CN110162692A - User tag determines method, apparatus, computer equipment and storage medium - Google Patents
User tag determines method, apparatus, computer equipment and storage medium Download PDFInfo
- Publication number
- CN110162692A CN110162692A CN201811503599.5A CN201811503599A CN110162692A CN 110162692 A CN110162692 A CN 110162692A CN 201811503599 A CN201811503599 A CN 201811503599A CN 110162692 A CN110162692 A CN 110162692A
- Authority
- CN
- China
- Prior art keywords
- user
- value
- label
- target labels
- active
- 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.)
- Granted
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention relates to a kind of user tags to determine method, apparatus, computer equipment and storage medium, which comprises obtains user's set, user's set includes the first child user set for having determined that label and the second child user set for not determining label;The corresponding initial labels value of each user is determined according to the corresponding label of user each in user set;According in user set between user the social degree of association and characteristic similarity the target labels point of reference between user is calculated;For the active user in the second child user set, according to the corresponding initial labels value of user in the target labels point of reference and user set between the user in the active user and user set, the corresponding target labels value of the active user is obtained;The corresponding target labels of the active user are determined according to the target labels value.The accuracy of label obtained by the above method is high and saves Internet resources.
Description
Technical field
The present invention relates to field of information processing, determine method, apparatus, computer equipment more particularly to user tag and deposit
Storage media.
Background technique
With the development of internet, the user on internet platform is more and more, in many cases it needs to be determined that user
Label, with according to user tag management and maintenance user.Currently, usually relying on user certainly in the label for determining user
Feel is filled in, but the data that user fills in internet are usually and imperfect, is needed to rely on and is manually labeled, and people is being carried out
When work marks, server sends the relevant information of user to client, determines user's by the relevant information manually according to user
Label is returned to server by client after label, and the relevant information that server retransmits next bit user is marked to client
Label judgement, however, artificial judgment label is influenced by people's subjective state, obtained label accuracy is lower, and terminal
It is repeatedly interacted with server needs, wastes Internet resources.
Summary of the invention
Based on this, it is necessary to it is lower for label accuracy obtained above, and also terminal and server need progress more
Secondary interaction, provides a kind of user tag and determines method, apparatus, computer equipment and storage medium the problem of wasting Internet resources.
A kind of user tag determines method, which comprises obtains user's set, user's set includes having determined that
First child user set of label and the second child user set for not determining label;According to each user in user set
Corresponding label determines the corresponding initial labels value of each user;According to the social pass in user set between user
The target labels point of reference between user is calculated in connection degree and characteristic similarity;For in the second child user set
Active user, according to the target labels point of reference and the use between the user in the active user and user set
The corresponding initial labels value of user in the set of family, obtains the corresponding target labels value of the active user;According to the target
Label value determines the corresponding target labels of the active user.
A kind of user tag determining device, described device includes: that user gathers acquisition module, for obtaining user's set,
User's set includes the first child user set for having determined that label and the second child user set for not determining label;Initially
Label value determining module, for determining that each user is corresponding according to the corresponding label of user each in user set
Initial labels value;Target labels point of reference computing module, for according to the social degree of association in user set between user
And the target labels point of reference between user is calculated in characteristic similarity;Target labels value computing module, for for institute
The active user in the second child user set is stated, according to the target between the user in the active user and user set
The corresponding initial labels value of user in Tag reference degree and user set, obtains the corresponding target of the active user
Label value;Target labels determining module, for determining the corresponding target labels of the active user according to the target labels value.
In one embodiment, the target labels point of reference includes first object Tag reference degree and the second target mark
Point of reference is signed, the target labels point of reference computing module is used for: according to the social association in user set between user
The first object Tag reference degree between user is calculated in degree, according to the characteristic similarity in user set between user
The second target labels point of reference between user is calculated;The target labels value computing module includes: the first influence label
It is worth computing unit, for for the active user in the second child user set, according to the active user and the user
Between user in set corresponding first object Tag reference degree and the user set in the corresponding initial mark of user
Label value, obtaining the active user corresponding first influences label value;Second influences label value computing unit, for for described
Active user in second child user set, according to corresponding the between the user in the active user and user set
The corresponding initial labels value of user in two target labels point of reference and user set, it is corresponding to obtain the active user
Second influence label value;Target labels value computing unit, for influencing label value according to the active user corresponding first
And second influence label value determine the corresponding target labels value of the active user.
In one embodiment, the second influence label value computing unit includes: characteristic pattern building subelement, and being used for will
As node, side right weight of the second target labels point of reference as side constructs each user in user's set
To characteristic pattern;Node obtains subelement, for obtaining the corresponding present node of active user described in the characteristic pattern and institute
State the adjacent node of present node;Subelement is updated, for using the corresponding initial labels value of each user as the spy
The current label value for levying corresponding node in figure, according to the current label value of adjacent node described in the characteristic pattern and described
The side right of present node and the adjacent node updates the current label value of the present node again;Subelement is returned to, for returning
Go back to the current label value of the adjacent node according to the characteristic pattern and the side of the present node and the adjacent node
Weight updates the step of current label value of the present node, until meeting the condition of convergence, it is corresponding to obtain the active user
Second influence label value.
In one embodiment, the target labels point of reference computing module includes: that object reference user gathers acquisition list
Member, for obtaining active user from the second child user set, according to the social incidence relation of the active user from institute
It states screening in user's set and obtains the corresponding object reference user set of the active user;Second target labels point of reference calculates
Unit, for the characteristic similarity according to the active user and each object reference user be calculated the active user with
The second target labels point of reference of the object reference user;The characteristic pattern building subelement is used for: the user is gathered
In each user as node, using the corresponding node of the object reference user as the adjacent node of the present node,
Weight of the second target labels point of reference of the active user and the object reference user as side, building obtain feature
Figure.
In one embodiment, the object reference user gathers acquiring unit and is used for: according to the society of the active user
It hands over incidence relation to obtain the active user corresponding first from user set and is directly linked user's set;According to described
The screening from the first direct correlation user's set of the characteristic similarity of active user and the first direct correlation user obtains the
One refers to user;First reference is obtained from user set according to the social incidence relation of the first reference user
User corresponding second is directly linked user's set;The feature phase for being directly linked user with second with reference to user according to described first
Like degree, screening obtains second with reference to user from the second direct correlation user's set;User and institute are referred to by described first
Second is stated with reference to user as the object reference user in object reference user set.
In one embodiment, the second target labels point of reference computing unit is used for: to the active user and respectively
The characteristic similarity of a object reference user counts, and obtains characteristic similarity statistical result;Obtain the current use
The current signature similarity at family and the object reference user is united according to the current signature similarity and the characteristic similarity
The ratio of meter result obtains the second target labels point of reference of the active user Yu the object reference user.
In one embodiment, the first influence label value computing unit is used for: by each user in user set
As node, side right weight of the first object Tag reference degree as side obtains social associated diagram;Obtain the social association
The adjacent node of the corresponding present node of active user described in figure and the present node;According in the social associated diagram
The side right of the label value of the adjacent node and the present node and the adjacent node updates the present node again
Label value;Return the adjacent node according to the social associated diagram label value and the present node with it is described adjacent
The side right of node updates the step of label value of the present node again, until meeting the condition of convergence, obtains the active user
Corresponding first influences label value.
In one embodiment, the target labels value computing unit is used for: it is corresponding to obtain the first influence label value
The first weight and described second influence corresponding second weight of label value;Label value and correspondence are influenced according to described first
The first weight, it is described second influence label value and corresponding second weight be weighted summation, obtain the second user
Corresponding target labels value.
In one embodiment, the initial labels value determining module includes: target labels classification acquiring unit, for obtaining
Take the corresponding target labels classification of each first user;First initial labels vector obtains unit, for according to each institute
It states the corresponding target labels classification of the first user and obtains the corresponding initial labels vector of first user;Second initial labels to
Unit is measured, for obtaining default label vector as the corresponding initial mark of second user in the second child user set
Vector is signed, the value of the default label vector is consistent.
In one embodiment, the first initial labels vector obtains unit and is used for: corresponding for first user
Each target labels classification, obtain corresponding first preset value as corresponding vector value, it is corresponding for first user
Each non-targeted label classification, obtain corresponding second preset value as corresponding vector value;By each target labels
The corresponding vector value of classification and each non-targeted corresponding vector value composition of label classification first user are corresponding
Initial labels vector, first preset value are different from second preset value.
A kind of computer equipment, including memory and processor are stored with computer program, the meter in the memory
When calculation machine program is executed by the processor, so that the processor executes the step of above-mentioned user tag determines method.
A kind of computer readable storage medium is stored with computer program on the computer readable storage medium, described
When computer program is executed by processor, so that the processor executes the step of above-mentioned user tag determines method.
Above-mentioned user tag determines method, apparatus, computer equipment and storage medium, in the label for determining user, root
It according to the user for having determined that the user of label goes prediction not determine label, and is the characteristic similarity between synthetic user and pass
Connection degree is calculated, and without manually marking the label of also available user, therefore, the accuracy of obtained label is high and saves
Save Internet resources.
Detailed description of the invention
Fig. 1 is the applied environment figure that the user tag provided in one embodiment determines method;
Fig. 2 is the flow chart that user tag determines method in one embodiment;
Fig. 3 is that the active user in the second child user set is collected according to active user and user in one embodiment
The corresponding initial labels value of each user, meter in corresponding target labels point of reference and user's set between each user in conjunction
Calculation obtains the flow chart of the corresponding target labels value of active user;
Fig. 4 A is in one embodiment according to corresponding second target mark between active user and the user in user's set
The corresponding initial labels value of user in point of reference and user's set is signed, obtaining active user corresponding second influences label value
Flow chart;
Fig. 4 B is the schematic diagram of characteristic pattern in one embodiment;
Fig. 5 be in one embodiment according to user gather in characteristic similarity between user be calculated between user
The flow chart of second target labels point of reference;
Fig. 6 A is that the active user in the second child user set is collected according to active user and user in one embodiment
Between user in conjunction corresponding first object Tag reference degree and user set in the corresponding initial labels value of user, obtain
The flow chart of label value is influenced to active user corresponding first;
Fig. 6 B is the schematic diagram of social associated diagram in one embodiment;
Fig. 7 A is the flow chart that user tag determines method in one embodiment;
Fig. 7 B is the schematic illustration that user tag determines method in one embodiment;
Fig. 8 is the structural block diagram of user tag determining device in one embodiment;
Fig. 9 is the structural block diagram of user tag determination unit in one embodiment;
Figure 10 is the internal structure block diagram of computer equipment in one embodiment.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
It is appreciated that term " first " used in this application, " second " etc. can be used to describe various elements herein,
But unless stated otherwise, these elements should not be limited by these terms.These terms are only used to by first element and another yuan
Part is distinguished.For example, in the case where not departing from scope of the present application, the first child user collection can be collectively referred to as to the second son and used
Family set, and similarly, the second child user collection can be collectively referred to as the first child user set.
