CN108052961A - A kind of Multifactor Decision Making method that activity social network user activity is recommended - Google Patents
A kind of Multifactor Decision Making method that activity social network user activity is recommended Download PDFInfo
- Publication number
- CN108052961A CN108052961A CN201711224583.6A CN201711224583A CN108052961A CN 108052961 A CN108052961 A CN 108052961A CN 201711224583 A CN201711224583 A CN 201711224583A CN 108052961 A CN108052961 A CN 108052961A
- Authority
- CN
- China
- Prior art keywords
- activity
- user
- sponsor
- participated
- movable
- 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
- 230000000694 effects Effects 0.000 title claims abstract description 212
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000010276 construction Methods 0.000 claims abstract description 10
- 238000003066 decision tree Methods 0.000 claims abstract description 6
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 5
- 230000014509 gene expression Effects 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 7
- 238000000205 computational method Methods 0.000 claims description 5
- 241000287196 Asthenes Species 0.000 claims description 3
- 230000002123 temporal effect Effects 0.000 abstract 1
- 238000001914 filtration Methods 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 4
- 235000010627 Phaseolus vulgaris Nutrition 0.000 description 3
- 244000046052 Phaseolus vulgaris Species 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 238000007418 data mining Methods 0.000 description 2
- 230000009916 joint effect Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 238000006424 Flood reaction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 238000005295 random walk Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2323—Non-hierarchical techniques based on graph theory, e.g. minimum spanning trees [MST] or graph cuts
-
- 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)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- General Engineering & Computer Science (AREA)
- Economics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Discrete Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Marketing (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 present invention is a kind of Multifactor Decision Making method that movable social network user activity is recommended, and its step are as follows:Facing movament social networks EBSN construction activities sponsor user's ESU graph models calculate the social influence power that activity generates user in figure;The activity description correlation that calculating activity and user participated in;The activity venue correlation that calculating activity and user participated in;The activity time correlation that calculating activity and user participated in;On the basis of aforementioned four factor, determine whether user can activity using classical J48 decision Tree algorithms.The method of the present invention proposes social influence power, content relevance, place correlation and the temporal correlation Multiple factors using activity and user, predict user whether activity, the feature of movable social networks EBSN is reasonably make use of, is suitble to the activity recommendation of activity social networks EBSN.
Description
Technical field
The present invention relates to a kind of information services, specifically, are related to a kind of movable social network user activity
The Multifactor Decision Making method of recommendation.
Background technology
In recent years, movable social networks (Event-based Social Networks, be abbreviated as EBSNs) caused and ground
The concern for the person of studying carefully is primarily due to its platform for providing tissue to the user by online mode, participating in and share social activity,
And user can also participate in real Below-the-line, such as Meetup, Plancast, Facebook Event and bean cotyledon it is same
City etc..The many activities held on EBSNs in different time and place, user need to take a significant amount of time and could find
To oneself interested activity.On EBSNs, realization activity is precisely recommended from trend user, and then user is allowed to enrich amateurish life
Living, expansion social networks and the amusement of enjoyment team have important practical significance.
Unlike previous social networks, EBSNs has the characteristics of its is exclusive:(1) activity is mostly is held by sponsor
, the activity held contain Activity Type, activity description, the activity time, activity venue and to activity it is interested/participate in
People etc. information, movable intension are more abundant;(2) there is substantial amounts of entity on EBSNs, including activity, user and sponsor etc.,
And distinctive complicated social networks are constructed between these entities;(3) user's activity is not only attracted by activity description,
And it is influenced by factors such as social networks, movable when and wheres.
In the proceeding that the U.S. in 2012 publishes:18th Knowledge Discovery in 2012 and data mining meeting (18th
ACM SIGKDD conference on Knowledge Discovery and Data Mining), it is entitled:It is movable social
Network:Connect the online and offline social activity world (Event-based social networks:linking the online
And offline social worlds), author is Liu X, He Q, Tian Y, Lee WC, McPherson J, Han J,
This article defines the concept of movable social networks EBSNs for the first time, it is believed that it is connected to the social world on line and on line, is one
The new social media of kind.
