CN108647334B - Video social network homology analysis method under spark platform - Google Patents

Video social network homology analysis method under spark platform Download PDF

Info

Publication number
CN108647334B
CN108647334B CN201810451383.2A CN201810451383A CN108647334B CN 108647334 B CN108647334 B CN 108647334B CN 201810451383 A CN201810451383 A CN 201810451383A CN 108647334 B CN108647334 B CN 108647334B
Authority
CN
China
Prior art keywords
video
user
similarity
social network
relationship
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.)
Active
Application number
CN201810451383.2A
Other languages
Chinese (zh)
Other versions
CN108647334A (en
Inventor
王勇
杨晓东
张译权
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Electronic Science and Technology of China
Original Assignee
University of Electronic Science and Technology of China
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by University of Electronic Science and Technology of China filed Critical University of Electronic Science and Technology of China
Priority to CN201810451383.2A priority Critical patent/CN108647334B/en
Publication of CN108647334A publication Critical patent/CN108647334A/en
Application granted granted Critical
Publication of CN108647334B publication Critical patent/CN108647334B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/01—Social networking

Abstract

The invention discloses a video social network homology analysis method under a Spark platform, which comprises the steps of modeling a video social network relationship graph, preprocessing video similarity based on a graph structure, calculating user similarity based on video weight and timestamp punishment and pruning results based on user relationship; the video social network relationship graph is modeled and used for modeling the structured video social network data into a graph form; the video similarity preprocessing based on the graph structure is used for obtaining global and dynamic video similarity; the user similarity calculation based on the video weight and the timestamp is used for calculating the similarity between users; and the result pruning based on the user relationship is used for screening the high-similarity user pairs and removing the noise users. The method provided by the invention can quickly and accurately find all homologous user pairs which possibly correspond to users in the same real world from massive video social network data.

