparatus, computer installation and storage medium | |
| Zygmunt | Role identification of social networkers | |
| CN116522131A (en) | Object representation method, device, electronic equipment and computer readable storage medium | |
| CN113763928A (en) | Audio category prediction method and device, storage medium and electronic equipment | |
| CN109547410A (en) | Request recognition methods, device, server and storage medium based on GCN |
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: 20190118 |