In the proceeding that the U.S. in 2015 publishes:General fit calculation meeting (ACM International in 2015
Joint Conference on Pervasive and Ubiquitous Computing), it is entitled:Whom I will invite participate in
PartyIt merges user preference and influences maximized social activity (Who should I invite for my party
Combining user preference and influence maximization for social events), author
It is Yu Z, Du R, Guo B, Xu H, Gu T, Wang Z, Zhang D, this article, which extends trust distribution model identification, most to be influenced
The invitee of power, it is contemplated that user maximizes the preference and social influence of activity.
In the periodical that Holland in 2017 publishes:Nerve calculates (Neurocomputing), entitled:Movable social networks is used
In mixing collaborative filtering model (the A hybrid collaborative filtering model for of social influence prediction
Social influence prediction in event-based socialnetworks), author is Li X, Cheng
X, Su S, Li SC, Yang JY, this article propose the neighbours of mixing collaborative filtering model M F-EUN, the activity of having merged and user,
In order to solve the openness of social influence matrix, it is proposed that the neighbor discovering method based on additional information.Its user to EBSNs
Between, the modeling between user and activity, between activity and the side of holding it is not accurate enough, and do not account for element of time to work
The dynamic influence recommended.
Inventor has found that the prior art is in terms of movable social network user activity recommendation, to movable social networks
Modeling it is not accurate enough, to the signature analysis of movable social networks using not comprehensive enough.
The content of the invention
Problem and shortage existing in the prior art in view of the above, the technical problem to be solved in the present invention is to provide activities
The Multifactor Decision Making method that social network user activity is recommended, this method more reasonably build movable social networks
Mould, the feature that comprehensively make use of movable social networks are more suitable for the recommendation of movable social network user activity.
To solve the above-mentioned problems, the present invention uses following technical proposals:
The present invention is the Multifactor Decision Making method that movable social network user activity is recommended, and is as follows:
A, facing movament social networks EBSN construction activities-sponsor-user's ESU graph models calculate activity e in figureiIt is right
User ujThe social influence power si (e of generationi, uj);
B, calculating activity eiWith user ujThe activity description correlation c (e participated ini, uj);
C, calculating activity eiWith user ujThe activity venue correlation l (e participated ini, uj);
D, calculating activity eiWith user ujThe activity time correlation t (e participated ini, uj);
E, on the basis of aforementioned four factor, user u is determined using classical J48 decision Tree algorithmsjWhether work can be participated in
Dynamic ei。
In the Multifactor Decision Making method that a kind of movable social network user activity of the present invention is recommended, the step
Suddenly facing movament social networks EBSN construction activities-sponsor-user's ESU graph models described in (A) calculate activity e in figurei
To user ujThe social influence power si (e of generationi, uj), preferably it is as follows:
A1, construction activities-sponsor-user's ESU graph models, ESU graph models are described as two tuples:ESU=
(Entity, Relation), wherein, Entity represents entity, collects U including active set E, sponsor collection S and user;Relation
Representation relation has sponsor-activity relationship collection SE, sponsor-customer relationship collection SU, user-activity relationship collection UE and user
Between set of relations UU;
A2, sponsor-activity relationship weight is calculated, makes UN (sj) it is sponsor sjComprising number of users, make UN (ei) be
sjIn to movable eiInterested/number of users for participating in, then sponsor sj- activity eiRelation weight be:
A3, sponsor-customer relationship weight is calculated, makes SN (ui) it is user uiSponsor's quantity of addition, EN (sj) based on
The side of doing sjThe amount of activity once held, SEN (ui) it is user uiThe sponsor s once participated injAmount of activity, then sponsor
sj- user uiRelation weight be:
A4, user-activity relationship weight is calculated, makes E (uj) it is user ujThe active set participated in, sjFor movable eiInstitute
Sponsor, ES (uj) it is user ujParticipated in sjThe active set held, then user uj- activity eiRelation weight be:
A5, user-user relation weight is calculated, it is the active set that user u is participated in make E (u), and E (v) participates in for user v
Active set, S (u) be user u add in sponsor set, S (v) be user v add in sponsor set, then user u- use
The relation weight of family v is:
A6, ESU Tu Zhong sponsors, user and the iterative calculation of the influence power of activity are respectively such as formula (1-1), (1-2), (1-3)
It is shown:
Sk=c (WSEEk-1+WSUUk-1)+(1-c)sej (1-1)
Uk=c (WSUSk-1+WUUUk-1)+(1-c)uej (1-2)
Ek=c (WSESk-1+WUEUk-1)+(1-c)eej (1-3)
Se in formula (1-1), (1-2), (1-3)j、uejAnd eejRespectively jth dimension is 1 unit vector, and 1-c is return
Node sej、uejAnd eejProbability, WSE、WSU、WUUAnd WUERespectively sponsor-activity, sponsor-user, user-user and
The relation weight of user-activity.