Fig. 1 is the applied environment figure that the user tag that provides determines method in one embodiment, as shown in Figure 1, answering at this
With in environment, including terminal 110 and server 120.It is corresponding each user in user's set has been can store on server 120
Target labels, server 120 can be gathered according to user in each user corresponding target labels user is managed,
Such as corresponding pushed information is sent to the corresponding terminal 110 of user according to the target labels of user, if the corresponding mesh of user
Mark label is to like automobile, then pushes the relevant information of automobile to terminal 110.If the corresponding target labels of user are to like trip
Trip then pushes the relevant information in tourist attractions to terminal 110.The corresponding target labels of each user are servers in user's set
120 user tags provided according to embodiments of the present invention determine what method obtained.
Server 120 can be independent physical server, be also possible to the server set that multiple physical servers are constituted
Group can be to provide the Cloud Server of the basic cloud computing service such as Cloud Server, cloud database, cloud storage and CDN.Terminal 110
It can be smart phone, tablet computer, laptop, desktop computer, intelligent sound box, smartwatch etc., but do not limit to
In this.Terminal 110 and server 120 can pass through bluetooth, USB (Universal Serial Bus, universal serial bus)
Or the communication connections mode such as network is attached, the present invention is herein with no restrictions.
As shown in Fig. 2, the present embodiment is mainly with this in one embodiment it is proposed that a kind of user tag determines method
Method is illustrated applied to server 120 in above-mentioned Fig. 1.It can specifically include following steps:
Step S202, obtain user set, user set include have determined that label the first child user set and not really
Calibrate the second child user set of label.
It specifically, include multiple users in user's set, it is multiple to refer to more than two (including two), the first child user collection
The quantity of user also may include multiple in conjunction and the second child user set, and particular number can according to need setting.It lifts a
Example, for social application wechat, it is assumed that have 800,000,000 users in wechat, then the number of user's set can be 800,000,000.Certainly,
Selected part user user's set can also be formed from this 800,000,000 user, such as the user in the same city is formed into user's collection
It closes.Label can be used for classifying to user for the characteristics of indicating user.The label of user may include it is a variety of, according to
Concrete condition can have different definition.Such as credit evaluation, then the label of user may include that credit is very good, credit
The classifications such as medium and credit is poor.And ad click is estimated, then the label of user may include clicking and not clicking two
Kind classification.Resource recommendation, such as finance product are recommended, then the label of user may include conservative financing user and emit
Dangerous type financing user.Have determined that the user of label refers to that the label of the user is known.Do not determine that the user of label refers to this
The label of user be it is unknown, need to further determine that.The label of each first user in first child user set is really
It is fixed, and the label of each second user in the second child user set be it is undetermined, need according to the first child user set
In the label of the first user be determined.
Step S204, according to user gather in the corresponding label of each user determine the corresponding initial labels of each user
Value.
Specifically, initial labels value is obtained according to the corresponding label of user, for having determined that the first user of label,
Different labels can correspond to different initial labels values, for not determining the second user of label, then available preset
Label value is as initial labels value.
In one embodiment, according to user gather in the corresponding label of each user determine that each user is corresponding initial
Label value includes: to obtain the corresponding target labels classification of each first user in the first child user set;It is used according to each first
The corresponding target labels classification in family obtains the corresponding initial labels vector of the first user;Default label vector is obtained as the second son
The corresponding initial labels vector of second user in user's set, the value for presetting label vector are consistent.
Specifically, it can be the vector value that a kind of label corresponds to predeterminated position in initial labels vector.For having determined that mark
The target user of label, the corresponding initial labels vector of different target label classification are different, in this way can be to the label of user
It distinguishes.And for not determining the user of label, then default label vector is obtained, the value for presetting label vector is consistent, indicates
Also user tag is not distinguished.Default label vector can be the vector that vector value is 0.
In one embodiment, according to the corresponding target labels classification of each first user obtain the first user it is corresponding just
Beginning label vector includes: each target labels classification corresponding for the first user, obtains corresponding first preset value conduct pair
The vector value answered, each non-targeted label classification corresponding for the first user obtain corresponding second preset value as correspondence
Vector value;By the corresponding vector value of each target labels classification and the corresponding vector value composition of each non-targeted label classification
The corresponding initial labels vector of first user.
Specifically, the first preset value and the second preset value can according to need setting, and the first preset value and second are preset
Value is different.In one embodiment, the first preset value can be 1, and the second preset value can be 0.For example, it is assumed that the first child user
Include L the first users in set, include U user in the second child user set, label has m classification, wherein m for greater than
Positive integer equal to 2 can then indicate that the corresponding label value of user, vector have m vector value, a kind of label pair with vector
The vector value of predeterminated position is answered, if the corresponding label of user is f class label, f-th of vector in initial labels vector
Value is the first preset value such as 1, other are the second preset value such as 0.For do not determine label second user, then it is corresponding to
Each vector value of amount can be the second preset value such as 0.For an actual example, it is assumed that then the corresponding label of party A-subscriber is
Third label, then the corresponding vector of party A-subscriber is (0,0,1,0).
In one embodiment, the matrix F of the corresponding initial labels vector composition of each user can be used for matrix (1) table
Show, in matrix (1), a line indicates that an initial labels vector, n are equal to the quantity L's of the first user and quantity U of second user
With the quantity of user in expression user's set.
Step S206, according to user gather in the social degree of association between user and characteristic similarity user is calculated
Between target labels point of reference.
Specifically, Tag reference degree indicates influence degree of the label to another user tag of a user, label ginseng
Degree of examining is bigger, then the label of a user is influenced bigger by the label of another user.Target mark between user a and user b
Label point of reference may include user a to the target labels point of reference of user b and user b to the target labels point of reference of user a,
User a can be identical or not to the target labels point of reference of user a to the target labels point of reference of user b and user b
Together.If it is different, then being joined when calculating the label of user a according to the label of user b using target labels of the user b to user a
Degree of examining is calculated;When calculating the label of user b according to the label of user a, joined using target labels of the user a to user b
Degree of examining is calculated.The social degree of association indicates the correlation degree of user socially, and the social degree of association is bigger, then it represents that association is got over
Closely.The social degree of association can be obtained according to incidence relation of the user on social networks, and incidence relation refers to user and user
Between on social networks exist connection, may include direct incidence relation and indirect incidence relation.Such as social activity is answered
With friend relation, the same social group user and the partnership of game player on gaming platform etc..Social network
Network refers to the relational network formed based on social activity, such as the relational network in wechat, QQ, microblogging and society.Social activity is answered
With can be instant messaging application, SNS (Social Network Services, social network service) application, discussion bar application etc..
Instant messaging application may include wechat, QQ and MSN etc..SNS application may include Renren Network and Facebook etc., but be not limited to
This.Characteristic similarity indicates that the similarity degree of user characteristically, characteristic similarity are bigger, then it represents that the feature of user gets over phase
Seemingly.Target labels point of reference between two users is according to the social degree of association and characteristic similarity between the two users
It is calculated.It can be the comprehensive social degree of association and characteristic similarity obtain a comprehensive target labels point of reference, when
Right target labels point of reference also may include first object Tag reference degree and the second target labels point of reference, wherein the first mesh
Marking Tag reference degree is obtained according to social calculation of relationship degree, and the social degree of association and first object Tag reference degree are positively correlated pass
System.Second target labels point of reference is calculated according to characteristic similarity, characteristic similarity and first object Tag reference degree
Positive correlation.Negative correlativing relation refers to: it is different that two variables change directions, another when the descending variation of a variable
A ascending variation of variable.Positive correlation refers to: two variables variation directions are identical, a descending variation of variable
When, the descending variation of another variable.
In one embodiment, the target labels in user and user's set between each user can be calculated to refer to
Degree, can also be with the target labels point of reference between calculating section user in order to reduce calculation amount.
In one embodiment, for active user, can be gathered according to the social incidence relation between user from user
Middle acquisition is used as with the user that active user has social incidence relation and refers to user, according to active user and with reference between user
The social degree of association and characteristic similarity be calculated active user and this with reference to the target labels point of reference between user, and
For the non-reference user of active user in user's set, then the target labels point of reference of active user and non-reference user can be with
It is preset value is, for example, 0.For an actual example, it is assumed that include 4 users: a1, a2, a3 and a4 in user's set.Then may be used
A1 and a2, a1 and a3, a1 and a4, a2 and a3, a2 and a4, a3 is calculated according to the social degree of association and characteristic similarity
Target labels point of reference between a4 user.But in order to reduce calculation amount, it is assumed that incidence relation or society is not present in a2 and a3
It hands over the degree of association smaller, does not then calculate the target labels point of reference of a2 and a3 according to the social degree of association and characteristic similarity, write from memory
The target labels point of reference recognized between a2 and a3 is that preset value is, for example, 0.
In one embodiment, it can calculate in the second user and the first child user set in the second child user set
Tag reference degree between first user obtains second with the label and corresponding target labels point of reference of the first user of reference
The label of user.For example, it is assumed that include 2 users: a1, a2 in the first child user set, wrapped in the second child user set
2 users: a3 and a4 are included, then can calculate the target labels between target labels point of reference, a3 and the a2 between a3 and a1
Point of reference calculates the target labels point of reference between target labels point of reference, a4 and the a2 between a4 and a1.
In one embodiment, can be chosen from the first child user set according to the social incidence relation of second user with
The first user that second user has social incidence relation, which is used as, refers to user, according to second user and with reference to the society between user
The degree of association and characteristic similarity is handed over first user to be calculated and with reference to the target labels point of reference between user.
In one embodiment, first in the second user and the first child user set in the second child user set is calculated
Second in the second user and the second child user set in Tag reference degree and the second child user set between user is used
Tag reference degree between family.It therefore, can be with reference to the user in the first child user set in the label for determining second user
Label and the label of the user in the second child user set obtain the label of second user.
In one embodiment, can be chosen from the first child user set according to the social incidence relation of second user with
The first user that second user has social incidence relation refers to user as first, according to the social incidence relation of second user
The second user that there is social incidence relation with second user is chosen from the second child user set as second with reference to user.Its
In, first can be the same or different with reference to the number of user and the number of the second reference user.According to active user and the
The social degree of association and characteristic similarity between one reference user are calculated between active user and the first reference user
Target labels point of reference, according to active user and second with reference between user the social degree of association and characteristic similarity calculate
Obtain the target labels point of reference between active user and the second reference user.
