CN107808168A - A kind of social network user behavior prediction method based on strong or weak relation - Google Patents
A kind of social network user behavior prediction method based on strong or weak relation Download PDFInfo
- Publication number
- CN107808168A CN107808168A CN201711050008.9A CN201711050008A CN107808168A CN 107808168 A CN107808168 A CN 107808168A CN 201711050008 A CN201711050008 A CN 201711050008A CN 107808168 A CN107808168 A CN 107808168A
- Authority
- CN
- China
- Prior art keywords
- user
- strong
- weak
- relation
- state
- 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
- 238000000034 method Methods 0.000 title claims abstract description 26
- 230000003993 interaction Effects 0.000 claims abstract description 9
- 230000008451 emotion Effects 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 7
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 claims description 3
- 230000002452 interceptive effect Effects 0.000 abstract description 4
- 230000015572 biosynthetic process Effects 0.000 abstract description 2
- 238000003786 synthesis reaction Methods 0.000 abstract description 2
- 210000003813 thumb Anatomy 0.000 abstract description 2
- 230000006399 behavior Effects 0.000 description 27
- 230000006870 function Effects 0.000 description 8
- 230000008901 benefit Effects 0.000 description 5
- 238000011160 research Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 3
- 230000006855 networking Effects 0.000 description 3
- 230000007935 neutral effect Effects 0.000 description 3
- 238000000926 separation method Methods 0.000 description 3
- 230000007704 transition Effects 0.000 description 3
- 230000003542 behavioural effect Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000003924 mental process Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000010924 continuous production Methods 0.000 description 1
- 230000001276 controlling effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 230000005059 dormancy Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
- G06F16/353—Clustering; Classification into predefined classes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F18/2148—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the process organisation or structure, e.g. boosting cascade
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/29—Graphical models, e.g. Bayesian networks
- G06F18/295—Markov models or related models, e.g. semi-Markov models; Markov random fields; Networks embedding Markov models
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Human Resources & Organizations (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Databases & Information Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides a kind of social network user behavior prediction method based on strong or weak relation, belongs to social networks studying technological domain.This method input data is social network user data, status information that personal information, user including user are delivered, user such as thumb up and commented at the interactive information, the behavior for user, specially each shape probability of state and ectocine and state are exported to the probability of behavior.This method, which by social networks strong or weak relation is measured and built, to be introduced the continuous time hidden Markov model of exogenousd variables and is finally completed the output of user behavior.This method structure considers the psychological factor of deep layer and the synthesis strong or weak relation model frameworks of semantic information such as interaction statistics between user property, user, speech interaction and image media, improves the relation power measure in the past based on simple data statistics.
Description
Technical field
The present invention relates to social networks studying technological domain, particularly relates to a kind of social network user based on strong or weak relation
Behavior prediction method.
Background technology
The research of psychology and sociology to social networks strong or weak relation has many wide variety of theories.Early in 1967
Year, social psychologist of Harvard University Stanley Milgram just propose six degree of separation theorems (Six Degrees of
Separation), it is indicated that the people that you are spaced between any one stranger not over six [1].Society of the U.S. in 1973
Can scholar, economist M.Granovetter define relation (tie) refer between interpersonal, tissue and tissue due to
Exchange and contact and a kind of existing tie contact, and propose binding time amount, emotion compactness, familiarity and reciprocal exchange
The relationship strength definition of four standards, and the overlapping degree [2] of a feasible determination methods, i.e. friend's circle.Nicholas
A.Christakis and James H.Fowler propose that within three degree be strong ties, and strong ties can trigger behavior；Meet
It is Weak link more than three degree, Weak link transmission information [3].A.L.Kavanaugh et al. is by analyzing Blacksburg groups of people
Group's survey data obtains, and weak relation can help individual to expand social scope and information exchange frequency [4] to a certain extent.
