CN109190033A - A kind of user's friend recommendation method and system - Google Patents
A kind of user's friend recommendation method and system Download PDFInfo
- Publication number
- CN109190033A CN109190033A CN201810967953.3A CN201810967953A CN109190033A CN 109190033 A CN109190033 A CN 109190033A CN 201810967953 A CN201810967953 A CN 201810967953A CN 109190033 A CN109190033 A CN 109190033A
- Authority
- CN
- China
- Prior art keywords
- user
- bridge
- relationship
- target user
- target
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Abstract
A kind of recommended method and system of user good friend, the method includes the two grade network relationship of target user and recommended candidate collection user are constructed by user behavior, relationship between integration objective user and recommended candidate collection user, according to user's needs, user's set that target user meets its interest preference is recommended.The dynamic adjustment that the time of the act factor carries out Generalization bounds is introduced, two degree of relationships based on user establish the friend recommendation method based on user interest using intermediate node as the bridge recommended.So that the intimate degree between each user has time dimension, so that evaluation data are truer, also can be more objective and close relationship between each user be really reacted, increase the accuracy of recommendation results, to recommend the interested user of its most probable for user, and finally use experimental verification, it was demonstrated that the accuracy and validity of recommendation results.
Description
Technical field
Present document relates to technical field of data processing, and in particular to a kind of user's friend recommendation method and system.
Background technique
Microblogging is maximum as current user base number, most rapid social network-i i-platform of growing up, and microblogging recommendation has become it
One of the core function for improving user experience and meeting user demand.And add good friend as the key activities in microblog,
How by the application to microblogging recommendation function, the recommendation material of user interest demand will be met, pass through most appropriate strategy side
Formula, and most suitable opportunity is selected to be recommended, help target user to reduce experience cost, and searching out for efficient quick is most felt
The problem of information of interest has become urgent need to resolve.It is investigated and summarizes by the recommendation function to microblog, such as Fig. 1
Shown, microblog co-exists in following 5 kinds of ways of recommendation at present: (1) recommendation based on attribute information, (2) are based on influence power
Recommend, (3) recommendation based on social relations, (4) recommendation, (5) packet-based recommendation based on content information, in which:
(1) will be had based on the recommendation of attribute information according to registration information of the user on microblogging, such as company, school
Same alike result and be located at same community network in recommended candidate recommend target user.Usually with " XX in microblog
The forms such as university alumnus " or " you come from XX company " show.
(2) based on the recommendation of influence power, microblog users interest demand performance sometimes is indefinite, and is based on each user couple
Popular message all has curious psychology, by the high and popular Candidate Recommendation of influence power in one period to target user.It pushes away
It recommends and shows reason for the forms such as " heat was searched in 24 hours ", " man of the hour ".
(3) recommendation based on social relations will be closed according to the complex network for the behavior building that microblog users generate with user
System is most close, the interaction highest candidate resource of the frequency recommends target user.This method recommends to show that reason is that " good friend's is good
The forms such as friend ", " concern of concern ".Wherein, the thought that the former expresses is user to letter interested to good friend in close relations
Breath may also be interested, and the strategy that " concern of concern " is conveyed is user to its interested key node interest subjected
The content of circle can also generate certain interest.And both modes are also exactly the embodiment of customer relationship type in social networks.
(4) blog article of user is analyzed by the technological means of word frequency statistics and semantic analysis based on the recommendation of content information
Content, carries out keyword extraction, and the interest tags possessed in conjunction with user itself, and the vector that building represents user interest indicates
Model, with label mapping classification system, the candidate shape with "-XX classification of having similar tastes and interests " having similar tastes and interests with target user
Formula recommends target user.
(5) a certain famous person is disclosed the friend information of recommendation according to good friend's grouping information of user by packet-based recommendation
Targetedly recommended in the field group belonging to it.Its thought to be expressed is user to its interested user
Think that the valuable and worth information shared similarly can be interested.It is consulted and is browsed in grouping, more there is needle
To property, and successfully improve user experience.
Although having their own characteristics each in above five kinds of recommended methods, all there is certain shortcoming.Although first two method
It can be good at solving the problems, such as there is cold start-up in recommender system, but user divides too coarseness, lacks apparent needle
To property;The recommendation carried out based on interest only extracts keyword progress interest tags from the blog article content and label information of user
Match, not from the relationship between the user that taps the latent power in dynamic social networks, effect is relatively simple.
Summary of the invention
The present invention provides a kind of user's friend recommendation method and system, and in order to solve the above-mentioned technical problem, the present invention adopts
Technical solution design are as follows:
A kind of recommended method of user good friend, which comprises target user is constructed by user behavior and recommends to wait
The two grade network relationship of the two grade network relationship of selected works user, the target user and recommended candidate collection user include target user
With the primary network station relationship of the primary network station relationship of bridge user and bridge user and recommended candidate collection user;
The target user and the bridge are obtained according to the primary network station relationship of the target user and the bridge user
The cohesion relation value of beam user, and obtained according to the primary network station relationship of the bridge user and the recommended candidate collection user
The similarity relation value of the bridge user and the recommended candidate collection user;
It is pushed away with the cohesion relation value of the bridge user and the bridge user with described according to the target user
The similarity relation value for recommending Candidate Set user obtains the recommendation of the recommended candidate collection user;
According to the recommendation of recommended candidate collection user, selection meets user's set of the interest preference of the target user.
The primary network station relationship according to target user and the bridge user obtains the target user and the bridge
The cohesion relation value of beam user includes:
The target user and the bridge user are obtained according to the cohesion of the target user and the bridge user
Cohesion relation value;
It is mutual with the bridge user that the cohesion of the target user and the bridge user include the target user
The time of the act relationship of fatigue resistance and the target user and the bridge user.
The mutual fatigue resistance of the target user and the bridge user include in set time period the target user with
Interaction quantity and mutual dynamic frequency between the bridge user.
The time of the act relationship of the target user and the bridge user are by including that the function of the time of the act factor is determined
It is fixed.
The similarity relation value of the bridge user and the recommended candidate collection user are pushed away by the bridge user with described
The common friend quantity for recommending Candidate Set user is obtained as measurement index.