In one embodiment, by the first child user set, it is greater than first with the social degree of association of second user and is associated with
The first user for spending threshold value refers to user as first.It is big with the social degree of association of second user by the second child user set
User is referred to as second in the second user of the second degree of association threshold value.Wherein, the second degree of association threshold value can be greater than first and close
Connection degree threshold value.Since the label of the first user is determining, and the label of second user is obtained with reference to the label of user, because
This, the accuracy of the label of the first user is higher, therefore the second degree of association threshold value is greater than the first degree of association threshold value, can make true
When determining the label of second user, more refer to has determined that label and with the social incidence relation of second user presence first
The label of user, the accuracy of the label improved.
In one embodiment, the social degree of association is obtained according to the incidence relation between user.Such as it can be according to user
Interbehavior and at least one of the length of social networks chain determine.Interbehavior may include the connection between user
Number and connection frequency etc..Social networks chain is according to the intermediate user between two users of connection, two users according to society
Association sequence is handed over successively to form.User is regarded as a node, is connected between node and node with direct correlation relationship
Line obtains social networks chain as path.The length of relation chain can be obtained with the number in path.For example, for a1 and a4, it is false
If the good friend of a1 is a2, the good friend of a2 is a3, and the good friend of a3 is a4, then a2, a3 are intermediate user, and relationship by objective (RBO) chain is a1 → a2
→ a3 → a4, relationship chain length are 3.If interbehavior indicates user, connection is closer, and the social degree of association is bigger, if closed
The length of tethers is shorter, then the social degree of association is bigger.If the social degree of association is for example, it is 1 that relationship chain length, which can be set,
1, if relationship chain length is 2, the social degree of association is 0.6.If 1 year connection frequency is 2 days primary between user,
The social degree of association is 0.8, if 1 year connection frequency is 10 days primary between user, the social degree of association is 0.3.
In one embodiment, for there are the users of friend relation, then otherwise it is 0 that the social degree of association, which is 1,.
In one embodiment, characteristic similarity is obtained according to the distance of a feature or multiple features between user,
Feature is used to indicate the attribute of user, and calculating feature corresponding to characteristic similarity can according to need determination.Such as it can select
The attributes such as age, gender and the level of consumption at family are taken as feature.Characteristic similarity can be used for measuring with characteristic distance
Whether image is similar.Characteristic distance is smaller, then image is more similar, and similarity is bigger, then image is more similar.Characteristic distance can be with
Characteristic similarity is negatively correlated relationship, i.e. characteristic distance is big, then characteristic similarity is small.Spy can be calculated according to characteristic distance
Levy similarity.For example, the inverse of characteristic distance can be characteristic similarity.In another example in the feature for calculating user i and user j
Similarity wijWhen corresponding formula can indicate as follows:Wherein exp refer in higher mathematics with
Natural constant e is the exponential function at bottom, | | xi-xj||2Refer to the comprehensive distance of the feature of user i and user j, comprehensive distance is for example
For user each characteristic distance and average value.The distance of feature can according to need setting, for example, use can be set
The age difference at family and the relationship of characteristic distance, age difference is bigger, then characteristic distance is bigger.Also the gender of user can be set
With the relationship of characteristic distance.After obtaining the corresponding distance of each feature, the sum of the corresponding distance such as distance of each feature of synthesis
As total distance.For example, if gender is identical, distance is 0, if gender is different, distance is 0.05.Assuming that i and j
Age difference is 10 years old, then characteristic distance can be 0.02, if i and j age difference are 20 years old, user distance can be
0.1.The value of a can according to need setting but cannot be 0, such as a can be 1.
In one embodiment, the social calculation of relationship degree in being gathered according to user between user obtains between user
First object Tag reference degree, such as can be using the social degree of association as first object Tag reference degree, can also be to social activity
The degree of association is further processed to obtain first object Tag reference degree, for example, the social degree of association is normalized to obtain
One target labels point of reference.For example, user j can use formula to the first object Tag reference degree of user i
It indicates, wherein wijIndicate the social degree of association of i and j, n is the number of users in user's set.Then by first object Tag reference
The matrix P1 of degree composition can be as shown in matrix (2), wherein in matrix (2), pijIndicate user j to the first object mark of user i
The probability that label point of reference, i.e. label are transferred to user i from user j.It is appreciated that the numerical value of row k kth column is used in matrix P1
Family to the Tag reference degree of oneself can be default value, for example, 1.
In one embodiment, the characteristic similarity in being gathered according to user between user is calculated between user
The second target labels point of reference.Such as can be using characteristic similarity as the second target labels point of reference, it can also be to feature
Similarity is further processed to obtain the second target labels point of reference, for example, characteristic similarity is normalized to obtain
Two target labels point of reference.For example, the second target labels point of reference of user j and user i can use formulaIt indicates, wherein wijThe characteristic similarity of ' expression i and j, n are the number of users in user's set.Then by
The matrix P2 of second target labels point of reference composition can be as shown in matrix (3), wherein in matrix (3), p 'ijIndicate j pairs of user
The probability that the second target labels point of reference of user i, i.e. label are transferred to user i from user j.It is appreciated that kth in matrix P2
Row kth column numerical value, that is, user to the Tag reference degree of oneself can be default value, for example, 1.
Step S208, for the active user in the second child user set, according to the use in active user and user's set
The corresponding initial labels value of user in target labels point of reference and user's set between family, it is corresponding to obtain active user
Target labels value.
Specifically, active user refers to second user current when calculating target labels value, due to the second child user set
Including multiple second users, therefore when determining the corresponding target labels value of a user, a user is active user, when determining that b uses
When the corresponding target labels value in family, b user is active user.When calculating target labels value, active user and user are utilized
Target labels value is calculated in target labels point of reference and the corresponding initial labels value of the user.If active user's is initial
Label value is a value, then the sum for the product that can be multiplied according to target labels point of reference with initial labels value obtains target
Label value, if the initial labels value of active user is the initial labels vector of multiple vector values composition, a kind of label is corresponding pre-
If the vector value of position, then target labels value is object vector value.
It in one embodiment, can be with reference to one or more users when determining the corresponding target labels value of active user
Initial labels value.It, then can be according to the initial labels value of user and should if it is the initial labels value with reference to a user
Target labels point of reference between user and active user obtains the corresponding target labels value of active user.If it is with reference to multiple
With reference to the initial labels value of user, then active user and each target labels point of reference and reference with reference between user are integrated
The corresponding target labels value of active user is calculated in the initial labels value of user.Such as it can calculate each with reference to the first of user
Beginning label value and corresponding active user and product with reference to the target labels point of reference between user, by each product addition,
Obtain the corresponding target labels value of active user.
For example, when that obtain is the second user in the second child user set and the first user in the first child user set
Between target labels point of reference when, can according between active user and the first user target labels point of reference with it is corresponding
The sum of the product of the initial labels value of first user obtains the corresponding target labels value of active user.It is false for an actual example
If the first user has 2: a2 and a3, the target labels point of reference between active user a1, a1 and a2, a1 and a3 is respectively
p12、p13, the initial labels value of a2 is h2, and the initial labels value of a3 is h3, then the corresponding target labels value of a1 can be calculated
For p12*h2+p13*h3。
For example, when that obtain is the second user in the second child user set and the first user in the first child user set
Between Tag reference degree and second user in the second child user set and the second user in the second child user set it
Between Tag reference when spending, can be according to target labels point of reference and corresponding first user between active user and the first user
Initial labels value product sum and active user and the second family between target labels point of reference with it is corresponding second use
The sum of the product of the initial labels value at family, obtains target labels value.
In one embodiment, when target labels point of reference includes first object Tag reference degree and the second target labels
When point of reference, it can be calculated according to first object Tag reference degree initial labels value corresponding with user in user's set
Active user corresponding first influences label value, according to corresponding second target between active user and the user in user's set
The corresponding initial labels value of user in Tag reference degree and user's set, active user corresponding second, which is calculated, to be influenced
Label value obtains target labels value according to the first influence label value and the second influence label value.
In one embodiment, when a kind of label corresponds to the vector value of initial labels vector predeterminated position, can will work as
Target labels point of reference vector value corresponding with each label of the user between preceding user and user is multiplied, and obtains product
Afterwards, product corresponding for same class label is added to obtain the corresponding vector of label of the category in target labels vector
Value.
For an actual example, it is assumed that with reference to user include a2, a3 and a4, active user a1, a1 and a2, a1 with
First object Tag reference degree between a3, a1 and a4 is respectively p12、p13And p14,, the classification of label is three classes, therefore just
The vector value of beginning label vector is 3, and first vector value is the corresponding value of the first label classification, and second vector value is second
The corresponding value of label classification, third vector value are the corresponding value of third label classification.Assuming that a1, a2, a3 and a4 are corresponding
Initial labels vector is respectively (b11, b12, b13), (b21, b22, b23), (b31, b32, b33) and (b41, b42, b43),
It is vector that then active user a1 corresponding first, which influences label value, is expressed as (e11, e12, e13), then e11=b11+p12*b21+
p13*b31+p14* b41, e12=b12+p12*b22+p13*b32+p14* b42, e13=b13+p12*b23+p13*b33+p14*b43。
It is vector that similarly also available active user a1 corresponding second, which influences label value, is expressed as (g11, g12, g13), then target
Label vector can be (e11+g11, e12+g12, e13+g13)
When the matrix of first object Tag reference degree composition is matrix P1, the matrix of the second target labels point of reference composition is
Matrix P2, is expressed as follows with matrix: matrix P1 being multiplied with matrix F, obtains first object label matrix, by matrix P2 and matrix
F is multiplied, and obtains the second target labels matrix, obtains target mark in conjunction with first object label matrix and the second target labels matrix
Matrix is signed, a line of target labels matrix can represent the corresponding target labels value of a user.When carrying out matrix multiple, root
The principle being divided by according to matrix, label value other for each tag class are according to target labels point of reference and same label
What label value was calculated.That is, being according to active user and user for the corresponding vector value of r-th of label of active user
Target labels point of reference and r-th of label of user corresponding vector value obtain.This can pass through the rule of matrix multiplication
It is embodied, such as matrix P1 is being multiplied with matrix F, when obtaining matrix F 1, the value F of h row r column in F1hrEqual to matrix
The matrix value of P1 h row and the product that is multiplied of matrix value arranged of r in matrix F and.
In one embodiment, matrix P1 is multiplied with matrix F, obtains first object label matrix, by matrix P2 and square
Battle array F is multiplied, and obtains the second target labels matrix.Target is obtained in conjunction with first object label matrix and the second target labels matrix
Label matrix.Obtain target labels matrix FMeshFormula can be as follows, wherein q1 and q2 be weight, specific value according to
It needs to be arranged, such as can be 1.
