CN107526850A - Social networks friend recommendation method based on multiple personality feature mixed architecture - Google Patents
Social networks friend recommendation method based on multiple personality feature mixed architecture Download PDFInfo
- Publication number
- CN107526850A CN107526850A CN201710944432.1A CN201710944432A CN107526850A CN 107526850 A CN107526850 A CN 107526850A CN 201710944432 A CN201710944432 A CN 201710944432A CN 107526850 A CN107526850 A CN 107526850A
- Authority
- CN
- China
- Prior art keywords
- user
- targeted customer
- random walk
- social networks
- cohesion
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 208000025967 Dissociative Identity disease Diseases 0.000 title claims abstract description 12
- 238000005295 random walk Methods 0.000 claims abstract description 44
- 239000011159 matrix material Substances 0.000 claims abstract description 18
- 238000012546 transfer Methods 0.000 claims abstract description 15
- 230000003993 interaction Effects 0.000 claims abstract description 13
- 230000002452 interceptive effect Effects 0.000 claims abstract description 11
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 10
- 238000012512 characterization method Methods 0.000 claims abstract description 9
- 230000000694 effects Effects 0.000 claims description 22
- 210000003813 thumb Anatomy 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 4
- 239000000284 extract Substances 0.000 claims description 3
- 238000013508 migration Methods 0.000 claims description 3
- 230000005012 migration Effects 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 230000003542 behavioural effect Effects 0.000 abstract description 3
- 238000011156 evaluation Methods 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 5
- 238000002474 experimental method Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 239000013065 commercial product Substances 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Marketing (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Computing Systems (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of social networks friend recommendation method based on multiple personality feature mixed architecture, interactive information between the attribute information of this method combination user and user, define cohesion feature, and cohesion probability transfer matrix is built, introduce random walk model and calculate level of interaction score value between user.On this basis, the personalized node diagnostic of user is defined, assigns different weights respectively using weighted mean method for it, finally using Random Walk Algorithm, the recommendation list for obtaining Top N is recommended.The present invention has taken into full account the attribute information and interactive information of user, consider the interest characteristics of user, and behavioural characteristic of the user in social networks, true good friend's demand of user can truly be reflected by these personalized network characterizations and node diagnostic, largely improve the accuracy rate of friend recommendation.
Description
Technical field
The present invention relates to the personalized friend recommendation model in social networks, more particularly to one kind are special based on multiple personalityization
Levy the social networks friend recommendation method of mixed architecture.Interactive information between the attribute information of this method combination user and user,
Cohesion feature is defined, and builds cohesion probability transfer matrix, random walk model is introduced and calculates level of interaction between user
Score value.On this basis, the personalized node diagnostic of user is defined, different power is assigned respectively for it using weighted mean method
Value, finally using Random Walk Algorithm, the recommendation list for obtaining Top-N is recommended.Based on this, it is unified personalized good to build
Friendly recommended models, to realize personalized friend recommendation.
Background technology
Recommendation is a kind of effective means for searching for information, if applied in can also attract customer in commercial product recommending;In society
Hand in network, friend recommendation is necessary, and the user for urgently wanting to improve sociability especially for those is very
Important.
In recent years, it has been suggested that a variety of recommendation methods, e.g., based on collaborative filtering, based on label, based on content etc..These
Method is widely used in major commending system, but these methods have ignored the interaction letter between the individualized feature of user and user
Breath, e.g. ,@, comment, forward, thumb up etc..
Dynamic social networks develops into the new environment for we providing a checking recommendation method, while also brings new
Challenge.Due to the dynamic of user behavior and the importance of user mutual, how according to customer interaction information progress friend recommendation
How individualized feature design effective ways progress friend recommendation is utilizedHow to consider above-mentioned factors and establish unified move
States model realizes personalized friend recommendationThese are all urgent problems to be solved.
The content of the invention
It is contemplated that using the interactive information between the behavioural information of user and user, structure is based on multiple features mixed structure
Unified dynamic social networks friend recommendation model.The model is proposed a kind of by introducing network characterization and user node feature
Random walk framework and the mixed structure of weighted mean method ignore user personalized information to make up existing friend recommendation method
Deficiency.The essential information and use of user in social networks can be taken into full account with reference to personalized network feature and user node feature
Interactive information between family, it is final to improve the accuracy recommended.