In the Multifactor Decision Making method that a kind of movable social network user activity of the present invention is recommended, the step
Suddenly the calculating activity e described in (B)iWith user ujThe activity description correlation c (e participated ini, uj), preferred specific steps
It is as follows:
B1, user ujThe active set once participated in is denoted as E (uj), use the phase of the type label calculating activity of activity
Like property, CF (ei) expression activity eiType label set, CF (E (uj)) represent user ujParticipated in the type label collection of activity
It closes;
B2, activity eiWith E (uj) similarity calculating method be
In the Multifactor Decision Making method that a kind of movable social network user activity of the present invention is recommended, the step
Suddenly the calculating activity e described in (C)iWith user ujThe activity venue correlation l (e participated ini, uj), preferred specific steps
It is as follows:
C1、Expression activity eiPlace,Represent user ujPlace,It representsWithDistance;
C2, activity eiPlace and user ujThe correlation calculations method in place is:
In the Multifactor Decision Making method that a kind of movable social network user activity of the present invention is recommended, the step
Suddenly the calculating calculating activity e described in (D)iWith user ujThe activity time correlation t (e participated ini, uj), it is preferred specific
Step is as follows:
It is D1, that user was once participated in movable eiThe active set of same type is denoted as E (uj) (each activity has class
Type label), movable eiTime of holding be denoted asThe same type activity e that user participated ink′∈E(uj) time of holding be denoted asThen activity eiTime and same type activity ek' time deviation be:
D2, activity eiWith user ujTime deviation computational methods be:
Min expressions are minimized.
Compared with prior art, the Multifactor Decision Making method that movable social network user activity of the invention is recommended,
Modeling, the feature that comprehensively make use of movable social networks have more reasonably been carried out to movable social networks, has been more suitable for movable society
Hand over the recommendation of network user's activity.
Description of the drawings
Fig. 1 is the flow chart for the Multifactor Decision Making method that the movable social network user activity of the present invention is recommended;
Fig. 2 is the facing movament social networks EBSN construction activities-sponsor-user ESU figures described in step 101 in Fig. 1
Model calculates activity e in figureiTo user ujThe social influence power s of generationi(ei, uj) flow chart.
Specific embodiment
The implementation process of the present invention is described in further detail with reference to the accompanying drawings and detailed description.
With reference to the Multifactor Decision Making method that Fig. 1, movable social network user activity of the invention recommend, this method bag
Include following steps:
Step 101, facing movament social networks EBSN construction activities-sponsor-user's ESU graph models calculate living in figure
Dynamic eiTo user ujThe social influence power si (e of generationi, uj), with reference to Fig. 2, it is as follows:
Step 201, construction activities-sponsor-user's ESU graph models, ESU graph models are described as two tuples:ESU=
(Entity, Relation), wherein, Entity represents entity, collects U including active set E, sponsor collection S and user;Relation
Representation relation has sponsor-activity relationship collection SE, sponsor-customer relationship collection SU, user-activity relationship collection UE and user
Between set of relations UU;
Step 202 makes UN (sj) it is sponsor sjComprising number of users, make UN (ei) it is sjIn to movable eiInterested/
The number of users of participation.Then sponsor sj- activity eiRelation weight be:
Formula (1) make use of sponsor sjMiddle activity eiNumber of users and sjIn all number of users measurements sponsor
Side is to the relation weight of activity.se(sj, ei) value is bigger, illustrate sponsor sjIn there is more user to participate in sjThe activity held.