The presence of weak relation compensate for the structural hole [5] between strong network of personal connections.In recent years, the research of annexation obtains in social networks
More and more extensive attention.Pasquale De Meo et al. show in the research to Facebook data in 2014, social network
Most of in network is all Weak link, and it is most important [6] to excavate these Weak links.Gillian M.Sandstrom et al. pass through the heart
Neo-Confucianism experiment, highlights the strength of Weak link, it was demonstrated that influence [7] of the Weak link to social individual.Jeffrey Boas et al.
The important function [8] of " dormancy relation " in society's connection is safeguarded was pointed out by psychology and sociological experimentation in 2015.This
A little psychology and sociological achievement in research are all the theoretical references of the present invention.
With the development of social networks, the gradual networking of social effectiveness [9].It is pre- that Gilbert et al. is based on social media
The power [10] of survey relation, it does not consider deep layer by obtaining relation power to some simple statistics amount linear combinations
Psychological factor and semantic information, simple linear combination be also difficult to fully describe the strong or weak relation of complexity.Onnela et al.
Structure, interactive mode based on the communication datas such as message registration, mail daily record prediction social networks in mobile communication internet
[11], the factors that it is considered such as intensity are single with interacting, it is difficult to weigh the power of relation comprehensively.Pappalardo et al. is based on
Online social networks proposes a kind of various dimensions of relationship strength and defines method [12], but its be more from network structure,
Consideration is lacked to semanteme for being exchanged between user etc..In terms of social network influence force modeling, existing researcher is according to mutual
The foundation such as user annotation, microblogging in networking, the article delivered influences model, and the propagation model [13,14,15] influenceed.
Markus Germar et al. trial establishes social influence with diffusion analysis and perceives decision model [16].These influence power moulds
Type often not from Psychological Angle, considers complicated between user itself psychology, ectocine and itself behavior
Relation, it is insufficient in terms of the degree of accuracy and interpretation, and these are exactly that the present invention is of concern.
The present invention refers to psychology and sociological achievement in research, according to the attribute of user and behavior, and two users
Between various interbehaviors, a series of Weak Classifier (Weak classifier) is built, using Gradient
Boosting algorithms are merged to obtain user's strong or weak relation data, consider user property, user mutual, multimedia messages
Deng factors.On this basis, study influence of the strong or weak relation to user behavior, comprehensive continuous time, mixed feeling etc. because
Element, it is proposed that introduce the continuous time hidden Markov model of exogenousd variables, can not only predict social network user row exactly
Also to have interpretation well.
Bibliography：
[1] [U.S.] Watts writes, and Chen Yu etc. translates six degree of separations of：The science [M] in the epoch of-individual interconnection, People's University
Publishing house, 2011.
[2]Granovetter M.S.:The Strength of Weak Ties[J].American Journal of
Sociology,78(6):1360-1380,1973.
[3] Scottus Te Laideng [is added] to write, Wei Wei translates the strong relations of：The key [M] of socialization sale to win, the people
University press, 2012.
[4]Kavanaugh A.L.,Reese D.D.,Carroll J.M.,Rosson M.B.:Weak Ties in
Networked Communities[J].The Information Society,21(2):119-131,2005.
[5]Ronald S.Burt.Structural Holes:The social Structure of Competition
[M].Harvard University Press,1992.
[6]Pasquale De Meo,Emilio Ferrara,Giacomo Fiumara,AlessandroProvetti:
On Facebook,most ties are weak.Communications of the ACM.57(11):78-84,2014.
[7]Gillian M.Sandstrom,Elizabeth W.Dunn:Social Interactions and Well-
Being The Surprising Power of Weak Ties.Personality and Social Psychology
Bulletin,2014.
[8]Jeffrey Boase,Tetsuro Kobayashi,Andrew Schrock,et al.Reconnecting
Here and There The Reactivation of Dormant Ties in the United States and
Japan.American Behavioral Scientist,2015.
[9]AnatoliyGruzd,Barry Wellman:Networked Influence in Social Media
Introduction to the Special Issue.American Behavioral Scientist,58(10):1251-
1259,2014.
[10]Gilbert E.,Karahalios K.:Predicting tie strength with social
media.SIGHCI,pp.211-220,2009.
[11]Onnela J.P.,Saramaki J.,Hyvonen J.,Szabo G.,Lazer D.,Kaski K.,
Kertesz J.,Barabasi A.L.:Structure and tie strengths in mobile communication
networks.Proceedings of the National Academy of Sciences of the United States
of America,104(18):7332-7336,2007.
[12]Pappalardo L.,Rossetti G.,Pedreschi D.:How Well Do We Know Each
OtherDetecting Tie Strength in Multidimensional Social Networks.IEEE/ACM
International Conference on Advances in Social Networks Analysis and Mining,
pp.1040-1045,2012.
[13]Anagnostopoulos A.,Kumar R.,Mahdian M.:Influence and correlation
in social networks.ACM SIGKDD,pp.7-15,2008.
[14]La Fond T.,Neville J.:Randomization tests for distinguishing
social influence and homophily effects.International Conference on World Wide
Web,pp.601–610,2010.
[15]Gomez-Rodriguez M.,Leskovec J.,Krause A.:Inferring networks of
diffusion and influence.Acm Transactions on Knowledge Discovery from Data,5
(4):21:1–21:37,2012.
[16]Markus Germar,Alexander Schlemmer,Kristine Krug,et al:Social
Influence and Perceptual Decision Making A Diffusion Model
Analysis.Personality and Social Psychology Bulletin,40(2):217-231,2014.
[17]Henry N.,Fekete J.,McGuffin M.J.:NodeTrix:a Hybrid Visualization
of Social Networks.IEEE Transactions on Visualization and Computer Graphics,
13(6):1302-1309,2007.
The content of the invention
The technical problem to be solved in the present invention is to provide a kind of social network user behavior prediction side based on strong or weak relation
Method.Modeling is predicted using behavior of the social networks strong or weak relation data to user.In social networks, the behavior of user by
To other users behavior and the influence of oneself state, and oneself state is except the behavioral implications by other users in network,
The influence of state before also suffering from.In other words, the various strong or weak relations in network pass through itself and the state for influenceing user
Behavior to user impacts.The purpose of the present invention be exactly by modeling of the social networks strong or weak relation to customer impact,
A kind of social network user behavior prediction method based on strong or weak relation is provided.
This method comprises the following steps：
(1) social networks strong or weak relation is measured：Social network user data are inputted, relation power is carried out using Weak Classifier
Prediction, after Weak Classifier is obtained, using Gradient boosting algorithms, using study sequence to each Weak Classifier
It is weighted；
(2) structure introduces the continuous time hidden Markov model of exogenousd variables：Set SiRepresent certain sampling time point i user
State, AiRepresent the behavior of time point user, FiUser is represented in an ectocine suffered by the i moment, wherein, Fi
And AiFor observable variable, SiFor unobservable state variable；Input the emotion and influence factor t of nearest a period of time1,F1,
S1,t2,F2,S2,…,tn,Fn,Sn, ti<ti+1, task is given tn+1>tn, and the S of a period of time recentlyi(i∈[1,n])
And Fi(i ∈ [1, n+1]), export it and correspond to the probability P (S of each staten+1|t1,F1,S1,t2,F2,S2,…,tn,Fn,Sn,
tn+1,Fn+1), and ectocine and state are to the probability function P (A of behaviorn+1|tn+1,Sn+1,Fn+1,tn,Fn,tn-1,Fn-1,…,
t1,F1,)；This mapping is completed using deep neural network；In implementation process, pass through the strong or weak relation obtained in step (1)
To FiThe weight that actually enters be adjusted, the training to whole model is using classical forward, backward algorithm, so as to export use
Family behavior.
Wherein, the state that personal information of the social network user data including user of input, user deliver in step (1)
Information, user are thumbed up and commented on.
Weak Classifier includes the relation grader based on user property, point based on interaction statistics between user in step (1)
Class device, the grader based on speech interaction and the grader based on image.
The state of user is the inherent state of user in step (2), including glad, surprised, sad, angry, frightened and detest
Dislike emotion.
The present invention considers the factors such as user property, user mutual, multimedia messages and carries out social networks power
The analysis of relation and measurement, Weak Classifier is built respectively to these factors, arranged using Gradient boosting algorithms and study
Sequence (Learn to rank) weights to Weak Classifier, obtains social networks strong or weak relation.Then, it is proposed that introduce exogenousd variables
Continuous time hidden Markov model carries out user's behavior prediction, and unlike traditional hidden Markov model, introduces
One external influence variable (observable variable), and from the time it is upper for, be a continuous process.The pass of this model
Key is the transition probability model for establishing User Status variable.Emotion is a very complicated mental process, and it is past that it shifts change
Toward the emotion for being related to nearest some time, simple first-order Markov process is often inadequate.In order to portray user
The distribution of emotion, the present invention is using all states in nearest a period of time, similar to high-order Markov model.Pass through depth
Neutral net is spent to portray the state transition function and behavioral implications function inside Markov model, investigates different depth nets
The concrete structure of the characteristics of network, the selection of the number of plies and network.
The above-mentioned technical proposal of the present invention has the beneficial effect that：
In such scheme, structure considers the deep layers such as interaction statistics between user property, user, speech interaction and image media
Psychological factor and semantic information synthesis strong or weak relation model framework, improve in the past based on simple data statistics relation it is strong
Weak measure；Structure introduces the continuous time hidden Markov model of exogenousd variables, carries out the social networks based on strong or weak relation
User's behavior prediction, combine legacy Markov model interpretable, expressivity of neutral net, take into full account user from body and mind
Between reason, ectocine and itself behavior many advantages, such as complicated relation.
Brief description of the drawings
Fig. 1 is the social network user behavior prediction method flow schematic diagram based on strong or weak relation of the present invention；
Fig. 2 is present invention introduces the schematic diagram of the continuous time hidden Markov model of exogenousd variables.
Embodiment
To make the technical problem to be solved in the present invention, technical scheme and advantage clearer, below in conjunction with accompanying drawing and tool
Body embodiment is described in detail.
The present invention provides a kind of social network user behavior prediction method based on strong or weak relation, as shown in figure 1, this method
Comprise the following steps：
(1) social networks strong or weak relation is measured：Social network user data are inputted, relation power is carried out using Weak Classifier
Prediction, after Weak Classifier is obtained, using Gradient boosting algorithms, using study sequence to each Weak Classifier
It is weighted；
(2) structure introduces the continuous time hidden Markov model of exogenousd variables：Set SiRepresent certain sampling time point i user
State, AiRepresent the behavior of time point user, FiUser is represented in an ectocine suffered by the i moment, wherein, Fi
And AiFor observable variable, SiFor unobservable state variable；By the strong or weak relation that is obtained in step (1) to FiReality
Input weight is adjusted, so as to export user behavior.
In specific operation process, input data is social network user data, including the personal information of user, Yong Hufa
The status information of table, user such as thumb up and commented at the interactive information, export the behavior for user, specially each shape probability of state and
The probability of ectocine and state to behavior.Concretely comprise the following steps：
A. social networks strong or weak relation is measured.The strong and weak prediction of relation, Weak Classifier bag are carried out using various Weak Classifiers
Include following four part：(1) the relation grader based on user property.Psychology shows that the attribute of two people in itself determines them
It can get close to [17].The attribute of user is collected in social networks, user property also includes social property as friend's circle,
Because the registration of the nearer friend's people circle of two relations is larger, a characteristic vector (Feature Vector) is organized into,
Then it is predicted with graders such as neutral nets, i.e. attribute X1, X2 for two given people, parent can be turned into by learning them
The probability F (X1, X2) of nearly good friend.(2) grader based on interaction statistics between user.The people that relation is got close to can possess more friendships
Mutually, two person-to-person information flows also tend to be two-way.Count time and secondary that they forward, the behavior such as comment occurs
Number, is organized into feature.Consider that relatively interactive intensity, such as the state the few that A is sent out are replied simultaneously, but B is often replied, then
B and A largely has strong relation, such as close friend.There is sequential in view of the behavior of user, it would be desirable to will be each
The feature at time point is decayed according to current time point, and this requires the selection for carefully studying decay factor.Then by these
The power of the simple classification device projected relationships such as feature input Logistics Regression.(3) classification based on speech interaction
Device.In order to accurately portray flexible and changeable internet vocabulary, using the theorem in Euclid space embedding grammar of vocabulary, by vocabulary and higher-dimension to
Amount is combined, because the distance in higher dimensional space largely reflects the correlation between two words.Utilize topic mould
The instruments such as type portray topic probability of each word under the distribution of certain text, and all these attributes combine, and original word one
Rise and form one " super word ".Huge advantage of the tensor function in terms of the geometric object relation such as vector is portrayed, so from traditional
Recurrent neural network for portraying syntactic structure is set out, and transmission function therein is expressed with tensor function, with reference to " super
Huge advantage of the word " relative to single vocabulary in information content, and then the affect structure of parsing sentence.(4) user is extracted in society
Hand over networking to say the features such as the dominant hue of blit picture, preceding background, color contrast, the emotion of prognostic chart picture, and then analyze in more matchmakers
The influence of emotion under body yardstick.After these Weak Classifiers are obtained, using Gradient boosting algorithms, study is utilized
Sequence (Learn to rank) is weighted to each Weak Classifier.
B. structure introduces the continuous time hidden Markov model of exogenousd variables, and schematic diagram is as shown in Figure 2.S in figurei(certain is adopted
Sample time point i) represents the state of user, AiRepresent the behavior of user, FiRepresent user in an extraneous shadow suffered by the i moment
Ring (such as certain friend has sent out a model).FiAnd AiFor observable variable, SiFor unobservable state variable.Here shape
State variable is the inherent state of user (used here as Ekman glad, surprised, sad, angry, frightened and detest proposed etc. six
Class emotion [33], other classification can also be selected according to actual conditions).Emotion is a very complicated mental process, and it is shifted
Change often relates to some emotions of nearest period, and simple first-order Markov process is inadequate.In order to portray use
The distribution of family emotion, it is as follows to formalize this process：Input is the emotion and influence factor t of nearest a period of time1,F1,S1,t2,
F2,S2,…,tn,Fn,Sn, ti<ti+1, it is allowed to some FiAnd AiVacancy, i.e., at the moment not by external influence, or do not have
Behavior.Task is given tn+1>tn, and the S of period recentlyi,Fi, export it and correspond to the probability P (S of each staten+1|
t1,F1,S1,t2,F2,S2,…,tn,Fn,Sn,tn+1,Fn+1), and ectocine and state are to the probability function P (A of behaviorn+1|
tn+1,Sn+1,Fn+1,tn,Fn,tn-1,Fn-1,…).Complexity in view of this process and non-linear, uses a depth god
This mapping is completed through network (Deep neural network).It is used for regulating and controlling F as the strong or weak relation obtained by (a)iInfluence
Size.During realization, by strong or weak relation come to FiThe weight that actually enters be adjusted.Training to whole model
Using the forward, backward algorithm (Forward-backward algorithm) of classics, its advantage is can be to arbitrary probability point
Cloth is modeled.After state transition model is obtained, we carry out the reasoning of inherent state by improved viterbi algorithm.
Because only adjacent state and ectocine is associated a state, it is possible to is efficiently asked by dynamic programming algorithm
Solution.Because whole calculating of the algorithm to different user is independent, therefore can be to whole network Parallel implementation.
Described above is the preferred embodiment of the present invention, it is noted that for those skilled in the art
For, on the premise of principle of the present invention is not departed from, some improvements and modifications can also be made, these improvements and modifications
It should be regarded as protection scope of the present invention.
Claims (4)
- A kind of 1. social network user behavior prediction method based on strong or weak relation, it is characterised in that：Comprise the following steps：(1) social networks strong or weak relation is measured：Social network user data are inputted, the pre- of relation power is carried out using Weak Classifier Survey, after Weak Classifier is obtained, using Gradient boosting algorithms, each Weak Classifier is carried out using study sequence Weighting；(2) structure introduces the continuous time hidden Markov model of exogenousd variables：Set SiRepresent the shape of certain sampling time point i user State, AiRepresent the behavior of time point user, FiUser is represented in an ectocine suffered by the i moment, wherein, FiAnd Ai For observable variable, SiFor unobservable state variable；Input the User Status and influence factor t of nearest a period of time1,F1, S1,t2,F2,S2,…,tn,Fn,Sn, ti<ti+1, task is given tn+1>tn, and the S of a period of time recentlyi(i∈[1,n]) And Fi(i ∈ [1, n+1]), export it and correspond to the probability P (S of each staten+1|t1,F1,S1,t2,F2,S2,…,tn,Fn,Sn, tn+1,Fn+1), and ectocine and state are to the probability function P (A of behaviorn+1|tn+1,Sn+1,Fn+1,tn,Fn,tn-1,Fn-1,…, t1,F1,)；This mapping is completed using deep neural network；In implementation process, pass through the strong or weak relation obtained in step (1) To FiThe weight that actually enters be adjusted, the training to whole model uses forward, backward algorithm, so as to export user's row For.
- 2. the social network user behavior prediction method according to claim 1 based on strong or weak relation, it is characterised in that：Institute Stating the social network user data of input in step (1) includes the personal information of user, the status information that user delivers, Yong Hudian Praise and comment on.
- 3. the social network user behavior prediction method according to claim 1 based on strong or weak relation, it is characterised in that：Institute Stating Weak Classifier in step (1) includes the relation grader based on user property, based on the grader of interaction statistics, base between user The grader interacted in speech and the grader based on image.
- 4. the social network user behavior prediction method according to claim 1 based on strong or weak relation, it is characterised in that：Institute State user in step (2) state be user inherent state, including it is glad, surprised, sad, angry, frightened and detest emotion.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711050008.9A CN107808168B (en) | 2017-10-31 | 2017-10-31 | Social network user behavior prediction method based on strong and weak relation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711050008.9A CN107808168B (en) | 2017-10-31 | 2017-10-31 | Social network user behavior prediction method based on strong and weak relation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107808168A true CN107808168A (en) | 2018-03-16 |
| CN107808168B CN107808168B (en) | 2023-08-01 |
Family
ID=61582619
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711050008.9A Active CN107808168B (en) | 2017-10-31 | 2017-10-31 | Social network user behavior prediction method based on strong and weak relation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107808168B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109165239A (en) * | 2018-07-06 | 2019-01-08 | 重庆邮电大学 | Modeling method is propagated based on the D2D information of user property and contents attribute |
| CN109299429A (en) * | 2018-07-11 | 2019-02-01 | 重庆邮电大学 | A kind of dynamic society's relationship modeling method based on Wiener-Hopf equation |
| CN110009093A (en) * | 2018-12-07 | 2019-07-12 | 阿里巴巴集团控股有限公司 | For analyzing the nerve network system and method for relational network figure |
| CN111414478A (en) * | 2020-03-13 | 2020-07-14 | 北京科技大学 | Social network emotion modeling method based on deep cycle neural network |
| WO2021081741A1 (en) * | 2019-10-29 | 2021-05-06 | 深圳大学 | Image classification method and system employing multi-relationship social network |
| CN117289940A (en) * | 2023-08-28 | 2023-12-26 | 深圳市众为精密科技有限公司 | Optimization method and system for Zeiss three-dimensional equipment based on off-board programming software |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101615186A (en) * | 2009-07-28 | 2009-12-30 | 东北大学 | A kind of BBS user's abnormal behaviour auditing method based on Hidden Markov theory |
| CN102117325A (en) * | 2011-02-24 | 2011-07-06 | 清华大学 | Method for predicting dynamic social network user behaviors |
| CN102270212A (en) * | 2011-04-07 | 2011-12-07 | 浙江工商大学 | User interest feature extraction method based on hidden semi-Markov model |
| JP2017509947A (en) * | 2014-01-27 | 2017-04-06 | アリババ・グループ・ホールディング・リミテッドＡｌｉｂａｂａ Ｇｒｏｕｐ Ｈｏｌｄｉｎｇ Ｌｉｍｉｔｅｄ | Obtaining the social relationship type of a network subject |
| CN106649659A (en) * | 2016-12-13 | 2017-05-10 | 重庆邮电大学 | Link prediction system and method for social network |
| CN106682770A (en) * | 2016-12-14 | 2017-05-17 | 重庆邮电大学 | Friend circle-based dynamic microblog forwarding behavior prediction system and method |
| CN106780055A (en) * | 2016-11-28 | 2017-05-31 | 西安电子科技大学 | Probability size verification method based on MSVL social network user relationship strengths |
-
2017
- 2017-10-31 CN CN201711050008.9A patent/CN107808168B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101615186A (en) * | 2009-07-28 | 2009-12-30 | 东北大学 | A kind of BBS user's abnormal behaviour auditing method based on Hidden Markov theory |
| CN102117325A (en) * | 2011-02-24 | 2011-07-06 | 清华大学 | Method for predicting dynamic social network user behaviors |
| CN102270212A (en) * | 2011-04-07 | 2011-12-07 | 浙江工商大学 | User interest feature extraction method based on hidden semi-Markov model |
| JP2017509947A (en) * | 2014-01-27 | 2017-04-06 | アリババ・グループ・ホールディング・リミテッドＡｌｉｂａｂａ Ｇｒｏｕｐ Ｈｏｌｄｉｎｇ Ｌｉｍｉｔｅｄ | Obtaining the social relationship type of a network subject |
| CN106780055A (en) * | 2016-11-28 | 2017-05-31 | 西安电子科技大学 | Probability size verification method based on MSVL social network user relationship strengths |
| CN106649659A (en) * | 2016-12-13 | 2017-05-10 | 重庆邮电大学 | Link prediction system and method for social network |
| CN106682770A (en) * | 2016-12-14 | 2017-05-17 | 重庆邮电大学 | Friend circle-based dynamic microblog forwarding behavior prediction system and method |
Non-Patent Citations (3)
| Title |
|---|
| HONGXIAO FEI等: "《The Searching Ranking Model Based on the Sharing and Recommending Mechanism of Social Network》", 《APSCC 2015》 * |
| WONJOON KANG等: "《Prediction of state of user"s behavior using Hidden Markov Model in ubiquitous home network》", 《2010 IEEE INTERNATIONAL CONFERENCE ON INDUSTRIAL ENGINEERING AND ENGINEERING MANAGEMENT》 * |
| 王晓慧等: "《基于深度学习的情感化设计》", 《包装工程》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109165239A (en) * | 2018-07-06 | 2019-01-08 | 重庆邮电大学 | Modeling method is propagated based on the D2D information of user property and contents attribute |
| CN109165239B (en) * | 2018-07-06 | 2021-05-18 | 重庆邮电大学 | D2D information propagation modeling method based on user attributes and content attributes |
| CN109299429A (en) * | 2018-07-11 | 2019-02-01 | 重庆邮电大学 | A kind of dynamic society's relationship modeling method based on Wiener-Hopf equation |
| CN110009093A (en) * | 2018-12-07 | 2019-07-12 | 阿里巴巴集团控股有限公司 | For analyzing the nerve network system and method for relational network figure |
| CN110009093B (en) * | 2018-12-07 | 2020-08-07 | 阿里巴巴集团控股有限公司 | Neural network system and method for analyzing relational network graph |
| WO2021081741A1 (en) * | 2019-10-29 | 2021-05-06 | 深圳大学 | Image classification method and system employing multi-relationship social network |
| CN111414478A (en) * | 2020-03-13 | 2020-07-14 | 北京科技大学 | Social network emotion modeling method based on deep cycle neural network |
| CN111414478B (en) * | 2020-03-13 | 2023-11-17 | 北京科技大学 | Social network emotion modeling method based on deep cyclic neural network |
| CN117289940A (en) * | 2023-08-28 | 2023-12-26 | 深圳市众为精密科技有限公司 | Optimization method and system for Zeiss three-dimensional equipment based on off-board programming software |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107808168B (en) | 2023-08-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107808168A (en) | A kind of social network user behavior prediction method based on strong or weak relation | |
| Wang et al. | A survey of hybrid human-artificial intelligence for social computing | |
| Smolensky et al. | Computer-aided reasoned discourse or, how to argue with a computer | |
| CN107153642A (en) | A kind of analysis method based on neural network recognization text comments Sentiment orientation | |
| Shi et al. | Symmetry in computer-aided music composition system with social network analysis and artificial neural network methods | |
| Zhang et al. | Internet of brain, thought, thinking, and creation | |
| Ma | [Retracted] Facing Big Data Information Fusion and Data Mining Technology to Construct College Physical Education Teaching Evaluation System | |
| Tang | Towards meta-synthetic support to unstructured problem solving | |
| Pawar | Theory building for hypothesis specification in organizational studies | |
| Wei et al. | LSTM-SN: complex text classifying with LSTM fusion social network | |
| CN109919047A (en) | A kind of mood detection method based on multitask, the residual error neural network of multi-tag | |
| Zhang et al. | The Study of a Five‐Dimensional Emotional Model for Facial Emotion Recognition | |
| CN109446405A (en) | Travel industry promotion method and system based on big data | |
| Mote et al. | New directions in the use of network analysis in research and product development evaluation | |
| Zhang et al. | Construction of higher education teaching quality evaluation model based on scientific computing | |
| Wang | [Retracted] Construction and Effect Analysis of College Students’ Physical Education Teaching Mode Based on Data Mining Algorithm | |
| Yang | [Retracted] Application of Multidirectional Mutation Genetic Algorithm and Its Optimization Neural Network in Intelligent Optimization of English Teaching Courses | |
| Li | Application of an Internet of things oriented network education platform in English language teaching | |
| Rozić et al. | Artificial Intelligence for Knowledge Visualization: An Overview | |
| Li et al. | Recurrent dirichlet belief networks for interpretable dynamic relational data modelling | |
| Praveena et al. | A review on deep neural network design and their applications | |
| Vinnakota | A conceptual framework for complex system design and design management | |
| Aurigi et al. | Introducing shaping smart for better cities | |
| Hoppenbrouwers et al. | Navigating the Methodology Jungle--The communicative role of modelling techniques in information system development | |
| Liu | Analysis of psychological characteristics and emotional expression based on deep learning in higher vocational music education |
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 | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20231107 Address after: 100083 No. 30, Haidian District, Beijing, Xueyuan Road Patentee after: University OF SCIENCE AND TECHNOLOGY BEIJING Patentee after: Beijing Times Lingyun Intelligent Testing Technology Co.,Ltd. Address before: 100083 No. 30, Haidian District, Beijing, Xueyuan Road Patentee before: University OF SCIENCE AND TECHNOLOGY BEIJING Patentee before: BEIJING TIME HIGH-TECHNOLOGY Ltd. |
|
| CP03 | Change of name, title or address | ||
| CP03 | Change of name, title or address |
Address after: 1036, 1st Floor, Building A, East Side, No. 58 Dongbei Wangxi Road, Haidian District, Beijing, 100193 Patentee after: Beijing Times Lingyun Intelligent Testing Technology Co.,Ltd. Patentee after: University OF SCIENCE AND TECHNOLOGY BEIJING Address before: 100083 No. 30, Haidian District, Beijing, Xueyuan Road Patentee before: University OF SCIENCE AND TECHNOLOGY BEIJING Patentee before: Beijing Times Lingyun Intelligent Testing Technology Co.,Ltd. |