Description

Video social network homology analysis method under spark platform
Technical Field
The invention relates to the technical field of homology analysis under a social network, in particular to a video social network homology analysis method under a Spark platform.
Background
Homology analysis is a technology for finding entities corresponding to the same real world in different social networks. The current methods mainly focus on the following aspects:
(1) homology analysis methods based on pairwise matching. The homology analysis method based on pairwise matching is to gather two data setsS 1 ={r 1 ,r 2 ,r 3 ,...}AndS2={c 1 ,c 2 ,c 3 ,...}by using a certain kind ofSimilarity matching algorithmf(r 1 ,c 1 )To perform a one-to-one match. The matching idea is that firstly, the similarity of each attribute between two records is calculated, and the similarities of all the attributes are combined into a similarity comparison vectorV={sim 1， sim 2， sim 3，。。。 }Whereinsim i And finally, judging whether the two records correspond to the same user in the real world or not by comparing the vectors. When matching in pairs, the accuracy of the obtained similarity comparison vector is often decisive for the matching result. The common attribute similarity calculation method can be divided into a similarity comparison algorithm based on characters and a similarity comparison algorithm based on tokens according to the difference of the character string division granularity. In homology analysis algorithms based on pairwise matching, the relationship between records is often not taken into account. Homology analysis algorithms based on pairwise matching are "static" in that the similarity of each pair of matching pairs depends only on the similarity value of the attributes between the matches, which is fixed and not affected by other records. However, in practical situations, the records often influence each other.
(2) Relationship-based homology analysis. Relationship-based homology analysis abstracts a social network into a graph consisting of a set of points V and a set of edges EG=(V，E). Depending on the type of relationship between records, this graph may be a directed graph or an undirected graph. After the social network relationship graph is constructed, similarity between records is calculated through analysis of relationships between vertexes. The homology analysis based on the relationship needs to provide a certain prior homologous user matching pair before calculation, and the quantity and quality of the provided prior homologous user matching pair have decisive influence on the analysis result. In the analysis across social networks, the difficulty is great and the analysis cannot be performed as a main part.
In addition, compared with the traditional social network, in the video social network, a video publisher is more concerned to attract fans through a high-quality video area, obtains more attention, and is relatively not concerned to find friends in the same city and the same hobby, so that in the video social network, the information loss rate of a user is high, and homology analysis is relatively difficult only by means of user information. And in the video social network, there is a special record: and (6) video. A video recording consists of two parts: video descriptions and video content files. Since the video description is only a simple summary of the video content, there may also be description errors or deviations. If analysis of the relational graph or analysis of attribute feature values is performed according to a conventional homology analysis method (where video description is used as an attribute of a user), the accuracy of the matching result is greatly affected. However, if a video content file is used as a reference for matching, since the video content file is as small as several MB and as large as hundreds of MB, no matter frame-by-frame analysis or analysis by extracting key frames, a lot of time is wasted, and there are great demands on machine performance and algorithm performance.
Disclosure of Invention
Aiming at the problems of particularity of a video social network and the existing homology analysis method, the invention provides a video social network homology analysis method under a Spark platform, which specifically comprises a video social network relationship graph modeling step S1, a graph structure-based video similarity preprocessing step S2, a video weight and timestamp punishment-based user similarity calculation step S3 and a user relationship-based result pruning step S4; for two video social networks with similar user figures, inputting structured video information data and user information data to a Spark platform, constructing and analyzing graphs, and outputting matching pairs which are homologous users in the two video social networks; wherein the content of the first and second substances,