F1=P1*F (4)
F2=P2*F (5)
FMesh=F1*q1+F2*q2 (6)
In one embodiment, according to corresponding target labels point of reference between active user and user in user's set
And the corresponding initial labels value of user in user's set, it can be when the corresponding target labels value of active user is calculated only
It is once calculated, successive ignition calculating can also be carried out, it, can be according to target mark corresponding between user in iterative calculation
The corresponding initial labels value of user obtains intermediate label value in label point of reference and user's set, further according to intermediate label value and uses
Target labels value is calculated in corresponding target labels point of reference between family.
Step S210 determines the corresponding target labels of active user according to target labels value.
Specifically, after obtaining target labels value, target labels are determined according to the size of target labels value.For example, for mesh
Mark label matrix, the corresponding target labels vector of available each active user, according to vector each in target labels vector
The size of value determines target labels.It can be using the corresponding label of vector value maximum in vector value as target labels, certainly
It is also possible to vector value being greater than the corresponding label of preset value such as 0.8 as target labels.For an actual example, it is assumed that when
The corresponding target labels vector of preceding user is (0.1,0.3,0.6), and 0.6 corresponding tag representation credit is poor, 0.3 corresponding label
Indicate that credit is medium, 0.1 corresponding tag representation information is good, then the corresponding target labels of active user are that credit is poor.
It in one embodiment, can be by second user and corresponding mesh after obtaining the corresponding target labels of second user
Mark label associated storage.
In one embodiment, can be believed according to the target labels of second user to the corresponding terminal push of second user
Breath.For example, the fund for then recommending financing risk low then recommends the financing user of radical type for the financing user of conservative
Risk of managing money matters is higher but returns also higher fund.
In one embodiment, different user management strategies can be implemented according to the target labels of second user, for example,
It is the user of credit difference for user tag, then the user is added to the blacklist of loan product.
User tag provided in an embodiment of the present invention determines method, in the label for determining user, according to having determined that label
User go prediction not determine the user of label, and be that characteristic similarity and calculation of relationship degree between synthetic user obtains
, without manually marking the label of also available user, therefore, the accuracy of obtained label is high and saves Internet resources.
In one embodiment, when target labels point of reference includes first object Tag reference degree and the second target labels
When point of reference, as shown in figure 3, step S208 according to active user and is used i.e. for the active user in the second child user set
Family set in user between corresponding target labels point of reference and user set in the corresponding initial labels value of user, obtain
Include: to the corresponding target labels value of active user
Step S302, for the active user in the second child user set, according to the use in active user and user's set
Between family corresponding first object Tag reference degree and user set in the corresponding initial labels value of user, currently used
Family corresponding first influences label value.
Specifically, for active user, the first object Tag reference degree between active user and user is obtained, obtaining should
The corresponding initial labels value of user obtains the user according to first object Tag reference degree initial labels value corresponding with the user
Influence label value to active user, the influence label value in conjunction with user to active user obtain the first influence label value, such as
What the sum that the first influence label value can be the influence label value according to each with reference to user to active user obtained.Calculating the
When one influence label value, take turns can be carried out more and iterated to calculate.Such as each active user, when calculate for the first time,
It can be according in corresponding first object Tag reference degree between active user and the user in user's set and user's set
The corresponding initial labels value of user update to obtain intermediate label value, then further according to the updated intermediate label value of user and
Corresponding first object Tag reference degree updates intermediate label value between user in active user and user's set, be repeated once or
In repeatedly being gathered according to the updated intermediate label value of active user and active user with user corresponding first between user
Target labels point of reference updates the step of intermediate label value, and obtaining active user corresponding first influences label value.Iterative calculation
Method be expressed as follows with matrix: for first object Tag reference degree composition matrix P1, initial labels vector composition just
P1 can be multiplied with F, obtain updated F, then P1 is multiplied with updated F by beginning label matrix F, repeat by P1 with
The step of updated F is multiplied, until meet the condition of convergence, wherein the condition of convergence include number of repetition reach preset times and
The variation of the preceding F once updated the and F currently updated is less than at least one of preset value.Before wherein the F that once updates with work as
The variation of the F of preceding update can be indicated with the quadratic sum of the difference of the matrix value of same position.Due to passing through successive ignition,
Can according to the social degree of association from the user for partially having label and without in tagging user study to useful information, to determine nothing
The label of tagging user.I.e. when calculating active user corresponding first influences label value, the first user and second user pair
The label answered will affect the label value of active user, and the influence degree of label value, that is, first object Tag reference degree is basis
What the social degree of association obtained, therefore, user more close for social connections, interactional degree is bigger, therefore passes through excessive
The propagation of secondary label value can make the label for being associated with close user same or similar.
Step S304, for the active user in the second child user set, according to the use in active user and user's set
Between family corresponding second target labels point of reference and user set in the corresponding initial labels value of user, currently used
Family corresponding second influences label value.
Specifically, for active user, the second target labels point of reference between active user and user is obtained, obtaining should
The corresponding initial labels value of user obtains the user according to the second target labels point of reference initial labels value corresponding with the user
Influence label value to active user, the influence label value in conjunction with user to active user obtain the second influence label value, such as
What the sum that the second influence label value can be the influence label value according to user to active user obtained.It influences to mark in calculating second
When label value, take turns can be carried out more and iterated to calculate.Such as it can basis when calculate for the first time for each active user
Active user and user set in user between corresponding second target labels point of reference and user set in user couple
The initial labels value answered updates to obtain intermediate label value, then further according to the updated intermediate label value of user and active user
Corresponding second target labels point of reference updates intermediate label value between the user in user's set, is repeated one or more times root
According to corresponding second target between the updated intermediate label value of active user and active user and the user in user's set
Tag reference degree updates the step of intermediate label value, and obtaining active user corresponding second influences label value.The side of iterative calculation
Method is expressed as follows with matrix: for the matrix P2 of the second target labels point of reference composition, the initial mark of initial labels vector composition
Matrix F is signed, P2 can be multiplied with F, obtain updated F, then P2 is multiplied with updated F, repeat P2 and is updated
The step of rear F is multiplied, until meet the condition of convergence, wherein the condition of convergence includes that number of repetition reaches preset times and previous
The variation of the F of the secondary update and F currently updated is less than at least one of preset value.Before wherein the F that once updates and it is current more
The variation of new F can be indicated with the quadratic sum of the difference of the matrix of same position.It, being capable of root due to passing through successive ignition
According to the social degree of association from the user for partially having label and without study in tagging user to useful information, used with determining without label
The label at family, i.e., when calculating active user corresponding second influences label value, the first user and the corresponding mark of second user
Label will affect the label value of active user, and the influence degree of label value i.e. the second target labels point of reference is according to feature phase
It is obtained like degree, therefore, user more similar for feature, interactional degree is bigger, therefore passes through multiple label value
It propagates, the label of the similar user of feature can be made same or similar.
In one embodiment, when updating matrix F every time, due to the first user for having determined that label, label
It is known that therefore when obtaining updated matrix F, it can be by the corresponding label vector weight of the first user in updated matrix F
It is set to initial labels vector.
Step S306 influences label value according to active user corresponding first and the second influence label value determines current use
The corresponding target labels value in family.
Specifically, after obtaining the first influence label value and the second label value, label value and second is influenced according to first
It influences label value and determines the corresponding target labels value of active user.For example, label value and the second shadow can be influenced according to first
The sum for ringing label value obtains target labels value, and can be influences the sum of label value and the second influence label value as mesh for first
Mark label value, be also possible to obtain the first influence label value and second influence label value and after, be added with default label value
Obtain target labels value.
In one embodiment, label value is influenced according to active user corresponding first and the second influence label value determines
The corresponding target labels value in current family includes: to obtain first to influence corresponding first weight of label value and the second influence label value
Corresponding second weight;Label value and corresponding first weight are influenced according to first, second influence label value and corresponding
Second weight is weighted summation, obtains the corresponding target labels value of active user.
Specifically, the first weight and the second weight can according to need setting, such as the first weight can be 0.6, the
Two weights can be 0.7.Obtain corresponding first weight of the first influence label value and the second influence label value corresponding second
After weight, label value and corresponding first weight, the second influence label value and corresponding second weight are influenced according to first
It is weighted summation, obtains the corresponding target labels value of active user.
In one embodiment, as shown in Figure 4 A, step S304 be according to active user and user set in user it
Between corresponding second target labels point of reference and user set in the corresponding initial labels value of user, current use is calculated
Family corresponding second influences label value
Step S402, using user gather in each user as node, power of the second target labels point of reference as side
Weight, building obtain characteristic pattern.
Specifically, figure is made of the side between node and node, using user as node, the second target labels ginseng
Weight of the degree of examining as side, obtains characteristic pattern.
It as shown in Figure 4 B, is the characteristic pattern in one embodiment, it is assumed that include 5 users: a1~a5 in user's set, then
There are 5 nodes in characteristic pattern, the line segment between 5 nodes indicates side.
Step S404 obtains the adjacent node of the corresponding present node of active user and present node in characteristic pattern.
Specifically, adjacent node refers to that there are the nodes of side connection, if the second target labels point of reference between node
It is not 0, then there are sides between node, if the second target labels point of reference between node is 0, are not present between node
Side.As shown in Figure 4 B, if present node is a1, the adjacent node of a1 is a2, a4 and a5.
Step S406, according to the current label value of adjacent node in characteristic pattern and the side right of present node and adjacent node
The current label value of present node is updated again.
Specifically, the corresponding label value of node in characteristic pattern can be updated by a wheel or the update more taken turns, every wheel
When the corresponding label value of each second user is updated.After one wheel of label value experience of the node in characteristic pattern updates,
Obtained label value is updated as current label value.When updating first time, using the corresponding initial labels value of user as feature
The current label value of corresponding node in figure.The label value of present node is according to the current label value of adjacent node and current
What the side right on the side between node and corresponding adjacent node was updated again, according to present node and adjacent node while while
Weight is weighted summation with the current label value of corresponding adjacent node, and it is updated current to obtain present node in characteristic pattern
Label value.After the corresponding node of second user each in characteristic pattern is used as present node to update, S408 is entered step.
Step S408 judges whether to meet the condition of convergence;If it is not, then return according in characteristic pattern adjacent node it is current
The side right of label value and present node and adjacent node updates the step of current label value of present node again, if so, into
Enter step S410, obtaining active user corresponding second influences label value.
Specifically, it updates in characteristic pattern after the label value of the corresponding present node of each second user, continues to return to basis
The current label value of adjacent node and the side right of present node and adjacent node update the current of present node again in characteristic pattern
The step of label value, the update of next round is carried out with the label value to the node in characteristic pattern, until meet the condition of convergence, it will most
The label value of the node updated afterwards influences label value as second, and wherein the condition of convergence may include that recycle time reaches in advance
If in number and characteristic pattern it is preceding once update while weight with currently update while the variation of weight be less than in preset value
At least one.In characteristic pattern it is preceding once update while weight with currently update while the variation of weight can use identical bits
The quadratic sum of the difference on the side set indicates.