Foregoing invention purpose is realized in order to solve above-mentioned technical problem, the present invention is real by the following technical programs
Existing:
A kind of social networks friend recommendation method based on multiple personality feature mixed architecture, this method content include with
Lower step:
S1, cohesion feature is defined, and build cohesion probability transfer matrix;
S2, introduce random walk model and calculate level of interaction score value between user;
S3, defines and extracts the node diagnostic of user on the basis of network characterization cohesion is defined, and structure multiple features turn
Move matrix;
S4, to the node diagnostic extracted, calculate its similarity and sort;
S5, Random Walk Algorithm is run, recommendation list is obtained according to random walk result, and to targeted customer's recommended candidate
User.
Further, it is described to define cohesion feature in S1, and cohesion probability transfer matrix is built, its constraints
Including:
A the essential information and its microblogging multidate information of targeted customer in social networks) is captured;The essential information includes using
Name in an account book, hobby, age this kind of attribute information;The microblogging multidate information includes thumbing up number, comment number and forwarding number;
B the essential information of targeted customer good friend) is extracted, and two degree of good friends of the targeted customer and three degree of good friends' is basic
Information;
C) extract user between interacting activity information, herein interacting activity information include:Targeted customer, targeted customer it is good
Friend, two degree of good friends of targeted customer, targeted customer three degree of good friends between interacting activity information, and social activity is built with this
The social networks of targeted customer in network;Interacting activity information includes a variety of interactive informations between the user, such as, comment, turn
Send out, thumb up and reply;
D network characterization) is defined:Cohesion, it defines such as formula (1)
Intimacy degree (u, v)=log (#uv+1.1) * log (#vu+1.1) (1)
Wherein u and v is respectively the user in system.
This feature has taken into full account that the attribute information of user (including@, comment, a variety of interactive shapes such as forwards, thumbed up, replying
Formula), frequency, the information such as form of its interaction of comprehensive analysis, the same formula with abstract are expressed, and structure is intimately accordingly
Probability transfer matrix is spent, recommendation list is obtained using random walk model.
Further, in S2, the random walk model, its constraints includes:
A) random walk model is that have Markov property according to the characteristic of targeted customer's state, i.e., in t+1 moment users
State only and the state of user during moment t has relation, be expressed as formula below:
Pr(Xu,t+1=i | Xu,t) (2)
B) according to the cohesion feature and node diagnostic of definition, multiple features probability transfer matrix is established, as probability
Distribution vector PI, jCarry out random walk;
C) input parameter of random walk process has four, is respectively:Adjacency matrixInitial score distribution vector r0,
Restart probability of happening c and ProbabilityDistribution Vector p;Scoring distribution vector after each migration is r0, then r calculation formula expression
For:
The r that formula (3) is calculated every time substitutes into r0, iterate until convergence, obtain Stationary Distribution probability vector
R, r are the similarity vector of recommendation results.
Further, in S3, the node diagnostic of the user, its constraints includes:
Define personalized node diagnostic:This feature is used to describing the attribute information of user, the attribute information include sex,
Age, hobby, the customized information of user is taken into full account, and the attribute information with targeted customer is represented with abstract formula
Similitude, and the similarity is ranked up.
Further, in S4, the calculate node characteristic similarity value simultaneously sorts, and its constraints includes:
A weighted mean method) is introduced, different weights is assigned for various personalized node diagnostics;
B the row of the personalized node diagnostic similitude of the user and targeted customer in network) is obtained using weighted mean method
Sequence value.
Further, in S5, the operation Random Walk Algorithm, its constraints includes:
A) the Stationary Distribution probability vector r obtained according to random walk, it is ranked up to obtain Top-N recommendation row
Table;
B) by the friend recommendation in recommendation list to targeted customer.
Due to there is such beneficial effect compared with prior art using above-mentioned technical proposal, the present invention:
1) present invention proposes a kind of novel personalized network feature, and is integrated into random walk model and applies
In friend recommendation;
2) present invention proposes two new personalized node/user characteristicses, and carries out mixed structure personalization good friend and push away
Recommend;
3) to realize friend recommendation, present invention design and implementation have the mixing friend recommendation frame of generality and scalability
Structure;
4) weighted mean method is integrated into random walk framework to realize personalized friend recommendation by the present invention;
5) present invention has taken into full account the attribute information and interactive information of user, it is contemplated that the interest characteristics of user, and
Behavioural characteristic of the user in social networks, it can truly reflect use by these personalized network characterizations and node diagnostic
True good friend's demand at family, largely improve the accuracy rate of friend recommendation.