s1: decomposing a video social network relationship graph into the following steps based on the user information data and the video information data structured by the video social network: reconstructing a video social network relationship graph by the video information subgraph, the video publishing subgraph and the user relationship subgraph; the step is used for modeling the structured video social network data into a graph form so as to reduce the analysis difficulty;
s2: based on the video information subgraph, carrying out video similarity preprocessing by using a distributed graph structure method; the step obtains global and dynamic video similarity through a graph structure so as to more accurately reflect the similarity difference between videos;
s3: based on the video distribution subgraph, calculating user similarity by using weighted video similarity processed by a timestamp penalty function; the user similarity is judged according to the video similarity after weighting and timestamp punishment so as to improve the accuracy of user similarity calculation;
s4: based on the user relation subgraph, pruning the high-score matching pairs by using a pruning method; the step prunes the retrieved high-similarity users and removes the noise users.
Further, the step S1 includes the following sub-steps:
s11: decomposition of video social network relationship diagram: the video social network comprises five important information of users, videos, user information, user relationship information and video information; decomposing the video social network relationship graph into: the video information subgraph, the video publishing subgraph and the user relationship subgraph are used for reducing the analysis difficulty;
s12: video information subgraph modeling: extracting video and its information label from structured video information data, using attribute set
Figure DEST_PATH_IMAGE002
Represents; extracting a relation pair set of video and information labels from an attribute set
Figure DEST_PATH_IMAGE004
By using
Figure DEST_PATH_IMAGE006
Representing videoiFor information labeljThe inclusion relationship of (1); each of which is provided with one
Figure 510357DEST_PATH_IMAGE006
Then establish a directed edgee=(m i ,t j )；
S13: modeling a video distribution subgraph: extracting user and video released by user from structured user information data, and releasing set by video
Figure DEST_PATH_IMAGE008
Represents; extracting a set of distribution relation pairs of users and videos from a video distribution attribute set
Figure DEST_PATH_IMAGE010
By using
Figure DEST_PATH_IMAGE012
Represents the useriFor videojThe publishing relationship of (1); each of which is provided with one
Figure 817711DEST_PATH_IMAGE012
Then establish a directed edgee=(u i ,m j )；
S14: modeling a user relation subgraph: extracting users and related users from structured user information data, and using relationship set
Figure DEST_PATH_IMAGE014
Representing, attention sets in a set of relationships
Figure DEST_PATH_IMAGE016
Vermicelli collection
Figure DEST_PATH_IMAGE018
Extracting a set of user relationship pairs from the set of user relationships
Figure DEST_PATH_IMAGE020
By using
Figure DEST_PATH_IMAGE022
Represents the useriAnd the userjThe attention relationship of (1); each of which is provided with one
Figure 784398DEST_PATH_IMAGE022
Then establish a directed edgee=(u i ,u j )。
Further, the step S2 includes the following sub-steps:
s21: creating a variable G for storing a point set, an edge set, the in-degree of each point and the out-degree of each point in the video information subgraph;
s22: reading video collections in sequenceM a Video inm i To makem i Information label corresponding to the samet k Composite edge(m i ,t k )Add to panel G;
s23: reading video sets in turn in the same wayM b Video inm j To makem j Information label corresponding to the samet h Composite edge(m j ,t h )Is added to the figureG；
S24: initializing a similarity matrixSAnd state transition matrixW；
S25: iterative computation of similarity matrixSAfter each iteration, a decision is madeSWhether to converge or not: (1) if the judgment result is true, the iteration of the similarity is not carried out, and the step goes to S26; (2) otherwise, continuing to iterate the similarity;
s26: the similarity of the video nodes in the converged similarity matrix S is screened out and stored inS m And outputting the obtained product.
Further, the step S3 includes the following sub-steps:
s31: extracting a set of usersU a User in (1)u i ；
S32: initializing a useru i Video weight set ofWIs empty, and extractedu i Distributed video for initializationu i Video collection ofM i ；
S33: in turn toM i Video inm k Calculating weight values and storing in a video weight setW；
S34: sequentially extracting user setsU b User in (1)u j Extracting users as wellu j Video collection ofM j Calculating andm k andm h time penalty value ofT loss And calculateu i Andu j the final similarity of (2);
s35: will be in the matrixS u In updating correspondingu i -u j Element (2)S u ] ij ；
S36: when in use
Figure DEST_PATH_IMAGE024
: repeating S31-S35; otherwise, the calculation is ended.