Step 203 makes SN (ui) it is user uiSponsor's quantity of addition, EN (sj) it is sponsor sjThe work once held
Dynamic quantity, SEN (ui) it is user uiThe sponsor s once participated injAmount of activity, then sponsor sj- user uiRelation weight
For:
Formula (2) first half illustrates user uiParticipate in sponsor sjMore, the sponsor s of activityjTo user uiInfluence get over
Greatly;Formula (2) latter half takes into account user uiOther sponsor's quantity added in, intuitively for, what user participated in
Sponsor is more, sponsor sjTo the influence power of the user with regard to smaller.
Step 204 makes E (uj) it is user ujThe active set participated in, sjFor movable eiThe sponsor at place, ES (uj) be
User ujParticipated in sjThe active set held.Then user uj- activity eiRelation weight be:
Formula (3) illustrates user ujParticipated in activity eiMore, the user u of activity of the sponsor at placejTo movable eiInfluence
It is bigger.
Step 205 makes E (u) be the active set that user u is participated in, and E (v) is the active set that user v is participated in, and S (u) is
Sponsor's set that user u is added in, S (v) are sponsor's set that user v is added in, then the relation weight of user u- user v is:
The joint activity and the situation of the common sponsor that adds in that the considerations of formula (4) is comprehensive user u and user v is participated in, α
And β is the parameter for balancing weight between joint activity and common sponsor.
It should be noted that in turn, user v is different to the weight of user u, because v and u may be to different number
Sponsor it is interested, it is also possible to participate in the activity of different number, but computational methods are similar with formula (4).
Step 206, ESU figures are digraph, and the importance of three classes entity embodies mode difference.Sponsor siImportance take
Certainly in siAll activities for holding and comprising the influence that is generated to it of all users, therefore siCalculating need to consider its out-degree
(Hubs), what PageRank considered is the in-degree of the page.The importance of movable e nodes depends on sponsor and user produces it
Raw influence.The importance of user's u node depends on sponsor and pays close attention to the influence that his user generates it.Therefore, ESU schemes
Activity and user's importance calculating, it is contemplated that its in-degree (Authorities).
Therefore, when calculating the importance of all kinds of nodes of ESU, it is necessary to be iterated from two angles of out-degree and in-degree
It calculates.For this purpose, we have proposed the two-way pitch point importance calculating side for restarting Random Walk Algorithm (BD-RWR) is based on ESU figures
Method.Importance, user and the iterative calculation of the importance of activity of campaign owners are respectively as shown in formula (5), (6), (7):
Sk=c (WSEEk-1+WSUUk-1)+(1-c)sej (5)
Uk=c (WSUSk-1+WUUUk-1)+(1-c)uej (6)
Ek=c (WSESk-1+WUEUk-1)+(1-c)eej (7)
Se in formula (5), (6), (7)j、uejAnd eejThe unit vector that respectively jth dimension is 1.1-c is return node
sej、uejAnd eejProbability.WSE、WSU、WUUAnd WUERespectively sponsor-activity, sponsor-user, user-user and user-
The relation weight of activity.
After iteration several times, each user u is obtainediPitch point importance, pitch point importance is bigger, illustrates movable ei
Bigger to the social influence power of user's generation on ESU figures, user is more possible to activity ei。
Step 102, calculating activity eiWith user ujThe activity description correlation c (e participated ini, uj):
Give a specific activities eiAfterwards, it is necessary to consider e when user oriented is recommendediActivity description and user ujOnce joined
Added the similitude of activity.Assuming that user ujThe active set once participated in is denoted as E (uj)。
Has method in calculating activity eiMovable E (the u participated in userj) similarity when, be mostly using LDA models
Theme is extracted from activity, and then uses the similitude of cosine similarity metrology activities.
But since the EBSNs means of activity are often more terse, this has aggravated the openness of text information, it is not easy to from text
In accurately extract theme.And all activities are mixed and extract theme, easily the small activity of user's participation amount
It floods.