Brief description of the drawings
Fig. 1 is that the present invention is based on multiple personality feature mixed architecture schematic diagram;
Fig. 2 is IN, ID, IA analysis result figure;
Fig. 3 is the tendency chart of weighted value impact evaluation index;
Fig. 4 is the result displaying figure for seeking micro- friend's application.
Embodiment
The present invention is described in further detail with embodiment below in conjunction with the accompanying drawings:
Following examples are used to illustrate the present invention, but are not limited to the scope of the present invention.
The data set used in following examples and experiment is as follows:
3 user data sets that data set size is 160,000 are randomly selected respectively carries out signature analysis, these three data sets
Respectively:DS1, DS2, DS3.
By recording the microblogging sum of user in each data set, thumbing up number, comment number, forwarding number etc., each data set is analyzed
Middle cohesion (IN), interest-degree (ID), interest activity (IA), its result is as shown in Fig. 2 abscissa represents number of users, unit:
Ten thousand, scale 2,4,6,8,10,12,14,16.Ordinate represents the forwarding of all users, the quantity sum commented on and thumbed up
SumIN, then take lg (denary logarithm).As can be seen from the figure IN is higher in DS1, and IA is higher in DS2, and in DS3
ID is higher.
Embodiment
1st, the embodiment of the present invention is with personalized node diagnostic:Interest-degree, interest activity, node sequencing method is with weighted average
It is that the present invention is based on multiple personality feature mixed architecture schematic diagram exemplified by method, shown in Fig. 1, implementation step is as follows:
First, cohesion feature is defined, and build cohesion probability transfer matrix;
Second, introduce Random Walk Algorithm and calculate level of interaction score value between user;
3rd, the node diagnostic of user is defined and extracted on the basis of network characterization cohesion is defined, and builds multiple features
Transfer matrix;
4th, weighted mean method is introduced to the node diagnostic extracted, is calculated its similarity and is sorted;
5th, random walk model is run, is recommended according to obtained recommendation list to targeted customer.
2nd, in embodiment 1, it is described to define cohesion feature, and cohesion probability transfer matrix is built, it is specific real
Now include:
A the essential information and its microblogging multidate information of the targeted customer in social networks) is captured, wherein, essential information bag
Include:The attribute informations such as user name, hobby, age;Microblogging multidate information includes:Thumb up number, comment number, forwarding number etc..
B the essential information of targeted customer two degree of good friends, three degree of good friends) are obtained;
C the interacting activity information in targeted customer and network between user) is obtained, is mainly included:@, comment, forward, thumb up,
Reply etc., and build with this social networks of targeted customer in social networks;
D network characterization) is defined:Cohesion, it defines such as formula (4)
Intimacy degree (u, v)=log (#uv+1.1) * log (#vu+1.1) (4)
User during u and v is respectively system in formula (4), #uv are the total degree of user u and v interaction, add and 1.1 are mainly
In order to which the result finally given is nonnegative number.
This feature has taken into full account interactive information (including@, comment, forwarding, the point between the attribute information of user and user
A variety of interactive forms such as praise, reply), the information such as the frequency of its interaction of comprehensive analysis, form is same to be carried out with the formula abstracted
Expression, and cohesion probability transfer matrix is built accordingly, obtain recommendation list using random walk model.
3rd, in embodiment 1, the random walk model, its specific implementation include:
A) random walk model is that have Markov property, the i.e. shape in t+1 moment users according to the characteristic of User Status
State only has relation with the state of user during moment t, with formula (5) explanation:
Pr(Xu,t+1=i | Xu,t) (5)
B) according to the cohesion feature and node diagnostic of definition, multiple features probability transfer matrix is established, as probability
Distribution vector Pi,jCarry out random walk;
C) input parameter of random walk process has four, is respectively:Adjacency matrixInitial score distribution vector r0,
Restart probability of happening c and ProbabilityDistribution Vector p;Scoring distribution vector after each migration is r0, then r calculation formula expression
For:
Formula is calculated to r every time and substitutes into r0, iterate until convergence, generally, iterations is at 50 times
Within be up to convergence, obtain the similarity vector that Stationary Distribution probability vector r, r are recommendation results.
