CN106354886B - The method of potential neighbor relational graph screening nearest-neighbors is utilized in recommender system - Google Patents
The method of potential neighbor relational graph screening nearest-neighbors is utilized in recommender system Download PDFInfo
- Publication number
- CN106354886B CN106354886B CN201610909600.9A CN201610909600A CN106354886B CN 106354886 B CN106354886 B CN 106354886B CN 201610909600 A CN201610909600 A CN 201610909600A CN 106354886 B CN106354886 B CN 106354886B
- Authority
- CN
- China
- Prior art keywords
- neighbors
- nearest
- relational graph
- potential neighbor
- target
- 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
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2462—Approximate or statistical queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/28—Databases characterised by their database models, e.g. relational or object models
- G06F16/284—Relational databases
- G06F16/285—Clustering or classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/60—Information retrieval; Database structures therefor; File system structures therefor of audio data
- G06F16/63—Querying
- G06F16/635—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/73—Querying
- G06F16/735—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computational Linguistics (AREA)
- Probability & Statistics with Applications (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Complex Calculations (AREA)
Abstract
The invention discloses the methods that potential neighbor relational graph screening nearest-neighbors are utilized in recommender system, comprising the following steps: (1) generating, there is the object gathering of redundancy properties to close C;(2) building gathering closes the corresponding potential neighbor relational graph of C;(3) quantify each edge weight in potential neighbor relational graph, while weight indicate while two objects being abutted become nearest-neighbors a possibility that;(4) potential neighbor relationship is cut, rejects extra comparison;(5) nearest-neighbors of the potential neighbor relational graph screening target after cutting are utilized.Recommendation based on complete large-scale dataset is mapped to the recommendation of a lesser data set of scale under the premise of ensureing the precision recommended by this screening technique, is reduced recommender system scale, is guaranteed the high efficiency of recommended method.
Description
Technical field
The present invention relates to recommended technology fields, screen arest neighbors using potential neighbor relational graph especially in recommender system
The method in residence.
Background technique
Current ever-increasing data volume makes user need to expend a great deal of time just find valuable information.Collaboration
Filtering is considered as one of the effective technology for solving information overload problem, is widely used to film, music, books, travelling, new
News etc. recommends field.The main thinking of collaborative filtering method is to the hobby prediction target user of project according to similar users to item
Purpose viewpoint, or implement to recommend for destination item according to opinion of the user to similar terms.Therefore, collaborative filtering needs to solve
A critical issue be: how the nearest-neighbors of effective selection target user or project.But recommender system possesses millions of
User and project, scale still sustainable growth are counted, nearest-neighbors is searched in mass data with conventional method and is difficult to ensure and closing
It manages and accurate recommendation is provided in the time.
The core of collaborative filtering method is to need to compare all items in recommender system (or user) two-by-two, is passed through
Similarity between calculating selects nearest-neighbors.The number compared two-by-two is more, and the efficiency of operation is lower, but it is searched for
A possibility that nearest-neighbors, is higher;Otherwise the number compared two-by-two is fewer, and operational efficiency is higher, but misses nearest-neighbors
Possibility is also higher, to influence recommendation precision.Data internal association is in close relations and complicated in recommender system in addition, valence
It is extremely unbalanced to be worth Density Distribution, the calculating of mass data in recommender system cannot be depended on as Small Sample Database collection to complete
The statistical analysis of office data and iterative calculation need the on-demand reduction method of heuristic data.
Summary of the invention
It is provided the technical problem to be solved by the present invention is to overcome the deficiencies in the prior art in recommender system using latent
In the method for neighborhood figure screening nearest-neighbors, the method for the present invention both ensures recommendation effect to reject redundancy and extra comparison
Rate, and recommendation precision is not sacrificed, it is ensured that accurate recommendation is provided within reasonable time.