The system comprises:
Data capture unit, the two grade network for constructing target user and recommended candidate collection user by user behavior close
The two grade network relationship of system, the target user and recommended candidate collection user include the primary network station of target user Yu bridge user
Relationship and the primary network station relationship of bridge user and recommended candidate collection user;
Data analysis unit, the two grade network relationship for being constructed according to data capture unit obtain each recommended candidate collection
The recommendation of user;It specifically includes:
First data analysis module, for being obtained according to the primary network station relationship of the target user and the bridge user
The cohesion relation value of the target user and the bridge user;
Second data analysis module, for being closed according to the primary network station of the bridge user and the recommended candidate collection user
System obtains the similarity relation value of the bridge user Yu the recommended candidate collection user;Third data analysis module is used for root
The similarity that the cohesion relation value and second data analysis module obtained according to first data analysis module obtains
Relation value obtains the recommendation of the recommended candidate collection user;
Data feedback unit, the recommended candidate collection user's for being obtained according to the third data analysis module pushes away
Value is recommended, selection meets user's set of the interest preference of the target user.
First data analysis module, specifically for being obtained according to the cohesion of the target user and the bridge user
To the cohesion relation value of the target user and the bridge user;
It is mutual with the bridge user that the cohesion of the target user and the bridge user include the target user
The time of the act relationship of fatigue resistance and the target user and the bridge user.
When the mutual fatigue resistance of target user described in first data analysis module and the bridge user include setting
Between interact quantity and mutual dynamic frequency between the target user and the bridge user in the period.
The time of the act relationship of target user described in first data analysis module and the bridge user by comprising
There is the function of the time of the act factor to determine.
The similarity relation value of bridge user described in second data analysis module and the recommended candidate collection user
It is obtained by the common friend quantity of the bridge user and the recommended candidate collection user as measurement index.
After adopting the above technical scheme, compared with the prior art, the invention has the following beneficial effects:
Firstly, this paper recommended method is from user behavior, by bridge user, determined from recommended candidate collection user with
Destination node user similarity, cohesion reach user's set of certain magnitude;Secondly, user behavior time factor is added herein
As evaluation goal node users and bridge user cohesion relationship factor of evaluation, in conjunction with the factor at any time of the behavior between user
Changing rule and mutual fatigue resistance the characteristics of, the comprehensive cohesion formula obtained between destination node user and bridge user,
So that the intimate degree between each user has time dimension, so that evaluation data are truer, it also can be more objective and true
Close relationship between real each user of reaction, increases the accuracy of recommendation results, to recommend its most probable interested for user
User, and finally use experimental verification, it was demonstrated that the accuracy and validity of recommendation results.
Detailed description of the invention
Wherein, symbol description is as follows in figure:
Fig. 1 is the recommended method summary figure that the present invention is directed to the prior art;
Fig. 2 is that the user of embodiment of the present invention 1 recommends degree relationship network;
Fig. 3 is the Ai Binsi forgetting curve figure of embodiment of the present invention 1;
Fig. 4 is the friend recommendation flow chart based on user interest model of embodiment of the present invention 1;
Fig. 5 is the number that 80 experiment users of embodiment of the present invention 1 pay close attention to the forwarding of user, comment on and refer to it
Statistical chart;
Fig. 6 is that the recommended method performance of the embodiment of the present invention 1 compares figure;
Fig. 7 is 1 method flow diagram of the embodiment of the present invention;
Fig. 8 is a kind of recommender system structural schematic diagram of the embodiment of the present invention 3.
Specific embodiment
For the deficiency of existing microblogging Generalization bounds, and the features such as combine the word of microblogging text few, lack of standardization, herein from society
The interest characteristics for excavating user in relational network are handed over, the dynamic adjustment that time factor carries out Generalization bounds are introduced, based on user's
Two degree of relationships establish the microblogging friend recommendation method and system based on user interest using intermediate node as the bridge recommended.
In order to enable those skilled in the art to better understand the solution of the present invention, below in conjunction in the embodiment of the present invention
Attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is only
The embodiment of a part of the invention, instead of all the embodiments.Based on the embodiments of the present invention, ordinary skill people
The model that the present invention protects all should belong in member's every other embodiment obtained without making creative work
It encloses.
Embodiment one:
First to proprietary vocabulary such as some proprietary vocabulary such as bridge cohesion, user's similarity, user's recommendation of the present invention
It illustrates one by one, then carries out embodiment explanation:
Firstly, bridge cohesion is interacted between bridge and target user and the important finger of relationship strength for measuring
Mark.If it is closer that target user with it pays close attention to the friend relation in list, more interactions can be generated therebetween.
User social contact attribute is represented in conjunction with such as bidirectional strength relationship in micro blog network, and unidirectional weak relationship is emerging more representative of user
The characteristics of interesting attribute, is chosen and forwards, comments between destination node and bridge user, referring to three kinds of movements as between analysis user
The behavioural characteristic of mutual fatigue resistance, defines, the mutual fatigue resistance between user i and its bridge user j is by shown in formula (1) herein:
Wherein, ni, j are the sum that destination node i is interacted with bridge user j and (forwarded, comments on, referring to), and ni, u are represented
Target user interacts total degree with its all bridge user's.As regulatory factor, the strong and weak pass in social networks is taken into consideration
Set type is set when between user being the strong relationship of two-way concernAnd the relationship type of user be unidirectionally pay close attention to it is weak
When relationship, setting
Above method only considered the behavial factor between user as the mutual fatigue resistance between two users of analysis.And it uses
The mutual-action behavior at family is closer apart from current time, more can accurately judge that the mutual dynamic frequency of two users this moment is strong and weak, user is very
The mutual-action behavior generated before long can not represent the interest of the user instantly, but with the continuous variation of time factor, it uses
Mutual fatigue resistance between family is constantly to decay.Attenuation law meets characteristic distributions (Fig. 3 institute of Ai Binsi forgetting curve
Show).
It was found from the figure that forgetting curve has followed the characteristic distributions of power function.Therefore, time factor good friend is introduced into push away
In recommending, the time of the act saturation between destination node and bridge user is set are as follows:
Wherein, θ is the parameter of user behavior factor variations rate at any time, sets θ=1 herein, that is, thinks all
User is identical to the speed of different behavior classifications factor variations at any time, tkRepresent the kth time mutual-action behavior distance of user now
Time, as unit of day, and n represents the total degree that user i interacts (forwarding and comment) with user j.
In conjunction with the changing rule of the factor and the characteristics of mutual fatigue resistance at any time of the behavior between user, synthesis obtains target
Cohesion formula between node i and bridge user j is such as shown in (3):
About the similarity 702 of user, as its name suggests, user's similarity 702 is for measuring between two user characteristics
The index of similar degree.If user i has paid close attention to user j, illustrate that i is interested in interest worlds belonging to user j.And to bridge
User j similarity degree is high and j also interested user y, can equally be full of interest, and then generate the desire paid attention in.User
Number of fans be exactly to be used to measure the most intuitive index of two user's similarity degrees.Therefore, bridge user j is defined herein and is recommended
Shown in similarity formula (4) between candidate y:
Wherein, common_cntj, y are the common number of fans between user j and user y, for measuring between two users
Similarity degree.