Use the similitude of the type label calculating activity of activity, movable eiWith E (uj) similarity calculating method such as formula
(8) shown in:
In formula (8), CF (ei) expression activity eiType label set, CF (E (uj)) represent user ujParticipated in activity
Type label set.
Step 103, calculating activity eiWith user ujThe activity venue correlation l (e participated ini, uj):
Activity venue refers to the true place held under active line.Unlike traditional social media, the act of Below-the-line
It is the characteristics of EBSNs is exclusive to do.Therefore, locality factors is to deciding whether to have important role to user's recommendation activity.
Movable eiPlace and user ujShown in the correlation calculations in place such as formula (9):
In formula (9),Expression activity eiPlace,Represent user ujPlace,It representsWithAway from
From.All there are one the distances of longitude and the numerical value of dimension, easily two places of calculating in place.
Step 104, calculating activity eiWith user ujThe activity time correlation t (e participated ini, uj);
People's activity is limited to the influence of time factor, normally behaves as the periodicity in day and week.For example, custom is every
Its After Hours activity or each weekend activity.Therefore, time factor to user can activity have it is important
It influences.
It is that user was once participated in movable eiThe active set of same type is denoted as E (uj) (each activity has type mark
Label), movable eiTime of holding be denoted asThe same type activity e that user participated ink′∈E(uj) time of holding be denoted asThen
Movable eiTime and same type activity ek' time deviation be:
Movable eiWith user ujTime deviation computational methods such as formula (11) shown in:
Formula (11) is in calculating activity eiWith user ujTime deviation when, on the one hand consider activity type because different
Time, the activity that user participates in is different, for example is attended a party at night, and weekend participates in tourism;On the other hand, using use
Family ujIn the activity participated in eiTime hithermost activity time deviation is held as module, it can be to avoid averaging
The deviation that value is brought.
Step 105, on the basis of aforementioned four factor, classical J48 decision Tree algorithms is used to determine user ujWhether can
Activity ei。
Recommend method using four kinds of different movable social network user activities, compare the effective of user's activity
Property.Four kinds of methods are as follows:
● basic matrix disassembling method, it only considers the participation of user-activity as a result, the weights of element use in matrix
The formula (3) of this specification calculates, and is abbreviated as MF;
● the relevant belief propagation of matrix decomposition has been used, user-activity has been incorporated and participates in result and simple social networks,
The structure of user social contact relation is abbreviated as SocialMF, parameter setting using the formula (4) of this specification;
● mixing collaborative filtering model is abbreviated as MF-EUN-ER, the threshold value of influence power is arranged to 0.5, movable neighbours k1
=50, the neighbours k of user2=60, region clustering number t=20;
● since traditional decision-tree readability is good, contribute to manual analysis, so being examined using classical J48 traditional decision-trees
The effect for putting forward 4 factors herein is tested, is abbreviated as F4-DT-ER;
The present invention chooses data of the bean cotyledon with city as experiment language material.Bean cotyledon has more than 3,000 ten thousand users, and line with city
The development liveness of lower activity is also very high.
In order to reduce the workload of data analysis, while in order to avoid some inactive users or activity are to overall experiment
As a result the very sluggish user of some of which and activity are deleted in influence.Interested/activity is less than the use of 5
Family is considered cold start-up user, accounts for the 9% of all total numbers of users;If some activity it is interested/number participated in is less than if 8
It is considered cold start-up activity, accounts for the 6% of all activity sums.After being pre-processed to the data of 2 months, it is obtained
8663 users, 17822 activities, 294 sponsors.
In EBSNs, user, with participating in or not participating in represent, can turn movable participation problem for some activity
It is changed to two classification problems.During experiment, 80% in data set is randomly selected for training, and others 20% are used to test.
Using F1-measure=(2 × P × R)/(P+R) as evaluation metrics, wherein P is recommended user's activity
Accuracy rate, R are the recall rates of recommended user's activity.
For the recommendation total result of user's activity, the results are shown in Table 1 for four kinds of methods acquisitions.