Further, the step S4 includes the following sub-steps:
s41: social networking over videoG a User set ofU a Get out the useru i And extractingu i Vermicelli collectionFans i And attention setFollow i ；
S42: from the user similarity matrix S m Is extracted fromu i The similarity is highest and is greater than the threshold valueδFront ofkThe users form the highest similarity setU top = {u 1 ,u 2 ,u 3 ,...u k }；
S43: from the highest similarity set in turnU top Extract the useru j And extractingu j Vermicelli collectionFans j And attention setFollow j ；
S44: computingu i Andu j vermicelli collectionFans i AndFans j then calculating the similarity ofu i Andu j attention set ofFollow i AndFollow j similarity of the cluster, and finally, calculating the similarity of the finished fan setsSim Fans Similarity to attention setSim Follow Utilizing computing usersu i Andu j the relationship similarity of (1);
s45: according to the similarity of the relationship calculated each time, finding the relationu i Adding the user with the highest relational similarity into the homologous user matching pair setS h ；
S46: has been calculatedU a All users inu i Then, outputting the matching pair set of the homologous usersS h 。
The invention has the beneficial effects that: the invention provides a high-accuracy homology analysis method, which can quickly and accurately find all homologous user pairs which possibly correspond to users in the same real world from massive video social network data.
Drawings
FIG. 1 is a processing flow of a video social network homology analysis method under a Spark platform;
FIG. 2 is a video information subgraph construction process;
FIG. 3 is a video distribution subgraph construction process;
FIG. 4 is a user relationship sub-graph construction process.
Detailed Description
The technical solution of the present invention is described in detail below with reference to the accompanying drawings, but the scope of the present invention is not limited to the following:
aiming at the particularity of the video social network and the problems of the traditional homology analysis method, the invention provides a video social network homology analysis method under a Spark platform.
As shown in fig. 1, a method for analyzing video social network homology under a Spark platform specifically includes a video social network relationship graph modeling step S1, a graph structure-based video similarity preprocessing step S2, a user similarity calculation step S3 based on video weight and timestamp penalty, and a user relationship-based result pruning step S4; for two video social networks with similar user figures, inputting structured video information data and user information data to a Spark platform, constructing and analyzing graphs, and outputting matching pairs which are homologous users in the two video social networks; wherein the content of the first and second substances,
s1: decomposing a video social network relationship graph into the following steps based on the user information data and the video information data structured by the video social network: reconstructing a video social network relationship graph by the video information subgraph, the video publishing subgraph and the user relationship subgraph; the step is used for modeling the structured video social network data into a graph form so as to reduce the analysis difficulty;
s2: based on the video information subgraph, carrying out video similarity preprocessing by using a distributed graph structure method; the step obtains global and dynamic video similarity through a graph structure so as to more accurately reflect the similarity difference between videos;
s3: based on the video distribution subgraph, calculating user similarity by using weighted video similarity processed by a timestamp penalty function; the user similarity is judged according to the video similarity after weighting and timestamp punishment so as to improve the accuracy of user similarity calculation;
s4: based on the user relation subgraph, pruning the high-score matching pairs by using a pruning method; the step prunes the retrieved high-similarity users and removes the noise users.
Further, the step S1 of building the video social network relationship graph includes the following sub-steps:
s11: decomposition of video social network relationship diagram: the video social network comprises five important information of users, videos, user information, user relation information and video information, the huge network is analyzed, records of the same user in the two social networks are found, and the video social network is difficult to research in the field of the existing graph theory. In addition, the user information in the video social network is seriously lost, and the user information is difficult to directly analyze. Thus, the video social network relationship graph is decomposed into 3 subgraphs, including: the video information subgraph, the video distribution subgraph and the user relationship subgraph.