4th, in embodiment 1, the node diagnostic, its constraints include:
A personalized node diagnostic) is defined:The interest of interest-degree (InterestDeg) description targeted customer and friend candidate
Interest between matching degree;And the interest similarity between user is calculated using cosine similarity, such as formula (8):
In formula,IuiTo describe useruiInterest keyword set, Ivj is the collection of description user vj interest keyword
Close.
B individualized feature) is defined:Interest activity (InterestActivity) description friend candidate issue it is newest on
The activity index of special interests microblogging;Calculated by formula (9):
In formulaMn(ui)Represent user uiIn the set for the special interests microblogging that nearest N number of middle of the month is issued, SoM (Mn(ui))
Represent Mn(ui) in all microbloggings forwarding, comment, thumb up, reply and@sum, w (n) n-th month, N represent nearest N number of
Month.
5th, in embodiment 1, described that node similitude is sorted, its specific implementation steps includes:
A weighted mean method) is introduced, different weights is assigned for the personalized node diagnostic of definition;
B the personalized node diagnostic similitude of the user and targeted customer in network) are obtained using weighted mean method
Ranking value;Its calculation formula such as formula (10) and formula (11):
RV(vj,ui)=d_weight × InterestDeg (vj,ui)+a_weight×InterestActivity(vj,
ui) (10)
RV(ui,vj)=d_weight × InterestDeg (ui,vj)+a_weight×InterestActivity(ui,
vj) (11)
Wherein d_weight and a_weight represents the weighted value of interest-degree and interest activity between user.
C optimal weighted value) is determined by example.
As weight a_weight is from 0 to 1, becoming in DS1, DS2, DS3 MAP, Recall and F1-Measure is obtained
Gesture, it can be found that for the big data set of interest matching degree from result, with a_weight, its performance also improves therewith, but
After it reaches peak performance, as weight a_weight increases, MAP, Recall and F1-Measure value reduce, that is,
Say the reduction of its performance.But for other data sets, with weight a_weight increase, it influences little, it might even be possible to neglects
Slightly its influence.That is, when randomly choosing data set, evaluation index is insensitive to weight a_weight.But work as a_
When weight is 0.5, behave oneself best in DS1.So it is 0.5 that weight a_weight is set in experiment afterwards.Work as weight
When a_weight is 0.5, each evaluation index of the inventive method in DS1, DS2, DS3 is as shown in table 1 below:
Table 1:Friend++ performance
From table 1 it follows that for each evaluation index, for DS1 than DS2, DS3 will be high, but all performance is good, therefore will
It is rational that weight a_weight, which is arranged to 0.5,.
6th, in embodiment 1, the operation random walk model obtains recommendation list, and its specific implementation steps includes:
A) the Stationary Distribution probability vector r obtained according to random walk, it is ranked up to obtain Top-N recommendation row
Table;
B) by the friend recommendation in recommendation list to targeted customer.
Below with one group of contrast experiment, to evaluate beneficial effects of the present invention.Select two baseline algorithms:Random walk
And the Semantic Similarity Measurement based on LDA (LDA-based Silmilarity, LDAS) (RW).
Random walk (RW):It make use of a new individualized feature cohesion.Specifically method is:From social networks
Middle extraction friends data and its interacting activity data, then calculate targeted customer and candidate using Random Walk Algorithm (RW)
Cohesion between friend's (candidate user), the recommendation list for finally obtaining Top-N targeted customers carry out friend recommendation.
Similitude (LDAS) based on LDA:It is recommended by the semantic similarity between calculating two users.Specifically
Way is:Nearly 3 months micro-blog informations are selected first, and higher-dimension theme vector is generated by LDA, then calculates vector similarity side by side
Sequence, Top-N friends candidate for finally obtaining targeted customer are recommended.
Using P@5, P@10, P@15, Recall, MAP and F1-Measure this six indexs, above three actual number is utilized
According to collection compared with RW and LDAS baseline algorithm.Table 2 lists the evaluation index result of distinct methods on three data sets.
The comparison of the results of property of table 2.
As can be seen from the table, according to all evaluation indexes, the recommendation precision of the inventive method be better than RW methods and
LDAS.Consider Recall indexs, in data set DS1, DS2 and DS3, Friend++ average behavior improves 24% than RW,
27% is improved than LDAS, so as to improve recommendation performance.Meanwhile two kinds of Baseline Methods compare Friend in terms of F1-Measure
++ 55% is have dropped, all these results have all absolutely proved that all improvement of the Friend++ to both approaches are all to have one
Determine meaning.