In one embodiment, the label value of the corresponding node of the first user can also be according to second user pair in characteristic pattern
The label value update method for the present node answered is updated can also be without updating.
In one embodiment, as shown in figure 5, according to user gather in characteristic similarity between user use is calculated
The second target labels point of reference between family includes:
Step S502 obtains active user from the second child user set, according to the social incidence relation of active user from
Screening obtains the corresponding object reference user set of active user in user's set.
Specifically, social incidence relation refers to that, there are social connections between user and user, social intercourse system is stored with social activity
Relation chain, the social incidence relation that social networks chained record user is established in social intercourse system, and record and have with user
The information of the other users of standby social activity incidence relation.Social networks chain is arranged successively composition according to according to user-association sequence.?
User regards a node as, and line is carried out between node and node with direct correlation relationship as path, obtains social pass
Tethers.For example, for a1 and a4, it is assumed that the good friend of a1 is a2, and the good friend of a2 is a3, and the good friend of a3 is a4, then social networks chain
For a1 → a2 → a3 → a4.Social incidence relation may include direct incidence relation and indirect incidence relation.Directly close
It is to be directly linked that connection relationship, which refers between user and user, is friend relation.Indirect relation refers between user and user
It is associated by one or more intermediate users, such as there are common good friends with user by user.Such as social networks chain is a1
In → a2 → a3 → a4, a1 and a2, a2 and a3, a3 and a4 are direct incidence relations, between a1 and a3, a1 and a4, a2 and a4 are
The incidence relation connect.Object reference user in object reference user set be screen and obtain from user's set, and with work as
There is the user of social incidence relation in preceding user.It can be and all there is social incidence relation in user's set with active user
User is also possible to the user that part has social incidence relation.For example, the object reference user in object reference user set
It can be the good friend within active user's M degree, social incidence relation can be indicated between user and user with expenditure, and degree is used with intermediate
The relationship at family is that the number of intermediate user adds 1 to be equal to degree.For example, if active user and certain user are good friend, intermediate user
It is 0, certain user is 1 degree of good friend of active user.If active user is the good friend of certain user good friend, intermediate user 1, certain
User is 2 degree of good friends of active user.
In one embodiment, according to the social incidence relation of active user, screening obtains active user from user's set
Corresponding object reference user set includes: that screening obtains currently from user's set according to the social incidence relation of active user
The corresponding initial reference user set of user, it is corresponding according to the corresponding current signature of active user and each initial reference user
Fixed reference feature the characteristic similarity of corresponding active user Yu initial reference user is calculated;According to active user with it is each
The screening from initial reference user set of the characteristic similarity of initial reference user obtains object reference user.
Specifically, current signature refers to the feature of active user, and fixed reference feature refers to the feature of initial reference user.Feature
Similarity can be calculated with the distance of feature, and the calculation method of the distance of feature can according to need setting, such as using
Euclidean distance algorithm is calculated.Characteristic similarity can be greater than to default characteristic similarity and/or characteristic similarity ranking position
Initial reference user within default ranking is as object reference user.Characteristic similarity ranking is according to carrying out from big to small
Sequence, characteristic similarity is bigger, then ranking is more forward.For example, the good friend within available active user M degree is as initial
With reference to user, characteristic similarity is obtained further according to the characteristic similarity calculating sifting of the good friend within active user and these M degree
The user of ranking top 10 is as object reference user.
In one embodiment, according to the social incidence relation of active user, screening obtains active user from user's set
Corresponding object reference user set includes: to obtain active user from user's set according to the social incidence relation of active user
Corresponding first is directly linked user's set;Characteristic similarity according to active user and the first direct correlation user is straight from first
It connects screening in association user set and obtains first with reference to user;Gathered according to the social incidence relation of the first reference user from user
It is middle to obtain first with reference to the corresponding second direct correlation user's set of user;It is directly linked and uses with second with reference to user according to first
The screening from second direct correlation user's set of the characteristic similarity at family obtains second with reference to user;By first with reference to user and
Second reference user is as the object reference user in object reference user set.
Specifically, the first direct correlation user in first direct correlation user's set is to exist directly to close with active user
The user of connection relationship after obtaining the first direct correlation user set, calculates active user and each first and is directly linked user's
Characteristic similarity is greater than default characteristic similarity and/or characteristic similarity ranking is located within default ranking by characteristic similarity
First be directly linked user as first refer to user.Characteristic similarity ranking be according to being ranked up from big to small, it is special
Sign similarity is bigger, then ranking is more forward.Default ranking can according to need setting, for example, 10.First is obtained with reference to user
Afterwards, direct correlation user of the acquisition first with reference to user, composition the second direct correlation user's set will be with the first reference user's
Characteristic similarity is greater than default characteristic similarity and/or characteristic similarity ranking is located at the second direct correlation preset within ranking
User refers to user as second, and presetting ranking for example can be 8, and presetting characteristic similarity for example can be 0.6.By the first ginseng
User and the second reference user are examined as object reference user, it will be understood that can also continue to obtain second with reference to user's
Be directly linked user, according to second with reference to user directly association user characteristic similarity screening third refer to user, will
Third is also used as object reference user with reference to user.For an actual example, 1 degree of good friend of available active user is as the
Then one direct association user is user's conduct that ranking is preceding 10 with the characteristic similarity of active user in 1 degree of good friend of acquisition
First refers to user, then finds this 10 first 1 degree of good friend with reference to each user in user.For this 10 first with reference to use
1 degree of good friend of each user in family, using with each first with reference to the characteristic similarity ranking of user be preceding 10 user as the
Two refer to user, acquire object reference user according to the algorithm, can reduce computation complexity, but can with reference to it is current
User characteristics are similar, and the label that there is the user of social incidence relation obtains the label of active user, obtained active user
Target labels accuracy it is high, can accurately classify to active user.
Step S504, according to the characteristic similarity of active user and each object reference user be calculated active user with
The second target labels point of reference of object reference user.
Specifically, the characteristic similarity of active user and object reference user can be referred to as the second target labels
Degree, characteristic similarity can also be further processed to obtain the second target labels point of reference, for example, to characteristic similarity into
Row normalization obtains the second target labels point of reference.
In one embodiment, the characteristic similarity of active user and each object reference user can be counted,
Obtain characteristic similarity statistical result;The current signature similarity for obtaining active user and object reference user, according to current spy
The ratio of sign similarity and characteristic similarity statistical result obtains active user and the second target labels of object reference user are joined
Degree of examining.Such as the object reference user of active user is a2, a3 and a4, active user and object reference user be a2, a3 with
And a4 similarity is respectively 0.8,0.7 and 0.6.Then similarity statistical result be 0.8+0.7+0.6=2.1, active user with
The second target labels point of reference between a2 can be 0.8/2.1 ≈ 0.38.
In one embodiment, using user gather in each user as node, the second target labels point of reference conduct
The weight on side, building obtain characteristic pattern include: using user gather in each user as node, by object reference user correspondence
Adjacent node of the node as present node, the second target labels point of reference of active user and object reference user are as side
Weight, building obtain characteristic pattern.
Specifically, after the second target labels point of reference for obtaining active user and object reference user, in construction feature figure
When, using the corresponding node of object reference user as the adjacent node of present node, the of active user and object reference user
Weight of the two target labels point of reference as side.
It is the social activity according to user when calculating the second target labels point of reference between user in the embodiment of the present invention
Incidence relation gathers this screen fraction user from user to calculate the second target labels point of reference between user, passes through choosing
It takes the user in the presence of social incidence relation to calculate characteristic similarity, and further sieves object reference user using characteristic similarity,
It is big due to there is social incidence relation and characteristic similarity while reducing the calculation amount of the second target labels point of reference
The reference value of user is stronger, therefore can also effectively be propagated label.
It is appreciated that since matrix P2 indicates the second target labels point of reference between user and user, characteristic pattern
The weight on middle side i.e. the second target labels point of reference can use matrix P2 and be indicated, wherein in matrix P2 j-th of the i-th row to
Magnitude indicates the weight on the side between i-node and j node, if vector value is not 0, then it represents that i-node and j node in characteristic pattern
It is adjacent node, if vector value is 0, then it represents that i-node and j node are not adjacent nodes in characteristic pattern.To in characteristic pattern
The renewal process of the label value of node can be indicated with following steps:
Step 11: carrying out label propagation: FMore=P2*FWhen;I.e. according to matrix P2 and current label matrix FWhenIt is calculated
Update label matrix FMore, wherein current label matrix is initial labels matrix when calculating for the first time.
Step 12: label matrix F will be updatedMoreIn the corresponding label vector of the first user be reduced to initial labels vector, obtain
To current label matrix FWhen, return step 11, until meeting the default condition of convergence.
In one embodiment, as shown in Figure 6A, step S304 is i.e. for the active user in the second child user set, root
According to corresponding first object Tag reference degree between active user and the user in user's set and the user in user's set
Corresponding initial labels value, obtaining the corresponding first influence label value of active user can specifically include following steps:
Step S602, using user gather in each user as node, first object Tag reference degree as while while
Weight obtains social associated diagram.
Specifically, figure is made of the side between node and node, using user as node, first object label ginseng
Weight of the degree of examining as side, obtains social associated diagram.
As shown in Figure 6B, it is the social associated diagram in one embodiment, it is assumed that include 5 users in user's set: a1~
A5 then has 5 nodes in social associated diagram, and the line segment between 5 nodes indicates side.
Step S604 obtains the adjacent segments of the corresponding present node of active user and present node in social associated diagram
Point.
Specifically, adjacent node refers to that there are the nodes of side connection, if the first object Tag reference degree between node
It is not 0, then there are sides between node, if the first object Tag reference degree between node is 0, are not present between node
Side.As shown in Figure 6B, if present node is a1, the adjacent node of a1 is a3 and a5.Can will exist with active user
Adjacent node of the corresponding node of the user of direct correlation relationship as present node.
Step S606, according to the current label value of adjacent node in social associated diagram and present node and adjacent node
Side right updates the current label value of present node again.
Specifically, the corresponding label value of node in social associated diagram can be by a wheel or the update more taken turns, every wheel
The corresponding label value of each second user can be updated when update.When the label value of the node in social associated diagram is undergone
After one wheel updates, the label value updated is as current label value.When updating first time, by the corresponding initial labels of user
It is worth the current label value as corresponding node in social associated diagram.The label value of present node is the label according to adjacent node
What the side right on the side between value and present node and corresponding adjacent node was updated again, according to present node and adjacent segments
The side right weight on the side of point with the label value of corresponding adjacent node is weighted summation, obtains in social associated diagram present node more
Label value after new.After the corresponding node of second user each in social associated diagram is used as present node to update, into
Enter step S608.