About user's recommendation, user's cohesion and the phase between microblog users different levels are above respectively defined
Like the calculation method of degree.And how target user will be recommended with the candidate resource of two degree of relationships with user, and by intimate
It spends highest bridge user to be shown, to attract user's to pay attention in the research emphasis that desire is this patent.
As shown in Fig. 2, user node U0 is collected by the behaviors such as concern, forwarding or comment and oneself interested user
U={ U1, U2 ... Ui } constructs social networks, to obtain corresponding information and meet affection need.By this and mesh
There is mark node the user of direct relation to be known as bridge user.And the use of interest to its interested good friend based on target user
Family, constructs the degree relationship network of target user, so that recommended candidate resource is passed through bridge at the characteristics of also generating certain interest
Beam user is that target user is targetedly recommended, and mainly includes following techniqueflow:
The shared w bridge user of setting destination node i and recommended candidate y (w1, w2 ..., wj), in conjunction with the parent between user
The calculation method of density and similarity defines the recommendation score that candidate user y is recommended to target user i are as follows:
Wherein, Xi, j be destination node and bridge user cohesion score, and Simj, y for bridge user with finally push away
The similarity score between candidate is recommended, in summary factor, has finally obtained the recommendation size of candidate resource.
A kind of method of microblog users friend recommendation, the method includes constructing target user by microblogging behavior and recommend
The two grade network relationship of Candidate Set user integrates the relationship between each user, recommends target user and meet its interest preference
User's set, such as TOP-N user's set is chosen according to recommendation size.
Step 702, the two grade network relationship of target user and recommended candidate collection user are constructed by user behavior, comprising:
Primary network station relationship between target user and bridge user and bridge user and recommended candidate collection user.
Step 704, the two grade network relationship that target user and recommended candidate collection user are constructed by user behavior, it is comprehensive
Close the relationship between each user, comprising: the cohesion relationship of the comprehensive target user and the bridge user are (referred to as intimate
Degree) and integrate the similarity relationship (abbreviation similarity) of the bridge user and the recommended candidate collection user and integrate institute
State the cohesion relationship of target user and the bridge user, the similarity of the bridge user and the recommended candidate collection user
Relationship between relationship.
The cohesion relationship of the destination node user and the bridge user includes:
The destination node and the Bridge Joints interact and relationship strength, including the destination node and the bridge
The mutual fatigue resistance and destination node of girder connection and the time of the act relationship of the Bridge Joints.
The mutual fatigue resistance of the destination node and the Bridge Joints include the destination node and the bridge user it
Between microblogging behavior act show the factor;The microblogging behavior act performance factor includes that the destination node and the bridge are used
Mutual dynamic frequency between interaction quantity, the destination node and the bridge user between family.
The similarity relationship for stating the bridge user and the recommended candidate collection user include: the bridge user with it is described
The similarity relationship of recommended candidate collection user using the common number of fans of the bridge user and the recommended candidate collection user as
Measurement index.
The recommendation of step 706, step 708, the recommended candidate collection user is used by the target user and the bridge
The cohesion score that the cohesion relationship at family obtains i.e. cohesion relation value and the bridge user and the recommended candidate collection
The similarity score of user, that is, similarity relation value codetermines;
The cohesion score that the cohesion relationship according to the destination node user and the bridge user obtains with
And calculated the recommendations time of similarity score that the similarity relationship of the bridge user and the recommended candidate collection user obtain
The recommendation of selected works user, according to user need to target user recommend meet its interest preference user set (such as according to
Recommendation must sort TOP-N user's set), user's friend recommendation is completed.
Embodiment two:
Firstly, obtaining true user information from microblog with crawler software.Raw data set is handled
It is found with after analysis, user can only pay close attention to interested a few users in list with it and generate mutual-action behavior, count destination node
Number is interacted between bridge user, calculates the cohesion score of the two.User according to number of fans in chapter 2 less than 50
It is not easy the characteristics of being paid attention in, user of the number of fans less than 50 in filtered recommendation candidate resource, and with bridge cohesion score
Highest user forms recommended candidate resource collection as displaying reason is recommended.
Secondly, choosing on the basis of analyzing the assessment indicator of existing personalized recommendation method at present and being used to verify the side this paper
The evaluation index of method, i.e. recall rate, precision rate, F1 mean value, and specifically give the expression way of this four indexs.
Finally, the accuracy and validity of contrived experiment method of proof.User in recommended candidate is worth according to recommendation
The height of score gives user to carry out TOP-N displaying, and recommendation results are met at user and are manually evaluated, and judges whether it can be to pushing away
The user recommended generates and pays attention in behavior, and at the same time, strategy on the line of the experimental result of this paper and current microblog is carried out
Comparative analysis, and evaluated by index of correlation;In addition, user is to recommendation in order to verify when user interest changes
As a result acceptance level generates different recommendation list informations, and to recommendation results by the way that four groups of different time factors are arranged
It compares and analyzes.
In order to evaluate the accuracy and recommendation of the personalized friend recommendation algorithm proposed in this paper based on microblog users interest
Effect, and verify when user interest changes, to the acceptance level of recommendation results;By the true use for acquiring microblog
User data collection devises following two experiments herein:
Experiment one: the accuracy of recommendation results is examined.It is verified the candidate resource information to be recommended that context of methods obtains
Whether the interest of user can be met.This experimental setup time factor T=30, i.e. collection target user pay close attention to the use in list with it
Interaction of the family in one month (forwards, comments on, referring to) information, and the relationship type between user is combined to determine regulatory factor
Value, obtain the cohesion between destination node and bridge user.And then using the biggish bridge user of cohesion score as
Recommend to show reason, calculates candidate resource collection to be recommended.It gives the user to carry out artificial division, is classified as " interested "
" loseing interest in " two class, is finally exported the user in recommended candidate, and referred to by model evaluation in the form of TOP-N
Mark is evaluated and is verified to experimental result.
In order to verify the accuracy of this paper result, need to evaluate model using reasonable evaluation index.And it is current
In the application scenarios of many personalized recommendation systems, focus of attention is no longer that system provides score of how accurately testing and assessing, and
It is whether the article that user recommends out to these is interested.And precision of classifying exactly is used to measure recommender system and helps user
Find the capacity of water for the candidate resource that it is really liked.Herein using the common counter such as accuracy rate of classification precision
(precision), recall rate (recall), F1 mean value and Average Accuracy (Average Precision) carry out experiment knot
The evaluation and verifying of fruit.