In Friend++, three kinds of saturations are considered:Cohesion (I), interest-degree (D) and interest activity (A).Delete one by one
Except these factors, to show its influence to recommending performance.Especially, interest-degree saturation is deleted, with Friend++-D tables
Show, delete interest activity saturation, represented with Friend++-A, delete cohesion saturation, represented with Friend++-I.
Fig. 3 shows the average F1-Measure of different saturations in DS1, DS2 and DS3, wherein, Friend++-I does not include intimate
The factor is spent, Friend++-D does not include the interest-degree factor, and Friend++-A does not include the interest activity factor.DS1 is that have height
IN, DS2 are that have high IA (meaning that user has more interest activities), and DS3 is the data set for having high ID.Ignore this every time
A certain specific factor in three factors, it was observed that average F1-Measure can be decreased obviously, this shows that the inventive method passes through
Considering these three different saturations has better performance.It can be seen that Friend++ is averaged in DS1
F1-Measure value is maximum, next to that two other data set.
Further result of the analysis in same data set, high IN in DS1:It is gradual to remove the influence of I, D and A to index
Reduce, therefore desired value increases successively;High IA in DS2:Because IA is small to Index Influence, therefore to Friend++-A value most
Greatly, Friend++-I and Friend++-D value is the 3rd and the 2nd successively;High ID in DS3, because ID has a great influence, therefore
Friend++-D value is minimum, and Friend++-I takes second place, and Friend++-A value is maximum;Then different pieces of information concentration is analyzed again
As a result, remove I, because in DS1 being high IN, therefore influence maximum, index is minimum;Remove D, because high ID in DS3, therefore index
It is minimum;Remove A, because high IA in DS2, therefore index is minimum.
Now, an application in microblogging is passed through:Micro- friend is sought to further illustrate Friend++ recommendation performance.The application
Program can provide friend recommendation service for microblog users, consider at 2 points:1) according to the good friend of targeted customer and two degree of good friends it
Between interaction and cohesion, list friend candidate that his most probable thinks chat, while calculate the recommendation index of each friend candidate;
2) according to his microblogging active level, the interaction index between he and he friend is calculated.As shown in figure 4, seeking micro- friend can have
Effect ground solves the problems, such as the recommendation of reality, and him is provided desired service for targeted customer.Seek micro- friend and additionally provide some users only
Want the Validity Index obtained.It can be found that with the increase of interaction index from Fig. 4, index is recommended also to increase.This can be with
It is construed to that there is similar interests or the user of activity more likely to turn into friend.
According to test result, it is found that recommended user will agree in three seconds, most slow also can be in ten hours
Meet with a response (including user not online situation), and about 97.73% user can give a response.Model of the present invention it is excellent
More property is far above in this, examines the result in figure.It will be apparent that the Top-N user in recommendation list can not only be recommended to
Targeted customer, and their friend can also be recommended to targeted customer (with recommended user's " merely getting "), and this is also one
The new social path of kind.From the above-mentioned availability and validity that can clearly illustrate the inventive method.
Claims (6)
- A kind of 1. social networks friend recommendation method based on multiple personality feature mixed architecture, it is characterised in that:This method Content comprises the following steps:S1, cohesion feature is defined, and build cohesion probability transfer matrix;S2, introduce random walk model and calculate level of interaction score value between user;S3, the node diagnostic of user, structure multiple features transfer square are defined and extracted on the basis of network characterization cohesion is defined Battle array;S4, to the node diagnostic extracted, calculate its similarity and sort;S5, Random Walk Algorithm is run, recommendation list is obtained according to random walk result, and use to targeted customer's recommended candidate Family.
- A kind of 2. social networks friend recommendation side based on multiple personality feature mixed architecture according to claim 1 Method, it is characterised in that:It is described to define cohesion feature in S1, and cohesion probability transfer matrix is built, its constraints bag Include:A the essential information and its microblogging multidate information of targeted customer in social networks) is captured;The essential information includes user Name, hobby, age this kind of attribute information;The microblogging multidate information includes thumbing up number, comment number and forwarding number;B the essential information of targeted customer good friend, and the basic letter of two degree of good friends of the targeted customer and three degree of good friends) are extracted Breath;C) extract user between interacting activity information, herein interacting activity information include:Targeted customer, targeted customer good friend, Interacting activity information between two degree of good friends of targeted customer, three degree of good friends of targeted customer, and social network is built with this The social networks of targeted customer in network;Interacting activity information includes a variety of interactive informations between the user, such as, comment, turn Send out, thumb up and reply;D network characterization) is defined:Cohesion, it defines such as formula (1)Intimacy degree (u, v)=log (#uv+1.1) * log (#vu+1.1) (1)Wherein u and v is respectively the user in system.