Step S608 judges whether to meet the condition of convergence, if it is not, then returning according to adjacent node in social associated diagram
Current label value and the side right of present node and adjacent node update the step of current label value of present node again, if
It is to enter step S610, obtaining active user corresponding first influences label value.
Specifically, after the label value for updating the corresponding present node of each second user in social associated diagram, continue to return
It is updated again according to the side right of the current label value of adjacent node in social associated diagram and present node and adjacent node and works as prosthomere
The step of current label value of point, carries out the update of next round with the label value to the node in social associated diagram, until meeting
The label value of the condition of convergence, the node that final updating is obtained influences label value as first, and wherein the condition of convergence can wrap
Include recycle time reach in preset times and social associated diagram it is preceding once update while weight with currently update while power
The variation of weight is less than at least one of preset value.In social associated diagram it is preceding once update while weight with currently update while
The variation of weight can be indicated with the quadratic sum of the difference on the side of same position.
In one embodiment, the label value of the corresponding node of the first user can also be used according to second in social associated diagram
The label value update method of the corresponding present node in family is updated can also be without updating.
It is appreciated that social activity is closed since matrix P1 indicates the first object Tag reference degree between user and user
The weight on side, that is, first object Tag reference degree can use matrix P1 and be indicated in connection figure, wherein the i-th row jth in matrix P1
A vector value indicates the weight on the side between i-node and j node, if vector value is not 0, then it represents that i is saved in social associated diagram
Point and j node are adjacent nodes, if vector value is 0, then it represents that i-node and j node are not adjacent nodes in social associated diagram.
The renewal process of the label value of node in social associated diagram can be indicated with following steps:
Step 21: carrying out label propagation: FMore=P1*FWhen;I.e. according to matrix P1 and current label matrix FWhenIt is calculated
Update label matrix FMore, wherein current label matrix is initial labels matrix when calculating for the first time.
Step 22: label matrix F will be updatedMoreIn the corresponding label vector of the first user be reduced to initial labels vector, obtain
To current label matrix FWhen, return step 21, until meeting the default condition of convergence.
Below in conjunction with Fig. 7 A, with a specific embodiment to user tag provided in an embodiment of the present invention determine method into
Row explanation:
Step S702, obtains the relation chain between the user's set and user of input, and user's set includes having determined that mark
First child user set of label and the second child user set for not determining label.
Wherein, user's set can be user whole in social application, is also possible to the social degree of association and is greater than default close
The set of user's composition of connection degree.
Step S704, according to user gather in the corresponding label of each user determine and each state the corresponding initial labels of user
Value;
Step S706 obtains the user in user's set in each user within M degree good friend, it is corresponding to form each user
Initial reference user set;
Wherein, M can according to need setting, for example, 2.
Step S708, being obtained in initial reference user set with the characteristic similarity of active user is a initial of preceding W
With reference to user as object reference user;
Step S710, according to the characteristic similarity of active user and each object reference user be calculated active user with
The second target labels point of reference of object reference user;
Step S712, using user gather in each user as node, phase of the object reference user as present node
Neighbors, weight of the second target labels point of reference as the side between corresponding node, obtains characteristic pattern;
Step S714 repeats to work as prosthomere according to the label value of adjacent node in characteristic pattern and the weight update on corresponding side
The step of label value of point, is one or many, until meeting the condition of convergence, obtains the second disturbance degree label value;
Step S716, according to user gather in the social calculation of relationship degree of active user and each user obtain active user
With the first object Tag reference degree of each user;
Step S718, using user gather in each user as node, first object Tag reference degree is as corresponding
The weight on the side between node obtains social associated diagram;
Step S720, repeatedly more according to the current label value of adjacent node in social associated diagram and the weight on corresponding side
The step of current label value of new present node, is one or many, until meeting the condition of convergence, obtains the first disturbance degree label value;
Step S722 obtains target labels value according to the first disturbance degree label value and the second disturbance degree label value;
Step S724 determines the corresponding target labels of active user according to target labels value.
In one embodiment, user tag provided in an embodiment of the present invention determines that the schematic illustration of method can be such as figure
Shown in 7B, it needs to be determined that user label when, the data of input include user identifier, user characteristics, social networks chain and
The corresponding label of certain customers.Server can construct initial labels matrix F based on user tag, and be looked into based on social networks chain
The good friend looked within the M degree of each user gathers, and obtains and the user characteristics in good friend's set within each user M degree
K most like user, after finding the similar association user of the corresponding K feature of each user, building is similar based on feature
The characteristic pattern of degree, the similar association user of the corresponding K feature of each user are the adjacent node of the user, be may then based on
Matrix P2 and characteristic pattern carry out label propagation, and the number that label is propagated can be multiple, until F restrains.In addition to this, base is gone back
In social networks chain building social activity associated diagram, there is the user of direct correlation relationship adjacent node each other, may then based on matrix
P2 and social associated diagram carry out label propagation, and the number that label is propagated can be multiple, until F restrains.Then it can merge
Two labels propagate as a result, obtaining target labels and exporting.Wherein, when fusion, can be will finally propagate two obtained
Target labels matrix is added.
As shown in figure 8, in one embodiment, providing a kind of user tag determining device, which determines dress
Setting can integrate in above-mentioned server 120, can specifically include user's set acquisition module 802, initial labels value determines
Module 804, target labels point of reference computing module 806, target labels value computing module 808 and target labels determining module
810。
User, which gathers, obtains module 802, and for obtaining user's set, user's set includes having determined that the first son of label is used
Family set and the second child user set for not determining label;
Initial labels value determining module 804 determines each use for the corresponding label of user each in gathering according to user
The corresponding initial labels value in family;
Target labels point of reference computing module 806, between user in being gathered according to user the social degree of association and
The target labels point of reference between user is calculated in characteristic similarity;
Target labels value computing module 808, for for the active user in the second child user set, according to active user
With each user is corresponding initial in corresponding target labels point of reference and user's set between each user in user's set
The corresponding target labels value of active user is calculated in label value;
Target labels determining module 810, for determining the corresponding target labels of active user according to target labels value.
In one embodiment, target labels point of reference includes first object Tag reference degree and the second target labels ginseng
Degree of examining, target labels point of reference computing module 806 are used for: according to user gather in social calculation of relationship degree between user obtain
First object Tag reference degree between user, according to user gather in characteristic similarity between user be calculated user it
Between the second target labels point of reference;
As shown in figure 9, target labels value computing module 808 includes:
First influences label value computing unit 808A, for for the active user in the second child user set, according to working as
Preceding user and user set in user between corresponding first object Tag reference degree and user set in user it is corresponding
Initial labels value, obtain active user it is corresponding first influence label value;
Second influences label value computing unit 808B, for for the active user in the second child user set, according to working as
Preceding user and user set in user between corresponding second target labels point of reference and user set in user it is corresponding
Initial labels value, obtain active user it is corresponding second influence label value;
Target labels value computing unit 808C, for influencing label value and the second shadow according to active user corresponding first
It rings label value and determines the corresponding target labels value of active user.
In one embodiment, the second influence label value computing unit 808 includes:
Characteristic pattern constructs subelement, for using user gather in each user as node, the second target labels refer to
The side right weight as side is spent, building obtains characteristic pattern;
Node obtains subelement, for obtaining the phase of the corresponding present node of active user in characteristic pattern and present node
Neighbors;
Subelement is updated, for according to the label value and present node of adjacent node in characteristic pattern and the side of adjacent node
The label value of weight update present node;
Subelement is returned to, for returning to label value and present node and adjacent node according to adjacent node in characteristic pattern
Side right the step of updating the label value of present node again obtain corresponding second shadow of active user until meeting the condition of convergence
Ring label value.
In one embodiment, target labels point of reference computing module 806 includes:
Object reference user gathers acquiring unit, for obtaining active user from the second child user set, according to current
The screening from user's set of the social incidence relation of user obtains the corresponding object reference user set of active user;
Second target labels point of reference computing unit, for the feature phase according to active user and each object reference user
The second target labels point of reference of active user Yu object reference user are calculated like degree;
Characteristic pattern building subelement is used for:
Using user gather in each user as node, using the corresponding node of object reference user as present node
Weight of the second target labels point of reference of adjacent node, active user and object reference user as side, building obtain feature
Figure.
In one embodiment, object reference user gathers acquiring unit and is used for:
Active user corresponding first is obtained from user's set according to the social incidence relation of active user to be directly linked
User's set;
It is sieved from first direct correlation user's set according to the characteristic similarity of active user and the first direct correlation user
Choosing obtains first with reference to user;
First, which is obtained, from user's set according to the social incidence relation of the first reference user refers to user corresponding second
It is directly linked user's set;
User's set is directly linked from second with reference to the characteristic similarity that user is directly linked user with second according to first
Middle screening obtains second with reference to user;
Using first with reference to user and the second reference user as the object reference user in object reference user set.
In one embodiment, the second target labels point of reference computing unit is used for:
The characteristic similarity of active user and each object reference user are counted, characteristic similarity statistics knot is obtained
Fruit;
The current signature similarity for obtaining active user and object reference user, according to current signature similarity and feature phase
The second target labels point of reference of active user Yu object reference user are obtained like the ratio of degree statistical result.
In one embodiment, the first influence label value computing unit 808A is used for:
Using user gather in each user as node, side right weight of the first object Tag reference degree as side obtains
Social associated diagram;
Obtain the adjacent node of the corresponding present node of active user and present node in social associated diagram;
It is updated and is worked as again according to the side right of the label value of adjacent node and present node and adjacent node in social associated diagram
The label value of front nodal point;
It returns according to the side right of the label value of adjacent node and present node and adjacent node weight in social associated diagram more
The step of label value of new present node, until meeting the condition of convergence, obtaining active user corresponding first influences label value.
In one embodiment, target labels value computing unit 808C is used for:
Obtaining first influences corresponding first weight of label value and corresponding second weight of the second influence label value;
Label value and corresponding first weight, the second influence label value and corresponding second weight are influenced according to first
It is weighted summation, obtains the corresponding target labels value of second user.
In one embodiment, initial labels value determining module 804 includes:
Target labels classification acquiring unit, for obtaining the corresponding target labels classification of each first user;
First initial labels vector obtains unit, for obtaining according to the corresponding target labels classification of each first user
The corresponding initial labels vector of one user;
Second initial labels vector obtains unit, for obtaining default label vector as in the second child user set
The corresponding initial labels vector of two users, the value for presetting label vector are consistent.