The global recommendation result that 1. 4 kinds of methods of table obtain
By table 1 as it can be seen that the result of method F4-DT-ER proposed by the present invention acquisitions is ideal, F1-measure=
0.86, followed by MF-EUN-ER methods, F1-measure=0.80.In four kinds of methods, the effect of MF is worst, F1-measure
=0.67, main cause is its only relation between user and activity, has ignored a lot of other passes on EBSNs
System, similar to the recommendation method between legacy user-project, this also illustrates that traditional recommendation method is joined for the user of EBSNs
It is poor to add activity recommendation effect.The effect of method SocialMF is promoted, F1-measure=0.72, the reason is that its
Take into account social networks, it is contemplated that the belief propagation between user, but it is only the social networks considered between user,
The relation other entities using not enough.Method MF-EUN-ER has the special feature that according to EBSNs, it is contemplated that activity description,
Activity venue and social networks, obtained result have and significantly improve, and F1-measure values improve 0.13 than NMF,
0.08 is improved than SocialMF.Social pass exclusive deep institute extracting method F4-DT-ER of the present invention analysis and utilization EBSNs
It is feature, it is proposed that the user force computational methods of ESU figures make use of content, place and the time factor of activity well,
Achieve best effect.
This explanation, it is necessary to the characteristics of catching its exclusive for EBSNs, it is comprehensive the considerations of many aspects factor, ability
Obtain better user's activity recommendation results.And the ESU charts representation model of the invention carried, social influence power calculating side
Method, activity recommendation Multifactor Decision Making model are all the characteristics of EBSNs institutes are exclusive.
Claims (5)
1. a kind of Multifactor Decision Making method that activity social network user activity is recommended, is as follows:
A, facing movament social networks EBSN construction activities-sponsor-user's ESU graph models calculate activity e in figureiTo user uj
The social influence power si (e of generationi, uj);
B, calculating activity eiWith user ujThe activity description correlation c (e participated ini, uj);
C, calculating activity eiWith user ujThe activity venue correlation l (e participated ini, uj);
D, calculating activity eiWith user ujThe activity time correlation t (e participated ini, uj);
E, on the basis of aforementioned four factor, user u is determined using classical J48 decision Tree algorithmsjWhether can activity ei。
2. the Multifactor Decision Making method that a kind of movable social network user activity according to claim 1 is recommended,
It is characterized in that, facing movament social networks EBSN construction activities-sponsor-user's ESU graph models described in step (A), counts
Activity e in nomogramiTo user ujThe social influence power si (e of generationi, uj), it is as follows:
A1, construction activities-sponsor-user's ESU graph models, ESU graph models are described as two tuples:ESU=(Entity,
Relation), wherein, Entity represents entity, collects U including active set E, sponsor collection S and user;Relation, which is represented, to close
System, has sponsor-activity relationship collection SE, sponsor-customer relationship collection SU, user-to be closed between activity relationship collection UE and user
Assembly UU;
A2, sponsor-activity relationship weight is calculated, makes UN (sj) it is sponsor sjComprising number of users, make UN (ei) it is sjIn
To movable eiInterested/number of users for participating in, then sponsor sj- activity eiRelation weight be:
A3, sponsor-customer relationship weight is calculated, makes SN (ui) it is user uiSponsor's quantity of addition, EN (sj) it is sponsor
sjThe amount of activity once held, SEN (ui) it is user uiThe sponsor s once participated injAmount of activity, then sponsor sj- use
Family uiRelation weight be:
A4, user-activity relationship weight is calculated, makes E (uj) it is user ujThe active set participated in, sjFor movable eiThe master at place
The side of doing, ES (uj) it is user ujParticipated in sjThe active set held, then user uj- activity eiRelation weight be:
A5, user-user relation weight is calculated, it is the active set that user u is participated in make E (u), and E (v) is the work that user v is participated in
Dynamic set, S (u) are sponsor's set that user u is added in, and S (v) is sponsor's set that user v is added in, then user u- user v
Relation weight be:
A6, ESU Tu Zhong sponsors, user and the iterative calculation of the influence power of activity are respectively as shown in formula (1-1), (1-2), (1-3):
Sk=c (WSEEk-1+WSUUk-1)+(1-c)sej (1-1)
Uk=c (WSUSk-1+WUUUk-1)+(1-c)uej (1-2)
Ek=c (WSESk-1+WUEUk-1)+(1-c)eej (1-3)
Se in formula (1-1), (1-2), (1-3)j、uejAnd eejRespectively jth dimension is 1 unit vector, and 1-c is return node
sej、uejAnd eejProbability, WSE、WSU、WUUAnd WUERespectively sponsor-activity, sponsor-user, user-user and user-
The relation weight of activity.