S12: video information sub-graph modeling as shown in fig. 2: extracting video and its information label from structured video information data, using attribute set
Figure 572750DEST_PATH_IMAGE002
Represents; extracting a relation pair set of video and information labels from an attribute set
Figure 652702DEST_PATH_IMAGE004
By using
Figure 944006DEST_PATH_IMAGE006
Representing videoiFor information labeljThe inclusion relationship of (1); each of which is provided with one
Figure 7777DEST_PATH_IMAGE006
Then establish a directed edgee=(m i ,t j )；
S13: video publishing subgraph modeling as shown in fig. 3: extracting user and video released by user from structured user information data, and releasing set by video
Figure 800152DEST_PATH_IMAGE008
Represents; extracting a set of distribution relation pairs of users and videos from a video distribution attribute set
Figure 937873DEST_PATH_IMAGE010
By using
Figure 524712DEST_PATH_IMAGE012
Represents the useriFor videojThe publishing relationship of (1); each occurrence of oneAn
Figure 747883DEST_PATH_IMAGE012
Then establish a directed edgee=(u i ,m j )；
S14: user relationship sub-graph modeling is shown in FIG. 4: extracting users and related users from structured user information data, and using relationship set
Figure 78370DEST_PATH_IMAGE014
Representing, attention sets in a set of relationships
Figure 70597DEST_PATH_IMAGE016
Vermicelli collection
Figure 93916DEST_PATH_IMAGE018
Extracting a set of user relationship pairs from the set of user relationships
Figure 69963DEST_PATH_IMAGE020
By using
Figure 938561DEST_PATH_IMAGE022
Represents the useriAnd the userjThe attention relationship of (1); each of which is provided with one
Figure 50874DEST_PATH_IMAGE022
Then establish a directed edgee=(u i ,u j )。
Further, the step S2 of performing video similarity preprocessing by using a distributed graph structure method includes the following sub-steps:
s21: creating a variable G for storing a point set, an edge set, the in-degree of each point and the out-degree of each point in the video information subgraph;
s22: reading video collections in sequenceM a Video inm i To makem i Information label corresponding to the samet k Composite edge(m i ,t k )Add to panel G;
s23: reading video sets in turn in the same wayM b Video inm j To makem j Information label corresponding to the samet h Composite edge(m j ,t h )Is added to the figureG；
S24: initializing a similarity matrixSAnd state transition matrixW；
S25: iterative computation of similarity matrixSAfter each iteration, a decision is madeSWhether to converge or not: (1) if the determination is true, no further iteration of the similarity is performed, and the process proceeds to S26. (2) Otherwise, continuing to iterate the similarity;
s26: the similarity of the video nodes in the converged similarity matrix S is screened out and stored in S m And outputting the obtained product.
Further, the method for calculating user similarity by using video similarity after weighting and processing by timestamp penalty function in step S3 includes the following steps:
s31: extracting a set of usersU a User in (1)u i ；
S32: initializing a useru i Video weight set ofWIs empty, and extractedu i Distributed video for initializationu i Video collection ofM i ；
S33: in turn toM i Video inm k Calculating weight values and storing in a video weight setW；
S34: sequentially extracting user setsU b User in (1)u j Extracting users as wellu j Video collection ofM j Calculating andm k andm h time penalty value ofT loss And calculateu i Andu j is the most important ofFinal similarity;
s35: will be in the matrixS u In updating correspondingui-ujElement (2)S u ] ij ；
S36: when in use
Figure 994164DEST_PATH_IMAGE024
: repeating S31-S35; otherwise, the calculation is ended.
Further, the step S4 is a method for pruning the high-score matching pairs by using a pruning method, and the method comprises the following steps:
s41: social networking over videoG a User set ofU a Get out the useru i And extractingu i Vermicelli collectionFans i And attention setFollow i ；
S42: from the user similarity matrix S m Is extracted fromu i The similarity is highest and is greater than the threshold valueδFront ofkThe users form the highest similarity setU top = {u 1 ,u 2 ,u 3 ,...u k }；
S43: from the highest similarity set in turnU top Extract the useru j And extractingu j Vermicelli collectionFans j And attention setFollow j ；
S44: computingu i Andu j vermicelli collectionFans i AndFans j then calculating the similarity ofu i Andu j attention set ofFollow i AndFollow j similarity of the cluster, and finally, calculating the similarity of the finished fan setsSim Fans Similarity to attention setSim Follow Utilizing computing usersu i Andu j the relationship similarity of (1);
s45: according to the similarity of the relationship calculated each time, finding the relationu i Adding the user with the highest relational similarity into the homologous user matching pair setS h ；
S46: has been calculatedU a All users inu i Then, outputting the matching pair set of the homologous usersS h 。