- A kind of 3. social networks friend recommendation side based on multiple personality feature mixed architecture according to claim 1 Method, it is characterised in that:In S2, the random walk model, its constraints includes:A) random walk model is that have Markov property, the i.e. shape in t+1 moment users according to the characteristic of targeted customer's state State only has relation with the state of user during moment t, is expressed as formula below:Pr(Xu,t+1=i | Xu,t) (2)B) according to the cohesion feature and node diagnostic of definition, multiple features probability transfer matrix is established, as initial probability distribution Vectorial Pi,jCarry out random walk;C) input parameter of random walk process has four, is respectively:Adjacency matrixInitial score distribution vector r0, restart hair Raw probability c and ProbabilityDistribution Vector p;Scoring distribution vector after each migration is r0, then r calculation formula be expressed as:<mrow> <mi>r</mi> <mo>=</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>-</mo> <mi>c</mi> <mo>)</mo> </mrow> <msup> <mover> <mi>W</mi> <mo>~</mo> </mover> <mi>T</mi> </msup> <msub> <mi>r</mi> <mn>0</mn> </msub> <mo>+</mo> <mi>c</mi> <mi>p</mi> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>3</mn> <mo>)</mo> </mrow> </mrow>The r that formula (3) is calculated every time substitutes into r0, iterate until convergence, obtaining Stationary Distribution probability vector r, r is For the similarity vector of recommendation results.
- A kind of 4. social networks friend recommendation side based on multiple personality feature mixed architecture according to claim 1 Method, it is characterised in that:In S3, the node diagnostic of the user, its constraints includes:Define personalized node diagnostic:This feature is used for the attribute information for describing user, and the attribute information includes sex, year Age, hobby, the customized information of user is taken into full account, and the attribute information phase with targeted customer is represented with abstract formula It is ranked up like property, and to the similarity.
- A kind of 5. social networks friend recommendation side based on multiple personality feature mixed architecture according to claim 1 Method, it is characterised in that:In S4, the calculate node characteristic similarity value simultaneously sorts, and its constraints includes:A weighted mean method) is introduced, different weights is assigned for various personalized node diagnostics;B the ranking value of the personalized node diagnostic similitude of the user and targeted customer in network) is obtained using weighted mean method.
- A kind of 6. social networks friend recommendation side based on multiple personality feature mixed architecture according to claim 1 Method, it is characterised in that:In S5, the operation Random Walk Algorithm, its constraints includes:A) the Stationary Distribution probability vector r obtained according to random walk, it is ranked up to obtain Top-N recommendation list;B) by the friend recommendation in recommendation list to targeted customer.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710944432.1A CN107526850A (en) | 2017-10-12 | 2017-10-12 | Social networks friend recommendation method based on multiple personality feature mixed architecture |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710944432.1A CN107526850A (en) | 2017-10-12 | 2017-10-12 | Social networks friend recommendation method based on multiple personality feature mixed architecture |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107526850A true CN107526850A (en) | 2017-12-29 |
Family
ID=60684720
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710944432.1A Pending CN107526850A (en) | 2017-10-12 | 2017-10-12 | Social networks friend recommendation method based on multiple personality feature mixed architecture |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107526850A (en) |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108595674A (en) * | 2018-05-02 | 2018-09-28 | 上海同砚信息科技有限公司 | Social networks recommend method |
CN109582953A (en) * | 2018-11-02 | 2019-04-05 | 中国科学院自动化研究所 | A kind of speech of information is according to support methods of marking, equipment and storage medium |