3. the Multifactor Decision Making method that a kind of movable social network user activity according to claim 1 is recommended,
It is characterized in that, the calculating activity e described in the step (B)iWith user ujThe activity description correlation c (e participated ini, uj),
It is as follows:
B1, user ujThe active set once participated in is denoted as E (uj), using activity type label calculating activity similitude,
CF(ei) expression activity eiType label set, CF (E (uj)) represent user ujParticipated in the type label set of activity;
B2, activity eiWith E (uj) similarity calculating method be
4. the Multifactor Decision Making method that a kind of movable social network user activity according to claim 1 is recommended,
It is characterized in that, the calculating activity e described in above-mentioned steps (C)iWith user ujThe activity venue correlation l (e participated ini, uj),
It is as follows:
C1、Expression activity eiPlace,Represent user ujPlace,It representsWithDistance;
C2, activity eiPlace and user ujThe correlation calculations method in place is:
5. the Multifactor Decision Making method that a kind of movable social network user activity according to claim 1 is recommended,
It is characterized in that, the calculating calculating activity e described in above-mentioned steps (D)iWith user ujThe activity time correlation t (e participated ini,
uj), it is as follows:
It is D1, that user was once participated in movable eiThe active set of same type is denoted as E (uj) (each activity has type mark
Label), movable eiTime of holding be denoted asThe same type activity e that user participated ink′∈E(uj) time of holding be denoted asThen
Movable eiTime and same type activity ek' time deviation be:
D2, activity eiWith user ujTime deviation computational methods be:
Min expressions are minimized.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711224583.6A CN108052961B (en) | 2017-11-28 | 2017-11-28 | Multi-factor decision-making method for activity recommendation of active social network users |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711224583.6A CN108052961B (en) | 2017-11-28 | 2017-11-28 | Multi-factor decision-making method for activity recommendation of active social network users |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108052961A true CN108052961A (en) | 2018-05-18 |
CN108052961B CN108052961B (en) | 2022-05-31 |
Family
ID=62121100
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711224583.6A Active CN108052961B (en) | 2017-11-28 | 2017-11-28 | Multi-factor decision-making method for activity recommendation of active social network users |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108052961B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110889434A (en) * | 2019-10-29 | 2020-03-17 | 东南大学 | Social network activity feature extraction method based on activity |
CN111143700A (en) * | 2018-10-16 | 2020-05-12 | 北京邮电大学 | Activity recommendation method and device, server and computer storage medium |
CN117217808A (en) * | 2023-07-21 | 2023-12-12 | 广州有机云计算有限责任公司 | Intelligent analysis and prediction method for activity invitation capability |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102594905A (en) * | 2012-03-07 | 2012-07-18 | 南京邮电大学 | Method for recommending social network position interest points based on scene |
CN106776959A (en) * | 2016-12-05 | 2017-05-31 | 东北大学 | A kind of activity recommendation system and method based on on-line off-line dual social networks |
CN106980659A (en) * | 2017-03-20 | 2017-07-25 | 华中科技大学鄂州工业技术研究院 | A kind of doings based on isomery graph model recommend method |
-
2017
- 2017-11-28 CN CN201711224583.6A patent/CN108052961B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102594905A (en) * | 2012-03-07 | 2012-07-18 | 南京邮电大学 | Method for recommending social network position interest points based on scene |
CN106776959A (en) * | 2016-12-05 | 2017-05-31 | 东北大学 | A kind of activity recommendation system and method based on on-line off-line dual social networks |
CN106980659A (en) * | 2017-03-20 | 2017-07-25 | 华中科技大学鄂州工业技术研究院 | A kind of doings based on isomery graph model recommend method |
Non-Patent Citations (1)
Title |
---|
TUAN-ANH NGUYEN PHAM等: "A General Graph-based Model for Recommendation in Event-based Social Networks", 《INSTITUTE ELECTRICAL AND ELECTRONICS ENGINEERS》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111143700A (en) * | 2018-10-16 | 2020-05-12 | 北京邮电大学 | Activity recommendation method and device, server and computer storage medium |
CN111143700B (en) * | 2018-10-16 | 2023-07-25 | 北京邮电大学 | Activity recommendation method, activity recommendation device, server and computer storage medium |
CN110889434A (en) * | 2019-10-29 | 2020-03-17 | 东南大学 | Social network activity feature extraction method based on activity |
CN110889434B (en) * | 2019-10-29 | 2023-04-21 | 东南大学 | Social network activity feature extraction method based on activity |
CN117217808A (en) * | 2023-07-21 | 2023-12-12 | 广州有机云计算有限责任公司 | Intelligent analysis and prediction method for activity invitation capability |
CN117217808B (en) * | 2023-07-21 | 2024-04-05 | 广州有机云计算有限责任公司 | Intelligent analysis and prediction method for activity invitation capability |
Also Published As
Publication number | Publication date |
---|---|
CN108052961B (en) | 2022-05-31 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Christakis et al. | An empirical model for strategic network formation | |
Zheng et al. | Trading data in the crowd: Profit-driven data acquisition for mobile crowdsensing | |
Pham et al. | A general graph-based model for recommendation in event-based social networks | |
Yuan et al. | Graph-based point-of-interest recommendation with geographical and temporal influences | |
CN105809554B (en) | Prediction method for user participating in hot topics in social network | |
Wang et al. | Sustainable collaborator recommendation based on conference closure | |
Li et al. | A hybrid collaborative filtering model for social influence prediction in event-based social networks | |
CN106682770A (en) | Friend circle-based dynamic microblog forwarding behavior prediction system and method | |
CN108052961A (en) | A kind of Multifactor Decision Making method that activity social network user activity is recommended | |
Li et al. | Social influence based community detection in event-based social networks | |
Xia et al. | Socially-aware venue recommendation for conference participants | |
Zhan et al. | A novel trust computing system for social networks | |
CN103530428A (en) | Same-occupation type recommendation method based on developer practical skill similarity | |
Liao et al. | An effective latent networks fusion based model for event recommendation in offline ephemeral social networks | |
CN111428127A (en) | Personalized event recommendation method and system integrating topic matching and two-way preference | |
Gu et al. | Context aware matrix factorization for event recommendation in event-based social networks | |
He et al. | Next point-of-interest recommendation via a category-aware Listwise Bayesian Personalized Ranking | |
Caso et al. | Users Ranking in Online Social Networks to Support POIs Selection in Small Groups. | |
Hexmoor | Computational network science: an algorithmic approach | |
CN108182268A (en) | A kind of collaborative filtering recommending method and system based on community network | |
Lu et al. | Collective human behavior in cascading system: discovery, modeling and applications | |
CN109684561A (en) | Based on user register Behavioral change Deep Semantics analysis point of interest recommended method | |
Lv et al. | Incentive networks | |
Teoman et al. | Group oriented trust-aware location recommendation for location-based social networks | |
Karimi et al. | Location recommendation based on location-based social networks for entertainment services |
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 | ||
CB02 | Change of applicant information | ||
CB02 | Change of applicant information |
Address after: 222000 zhongzhaoman transfer of computer school of Huaihai Institute of technology, No. 59 Cangwu Road, Haizhou District, Lianyungang City, Jiangsu Province Applicant after: Jiangsu Ocean University Address before: 222000 zhongzhaoman transfer of computer school of Huaihai Institute of technology, No. 59 Cangwu Road, Haizhou District, Lianyungang City, Jiangsu Province Applicant before: HUAIHAI INSTITUTE OF TECHNOLOGY |
|
GR01 | Patent grant | ||
GR01 | Patent grant |