The formula of accuracy rate is general interested to user for measuring the recommendation results that context of methods obtains as indicated with 6
Rate size:
Wherein, what molecule indicated is the interested user of the really user annotation of the good friend of recommender system recommendation, and divides
What matrix showed is that system actually recommends good friend's number out;
And recall rate expression is user's probability that actually good friend interested is pushed out.It is for measuring recommendation results
No to cover all good friend's numbers for being labeled as " interested " of user enough comprehensively, formula such as (7) is shown:
Wherein, the interested good friend's quantity of user in molecules present recommendation list, and denominator is that user is actually interested
Good friend's quantity.
Coordinate mean value F1 such as formula (8), be the resultant effect index for investigating recall ratio and precision ratio:
Experimental data needed for this paper is all from actual microblog, with the crawler software API open from microblog
Middle acquisition data.By way of breadth first search, first since a certain target user's node U0, his concern column are collected
All users set of table information, i.e. user U0 concern, statistics U0 pays close attention to interacting for each user in list with it and (turns respectively
Send out, comment on, refer to) sum.Then, to the user (U1, U2 ..., Ui ...) of user U0 concern, i.e., as the bridge of rationale for the recommendation
User counts they and its bean vermicelli, that is, the common number of fans of the user with U0 with two degree of relationships, to calculate respectively
The consequently recommended value score of user to be recommended.
Data collection mode in this way can be obtained by two degree of relation data information of destination node user.Herein with
80 users for carrying out user interest analysis collect as target user.First by manually obtaining (click browsing record) and software
The mode of processing obtains the concern list information of target user, as the bridge data in two degree of relationships.Then crawler work is used
Tool obtains the initial data of the mutual-action behavior and concern relation generated between user, and finally export target user forwards, comments
By, refer to the total degree for each bridge user that he pays close attention to and the common number of fans of bridge user and final candidate to be recommended.
Finally, calculating separately out cohesion and bean vermicelli similarity between different levels user using formula (1-4), and pass through formula
(5) the consequently recommended value score of candidate resource is obtained.
Mutual-action behavior information of the experiment user good friend of interest with it in one month is had collected herein.To these users
Summarized (as shown in Figure 5) to the number of generated interaction (forward, comment on, refer to) to find afterwards, about 59.4% user
Between once interact and all do not generate.The number that most of bridge users are forwarded by experiment user, comment on, referring to is less,
Number of users of the interaction number less than or equal to 5 accounts for about overall 74.3%.Only about 7.9% user interaction number is greater than 20 times,
And always interact number caused by this certain customers group and be up to 2546 times, account for about overall 78.6%.Therefore, it is choosing herein
When bridge information, emphasis that will interact the high bridge user of number as displaying reason is recommended, at the same time, filter out experiment and use
Do not have to generate the number of users of mutual-action behavior in the concern list at family.
Above data illustrates user only and can pay close attention to it in list that a small number of interested good friends generate and interact, and with the use
Family shows information as the recommendation of candidate user, so that improves user pays attention in efficiency.And pass is added according to the user of chapter 2
Demand analysis is infused it is found that when the number of fans of user is less than 50, the probability paid attention in by other users is very low.Therefore, it is calculating
When bridge user pays close attention to the bean vermicelli similarity of the user in list with it, use of the number of fans less than 50 in recommended candidate is filtered out
Family.For the ease of following analysis and verifying, when calculating the consequently recommended value score of candidate resource user, score value is chosen
Come the candidate resource collection of first 20 user to be recommended composition target users.
In order to describe conveniently, and the true interest of the consequently recommended candidate resource to target user and user is compared
Compared with, herein to be illustrated for carrying out the experiment user B of user interest analysis in chapter 3, the personalized good friend of other users
Recommended method is similar with its.Wherein, the attention number of user B is 108, and user B is paid close attention in list with it at time factor T=30
44 users generate interaction, according to personalized friend recommendation method shown in this article, the bridge user of experiment user B choose and
It is as follows shown in the calculating step of the recommendation value score of end user:
Firstly, obtaining the bridge user behavior data being directly linked with experiment user, the candidate collection of bridge user is formed;
Including using experiment user B as node, counting its total degree for interacting between the good friend of concern and (forwarding, comment on and refer to);
The user for interacting that number is 0 at time factor T=30 is filtered, the candidate collection of bridge user is formed;
Secondly, calculating the intimate of the bridge user and experiment user being directly linked is worth and carries out cohesion value sequence, including
According to the relationship type between hard objectives node and bridge user, regulatory factor is determinedAnd it is generated according to each mutual-action behavior
When apart from the current time, the cohesion score of destination node and bridge user is calculated according to formula (3), and just according to score value
It is ranked up output;
Then, the behavioral data for the recommended candidate user being directly linked with bridge user is obtained, and is used according to recommended candidate
The similarity at family and bridge user obtains the similar value of recommended candidate user Yu bridge user, including using bridge user as node,
Filter the concern list of each bridge user, i.e., number of users of the number of fans less than 50 in recommended candidate, and calculated according to formula (4)
The similarity score of recommended candidate and bridge user out;
Finally, being worth according to the recommendation that above-mentioned steps scores calculate recommended candidate user, according to user demand, output
Recommendation results.Including statistics, the bridge number of each candidate resource of merging according to above-mentioned steps, and according to formula (5), calculate
The consequently recommended value score of the candidate resource of experiment user B out, and according to user demand, TOP-N exports recommendation results;
The interest model of experiment user B and bridge user, the information of recommended user and score value are summarized, such as table 1
It is shown:
The recommended candidate and its weight of 1 experiment user B of table
By field classification belonging to the recommended candidate of experiment user B and recommend value and the user interest information in chapter 3
Summarized, as shown in table 2:
The comparison of the value score of 2 user interest of table and recommended candidate
Cuisines | Fashion | Joke | Film | Caricature | Travelling | It is economical | |
Interest weight | 0.329 | 0.228 | 0.181 | 0.087 | 0.074 | 0.067 | 0.034 |
Recommend value | 0.282 | 0.243 | 0.163 | 0.122 | 0.085 | 0.054 | 0.051 |
Significant t-test is carried out to the data in table 2, t=0 is obtained, illustrates that the experimental result obtained twice has no significantly
Sex differernce;Correlation analysis is carried out to it, Pearson correlation coefficient r=0.974 illustrates that there are very high correlations for two groups of data
Property.