CN109658277A (en) * | 2018-11-30 | 2019-04-19 | 华南师范大学 | A kind of science social networks friend recommendation method, system and storage medium |
CN109933666A (en) * | 2019-03-18 | 2019-06-25 | 西安电子科技大学 | A kind of good friend's automatic classification method, device, computer equipment and storage medium |
CN110955840A (en) * | 2018-09-27 | 2020-04-03 | 微软技术许可有限责任公司 | Joint optimization of notifications and pushes |
CN111143701A (en) * | 2019-12-13 | 2020-05-12 | 中国电子科技网络信息安全有限公司 | Social network user recommendation method and system based on multiple dimensions |
CN111368217A (en) * | 2020-03-19 | 2020-07-03 | 淄博聚德网络科技有限公司 | Social method and system for promoting strangers to develop to friends |
CN111475724A (en) * | 2020-04-01 | 2020-07-31 | 上海硕恩网络科技股份有限公司 | Random walk social network event recommendation method based on user similarity |
CN111651671A (en) * | 2020-05-27 | 2020-09-11 | 腾讯科技(深圳)有限公司 | User object recommendation method and device, computer equipment and storage medium |
CN112100489A (en) * | 2020-08-27 | 2020-12-18 | 北京百度网讯科技有限公司 | Object recommendation method, device and computer storage medium |
CN112487303A (en) * | 2020-11-26 | 2021-03-12 | 杭州电子科技大学 | Topic recommendation method based on social network user attributes |
CN112507247A (en) * | 2020-12-15 | 2021-03-16 | 重庆邮电大学 | Cross-social network user alignment method fusing user state information |
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 |
CN113360764A (en) * | 2021-06-23 | 2021-09-07 | 甄付(上海)网络科技有限公司 | Social network construction method and system based on life service consumption scene |
CN113836444A (en) * | 2021-09-30 | 2021-12-24 | 中国科学院深圳先进技术研究院 | Linear time friend recommendation method, system, terminal and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102779182A (en) * | 2012-07-02 | 2012-11-14 | 吉林大学 | Collaborative filtering recommendation method for integrating preference relationship and trust relationship |
CN103379158A (en) * | 2012-04-24 | 2013-10-30 | 中国移动通信集团公司 | Method and system for recommending friend information in social network |
CN103559407A (en) * | 2013-11-14 | 2014-02-05 | 北京航空航天大学深圳研究院 | Recommendation system and method for measuring node intimacy in weighted graph with direction |
US20140244664A1 (en) * | 2013-02-25 | 2014-08-28 | Telefonaktiebolaget L M Ericsson (Publ) | Method and Apparatus For Determining Similarity Information For Users of a Network |
-
2017
- 2017-10-12 CN CN201710944432.1A patent/CN107526850A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103379158A (en) * | 2012-04-24 | 2013-10-30 | 中国移动通信集团公司 | Method and system for recommending friend information in social network |
CN102779182A (en) * | 2012-07-02 | 2012-11-14 | 吉林大学 | Collaborative filtering recommendation method for integrating preference relationship and trust relationship |
US20140244664A1 (en) * | 2013-02-25 | 2014-08-28 | Telefonaktiebolaget L M Ericsson (Publ) | Method and Apparatus For Determining Similarity Information For Users of a Network |
CN103559407A (en) * | 2013-11-14 | 2014-02-05 | 北京航空航天大学深圳研究院 | Recommendation system and method for measuring node intimacy in weighted graph with direction |
Cited By (23)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108595674A (en) * | 2018-05-02 | 2018-09-28 | 上海同砚信息科技有限公司 | Social networks recommend method |
CN110955840B (en) * | 2018-09-27 | 2023-10-27 | 微软技术许可有限责任公司 | Joint optimization of notifications and pushes |
CN110955840A (en) * | 2018-09-27 | 2020-04-03 | 微软技术许可有限责任公司 | Joint optimization of notifications and pushes |
CN109582953B (en) * | 2018-11-02 | 2023-04-07 | 中国科学院自动化研究所 | Data support scoring method and equipment for information and storage medium |
CN109582953A (en) * | 2018-11-02 | 2019-04-05 | 中国科学院自动化研究所 | A kind of speech of information is according to support methods of marking, equipment and storage medium |
CN109658277A (en) * | 2018-11-30 | 2019-04-19 | 华南师范大学 | A kind of science social networks friend recommendation method, system and storage medium |
CN109658277B (en) * | 2018-11-30 | 2022-12-27 | 华南师范大学 | Academic social network friend recommendation method, system and storage medium |
CN109933666A (en) * | 2019-03-18 | 2019-06-25 | 西安电子科技大学 | A kind of good friend's automatic classification method, device, computer equipment and storage medium |
CN109933666B (en) * | 2019-03-18 | 2022-12-30 | 西安电子科技大学 | Automatic friend classification method and device, computer equipment and storage medium |