In one embodiment, the first initial labels vector obtains unit and is used for:
Each target labels classification corresponding for the first user obtains corresponding first preset value as corresponding vector
Value, each non-targeted label classification corresponding for the first user obtain corresponding second preset value as corresponding vector value,
First preset value is different from the second preset value;
By the corresponding vector value of each target labels classification and the corresponding vector value composition of each non-targeted label classification
The corresponding initial labels vector of first user.
Figure 10 shows the internal structure chart of computer equipment in one embodiment.The computer equipment specifically can be figure
Server 120 in 1.As shown in Figure 10, it includes the place connected by system bus which, which includes the computer equipment,
Manage device, memory and network interface.Wherein, memory includes non-volatile memory medium and built-in storage.The computer equipment
Non-volatile memory medium be stored with operating system, can also be stored with computer program, which is held by processor
When row, processor may make to realize that user tag determines method.Computer program can also be stored in the built-in storage, the calculating
When machine program is executed by processor, processor may make to execute user tag and determine method.
It will be understood by those skilled in the art that structure shown in Figure 10, only part relevant to application scheme
The block diagram of structure, does not constitute the restriction for the computer equipment being applied thereon to application scheme, and specific computer is set
Standby may include perhaps combining certain components or with different component layouts than more or fewer components as shown in the figure.
In one embodiment, user tag determining device provided by the present application can be implemented as a kind of computer program
Form, computer program can be run in computer equipment as shown in Figure 10.Group can be stored in the memory of computer equipment
At each program module of the user tag determining device, for example, user shown in Fig. 8, which gathers, obtains module 802, initial labels
It is worth determining module 804, target labels point of reference computing module 806, target labels value computing module 808 and target labels to determine
Module 810.It is each that the computer program that each program module is constituted makes processor execute the application described in this specification
The user tag of embodiment determines the step in method.
For example, computer equipment shown in Fig. 10 can pass through the user in user tag determining device as shown in Figure 8
Set obtain module 802 in obtain user set, user set include have determined that label the first child user set and not really
Calibrate the second child user set of label;By initial labels value determining module 804 according to user gather in each user it is corresponding
Label determines the corresponding initial labels value of each user;In being gathered by target labels point of reference computing module 806 according to user
The target labels point of reference between user is calculated in the social degree of association and characteristic similarity between user;Pass through target mark
Label value computing module 808 is for the active user in the second child user set, according to each use in active user and user's set
The corresponding initial labels value of each user in corresponding target labels point of reference and user's set, is calculated current between family
The corresponding target labels value of user;Determine that active user is corresponding according to target labels value by target labels determining module 810
Target labels.
In one embodiment, a kind of computer equipment, including memory and processor are provided, memory is stored with meter
Calculation machine program, when computer program is executed by processor, so that processor executes the step of above-mentioned user tag determines method.This
The user tag that the step of place's user tag determines method can be above-mentioned each embodiment determines the step in method.
In one embodiment, a kind of computer readable storage medium is provided, computer program, computer journey are stored with
When sequence is executed by processor, so that processor executes the step of above-mentioned user tag determines method.User tag determination side herein
The user tag that the step of method can be above-mentioned each embodiment determines the step in method.
Although should be understood that various embodiments of the present invention flow chart in each step according to arrow instruction successively
It has been shown that, but these steps are not that the inevitable sequence according to arrow instruction successively executes.Unless expressly state otherwise herein,
There is no stringent sequences to limit for the execution of these steps, these steps can execute in other order.Moreover, each embodiment
In at least part step may include that perhaps these sub-steps of multiple stages or stage are not necessarily multiple sub-steps
Completion is executed in synchronization, but can be executed at different times, the execution in these sub-steps or stage sequence is not yet
Necessarily successively carry out, but can be at least part of the sub-step or stage of other steps or other steps in turn
Or it alternately executes.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, computer program can be stored in a non-volatile computer and can be read
In storage medium, the computer program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, the application
To any reference of memory, storage, database or other media used in provided each embodiment, may each comprise non-
Volatibility and/or volatile memory.Nonvolatile memory may include read-only memory (ROM), programming ROM (PROM),
Electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include arbitrary access
Memory (RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms, such as static
RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate sdram (DDRSDRAM), enhanced SDRAM
(ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) directly RAM (RDRAM), straight
Connect memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM) etc..
Each technical characteristic of embodiment described above can be combined arbitrarily, for simplicity of description, not to above-mentioned reality
It applies all possible combination of each technical characteristic in example to be all described, as long as however, the combination of these technical characteristics is not deposited
In contradiction, all should be considered as described in this specification.
The embodiments described above only express several embodiments of the present invention, and the description thereof is more specific and detailed, but simultaneously
Limitations on the scope of the patent of the present invention therefore cannot be interpreted as.It should be pointed out that for those of ordinary skill in the art
For, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to guarantor of the invention
Protect range.Therefore, the scope of protection of the patent of the invention shall be subject to the appended claims.
Claims (15)
1. a kind of user tag determines method, which comprises
Obtain user's set, user set includes having determined that the first child user set of label and not determining the of label
Two child user set;
The corresponding initial labels value of each user is determined according to the corresponding label of user each in user set;
According in user set between user the social degree of association and characteristic similarity the mesh between user is calculated
Mark Tag reference degree;
For the active user in the second child user set, according to the user in the active user and user set
Between target labels point of reference and the user set in the corresponding initial labels value of user, obtain the active user
Corresponding target labels value;
The corresponding target labels of the active user are determined according to the target labels value.
2. the method according to claim 1, wherein the target labels point of reference includes first object label ginseng
Degree of examining and the second target labels point of reference, the social degree of association and feature according in user set between user
Similarity calculation obtains the target labels point of reference between user
The first object Tag reference between user is obtained according to the social calculation of relationship degree in user set between user
The reference of the second target labels between user is calculated according to the characteristic similarity in user set between user in degree
Degree;
It is described for being gathered according to the active user and the user for the active user in the second child user set
In user between target labels point of reference and user set in the corresponding initial labels value of user, obtain described
The corresponding target labels value of active user includes:
For the active user in the second child user set, according to the user in the active user and user set
Between the corresponding initial labels value of user in corresponding first object Tag reference degree and user set, obtain described
Active user corresponding first influences label value;
For the active user in the second child user set, according to the user in the active user and user set
Between the corresponding initial labels value of user in corresponding second target labels point of reference and user set, obtain described
Active user corresponding second influences label value;
The active user couple is determined according to the corresponding first influence label value of the active user and the second influence label value
The target labels value answered.
3. according to the method described in claim 2, it is characterized in that, described according in the active user and user set
User between corresponding second target labels point of reference and the user set in the corresponding initial labels value of user, meter
Calculation obtains the corresponding second influence label value of the active user
Using each user in user set as node, side right weight of the second target labels point of reference as side,
Building obtains characteristic pattern;
Obtain the adjacent node of the corresponding present node of active user described in the characteristic pattern and the present node;
Using the corresponding initial labels value of each user as the current label value of corresponding node in the characteristic pattern, according to
The current label value of adjacent node described in the characteristic pattern and the side right of the present node and adjacent node weight are more
The current label value of the new present node;
The current label value of return adjacent node according to the characteristic pattern and the present node and the adjacent segments
The step of side right of point updates the current label value of the present node again obtains the current use until meeting the condition of convergence
Family corresponding second influences label value.
4. according to the method described in claim 3, it is characterized in that, the feature according in user set between user
Similarity calculation obtains the second target labels point of reference between user
Active user is obtained from the second child user set, according to the social incidence relation of the active user from the use
Screening obtains the corresponding object reference user set of the active user in the set of family;
According to the characteristic similarity of the active user and each object reference user be calculated the active user with it is described
The second target labels point of reference of object reference user;
Each user using in user set is as node, power of the second target labels point of reference as side
Weight, building obtain characteristic pattern and include:
Using each user in user set as node, the corresponding node of the object reference user is worked as described in
Power of the second target labels point of reference of the adjacent node of front nodal point, the active user and the object reference user as side
Weight, building obtain characteristic pattern.
5. according to the method described in claim 4, it is characterized in that, the social incidence relation according to the active user from
Screening obtains the corresponding object reference user set of the active user and includes: in user set
The active user corresponding first is obtained from user set according to the social incidence relation of the active user
It is directly linked user's set;
User's set is directly linked from described first according to the characteristic similarity that the active user is directly linked user with first
Middle screening obtains first with reference to user;
Described first, which is obtained, from user set according to the social incidence relation of the first reference user refers to user couple
Second answered is directly linked user's set;
User is directly linked from described second with reference to the characteristic similarity that user is directly linked user with second according to described first
Screening obtains second with reference to user in set;
Described first is joined with reference to user as the target in object reference user set with reference to user and described second
Examine user.
6. according to the method described in claim 4, it is characterized in that, described join according to the active user with each target
The second target labels point of reference of the active user Yu the object reference user is calculated in the characteristic similarity for examining user
Include:
The characteristic similarity of the active user and each object reference user are counted, characteristic similarity system is obtained
Count result;
The current signature similarity for obtaining the active user Yu the object reference user, according to the current signature similarity
The second target mark of the active user Yu the object reference user are obtained with the ratio of the characteristic similarity statistical result
Sign point of reference.
7. according to the method described in claim 2, it is characterized in that, described according in the active user and user set
User between corresponding first object Tag reference degree and the user set in the corresponding initial labels value of user, meter
Calculation obtains the corresponding first influence label value of the active user
Using each user in user set as node, side right weight of the first object Tag reference degree as side,
Obtain social associated diagram;
Obtain the adjacent node of the corresponding present node of active user described in the social associated diagram and the present node;
According to the label value of adjacent node described in the social associated diagram and the present node and the adjacent node
Side right updates the label value of the present node again;
The label value and the present node of return adjacent node according to the social associated diagram and the adjacent segments
The step of side right of point updates the label value of the present node again obtains the active user couple until meeting the condition of convergence
First answered influences label value.
8. according to the method described in claim 2, it is characterized in that, described influence to mark according to the active user corresponding first
Label value and the second influence label value determine that the corresponding target labels value of the active user includes:
Obtaining described first influences corresponding first weight of label value and corresponding second weight of the second influence label value;
Label value and corresponding first weight, the second influence label value and corresponding second are influenced according to described first
Weight is weighted summation, obtains the corresponding target labels value of the second user.
9. the method according to claim 1, wherein described corresponding according to each user in user set
Label determines that the corresponding initial labels value of each user includes:
Obtain the corresponding target labels classification of each first user in the first child user set;
The corresponding initial labels vector of first user is obtained according to the corresponding target labels classification of each first user;
The default label vector of acquisition is described as the corresponding initial labels vector of second user in the second child user set
The value of default label vector is consistent.