By being analyzed above it is found that experiment user B is interested in the information of cuisines, fashion, joke and film etc.,
And the bridge selected such as fashion street is clapped, bad joke is selected and the caricature retainer of a big family to be exactly that user B is referred to, forwards and comment on one morning secondary
The most users of number, as recommending to show reason, it is multifarious simultaneously to meet recommendation results, also largely promotees
Into being successfully established for relationship.Moreover, user's fields in candidate resource also cover the interest neck of user B substantially
Domain classification, and by the verifying of index of correlation, both discoveries have very high correlation.Therefore, proposed based on use
The personalized friend recommendation method of family interest has good applicability.
It just tests analysis to the accuracy of experiment recommendation results below:
Although being recommended between candidate user and the interest of experiment user B out by methods herein with very big related
Property, but and target user is not known about to behavior caused by recommendation results.Therefore, the index that text herein above is proposed is used as and comments
Price card is quasi-, and when verifying generates recommendation to target user in the form of TOP-N, whether target user can generate the behavior of paying attention in.
It transfers to the user manually to be marked 20 recommended candidates of experiment user B, is classified as " interested " and " no
It is interested " two classes.According to the score sequence for recommending value, recommend@3 ,@5,15@7 ,@users, statistics to target user every time
The results are shown in Table 3 for each assessment indicator:
The inspection result statistical form of 3 experiment user B of table
Assessment indicator | P@3 | P@5 | P@7 | P@15 |
Recall ratio | 0.143 | 0.286 | 0.357 | 0.429 |
Precision ratio | 0.667 | 0.800 | 0.714 | 0.400 |
F1 | 0.235 | 0.421 | 0.476 | 0.414 |
80 experiment users are done into same processing, finally summarize the average value of each test rating, it is as follows
Shown in table:
The gross examination result statistical form of 4 experiment user of table
Assessment indicator | P@3 | P@5 | P@7 | P@15 |
Recall ratio | 0.169 | 0.273 | 0.336 | 0.462 |
Precision ratio | 0.703 | 0.756 | 0.683 | 0.429 |
F1 | 0.272 | 0.401 | 0.450 | 0.445 |
As can be seen that the continuous increase of good friend's number N with recommendation, precision ratio can reduce from result above, and look into complete
In the continuous improvement, the application of F1 mean value specification can just obtain balance to rate between.As N=7, the numerical value of the index
It is maximum.I.e. every time to user recommend 7 good friends when, the effect reached is best.
In order to compare and analyze, and verify the recommendation effect of context of methods.With the existing 3 kinds of recommendations of current microblog
Method is tested as a control group, and the data decimation process of control group is as follows:
(1) the popular user that before choosing ranking 20 are searched in list from 24 hours heat forms recommended candidate;
(2) it is formed from target user with 20 users are randomly selected in same alike result information (same to company, same to school)
Recommended candidate;
(3) it using the good friend of good friend as point of penetration, i.e., by strong relationship type, is chosen altogether in two degree of relation datas of user
With good friend's number, most preceding 20 users form recommended candidate.
The data set of three of the above Generalization bounds selected manually is marked in the same way, is divided into " interested "
" loseing interest in " two class, and output displaying is carried out to user in the form of TOP N, Fig. 6 shows four kinds of different recommendation plans
Performance comparable situation slightly.
Experiment two:
The recommendation results that verifying analytic explanation above is obtained by context of methods are largely with user's
Interest has correlation.And compare and analyze, discovery context of methods is superior to other classics in each evaluation performance indicator
The effect of recommended method.However, user interest can change with the variation of time.Therefore, to consider in different times
Under Effects of Factors, whether user changes to the acceptance level of the good friend in its recommendation list.
The time window factor is adjusted to T1=5, T2=10, T3=15, T4=30, i.e., count respectively different time because
Number is interacted between target user and bridge user under child window, calculates cohesion score between the two, and then will be close
The similar users of the highest bridge user of density score recommend target user.Herein to experiment user recommend different time because
The candidate resource of the lower TOP7 of son, table 5 show experiment user B to the user information acceptance level in recommendation list.
User B receives situation to the user in recommendation list under the 5 different time factor of table
T1=5 | T2=10 | T3=15 | T4=30 |
Cuisines reviewer (√) | Cuisines reviewer (√) | Fashion entertaining shop (√) | Cuisines reviewer (√) |
Caricature shop (√) | Miss bear (√) | It buys at Beijing (√) | Caricature shop |
100000 bad jokes (√) | It buys at Beijing (√) | The bright bean vermicelli group (√) of Pan Yue | Financial and economic news (√) |
The bright bean vermicelli group (√) of Pan Yue | Anecdote makes laughs (√) | Miss bear (√) | Economic Observer (√) |
Miss bear (√) | The new information of film (√) | Jimmy (hairstylist) | Jimmy (hairstylist) |
Anecdote makes laughs (√) | Financial and economic news | Caricature shop (√) | Humorous huge explosion |
Shanghai Jiang Yu (√) | Good-for-nothing team (√) | Economic Observer | Xie Guozhong |
Note 1): that (√) is indicated is user B to recommending user information out interested
From this table it can be seen that no matter how time factor changes, used when to user's recommendation microblogging related with cuisines
When family, which can all give expression to the desire paid attention in, i.e. the cuisines Long-term Interest that is user.As time factor T4=30,
Recommend microblog users relevant to finance and economics information class to user, which all can be interested in it, and then generates the row paid attention in
For;However as T2=10, T3=15, user but pays attention in microblogging relevant to finance and economic in recommendation list there is no selection and uses
Family.I.e. with the variation of time, the interest of user can also be changed.
In order to examine recommendation effect of the experiment user B in different times under the factor, test to the data in table 5,
And similarly handled remaining experiment user, specific inspection result is as shown in table 6:
The inspection result statistical form of 6 experiment user of table
From this table it can be seen that every test rating of experiment user is constantly reduced with the variation of time factor,
I.e. when the interest of user generates drift, the variation of user interest is accurately captured, is produced according to the instant Characteristic of Interest of user
It is raw targetedly to recommend, while meeting user and paying attention in demand, the pushing efficiency of microblogging is also greatly improved.
Embodiment three:
A kind of user's friend recommendation system, the system comprises: data capture unit 801, data analysis unit 802, number
According to feedback unit 803.
The data capture unit 801, for constructing the two of target user and recommended candidate collection user by user behavior
The two grade network relationship of grade cyberrelationship, the target user and recommended candidate collection user include target user and bridge user
Primary network station relationship and the primary network station relationship of bridge user and recommended candidate collection user.
The data analysis unit 802, the two grade network relationship for being constructed according to data capture unit, obtains each recommendation
The recommendation of Candidate Set user;It specifically includes: the first data analysis module, for being used according to the target user and the bridge
The primary network station relationship at family obtains the cohesion relation value of the target user Yu the bridge user;Second data analyze mould
Block, for obtaining the bridge user and institute according to the primary network station relationship of the bridge user and the recommended candidate collection user
State the similarity relation value of recommended candidate collection user;Third data analysis module, for according to first data analysis module
The similarity relation value that obtained cohesion relation value and second data analysis module obtains obtains the recommended candidate
Collect the recommendation of user;
The data feedback unit 803, the recommended candidate collection user for being obtained according to the third data analysis module
Recommendation, selection meet the target user interest preference user set, can specifically be chosen according to recommendation size
TOP-N user's set.
The two grade network that target user and recommended candidate collection user are constructed by user behavior of the data capture unit
The two grade network relationship of relationship, the target user and recommended candidate collection user include the level-one net of target user Yu bridge user
Network relationship and the primary network station relationship of bridge user and recommended candidate collection user particularly refer to that target user and bridge use
Comment between family and bridge user and recommended candidate collection user, hop count, forwarding frequency, the number of comment and comment
Frequency etc..
The data analysis unit is respectively pushed away for analyzing the two grade network relationship that data capture unit constructs
Recommend the recommendation of candidate user;It include: one according to the target user with the bridge user of the first data analysis module
Grade cyberrelationship obtains the cohesion relation value of the target user Yu the bridge user;
The primary network station relationship according to the bridge user and the recommended candidate collection user of second data analysis module
Obtain the similarity relation value of the bridge user Yu the recommended candidate collection user;
The target user of third data analysis module obtained according to first data analysis module and the bridge
The bridge user and the recommended candidate collection that the cohesion relation value of beam user and the second data analysis module obtain use
The similarity relation value at family obtains the recommendation of the recommended candidate collection user.
Wherein, the first data analysis module, specifically for the cohesion according to the target user and the bridge user
Obtain the cohesion relation value of the target user Yu the bridge user;The target user is intimate with the bridge user's
Degree includes the target user and the mutual fatigue resistance of the bridge user and the row of the target user and the bridge user
For time relationship.The mutual fatigue resistance of the target user and the bridge user include the target user in set time period
Quantity and mutual dynamic frequency are interacted between the bridge user.
The time of the act relationship of the target user and the bridge user are by including that the function of the time of the act factor is determined
It is fixed.
The similarity relation value of the bridge user and the recommended candidate collection user are pushed away by the bridge user with described
The common friend quantity for recommending Candidate Set user is obtained as measurement index.
The recommendation of the recommended candidate collection user of the data analysis unit analysis is by the target user and the bridge
The cohesion relation value of user and the similarity relation value of the bridge user and the recommended candidate collection user codetermine.
The recommendation for the recommended candidate collection user that the data feedback unit is analyzed according to the data analysis unit, root
It is needed according to user, recommends the user's set for meeting its interest preference to target user, can specifically be selected according to recommendation size
It takes TOP-N user to gather, completes user's friend recommendation.
It should be understood that the particular order or level of the step of during disclosed are the examples of illustrative methods.Based on setting
Count preference, it should be appreciated that in the process the step of particular order or level can be in the feelings for the protection scope for not departing from the disclosure
It is rearranged under condition.Appended claim to a method is not illustratively sequentially to give the element of various steps, and not
It is to be limited to the particular order or level.
In above-mentioned detailed description, various features are combined together in single embodiment, to simplify the disclosure.No
This published method should be construed to reflect such intention, that is, the embodiment of theme claimed needs to compare
The more features of the feature clearly stated in each claim.On the contrary, as appended claims is reflected
Like that, the present invention is in the state fewer than whole features of disclosed single embodiment.Therefore, appended claims
It is hereby expressly incorporated into detailed description, wherein each claim is used as alone the individual preferred embodiment of the present invention.
For can be realized any technical staff in the art or using the present invention, above to disclosed embodiment into
Description is gone.To those skilled in the art;The various modifications mode of these embodiments will be apparent from, and this
The General Principle of text definition can also be suitable for other embodiments on the basis of not departing from the spirit and scope of the disclosure.
Therefore, the disclosure is not limited to embodiments set forth herein, but most wide with principle disclosed in the present application and novel features
Range is consistent.
Description above includes the citing of one or more embodiments.Certainly, in order to describe above-described embodiment and description portion
The all possible combination of part or method is impossible, but it will be appreciated by one of ordinary skill in the art that each implementation
Example can do further combinations and permutations.Therefore, embodiment described herein is intended to cover fall into the appended claims
Protection scope in all such changes, modifications and variations.In addition, with regard to term used in specification or claims
The mode that covers of "comprising", the word is similar to term " includes ", just as " including " solved in the claims as transitional word
As releasing.In addition, the use of any one of specification in claims term "or" being to indicate " non-exclusionism
Or ".
Those skilled in the art will also be appreciated that the various illustrative components, blocks that the embodiment of the present invention is listed
(illustrative logical block), unit and step can by electronic hardware, computer software, or both knot
Conjunction is realized.For the replaceability (interchangeability) for clearly showing that hardware and software, above-mentioned various explanations
Property component (illustrative components), unit and step universally describe their function.Such function
It can be that the design requirement for depending on specific application and whole system is realized by hardware or software.Those skilled in the art
Can be can be used by various methods and realize the function, but this realization is understood not to for every kind of specific application
Range beyond protection of the embodiment of the present invention.
Various illustrative logical blocks or unit described in the embodiment of the present invention can by general processor,
Digital signal processor, specific integrated circuit (ASIC), field programmable gate array or other programmable logic devices, discrete gate
Or transistor logic, discrete hardware components or above-mentioned any combination of design carry out implementation or operation described function.General place
Managing device can be microprocessor, and optionally, which may be any traditional processor, controller, microcontroller
Device or state machine.Processor can also be realized by the combination of computing device, such as digital signal processor and microprocessor,
Multi-microprocessor, one or more microprocessors combine a digital signal processor core or any other like configuration
To realize.
The step of method described in the embodiment of the present invention or algorithm can be directly embedded into hardware, processor execute it is soft
The combination of part module or the two.Software module can store in RAM memory, flash memory, ROM memory, EPROM storage
Other any form of storaging mediums in device, eeprom memory, register, hard disk, moveable magnetic disc, CD-ROM or this field
In.Illustratively, storaging medium can be connect with processor, so that processor can read information from storaging medium, and
It can be to storaging medium stored and written information.Optionally, storaging medium can also be integrated into the processor.Processor and storaging medium can
To be set in asic, ASIC be can be set in user terminal.Optionally, processor and storaging medium also can be set in
In different components in the terminal of family.
In one or more exemplary designs, above-mentioned function described in the embodiment of the present invention can be in hardware, soft
Part, firmware or any combination of this three are realized.If realized in software, these functions be can store and computer-readable
On medium, or it is transferred on a computer readable medium in the form of one or more instructions or code forms.Computer readable medium includes electricity
Brain storaging medium and convenient for so that computer program is allowed to be transferred to from a place telecommunication media in other places.Storaging medium can be with
It is that any general or special computer can be with the useable medium of access.For example, such computer readable media may include but
It is not limited to RAM, ROM, EEPROM, CD-ROM or other optical disc storages, disk storage or other magnetic storage devices or other
What can be used for carry or store with instruct or data structure and it is other can be by general or special computer or general or specially treated
The medium of the program code of device reading form.In addition, any connection can be properly termed computer readable medium, example
Such as, if software is to pass through a coaxial cable, fiber optic cables, double from a web-site, server or other remote resources
Twisted wire, Digital Subscriber Line (DSL) are defined with being also contained in for the wireless way for transmitting such as example infrared, wireless and microwave
In computer readable medium.The disk (disk) and disk (disc) includes compress disk, radium-shine disk, CD, DVD, floppy disk
And Blu-ray Disc, disk is usually with magnetic replicate data, and disk usually carries out optically replicated data with laser.Combinations of the above
Also it may be embodied in computer readable medium.
Above-described specific embodiment has carried out further the purpose of the present invention, technical scheme and beneficial effects
It is described in detail, it should be understood that being not intended to limit the present invention the foregoing is merely a specific embodiment of the invention
Protection scope, all within the spirits and principles of the present invention, any modification, equivalent substitution, improvement and etc. done should all include
Within protection scope of the present invention.
Claims (10)
1. a kind of recommended method of user good friend, which is characterized in that the described method includes:
The two grade network relationship of target user and recommended candidate collection user, the target user and recommendation are constructed by user behavior
The two grade network relationship of Candidate Set user include target user and bridge user primary network station relationship and bridge user with push away
Recommend the primary network station relationship of Candidate Set user;
The target user is obtained according to the primary network station relationship of the target user and the bridge user and the bridge is used
The cohesion relation value at family, and obtained according to the primary network station relationship of the bridge user and the recommended candidate collection user described
The similarity relation value of bridge user and the recommended candidate collection user;
It is waited according to the cohesion relation value and the bridge user of the target user and the bridge user and the recommendation
The similarity relation value of selected works user obtains the recommendation of the recommended candidate collection user;
According to the recommendation of recommended candidate collection user, selection meets user's set of the interest preference of the target user.
2. the method according to claim 1, wherein the level-one according to target user and the bridge user
The cohesion relation value that cyberrelationship obtains the target user and the bridge user includes:
The parent of the target user Yu the bridge user are obtained according to the cohesion of the target user and the bridge user
Density relationship value;
The target user includes that the target user and interacting for the bridge user are strong with the cohesion of the bridge user
The time of the act relationship of degree and the target user and the bridge user.
3. according to the method described in claim 2, it is characterized in that, the mutual fatigue resistance of the target user and the bridge user
Including interacting quantity and mutual dynamic frequency between target user described in set time period and the bridge user.
4. according to the method described in claim 2, it is characterized in that, the time of the act of the target user and the bridge user
Relationship is by including that the function of the time of the act factor determines.
5. the method according to claim 1, wherein the phase of the bridge user and the recommended candidate collection user
It is obtained by the common friend quantity of the bridge user and the recommended candidate collection user as measurement index like degree relation value.
6. a kind of user's friend recommendation system, which is characterized in that the system comprises:
Data capture unit, for constructing the two grade network relationship of target user and recommended candidate collection user by user behavior,
The two grade network relationship of the target user and recommended candidate collection user include the primary network station pass of target user and bridge user
The primary network station relationship of system and bridge user and recommended candidate collection user;
Data analysis unit, the two grade network relationship for being constructed according to data capture unit obtain each recommended candidate collection user
Recommendation;It specifically includes:
First data analysis module, it is described for being obtained according to the primary network station relationship of the target user and the bridge user
The cohesion relation value of target user and the bridge user;
Second data analysis module, for being obtained according to the primary network station relationship of the bridge user and the recommended candidate collection user
To the similarity relation value of the bridge user and the recommended candidate collection user;
Third data analysis module, the cohesion relation value for being obtained according to first data analysis module and described
The similarity relation value that two data analysis modules obtain obtains the recommendation of the recommended candidate collection user;
Data feedback unit, the recommendation of the recommended candidate collection user for being obtained according to the third data analysis module
Value, selection meet user's set of the interest preference of the target user.
7. system according to claim 6, which is characterized in that first data analysis module is specifically used for according to institute
The cohesion for stating target user and the bridge user obtains the cohesion relation value of the target user Yu the bridge user;
The target user includes that the target user and interacting for the bridge user are strong with the cohesion of the bridge user
The time of the act relationship of degree and the target user and the bridge user.
8. system according to claim 7, which is characterized in that target user described in first data analysis module with
The mutual fatigue resistance of the bridge user includes interacting between the target user and the bridge user in set time period
Quantity and mutual dynamic frequency.
9. according to system as claimed in claim 7, which is characterized in that target user described in first data analysis module and institute
The time of the act relationship of bridge user is stated by including that the function of the time of the act factor determines.