Claims (5)

1. A video social network homology analysis method under a Spark platform is characterized by comprising a video social network relationship graph modeling step S1, a graph structure-based video similarity preprocessing step S2, a user similarity calculation step S3 based on video weight and timestamp penalty, and a user relationship-based result pruning step S4; for two video social networks with similar user figures, inputting structured video information data and user information data to a Spark platform, constructing and analyzing graphs, and outputting matching pairs which are homologous users in the two video social networks; wherein the content of the first and second substances,
s1: decomposing a video social network relationship graph into the following steps based on the user information data and the video information data structured by the video social network: reconstructing a video social network relationship graph by the video information subgraph, the video publishing subgraph and the user relationship subgraph; the step is used for modeling the structured video social network data into a graph form so as to reduce the analysis difficulty;
s2: based on the video information subgraph, carrying out video similarity preprocessing by using a distributed graph structure method; the step obtains global and dynamic video similarity through a graph structure so as to more accurately reflect the similarity difference between videos;
s3: based on the video distribution subgraph, calculating user similarity by using weighted video similarity processed by a timestamp penalty function; the user similarity is judged according to the video similarity after weighting and timestamp punishment so as to improve the accuracy of user similarity calculation;
s4: based on the user relation subgraph, pruning the high-score matching pairs by using a pruning method; the step prunes the retrieved high-similarity users and removes the noise users.
2. The method for analyzing social network homology of video under Spark platform as claimed in claim 1, wherein said step S1 comprises the following sub-steps:
s11: decomposition of video social network relationship diagram: the video social network comprises five important information of users, videos, user information, user relationship information and video information; decomposing the video social network relationship graph into: the video information subgraph, the video publishing subgraph and the user relationship subgraph are used for reducing the analysis difficulty;
s12: video information subgraph modeling: extracting video and its information label from structured video information data, using attribute set m ═ t1，t2，t3,.. }; extracting a relation pair set H ═ H of the video and the information labels from the attribute set1,h2,h3,., using hi＝(mi,tj) Representing the inclusion relationship of the video i to the information label j; each occurrence of hi＝(mi,tj) Then, a directed edge e ═ m is establishedi,tj)；
S13: modeling a video distribution subgraph: extracting the user and the video distributed by the user from the structured user information data, and distributing the set u-m by using the video1，m2，m3,.. }; extracting a distribution relation pair set P ═ P between the user and the video from the video distribution set1,p2,p3,., by p ═ u (u)i,mj) Representing the distribution relation of the user i to the video j; each occurrence of p ═ ui,mj) Then, a directed edge e is established (u ═ u)i,mj)；
S14: modeling a user relation subgraph: extracting users and related users from structured user information data, and using relation set Urelation(Follow, Fans) in the set of relationshipsFocus set of (1) { focus ═ focus }1，follow2，follow3,., fan set Fans ═ Fans1，fans2，fans3,., extracting a user relationship pair set R (R) from the user relationship set1，r2，r3,., using r ═ u (u)i，uj) Representing the attention relationship between the user i and the user j; each occurrence of r ═ ui，uj) Then, a directed edge e is established (u ═ u)i，uj)。
3. The method for analyzing social network homology of video under Spark platform as claimed in claim 1, wherein said step S2 comprises the following sub-steps:
s21: creating a variable G for storing a point set, an edge set, the in-degree of each point and the out-degree of each point in the video information subgraph;
s22: reading video set M in sequenceaVideo m iniLet m beiInformation label t corresponding theretokComposition edge (m)i,tk) Add to panel G;
s23: reading video set M in turn in the same waybVideo m injLet m bejInformation label t corresponding theretohComposition edge (m)j,th) Add to panel G;
s24: initializing a similarity matrix S and a state transition matrix W;
s25: iteratively calculating a similarity matrix S, and judging whether S is converged after each iteration: (1) if the judgment result is true, the iteration of the similarity is not carried out, and the step goes to S26; (2) otherwise, continuing to iterate the similarity;
s26: the similarity of the video nodes in the converged similarity matrix S is screened out and stored in SmAnd outputting the obtained product.
4. The method for analyzing social network homology of video under Spark platform as claimed in claim 1, wherein said step S3 comprises the following sub-steps:
s31: extracting a user set UaUser u ini；
S32: initializing user uiIs empty and u is extractediThe published video is used to initialize uiVideo set Mi；
S33: in sequence to MiVideo m inkCalculating a weight value and storing the weight value into a video weight set W;
s34: sequentially extracting user set UbUser u injAlso extract user ujVideo set MjCalculating and mkAnd mhIs a time penalty value TlossAnd calculate uiAnd ujThe final similarity of (2);
s35: the corresponding u will be updated in the similarity matrix Si-ujElement [ S ] of]ij；
S36: when i < UaSize (): repeating S31-S35; otherwise, the calculation is ended.
5. The method for analyzing social network homology of video under Spark platform as claimed in claim 1, wherein said step S4 comprises the following sub-steps:
s41: social network G on videoaUser set U ofaTake out user uiAnd extracting uiVermicelli incorporating FansiAnd focus set Followi；
S42: from the user similarity matrix SmIs extracted fromiThe first k users with the highest similarity and larger than the threshold value delta form a highest similarity set Utop＝{u1,u2,u3,...uk}；
S43: sequentially collecting U from the highest similaritytopExtracts user u fromjAnd extracting ujVermicelli incorporating FansjAnd focus set Followj；
S44: calculating uiAnd ujVermicelli incorporating FansiAnd FansjThen calculating uiAnd ujFocus set ofiAnd FollowjSimilarity of the cluster, and finally, calculating the similarity Sim of the finished fan clusterFansAnd attention set similarity SimFollowComputing user uiAnd ujThe relationship similarity of (1);
s45: according to the similarity of the relationship calculated each time, the u and the u are foundiAdding the user with the highest relational similarity into the homologous user matching pair set Sh；
S46: after U is calculatedaAll users u iniThen, outputting the matching pair set S of the homologous usersh。
CN201810451383.2A 2018-05-11 2018-05-11 Video social network homology analysis method under spark platform Active CN108647334B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810451383.2A CN108647334B (en) 2018-05-11 2018-05-11 Video social network homology analysis method under spark platform