CN111143701A (en) * | 2019-12-13 | 2020-05-12 | 中国电子科技网络信息安全有限公司 | Social network user recommendation method and system based on multiple dimensions |
CN111368217A (en) * | 2020-03-19 | 2020-07-03 | 淄博聚德网络科技有限公司 | Social method and system for promoting strangers to develop to friends |
CN111475724A (en) * | 2020-04-01 | 2020-07-31 | 上海硕恩网络科技股份有限公司 | Random walk social network event recommendation method based on user similarity |
CN111651671B (en) * | 2020-05-27 | 2023-11-21 | 腾讯科技(深圳)有限公司 | User object recommendation method, device, computer equipment and storage medium |
CN111651671A (en) * | 2020-05-27 | 2020-09-11 | 腾讯科技(深圳)有限公司 | User object recommendation method and device, computer equipment and storage medium |
CN112100489B (en) * | 2020-08-27 | 2022-07-15 | 北京百度网讯科技有限公司 | Object recommendation method, device and computer storage medium |
CN112100489A (en) * | 2020-08-27 | 2020-12-18 | 北京百度网讯科技有限公司 | Object recommendation method, device and computer storage medium |
CN112487303A (en) * | 2020-11-26 | 2021-03-12 | 杭州电子科技大学 | Topic recommendation method based on social network user attributes |
CN112487303B (en) * | 2020-11-26 | 2022-04-22 | 杭州电子科技大学 | Topic recommendation method based on social network user attributes |
CN112507247A (en) * | 2020-12-15 | 2021-03-16 | 重庆邮电大学 | Cross-social network user alignment method fusing user state information |
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 |
CN113360764A (en) * | 2021-06-23 | 2021-09-07 | 甄付(上海)网络科技有限公司 | Social network construction method and system based on life service consumption scene |
CN113836444A (en) * | 2021-09-30 | 2021-12-24 | 中国科学院深圳先进技术研究院 | Linear time friend recommendation method, system, terminal and storage medium |
CN113836444B (en) * | 2021-09-30 | 2024-04-19 | 中国科学院深圳先进技术研究院 | Linear time friend recommendation method, system, terminal and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107526850A (en) | Social networks friend recommendation method based on multiple personality feature mixed architecture | |
US11070643B2 (en) | Discovering signature of electronic social networks | |
US8775418B2 (en) | Method for social search | |
CN108763314A (en) | A kind of interest recommends method, apparatus, server and storage medium | |
Pei-Lee et al. | Do the big five personality factors affect knowledge sharing behaviour? a study of malaysian univeristies | |
US8359276B2 (en) | Identifying influential persons in a social network | |
CN104778173B (en) | Target user determination method, device and equipment | |
CN103188139B (en) | A kind of information displaying method of recommending friends and device | |
CN111428147A (en) | Social recommendation method of heterogeneous graph volume network combining social and interest information | |
CN109241440A (en) | It is a kind of based on deep learning towards implicit feedback recommended method | |
CN107424043A (en) | A kind of Products Show method and device, electronic equipment | |
CN106886518A (en) | A kind of method of microblog account classification | |
CN103823888B (en) | Node-closeness-based social network site friend recommendation method | |
US20130024813A1 (en) | Method, system, and means for expressing relative sentiments towards subjects and objects in an online environment | |
CN110162717A (en) | A kind of method and apparatus of commending friends | |
JP2012168653A (en) | Information providing system | |
CN107767279A (en) | A kind of average weighted personalized friend recommendation method based on LDA | |
CN102332025A (en) | Intelligent vertical search method and system | |
CN107562947A (en) | A kind of Mobile Space-time perceives the lower dynamic method for establishing model of recommendation service immediately | |
CN103353920A (en) | Method and device for recommending games based on SNS | |
CN109726747A (en) | Recommend the data fusion sort method of platform based on social networks | |
CN107657043A (en) | A kind of combination chart model image based on content recommends method | |
CN105045822A (en) | Method for monitoring similar users of specific user in microblog | |
CN111831918A (en) | Mixed recommendation system in online education scene | |
Michail et al. | Bucketfood: A crowdsourcing platform for promoting gastronomic tourism |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20171229 |