10. according to the method described in claim 9, it is characterized in that, described according to the corresponding target of each first user
Label classification obtains the corresponding initial labels vector of first user
Each target labels classification corresponding for first user obtains corresponding first preset value as corresponding vector
Value, each non-targeted label classification corresponding for first user, obtain corresponding second preset value as it is corresponding to
Magnitude, first preset value are different from second preset value;
By the corresponding vector value of each target labels classification and the corresponding vector value of each non-targeted label classification
Form the corresponding initial labels vector of first user.
11. a kind of user tag determining device, described device include:
User, which gathers, obtains module, and for obtaining user's set, user's set includes the first child user for having determined that label
Set and the second child user set for not determining label;
Initial labels value determining module, for determining each use according to the corresponding label of user each in user set
The corresponding initial labels value in family;
Target labels point of reference computing module, for according to the social degree of association and feature in user set between user
Similarity calculation obtains the target labels point of reference between user;
Target labels value computing module, for for the active user in the second child user set, according to the current use
The user in target labels point of reference and user set between family and the user in user set is corresponding just
Beginning label value obtains the corresponding target labels value of the active user;
Target labels determining module, for determining the corresponding target labels of the active user according to the target labels value.
12. device according to claim 11, which is characterized in that the target labels point of reference includes first object label
Point of reference and the second target labels point of reference, the target labels point of reference computing module are used for:
The first object Tag reference between user is obtained according to the social calculation of relationship degree in user set between user
The reference of the second target labels between user is calculated according to the characteristic similarity in user set between user in degree
Degree;
The target labels value computing module includes:
First influences label value computing unit, for being worked as according to described for the active user in the second child user set
During corresponding first object Tag reference degree and the user gather between user in preceding user and user set
The corresponding initial labels value of user, obtaining the active user corresponding first influences label value;
Second influences label value computing unit, for being worked as according to described for the active user in the second child user set
During corresponding second target labels point of reference and the user gather between user in preceding user and user set
The corresponding initial labels value of user, obtaining the active user corresponding second influences label value;
Target labels value computing unit, for influencing label value and the second influence mark according to the active user corresponding first
Label value determines the corresponding target labels value of the active user.
13. device according to claim 12, which is characterized in that described second, which influences label value computing unit, includes:
Characteristic pattern constructs subelement, for using each user in user set as node, second target labels
Side right weight of the point of reference as side, building obtain characteristic pattern;
Node obtains subelement, for obtaining the corresponding present node of active user described in the characteristic pattern and described current
The adjacent node of node;
Subelement is updated, for using the corresponding initial labels value of each user as corresponding node in the characteristic pattern
Current label value, according to the current label value of adjacent node described in the characteristic pattern and the present node with it is described adjacent
The side right of node updates the current label value of the present node again;
Subelement is returned to, for returning to the current label value of the adjacent node according to the characteristic pattern and described working as prosthomere
The step of side right of point and the adjacent node updates the current label value of the present node again, until meet the condition of convergence,
Obtaining the active user corresponding second influences label value.
14. a kind of computer equipment, which is characterized in that including memory and processor, be stored with computer in the memory
Program, when the computer program is executed by the processor, so that the processor perform claim requires any one of 1 to 10
User tag described in claim determines the step of method.
15. a kind of computer readable storage medium, which is characterized in that be stored with computer on the computer readable storage medium
Program, when the computer program is executed by processor, so that the processor perform claim requires any one of 1 to 10 right
It is required that the step of user tag determines method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811503599.5A CN110162692B (en) | 2018-12-10 | 2018-12-10 | User label determination method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811503599.5A CN110162692B (en) | 2018-12-10 | 2018-12-10 | User label determination method and device, computer equipment and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110162692A true CN110162692A (en) | 2019-08-23 |
CN110162692B CN110162692B (en) | 2021-05-25 |
Family
ID=67645265
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811503599.5A Active CN110162692B (en) | 2018-12-10 | 2018-12-10 | User label determination method and device, computer equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110162692B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111339425A (en) * | 2020-03-05 | 2020-06-26 | 拉扎斯网络科技(上海)有限公司 | Object marking method, device, server and storage medium |
CN111552799A (en) * | 2020-04-30 | 2020-08-18 | 腾讯科技(深圳)有限公司 | Information processing method, information processing device, electronic equipment and storage medium |
CN112685626A (en) * | 2019-10-17 | 2021-04-20 | 北京星选科技有限公司 | Method and device for determining object type, electronic equipment and storage medium |
CN113159927A (en) * | 2021-04-30 | 2021-07-23 | 中国银行股份有限公司 | Method and device for determining client label |
CN114596097A (en) * | 2022-05-10 | 2022-06-07 | 富算科技(上海)有限公司 | User identification method, device, electronic equipment and computer readable storage medium |
US12020267B2 (en) | 2020-02-07 | 2024-06-25 | Tencent Technology (Shenzhen) Company Limited | Method, apparatus, storage medium, and device for generating user profile |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105069129A (en) * | 2015-06-24 | 2015-11-18 | 合肥工业大学 | Self-adaptive multi-label prediction method |
CN107220852A (en) * | 2017-05-26 | 2017-09-29 | 北京小度信息科技有限公司 | Method, device and server for determining target recommended user |
CN107368499A (en) * | 2016-05-12 | 2017-11-21 | 中国移动通信集团广东有限公司 | A kind of client's tag modeling and recommendation method and device |
CN107368534A (en) * | 2017-06-21 | 2017-11-21 | 南京邮电大学 | A kind of method for predicting social network user attribute |
WO2018077039A1 (en) * | 2016-10-27 | 2018-05-03 | 腾讯科技(深圳)有限公司 | Community discovery method, apparatus, server, and computer storage medium |
CN108307318A (en) * | 2017-12-21 | 2018-07-20 | 北京三快在线科技有限公司 | A kind of user information acquiring method and device, electronic equipment |
CN108921189A (en) * | 2018-05-23 | 2018-11-30 | 北京航空航天大学 | For the deduction method and device of social network user attribute |
-
2018
- 2018-12-10 CN CN201811503599.5A patent/CN110162692B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105069129A (en) * | 2015-06-24 | 2015-11-18 | 合肥工业大学 | Self-adaptive multi-label prediction method |
CN107368499A (en) * | 2016-05-12 | 2017-11-21 | 中国移动通信集团广东有限公司 | A kind of client's tag modeling and recommendation method and device |
WO2018077039A1 (en) * | 2016-10-27 | 2018-05-03 | 腾讯科技(深圳)有限公司 | Community discovery method, apparatus, server, and computer storage medium |
CN107220852A (en) * | 2017-05-26 | 2017-09-29 | 北京小度信息科技有限公司 | Method, device and server for determining target recommended user |
CN107368534A (en) * | 2017-06-21 | 2017-11-21 | 南京邮电大学 | A kind of method for predicting social network user attribute |
CN108307318A (en) * | 2017-12-21 | 2018-07-20 | 北京三快在线科技有限公司 | A kind of user information acquiring method and device, electronic equipment |
CN108921189A (en) * | 2018-05-23 | 2018-11-30 | 北京航空航天大学 | For the deduction method and device of social network user attribute |
Non-Patent Citations (1)
Title |
---|
马超: "基于主题模型的社交网络用户画像分析方法", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112685626A (en) * | 2019-10-17 | 2021-04-20 | 北京星选科技有限公司 | Method and device for determining object type, electronic equipment and storage medium |
US12020267B2 (en) | 2020-02-07 | 2024-06-25 | Tencent Technology (Shenzhen) Company Limited | Method, apparatus, storage medium, and device for generating user profile |
CN111339425A (en) * | 2020-03-05 | 2020-06-26 | 拉扎斯网络科技(上海)有限公司 | Object marking method, device, server and storage medium |
CN111339425B (en) * | 2020-03-05 | 2021-07-23 | 拉扎斯网络科技(上海)有限公司 | Object marking method, device, server and storage medium |
CN111552799A (en) * | 2020-04-30 | 2020-08-18 | 腾讯科技(深圳)有限公司 | Information processing method, information processing device, electronic equipment and storage medium |
CN111552799B (en) * | 2020-04-30 | 2021-07-09 | 腾讯科技(深圳)有限公司 | Information processing method, information processing device, electronic equipment and storage medium |
CN113159927A (en) * | 2021-04-30 | 2021-07-23 | 中国银行股份有限公司 | Method and device for determining client label |
CN114596097A (en) * | 2022-05-10 | 2022-06-07 | 富算科技(上海)有限公司 | User identification method, device, electronic equipment and computer readable storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN110162692B (en) | 2021-05-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110162692A (en) | User tag determines method, apparatus, computer equipment and storage medium | |
US11659050B2 (en) | Discovering signature of electronic social networks | |
CN111309824B (en) | Entity relationship graph display method and system | |
WO2018077039A1 (en) | Community discovery method, apparatus, server, and computer storage medium | |
CN111160954B (en) | Recommendation method facing group object based on graph convolution network model | |
CN109906451A (en) | Use the similarity searching of polyphone | |
Topirceanu et al. | Weighted betweenness preferential attachment: A new mechanism explaining social network formation and evolution | |
CN111177473B (en) | Personnel relationship analysis method, device and readable storage medium | |
CN111444395A (en) | Method, system and equipment for obtaining relation expression between entities and advertisement recalling system | |
CN113255895B (en) | Structure diagram alignment method and multi-diagram joint data mining method based on diagram neural network representation learning | |
CN107003834B (en) | Pedestrian detection device and method | |
CN107895038A (en) | A kind of link prediction relation recommends method and device | |
CN107729519A (en) | Appraisal procedure and device, terminal based on multi-source multidimensional data | |
CN110232638A (en) | A kind of overlapping community discovery method based on node importance with local expansion | |
CN113821827B (en) | Combined modeling method and device for protecting multiparty data privacy | |
CN107784511A (en) | A kind of customer loss Forecasting Methodology and device | |
CN104484365B (en) | In a kind of multi-source heterogeneous online community network between network principal social relationships Forecasting Methodology and system | |
CN103559318B (en) | The method that the object containing heterogeneous information network packet is ranked up | |
Mahyar et al. | CS-ComDet: A compressive sensing approach for inter-community detection in social networks | |
CN110442800A (en) | A kind of semi-supervised community discovery method of aggregators attribute and graph structure | |
CN108470251A (en) | Community based on Average Mutual divides quality evaluating method and system | |
CN111143700B (en) | Activity recommendation method, activity recommendation device, server and computer storage medium | |
CN110457387A (en) | A kind of method and relevant apparatus determining applied to user tag in network | |
CN113395172B (en) | Important user discovery and behavior prediction method based on communication network | |
CN115345635A (en) | Processing method and device for recommended content, computer equipment and storage medium |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |