CN109635989A - A kind of social networks link prediction method based on multi-source heterogeneous data fusion - Google Patents
A kind of social networks link prediction method based on multi-source heterogeneous data fusion Download PDFInfo
- Publication number
- CN109635989A CN109635989A CN201810999492.8A CN201810999492A CN109635989A CN 109635989 A CN109635989 A CN 109635989A CN 201810999492 A CN201810999492 A CN 201810999492A CN 109635989 A CN109635989 A CN 109635989A
- Authority
- CN
- China
- Prior art keywords
- user
- link prediction
- social networks
- vector
- heterogeneous data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a kind of methods of social networks link prediction based on multi-source heterogeneous data fusion, and using registering comprising customer relationship topological diagram and user, the social networks based on geographical location information for recording both heterogeneous data sources carries out link prediction.The present invention proposes a kind of mixed frame, customer relationship topological diagram and user in the social networks based on geographical location information is sufficiently captured by model AL to register the association recorded between both heterogeneous data sources, when overcoming single data source progress link prediction in social networks of the use based on geographical location information, the problem of prediction result inaccuracy, is effectively promoted the effect of link prediction.The calculating speed that is trained of deep learning is improved using local sensitivity Hash simultaneously and reduces storage overhead.
Description
Technical field
The invention belongs to the field of neural networks in machine learning, are a kind of methods based on deep learning, especially sharp
With deep learning in the social networks (Local Based Social Networks, LBSN) based on geographical location information
Customer relationship topological diagram and user, which register, to be recorded both isomeric datas and merges, and realizes social networks link prediction, and make
The calculating speed that deep learning is trained is improved with local sensitivity Hash (Locality Sensitive Hashing, LSH)
And reduce storage overhead.
Background technique
Social networks link prediction (Link Prediction, LP), abbreviation link prediction, it is intended to be closed from one by good friend
Be constitute customer relationship topological diagram in find out lost in the figure while or in the future will appear while.With social networking service
(Social Network Service, SNS) and other network applications increase rapidly, and network data is ubiquitous.It obtains
This kind of network data of friend relation on the APP such as Facebook, QQ, can construct customer relationship topological diagram, which opens up
Flutterring figure can be used for social networks link prediction.Meanwhile with the development of location technology, the GPS positioning function of mobile device is utilized
The location information of user can be acquired, this kind of location information combines the time of positioning that can form user and registers record.Many researchs
Show that user's record of registering also contributes to social networks link prediction.
Link prediction plays important role in information recommendation system, is mainly used in social network analysis, passes through
Link prediction can obtain the higher good friend of confidence level, recommend user can knowable people, the society of user can be significantly improved
Experience and loyalty are handed over, and brings huge economic benefit for enterprise.It is closed in addition to predicting the user in customer relationship topological diagram
Connection is outer, and the method and thought of link prediction can also be used to predict no label node in the network of known portions node type
Type, this has substantial worth for the optimization of network reconfiguration and structure function.
In traditional link prediction method, Jaccard, Euclidean distance or cosine value are generallyd use to measure two users
The similarity of node determines whether that there are the links with this.And these methods are all inflexible.If having changed new data set,
Or raw data set is increased or deletes data and then needs to recalculate all data, consume a large amount of calculate
Storage resource.Deep learning can flexibly handle mass data.The link prediction model that method based on deep learning is built, can
Optimized to the parameter of model by the training data of input magnanimity, prediction work is carried out to obtain trained model.
Summary of the invention
The LBSN data set that the present invention uses includes that customer relationship topological diagram and user register and record both different structures
Data source.Customer relationship topological diagram is made of the relationship between user, and the relationship between user is known as link (i.e. point to), and every
Chain routes the relationship composition of two user nodes.User registers record by the user node, sign-in desk longitude, sign-in desk latitude registered
It spends, register time and point of interest (Point-of-Interest, POI) composition.
The purpose of the present invention is intended to overcome when carrying out link prediction using single data source in LBSN, prediction result inaccuracy
Problem.Basic ideas of the invention are to propose a kind of mixed frame, and customer relationship topological diagram and user in LBSN are registered note
It records both isomeric datas and carries out fusion realization link prediction, enhance the prediction effect of existing link prediction method.It uses simultaneously
LSH promotes the performance for calculating and storing.
Based on foregoing invention thinking, the present invention provides a kind of social networks link prediction based on multi-source heterogeneous data fusion
Method comprising following steps:
S1, Data_process (G) → Tra, Tes: training set Tra is extracted from customer relationship topological diagram G=(V, E)
With test set Tes.Wherein V indicates the set of user node in topological diagram, and E indicates the set on side in topological diagram.If two in G
User uiAnd ujThere are social networks, then there are a lines between them, are denoted as eij=(ui,uj)；
S2,Using network representation learning method, learns and obtain from the positive sample G' of Tra
The social network user vector for taking V, is denoted asWherein d isDimension；
S3,It is registered and is recorded S=(U, L) according to user, building user-position is registered frequency matrixWherein U and L respectively indicates the set of the user in S and point set of registering, N are the numbers of users in U, and M is in L
Sign-in desk quantity.It recycles Poisson matrix decomposition to obtain the user in low-dimensional vector space and accesses preference vector, be denoted asWherein D isDimension；
S4,In order to capture the association of these two types of data sources in LBSN, similar anchor chain is connect
The mode of (anchor link) designs an improved deep learning model, referred to as AL.As sample,As sample
Corresponding label, two kinds of vectors, which are input to together in AL, carries out more wheel training.It has been merged in G using final trained AL generation
The new user of topology information accesses preference vector
S5,It willAnd ui 'vIt is merged again, is input to a convolution mind
Through being trained in network (Convolutional Neural Network, CNN).Tes is finally input to trained CNN
Middle carry out link prediction obtains prediction result result.
The method of the above-mentioned social networks link prediction based on multi-source heterogeneous data fusion, the step S1, it is therefore intended that
Obtain Tra and Tes.Link prediction can be regarded as two classification problems, and link present in G is considered as positive sample, without depositing
Link be considered as negative sample.Positive sample in Tra has been missing from the customer relationship topological diagram G' ∈ G of part of links, and these
The link of missing is by the positive sample as Tes.Specifically include it is following step by step:
S11 carries out data cleansing, so that the user in LBSN in G and S is consistent；
S12 selects some links as Tes positive sample from G.Guarantee after getting rid of Tes positive sample in G simultaneously
G' ∈ G is connection；Using G' as the positive sample of Tra；
S13, randomly chooses some links being not present as negative sample from G, by predefined pro rate to Tra and
In Tes.
The method of the above-mentioned social networks link prediction based on multi-source heterogeneous data fusion, the step S3, it is therefore intended that
It obtainsSpecifically include it is following step by step:
S31 constructs H using S.Wherein, the row of H indicates user, and column indicate POI, the value of H corresponding POI accessible by user
Number filling；
S32 carries out Poisson matrix decomposition to H, and available reflection user accesses the matrix of preferenceWith POI spy
Levy matrixPOI eigenmatrix can reflect the case where a certain POI is accessed by the user.UsRow conduct
The method of the above-mentioned social networks link prediction based on multi-source heterogeneous data fusion, the step S4, it is therefore intended that
The association in LBSN between G and S is captured, realizes fusion.In order to capture this association, the training of model AL specifically includes following
Step by step:
S41 is utilizedThe user node in Tra is represented, the cosine mean value cos of user's point pair in Tra is calculatedori；
S42 captures the association between G and S using the one-to-one relationship of user in V and U.By sampleAnd corresponding mark
LabelIt is divided into multiple batches (batch) and recycles and be input in multi-layer perception (MLP) (Multilayer perception, MLP)
It is trained；
S43, by more taking turns training realization to the tuning of the parameter in model AL.It, will after AL is trainedInput AL, output
ui 'v。
Two calculating functions involved in the implementation method step S42 of above-mentioned model AL.First calculating function be capture V and
The mapping function of user's one-to-one relationship, is denoted as in UThe corresponding loss function of the mapping function isWherein x indicates that sample, y indicate that true value, a indicate the output valve of model, and n indicates the quantity of sample.
Stochastic gradient descent algorithm optimization overall situation weight parameter W and global straggling parameter b are called, which is denoted as respectivelyWherein σ indicates activation primitive, and z is the input of neuron, is expressed as
It is the u in order to guarantee to generate that second, which calculates function,i 'vOffset will not be generated, that is, uses ui 'vIndicate user in Tra
It puts the cosine mean value to calculating otherwise is less than cosori.Therefore the limitation of cosine average value constraint is introduced, is denoted asWhereinWithRespectively indicate user umWith user un's
User accesses preference vector, and there are e in Gmn.N (U) indicates the number of users in U.The cosine average value constraint limits corresponding loss
Function isThe evolutionary process point of global weight parameter W and global straggling parameter b
It is not denoted as
The method of the above-mentioned social networks link prediction based on multi-source heterogeneous data fusion, the step S5, it is therefore intended that
G and S are merged to the link prediction for realizing that storage consumption is low, calculating speed is fast once again.Specifically include it is following step by step:
S51, willAnd ui 'vIt is spliced into a vector
S52, will using LSHProject to a binary vector mi∈{0,1}mOn, user uiUse miIt indicates；
S53, for any bar side e in Gij=(ui,uj), m is obtained using same procedurejAs user ujExpression；
S54, by miAnd mjSplicing is to obtain side eijBinary vector indicate mij(∈{0,1}2m)=[mi；mj]；
S55, the vector m for being 2m by lengthijIn element to successively fill size by the sequence of row major be n × n
In square matrix, this process is known as remolding.Then this square matrix is input to a convolutional neural networks (Convolutional
Neural Network, CNN) in be trained；
Tes is input in trained CNN and carries out link prediction by S56.
Compared with prior art, the invention has the following advantages:
1, it the present invention is based on the method for the social networks link prediction of multi-source heterogeneous data fusion, is connect using similar to anchor chain
Model AL, the new user for generating fusion customer relationship topological diagram accesses preference vector, can sufficiently capture social networks use
Family vector sum user accesses the association between preference vector.AL can not only be by the social network user vector sum user of same user
Access the two vectors of preference vector are aligned, also by introducing the tied mechanism of cosine mean value, to ensure two data sources
In user vector fusion after the new user that generates access preference vector and will not shift.
2, the present invention is based on the methods of the social networks link prediction of multi-source heterogeneous data fusion, by two numbers in LBSN
The social network user vector sum user that obtains respectively according to source accesses preference vector and splices, and application LSH by splicing to
Amount projects on a binary vector, then using the binary vector as the final expression vector of user node in LBSN, and
For the point for including according to link in customer relationship topological diagram to relationship, carrying out splicing using the final vector of user node indicates point pair
Relationship.It is finally a square matrix by the vector remodeling of this splicing, is input in CNN, realizes link prediction.LSH's applies energy
It improves the calculating speed that deep learning is trained and reduces storage overhead.
Detailed description of the invention
Fig. 1 is to capture to close between social network user vector sum user access preference vector based on multi-layer perception (MLP) (MLP)
The class anchor link model AL of connection.
Fig. 2 is the overall model framework of the method for the social networks link prediction based on multi-source heterogeneous data fusion.It utilizes
The new user for having merged customer relationship topological diagram generated in Fig. 1 accesses preference vector and spells with social network user vector
It connects, and application LSH is by the vector projection of splicing a to binary vector, then using the binary vector as user in LBSN
The final expression vector of node, and the point for including according to link in customer relationship topological diagram is to relationship, most using user node
Whole vector, which carries out splicing, indicates point to relationship.It is finally a square matrix by the vector remodeling of this splicing, is input in CNN.
When Fig. 3 is without using LSH and using LSH, the side of the social networks link prediction based on multi-source heterogeneous data fusion
The performance comparison of method.Wherein (a) is the comparison of memory consumption, is (b) comparison of CPU consumption, is (c) comparison of GPU consumption,
It (d) is the comparison of time loss.(d) vector dimension in is the vector dimension for being input to CNN.
Term is explained
LBSN is the abbreviation of Location-based Social Network, is indicated " location-based community network ".
LBSN is in comprising traditional society's network other than the connection of person to person, the also letter such as record time for having user to register and geographical location
Breath.
POI is the abbreviation of Point-of-Interest, is indicated " point of interest ".In LBSN, a POI is exactly user's label
The one place arrived.
LSH is the abbreviation of Locality Sensitive Hashing, is indicated " local sensitivity Hash ".It is that one kind is directed to
The quick closest lookup algorithm of magnanimity high dimensional data.
Specific embodiment
Below in conjunction with attached drawing, the invention will be further described.
Embodiment
The method of social networks link prediction provided by the embodiment based on multi-source heterogeneous data fusion, can be used for containing
Customer relationship topological diagram and user, which register, records the data set of both data sources.With the LBSN number of real world shown in table 1
According to collection, as Foursquare (can be fromhttp://snap.stanford.eduObtain) for tested.
Table 1: the relevant information of the social link prediction training set of multi-source heterogeneous data fusion
| Dataset | #check_ins | #POIs | #edges | #users |
| Foursquare@NYC | 22,563 | 1,992 | 5,810 | 588 |
| Foursquare@TKY | 38,742 | 2,212 | 9,624 | 1,055 |
| Gowalla@DC | 13,594 | 4,795 | 5,826 | 880 |
| Gowalla@CHI | 10,314 | 3,269 | 2,542 | 627 |
| Brightkite | 75,522 | 4,038 | 33,008 | 1,502 |
Fig. 1 is to capture to close between social network user vector sum user access preference vector based on multi-layer perception (MLP) (MLP)
The class anchor link model AL of connection.
As shown in Figure 1, substituting into step S1:Data_ using the customer relationship topological diagram G in Foursquare first
Process (G) → Tra, Tes obtains training set Tra and test set Tes.The positive sample G' in Tra is substituted into step S2 again:Common e-learning representation method wherein can be used, such as node2vec obtains social networks and uses
Family vectorNext it is first registered using the user in Foursquare and records S substitution step S3:
It obtains user and accesses preference vectorAgain by the output of step S2 and S3WithIt is input in model AL, uses step S4:Obtain ui 'v。
Table 2: the effect of social link prediction is carried out in three kinds of real data sets
Fig. 2 is the overall model framework of the method for the social networks link prediction based on multi-source heterogeneous data fusion.
As shown in Fig. 2, by the output of step S2 and S4And ui 'v, it inputs into prediction model CNN, uses step S5:Obtain final prediction result result.The link prediction effect of mixed model is shown in
Table 2.
#check_ins indicates that user registers and records quantity；
#POIs indicate user register record in different POI quantity；
#edges indicates the number of links in customer relationship topological diagram；
#users indicates the quantity of user in customer relationship topological diagram (or user register record)；
Foursquare@NYC indicates that region is the data of New York in data set Foursquare；
Foursquare@TKY indicates that region is the data in Tokyo in data set Foursquare；
Gowalla@DC indicates that region is the data in Washington in data set Gowalla；
Gowalla@CHI indicates that region is the data in Chicago in data set Gowalla；
Vec2link- is not using the method for the social networks link prediction based on multi-source heterogeneous data fusion of LSH；
Vec2link+ is the method for having used the social networks link prediction based on multi-source heterogeneous data fusion of LSH, with
Vec2link- is compared, and after embodiment has used LSH, storage is occupied and is lower, the advantage that calculating speed is promoted；
Average, Hadamard, Weighted-L1, Weighted-L2 are the control methods of vec2link+, are only made
With the information of customer relationship topological diagram in LBSN, implementation can be with bibliography [Grover, Aditya, and Jure
Leskovec."node2vec:Scalable feature learning for networks."Proceedings of the
22nd ACM SIGKDD international conference on Knowledge discovery and data
mining.ACM,2016.]；
Jaccard is the control methods of vec2link+, for comparing similitude and otherness between finite sample collection；
Walk2friend is the control methods of vec2link+, only uses user in LBSN and registers the data of record, real
Existing scheme can be with bibliography [Backes, Michael, et al. " walk2friends:Inferring Social Links
from Mobility Profiles."Proceedings of the 2017 ACM SIGSAC Conference on
Computer and Communications Security.ACM,2017.】。
It can be seen that the social network used based on multi-source heterogeneous data fusion in the present invention from the test result in table 2
The method of network link prediction, prediction effect will be comprehensively better than the effects that single data source progress link prediction is used only.
It follows that the present invention can be predicted efficiently against when carrying out link prediction using single data source in LBSN
As a result inaccurate problem realizes the promotion of link prediction effect.The method that the present invention uses deep learning, by user in LBSN
Relationship topology figure and user register record both isomeries data source carry out fusion realize link prediction.LSH is used simultaneously, with
Discrete binary set indicates user node, and the calculating speed of acceleration model simultaneously saves storage overhead.Invention achieves calculating
Speed is fast, storage consumption is few, link prediction effect is better than the purpose of single source data prediction effect.
Those of ordinary skill in the art will understand that the embodiments described herein, which is to help reader, understands this hair
Bright principle, it should be understood that protection scope of the present invention is not limited to such specific embodiments and embodiments.This field
Those of ordinary skill disclosed the technical disclosures can make according to the present invention and various not depart from the other each of essence of the invention
The specific variations and combinations of kind, these variations and combinations are still within the scope of the present invention.
Claims (5)
1. a kind of method of the social networks link prediction based on multi-source heterogeneous data fusion, it is characterised in that including following step
It is rapid:
S1, Data_process (G) → Tra, Tes: training set Tra and survey are extracted from customer relationship topological diagram G=(V, E)
Examination collection Tes；Wherein V indicates the set of user node in topological diagram, and E indicates the set on side in topological diagram；If two users in G
uiAnd ujThere are social networks, then there are a lines between them, are denoted as eij=(ui,uj)；
S2,Using network representation learning method, learns from the positive sample G' of Tra and obtain V's
Social network user vector, is denoted asWherein d isDimension；
S3,It is registered and is recorded S=(U, L) according to user, building user-position is registered frequency matrixWherein U and L respectively indicates the set of the user in S and point set of registering, N are the numbers of users in U, and M is in L
Sign-in desk quantity；It recycles Poisson matrix decomposition to obtain the user in low-dimensional vector space and accesses preference vector, be denoted asWherein D isDimension；
S4,In order to capture the association of G and S these two types data source in LBSN, similar anchor chain is connect
The mode of (anchor link) designs an improved deep learning model, referred to as AL；As sample,As sample
Corresponding label, two kinds of vectors, which are input to together in AL, carries out more wheel training；It has been merged in G using final trained AL generation
The new user of topology information accesses preference vector
S5,It willWithIt is merged again, is input to a convolutional Neural net
It is trained in network (Convolutional Neural Network, CNN)；Finally by Tes be input in trained CNN into
Line link prediction, obtains prediction result result.
2. the method for the social networks link prediction based on multi-source heterogeneous data fusion, feature exist according to claim 1
In the step S1 include it is following step by step:
S11 carries out data cleansing, so that the user in LBSN in G and S is consistent；
S12 selects some links as Tes positive sample from G；Guarantee the G' ∈ after getting rid of Tes positive sample in G simultaneously
G is connection；Using G' as the positive sample of Tra；
S13 randomly chooses some links being not present as negative sample, by predefined pro rate to Tra and Tes from G
In.
3. the method for the social networks link prediction based on multi-source heterogeneous data fusion, feature exist according to claim 1
In the step S3 include it is following step by step:
S31 constructs H using S；Wherein, the row of H indicates user, and column indicate POI, time of the value of H corresponding POI accessible by user
Number filling.
S32 carries out Poisson matrix decomposition to H, and available reflection user accesses the matrix of preferenceWith POI feature square
Battle arrayPOI eigenmatrix can reflect the case where a certain POI is accessed by the user；UsRow conduct
4. the method for the social networks link prediction based on multi-source heterogeneous data fusion, feature exist according to claim 1
In the step S4 include it is following step by step:
S41 is utilizedThe user node in Tra is represented, the cosine mean value cos of user's point pair in Tra is calculatedori；
S42 captures the association between G and S using the one-to-one relationship of user in V and U；By sampleAnd corresponding label
It is divided into multiple batches (batch) and recycles to be input in multi-layer perception (MLP) (Multilayer perception, MLP) and carry out
Training；
S43, by more taking turns training realization to the tuning of the parameter in model AL；It, will after AL is trainedInput AL, output
5. the method for the social networks link prediction based on multi-source heterogeneous data fusion, feature exist according to claim 1
In the step S5 include it is following step by step:
S51, willWithIt is spliced into a vector
S52, will using LSHProject to a binary vector mi∈{0,1}mOn, user uiUse miIt indicates；
S53, for any bar side e in Gij=(ui,uj), m is obtained using same procedurejAs user ujExpression；
S54, by miAnd mjSplicing is to obtain side eijBinary vector indicate mij(∈{0,1}2m)=[mi；mj]；
S55, the vector m for being 2m by lengthijIn element by the sequence of row major successively fill a size be n × n square matrix
In, this process is known as remolding；Then this square matrix is input to a convolutional neural networks (Convolutional Neural
Network, CNN) in be trained；
Tes is input in trained CNN and carries out link prediction by S56.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810999492.8A CN109635989B (en) | 2018-08-30 | 2018-08-30 | Social network link prediction method based on multi-source heterogeneous data fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810999492.8A CN109635989B (en) | 2018-08-30 | 2018-08-30 | Social network link prediction method based on multi-source heterogeneous data fusion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109635989A true CN109635989A (en) | 2019-04-16 |
| CN109635989B CN109635989B (en) | 2022-03-29 |
Family
ID=66066288
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810999492.8A Active CN109635989B (en) | 2018-08-30 | 2018-08-30 | Social network link prediction method based on multi-source heterogeneous data fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109635989B (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110335165A (en) * | 2019-06-28 | 2019-10-15 | 京东数字科技控股有限公司 | A kind of link prediction method and apparatus |
| CN110442802A (en) * | 2019-08-06 | 2019-11-12 | 中国科学技术大学 | A kind of more Behavior preference prediction techniques of social activity user |
| CN110543943A (en) * | 2019-09-10 | 2019-12-06 | 北京百度网讯科技有限公司 | Network convergence method and device, electronic equipment and storage medium |
| CN110598130A (en) * | 2019-09-30 | 2019-12-20 | 重庆邮电大学 | Movie recommendation method integrating heterogeneous information network and deep learning |
| CN111209943A (en) * | 2019-12-30 | 2020-05-29 | 广州高企云信息科技有限公司 | Data fusion method and device and server |
| CN111475739A (en) * | 2020-05-22 | 2020-07-31 | 哈尔滨工程大学 | Heterogeneous social network user anchor link identification method based on meta-path |
| CN111476673A (en) * | 2020-04-02 | 2020-07-31 | 中国人民解放军国防科技大学 | Method, device and medium for aligning users among social networks based on neural network |
| CN112446542A (en) * | 2020-11-30 | 2021-03-05 | 西安电子科技大学 | Social network link prediction method based on attention neural network |
| CN112569608A (en) * | 2020-12-22 | 2021-03-30 | 内蒙古工业大学 | Table game hybrid recommendation method based on multi-source heterogeneous data |
| CN112700056A (en) * | 2021-01-06 | 2021-04-23 | 中国互联网络信息中心 | Complex network link prediction method, complex network link prediction device, electronic equipment and medium |
| CN112994940A (en) * | 2019-05-29 | 2021-06-18 | 华为技术有限公司 | Network anomaly detection method and device |
| CN113298321A (en) * | 2021-06-22 | 2021-08-24 | 深圳市查策网络信息技术有限公司 | User intention prediction method based on multi-data fusion |
| CN115145991A (en) * | 2022-08-31 | 2022-10-04 | 南京三百云信息科技有限公司 | Data processing method and system suitable for heterogeneous data |
| CN116206453A (en) * | 2023-05-05 | 2023-06-02 | 湖南工商大学 | Traffic flow prediction method and device based on transfer learning and related equipment |
| CN117312281A (en) * | 2023-06-30 | 2023-12-29 | 江苏中科西北星信息科技有限公司 | Automatic fusion method, system, equipment and storage medium for multi-source heterogeneous data |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130262653A1 (en) * | 2012-03-28 | 2013-10-03 | Yan Qu | Construction Of Social Structures |
| US20160275401A1 (en) * | 2015-03-20 | 2016-09-22 | Fuji Xerox Co., Ltd. | Methods and systems of venue inference for social messages |
| CN106503859A (en) * | 2016-10-28 | 2017-03-15 | 国家计算机网络与信息安全管理中心 | A kind of message propagation prediction method and device based on online social relation network |
| CN107784124A (en) * | 2017-11-23 | 2018-03-09 | 重庆邮电大学 | A kind of LBSN super-networks link Forecasting Methodology based on time-space relationship |
-
2018
- 2018-08-30 CN CN201810999492.8A patent/CN109635989B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130262653A1 (en) * | 2012-03-28 | 2013-10-03 | Yan Qu | Construction Of Social Structures |
| US20160275401A1 (en) * | 2015-03-20 | 2016-09-22 | Fuji Xerox Co., Ltd. | Methods and systems of venue inference for social messages |
| CN106503859A (en) * | 2016-10-28 | 2017-03-15 | 国家计算机网络与信息安全管理中心 | A kind of message propagation prediction method and device based on online social relation network |
| CN107784124A (en) * | 2017-11-23 | 2018-03-09 | 重庆邮电大学 | A kind of LBSN super-networks link Forecasting Methodology based on time-space relationship |
Non-Patent Citations (3)
| Title |
|---|
| JIAWEI ZHANG等: "Transferring heterogeneous links across location-based social networks", 《PROCEEDINGS OF THE 7TH ACM INTERNATIONAL CONFERENCE ON WEB SEARCH AND DATA MINING》 * |
| ZIH-SYUAN WANG等: "Predicting POI visits with a heterogeneous information network", 《2015 CONFERENCE ON TECHNOLOGIES AND APPLICATIONS OF ARTIFICIAL INTELLIGENCE》 * |
| 吴帮莹: "基于多源异构数据融合的社交网络链路数据预测研究", 《中国优秀博硕士学位论文全文数据库(硕士) 信息科技辑》 * |
Cited By (27)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112994940A (en) * | 2019-05-29 | 2021-06-18 | 华为技术有限公司 | Network anomaly detection method and device |
| CN110335165A (en) * | 2019-06-28 | 2019-10-15 | 京东数字科技控股有限公司 | A kind of link prediction method and apparatus |
| CN110442802A (en) * | 2019-08-06 | 2019-11-12 | 中国科学技术大学 | A kind of more Behavior preference prediction techniques of social activity user |
| CN110442802B (en) * | 2019-08-06 | 2022-10-28 | 中国科学技术大学 | Multi-behavior preference prediction method for social users |
| CN110543943A (en) * | 2019-09-10 | 2019-12-06 | 北京百度网讯科技有限公司 | Network convergence method and device, electronic equipment and storage medium |
| US11368855B2 (en) | 2019-09-10 | 2022-06-21 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Network convergence method and device, electronic apparatus, and storage medium |
| CN110543943B (en) * | 2019-09-10 | 2022-03-25 | 北京百度网讯科技有限公司 | Network convergence method and device, electronic equipment and storage medium |
| CN110598130A (en) * | 2019-09-30 | 2019-12-20 | 重庆邮电大学 | Movie recommendation method integrating heterogeneous information network and deep learning |
| CN110598130B (en) * | 2019-09-30 | 2022-06-24 | 重庆邮电大学 | Movie recommendation method integrating heterogeneous information network and deep learning |
| CN111209943A (en) * | 2019-12-30 | 2020-05-29 | 广州高企云信息科技有限公司 | Data fusion method and device and server |
| CN111209943B (en) * | 2019-12-30 | 2020-08-25 | 广州高企云信息科技有限公司 | Data fusion method and device and server |
| CN111476673A (en) * | 2020-04-02 | 2020-07-31 | 中国人民解放军国防科技大学 | Method, device and medium for aligning users among social networks based on neural network |
| CN111475739A (en) * | 2020-05-22 | 2020-07-31 | 哈尔滨工程大学 | Heterogeneous social network user anchor link identification method based on meta-path |
| CN111475739B (en) * | 2020-05-22 | 2022-07-29 | 哈尔滨工程大学 | Heterogeneous social network user anchor link identification method based on meta-path |
| CN112446542A (en) * | 2020-11-30 | 2021-03-05 | 西安电子科技大学 | Social network link prediction method based on attention neural network |
| CN112446542B (en) * | 2020-11-30 | 2023-04-07 | 山西大学 | Social network link prediction method based on attention neural network |
| CN112569608A (en) * | 2020-12-22 | 2021-03-30 | 内蒙古工业大学 | Table game hybrid recommendation method based on multi-source heterogeneous data |
| CN112700056A (en) * | 2021-01-06 | 2021-04-23 | 中国互联网络信息中心 | Complex network link prediction method, complex network link prediction device, electronic equipment and medium |
| CN112700056B (en) * | 2021-01-06 | 2023-09-15 | 中国互联网络信息中心 | Complex network link prediction method, device, electronic equipment and medium |
| CN113298321A (en) * | 2021-06-22 | 2021-08-24 | 深圳市查策网络信息技术有限公司 | User intention prediction method based on multi-data fusion |
| CN113298321B (en) * | 2021-06-22 | 2022-03-11 | 深圳市查策网络信息技术有限公司 | User intention prediction method based on multi-data fusion |
| CN115145991B (en) * | 2022-08-31 | 2022-11-15 | 南京三百云信息科技有限公司 | Data processing method and system suitable for heterogeneous data |
| CN115145991A (en) * | 2022-08-31 | 2022-10-04 | 南京三百云信息科技有限公司 | Data processing method and system suitable for heterogeneous data |
| CN116206453A (en) * | 2023-05-05 | 2023-06-02 | 湖南工商大学 | Traffic flow prediction method and device based on transfer learning and related equipment |
| CN116206453B (en) * | 2023-05-05 | 2023-08-11 | 湖南工商大学 | Traffic flow prediction method and device based on transfer learning and related equipment |
| CN117312281A (en) * | 2023-06-30 | 2023-12-29 | 江苏中科西北星信息科技有限公司 | Automatic fusion method, system, equipment and storage medium for multi-source heterogeneous data |
| CN117312281B (en) * | 2023-06-30 | 2024-05-24 | 江苏中科西北星信息科技有限公司 | Automatic fusion method, system, equipment and storage medium for multi-source heterogeneous data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109635989B (en) | 2022-03-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109635989A (en) | A kind of social networks link prediction method based on multi-source heterogeneous data fusion | |
| Wang et al. | Recursive fact-finding: A streaming approach to truth estimation in crowdsourcing applications | |
| Joe et al. | Location-based orientation context dependent recommender system for users | |
| Zeng et al. | Latency-oriented task completion via spatial crowdsourcing | |
| CN112084422B (en) | Account data intelligent processing method and device | |
| CN104995870B (en) | Multiple target server arrangement determines method and apparatus | |
| US9665663B2 (en) | Generating user recommendations | |
| US20160105801A1 (en) | Geo-based analysis for detecting abnormal logins | |
| Pham et al. | A general model for out-of-town region recommendation | |
| CN104869581B (en) | It is propagated via the label in wireless sensor network and determines malfunctioning node | |
| US11943681B2 (en) | Human mobility measuring method | |
| US20170357903A1 (en) | Prediction System for Geographical Locations of Users Based on Social and Spatial Proximity, and Related Method | |
| Wu et al. | Adaptive lookup of open WiFi using crowdsensing | |
| Ding et al. | Spatial-temporal distance metric embedding for time-specific POI recommendation | |
| Kang et al. | How urban places are visited by social groups? Evidence from matrix factorization on mobile phone data | |
| EP4205039A1 (en) | Reinforced federated learning utilizing multiple specialized machine learning agents | |
| Lai et al. | Multi-view spatial-temporal enhanced hypergraph network for next poi recommendation | |
| Wang et al. | Connecting the hosts: Street-level ip geolocation with graph neural networks | |
| Yang et al. | Targeted pandemic containment through identifying local contact network bottlenecks | |
| Wang et al. | Online POI recommendation: Learning dynamic geo-human interactions in streams | |
| Rong et al. | GODDAG: generating origin-destination flow for new cities via domain adversarial training | |
| CN113158030B (en) | Recommendation method and device for remote interest points, electronic equipment and storage medium | |
| Liu et al. | LightTR: A Lightweight Framework for Federated Trajectory Recovery | |
| Jia et al. | A spatial–temporal hypergraph based method for service recommendation in the Mobile Internet of Things-enabled service platform | |
| Wang et al. | An optimization method for the geolocation databases of Internet hosts based on machine learning |
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 |