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810451383.2A CN108647334B (en) 2018-05-11 2018-05-11 Video social network homology analysis method under spark platform

Publications (2)

Publication Number Publication Date
CN108647334A CN108647334A (en) 2018-10-12
CN108647334B true CN108647334B (en) 2021-10-19

Family

ID=63754693

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810451383.2A Active CN108647334B (en) 2018-05-11 2018-05-11 Video social network homology analysis method under spark platform

Country Status (1)

Country Link
CN (1) CN108647334B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113556368A (en) * 2020-04-23 2021-10-26 北京达佳互联信息技术有限公司 User identification method, device, server and storage medium
CN112069156A (en) * 2020-07-14 2020-12-11 武汉白虹软件科技有限公司 Algorithm model for obtaining target object character relationship and relationship intimacy
CN114513337B (en) * 2022-01-20 2023-04-07 电子科技大学 Privacy protection link prediction method and system based on mail data

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104053023A (en) * 2014-06-13 2014-09-17 海信集团有限公司 Method and device for determining video similarity
CN104133897A (en) * 2014-08-01 2014-11-05 哈尔滨工程大学 Micro blog topic source tracing method based on topic influence
CN104956357A (en) * 2012-12-31 2015-09-30 谷歌公司 Creating and sharing inline media commentary within a network
CN107480714A (en) * 2017-08-09 2017-12-15 东北大学 Across social network user recognition methods based on full visual angle characteristic
CN107818183A (en) * 2017-12-05 2018-03-20 云南大学 A kind of Party building video pushing method based on three stage combination recommended technologies

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10157430B2 (en) * 2015-05-27 2018-12-18 Intellidimension Inc. Simulated network system and method having simulated user profiles and item profiles based on the same vocabulary for information integration with real world e-commerce and other user network systems
US20180121550A1 (en) * 2016-10-28 2018-05-03 Facebook, Inc. Ranking Search Results Based on Lookalike Users on Online Social Networks

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104956357A (en) * 2012-12-31 2015-09-30 谷歌公司 Creating and sharing inline media commentary within a network
CN104053023A (en) * 2014-06-13 2014-09-17 海信集团有限公司 Method and device for determining video similarity
CN104133897A (en) * 2014-08-01 2014-11-05 哈尔滨工程大学 Micro blog topic source tracing method based on topic influence
CN107480714A (en) * 2017-08-09 2017-12-15 东北大学 Across social network user recognition methods based on full visual angle characteristic
CN107818183A (en) * 2017-12-05 2018-03-20 云南大学 A kind of Party building video pushing method based on three stage combination recommended technologies

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
"Potential Friend Recommendation in Online Social Network";Xing Xie;《2010 IEEE/ACM Int"l Conference on Green Computing and Communications & Int"l Conference on Cyber, Physical and Social Computing》;20110307;第831-835页 *
"基于运营商跨平台同源数据的购物决策路径研究";钱兵 等;《电信技术》;20170625(第06期);第13-18页 *
"跨社交媒体网络大数据下的用户建模";项连城 等;《大数据》;20160920;第2卷(第05期);第32-42页 *

Also Published As

Publication number Publication date
CN108647334A (en) 2018-10-12

Similar Documents

Publication Publication Date Title
Wu et al. A novel clustering method for static video summarization
CN112434169B (en) Knowledge graph construction method and system and computer equipment thereof
CN108647334B (en) Video social network homology analysis method under spark platform
CN106599686A (en) Malware clustering method based on TLSH character representation
CN110880019B (en) Method for adaptively training target domain classification model through unsupervised domain
CN110555455A (en) Online transaction fraud detection method based on entity relationship
CN108470052B (en) Anti-trust attack recommendation algorithm based on matrix completion
WO2022088390A1 (en) Image incremental clustering method and apparatus, electronic device, storage medium and program product
CN112380453B (en) Article recommendation method and device, storage medium and equipment
CN114172688B (en) Method for automatically extracting key nodes of network threat of encrypted traffic based on GCN-DL (generalized traffic channel-DL)
Panda et al. Scalable video summarization using skeleton graph and random walk
CN115510981A (en) Decision tree model feature importance calculation method and device and storage medium
CN112036659A (en) Social network media information popularity prediction method based on combination strategy
Abirami et al. Performance analysis of K-means and bisecting K-means algorithms in Weblog data
Li et al. Superpixel segmentation based on spatially constrained subspace clustering
CN109542949B (en) Formal vector-based decision information system knowledge acquisition method
CN110688150B (en) Binary file code search detection method and system based on tensor operation
WO2016086731A1 (en) Multi-level parallel key frame cloud extraction method and system
Ning et al. Missing data of quality inspection imputation algorithm base on stacked denoising auto-encoder
Zhou et al. Image inpainting with contrastive relation network
Fu et al. An image topic model for image denoising
Dhoot et al. Efficient Dimensionality Reduction for Big Data Using Clustering Technique
CN113190841A (en) Method for defending graph data attack by using differential privacy technology
Huo et al. Encrypted traffic identification method based on multi-scale spatiotemporal feature fusion model with attention mechanism
Venkatavara Prasad et al. Sequential mining of real time moving object by using fast frequence pattern algorithm

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