10. system according to claim 6, which is characterized in that bridge user described in second data analysis module
With the similarity relation value of the recommended candidate collection user by the common good of the bridge user and the recommended candidate collection user
Friendly quantity is obtained as measurement index.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810967953.3A CN109190033B (en) | 2018-08-23 | 2018-08-23 | User friend recommendation method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810967953.3A CN109190033B (en) | 2018-08-23 | 2018-08-23 | User friend recommendation method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109190033A true CN109190033A (en) | 2019-01-11 |
CN109190033B CN109190033B (en) | 2022-01-11 |
Family
ID=64919263
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810967953.3A Active CN109190033B (en) | 2018-08-23 | 2018-08-23 | User friend recommendation method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109190033B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110674413A (en) * | 2019-09-09 | 2020-01-10 | 平安科技(深圳)有限公司 | User relationship mining method, device, equipment and storage medium |
CN111597454A (en) * | 2020-04-02 | 2020-08-28 | 微梦创科网络科技(中国)有限公司 | Account recommendation method and device |
CN112380423A (en) * | 2020-10-14 | 2021-02-19 | 广州市百果园网络科技有限公司 | Network object pushing method and device, readable storage medium and electronic equipment |
CN112559893A (en) * | 2021-01-11 | 2021-03-26 | 引粒网络科技(上海)有限公司 | User recommendation method in internet matching social scene |
CN113076450A (en) * | 2021-03-15 | 2021-07-06 | 北京明略软件系统有限公司 | Method and device for determining target recommendation list |
CN115186664A (en) * | 2022-09-13 | 2022-10-14 | 深圳市爱聊科技有限公司 | Method and system for measuring and calculating degree of coincidence between subjects based on multiple dimensions |
CN115357632A (en) * | 2022-07-21 | 2022-11-18 | 广东技术师范大学 | Learner personalized interpretable potential friend recommendation method and device |
TWI828186B (en) * | 2021-06-18 | 2024-01-01 | 韓商連加股份有限公司 | Method, computer device, and computer readable recording medium for friend recommendation |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102637183A (en) * | 2011-02-12 | 2012-08-15 | 北京千橡网景科技发展有限公司 | Method and device for recommending friends to user in social network |
CN103294733A (en) * | 2012-03-05 | 2013-09-11 | 北京千橡网景科技发展有限公司 | Method and system used for recommending friends to users |
EP2701103A2 (en) * | 2012-08-24 | 2014-02-26 | Samsung Electronics Co., Ltd | Method of recommending friends, and server and terminal therefor |
CN106326228A (en) * | 2015-06-17 | 2017-01-11 | 富士通株式会社 | Method and device for evaluating interest tendency of user |
CN106649781A (en) * | 2016-12-28 | 2017-05-10 | 北京小米移动软件有限公司 | Application recommendation method and device |
CN107066476A (en) * | 2016-12-13 | 2017-08-18 | 江苏途致信息科技有限公司 | A kind of real-time recommendation method based on article similarity |
-
2018
- 2018-08-23 CN CN201810967953.3A patent/CN109190033B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102637183A (en) * | 2011-02-12 | 2012-08-15 | 北京千橡网景科技发展有限公司 | Method and device for recommending friends to user in social network |
CN103294733A (en) * | 2012-03-05 | 2013-09-11 | 北京千橡网景科技发展有限公司 | Method and system used for recommending friends to users |
EP2701103A2 (en) * | 2012-08-24 | 2014-02-26 | Samsung Electronics Co., Ltd | Method of recommending friends, and server and terminal therefor |
CN106326228A (en) * | 2015-06-17 | 2017-01-11 | 富士通株式会社 | Method and device for evaluating interest tendency of user |
CN107066476A (en) * | 2016-12-13 | 2017-08-18 | 江苏途致信息科技有限公司 | A kind of real-time recommendation method based on article similarity |
CN106649781A (en) * | 2016-12-28 | 2017-05-10 | 北京小米移动软件有限公司 | Application recommendation method and device |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110674413A (en) * | 2019-09-09 | 2020-01-10 | 平安科技(深圳)有限公司 | User relationship mining method, device, equipment and storage medium |
CN110674413B (en) * | 2019-09-09 | 2022-03-25 | 平安科技(深圳)有限公司 | User relationship mining method, device, equipment and storage medium |
CN111597454A (en) * | 2020-04-02 | 2020-08-28 | 微梦创科网络科技(中国)有限公司 | Account recommendation method and device |
CN112380423A (en) * | 2020-10-14 | 2021-02-19 | 广州市百果园网络科技有限公司 | Network object pushing method and device, readable storage medium and electronic equipment |
CN112559893A (en) * | 2021-01-11 | 2021-03-26 | 引粒网络科技(上海)有限公司 | User recommendation method in internet matching social scene |
CN113076450A (en) * | 2021-03-15 | 2021-07-06 | 北京明略软件系统有限公司 | Method and device for determining target recommendation list |
CN113076450B (en) * | 2021-03-15 | 2024-03-22 | 北京明略软件系统有限公司 | Determination method and device for target recommendation list |
TWI828186B (en) * | 2021-06-18 | 2024-01-01 | 韓商連加股份有限公司 | Method, computer device, and computer readable recording medium for friend recommendation |
CN115357632A (en) * | 2022-07-21 | 2022-11-18 | 广东技术师范大学 | Learner personalized interpretable potential friend recommendation method and device |
CN115186664A (en) * | 2022-09-13 | 2022-10-14 | 深圳市爱聊科技有限公司 | Method and system for measuring and calculating degree of coincidence between subjects based on multiple dimensions |
CN115186664B (en) * | 2022-09-13 | 2023-01-13 | 深圳市爱聊科技有限公司 | Method and system for measuring and calculating coincidence degree between subjects based on multiple dimensions |
Also Published As
Publication number | Publication date |
---|---|
CN109190033B (en) | 2022-01-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109190033A (en) | A kind of user's friend recommendation method and system | |
McKenzie et al. | Weighted multi-attribute matching of user-generated points of interest | |
Aliannejadi et al. | A joint two-phase time-sensitive regularized collaborative ranking model for point of interest recommendation | |
CN103678618B (en) | Web service recommendation method based on socializing network platform | |
He et al. | Study on SINA micro-blog personalized recommendation based on semantic network | |
CN105787068B (en) | The academic recommended method and system analyzed based on citation network and user's proficiency | |
CN105849764A (en) | Systems and methods for identifying influencers and their communities in a social data network | |
US20130132481A1 (en) | Method, apparatus and system for providing social network service using social activities | |
EP2646973A2 (en) | Real-time personalized recommendation of location-related entities | |
CN104239399A (en) | Method for recommending potential friends in social network | |
Carmagnola et al. | User data distributed on the social web: how to identify users on different social systems and collecting data about them | |
CN109325171A (en) | User interest analysis method and system based on domain knowledge | |
CN105740448A (en) | Topic-oriented multi-microblog time sequence abstracting method | |
CN109684548A (en) | A kind of data recommendation method based on user's map | |
Miah et al. | A social media analytics perspective for human‐oriented smart city planning and management | |
Saraswathi et al. | Deep Learning Enabled Social Media Recommendation Based on User Comments. | |
Yu et al. | Point-of-interest recommendation based on user contextual behavior semantics | |
CN110543601B (en) | Method and system for recommending context-aware interest points based on intelligent set | |
Longley et al. | Classification through consultation: Public views of the geography of the e-society | |
Yang et al. | Personalized recommendation based on collaborative filtering in social network | |
Jasbick et al. | Some branches may bear rotten fruits: Diversity browsing vp-trees | |
Yang et al. | Personalized mobile application discovery | |
Bhutani et al. | WSEMQT: a novel approach for quality‐based evaluation of web data sources for a data warehouse | |
CN108959579A (en) | A kind of system obtaining user and Document personalization feature | |
Katsaliaki et al. | A profile of OR research and practice published in the Journal of the Operational Research Society |
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 |