The present invention uses following technical scheme to solve above-mentioned technical problem:
The method for screening nearest-neighbors using potential neighbor relational graph in recommender system proposed according to the present invention, including
Following steps:
Step 1, to set i ∈ O, O be the object set for needing to screen nearest-neighbors, and i is object, using Fuzzy clustering techniques foundation
Object i is assigned in multiple clusters by preset probability, thus generates the cluster containing K object cluster by the feature vector of object
Set C;
Step 2, building gathering close the corresponding potential neighbor relational graph G of CC={ VC,EC, wherein VCIt is vertex set, ECIt is
Undirected line set;It is specific as follows:
If object i and object j appears in gathering simultaneously and closes in the same cluster c of C, object i and j be referred to as co-occurrence to and remember
For < i, j >;For each pair of co-occurrence in gathering conjunction C to < i, j >, first by the corresponding vertex v of object i and jiAnd vjIt is added to figure
GCIn, if not there is nonoriented edge between two objects i and j, use side ei,jConnect vertex viAnd vj;Wherein, scheme GCIn every
Side ei,jIndicate a potential neighborhood, side ei,jTwo adjacent vertex vsiAnd vjCorresponding object i and j is known as contiguous object,
J ∈ O, c ∈ C, vi∈VC, vj∈VC, ei,j∈EC;
Step 3, quantization figure GCThe weight of middle each edge;
Step 4, to figure GCIt is cut, deletes potential neighbor relational graph GCThe weight on middle side is lower than wminSide, remaining side
Constitute a new figure GC';Wherein, wminFor the minimal weight threshold value of setting;
Step 5 chooses object i as target, utilizes the potential neighbor relational graph G after cuttingC' screening target arest neighbors
It occupies, for GC' in figure target i every adjacent side ei,j, relatively and calculate utility vector RiWith RjBetween similarity, then according to
Its nearest-neighbors is screened in all of its neighbor object of target i according to neighbour's alternative condition;Wherein, RiIndicate object i effectiveness to
Amount, RjIndicate the utility vector of object j.
As it is of the present invention in recommender system using the method for potential neighbor relational graph screening nearest-neighbors into one
Prioritization scheme is walked, uses following formula to calculate side e in the step 3i,jWeight ei,j.weight:
Wherein, CiIndicate that the gathering that object i is subordinate to is closed, CjIndicate that the gathering that object j is subordinate to is closed, Ci,jIt is shared for object i and j
Set,| * | for membership in set *, d (*) indicates the degree of vertex *.
As it is of the present invention in recommender system using the method for potential neighbor relational graph screening nearest-neighbors into one
Prioritization scheme is walked, neighbour's alternative condition refers to that the maximum preceding k object of the similarity of selection and target constitutes mesh in the step 5
Target neighbour collection.
As it is of the present invention in recommender system using the method for potential neighbor relational graph screening nearest-neighbors into one
Walk prioritization scheme, K >=1.
As it is of the present invention in recommender system using the method for potential neighbor relational graph screening nearest-neighbors into one
Walk prioritization scheme, side ei,jWeight and relational graph GCMiddle side ei,jThe object i and j abutted shared cluster is related.
The invention adopts the above technical scheme compared with prior art, has following technical effect that
Under the premise of ensureing recommendation precision, it is lesser that the recommendation based on complete large-scale dataset is mapped to a scale
The recommendation of data set, and recommender system scale is significantly reduced, guarantee the high efficiency of recommended method.Specific manifestation is in the following areas:
(1) object is assigned in multiple clusters by certain probability using Fuzzy clustering techniques, this redundancy properties provide
A kind of indispensable, reliable, method for avoiding missing nearest-neighbors;
(2) in cluster process, by the high object assignment of similitude into the same cluster, the object that peels off individually is excluded.This
It is not only advantageous for improving and recommends precision, while decreasing the extra comparison of object bring that peels off, improve and recommend efficiency;
(3) in the building process of potential neighbor relational graph, each pair of co-occurrence is to a line at most corresponding diagram, each edge table
Show that a kind of potential neighborhood, subsequent step are only compared the contiguous object of each edge, has calculated the similar of them
Degree;Therefore, redundancy and extra co-occurrence pair are eliminated, the number compared two-by-two in object set is reduced, improves the effect of recommendation
Rate;
(4) by the weight threshold on the appropriate side of setting, the side for being lower than weight in potential neighbor weighted connections figure is rejected, into
One step eliminates extra comparison.
Detailed description of the invention
Fig. 1 is process flow of the invention.
Specific embodiment
Technical solution of the present invention is described in further detail with reference to the accompanying drawing:
If Fig. 1 is process flow of the invention, following steps: step (1) generates the object gathering conjunction with redundancy properties
C;Step (2), building gathering close the corresponding potential neighbor relational graph of C;Each edge in step (3), quantization potential neighbor relational graph
Weight, while weight indicate while two objects being abutted become nearest-neighbors a possibility that;Step (4) closes potential neighbor
System is cut, and extra comparison is rejected;Step (5) utilizes the arest neighbors of the potential neighbor relational graph screening target after cutting
It occupies.
[embodiment 1]
In a film recommender system, it is known that user collects U, and Item Sets I, each project is a film;FiIt indicates
The feature vector of project i, for describing film school, prize-winning type, performer and director etc.;RiThe effectiveness of expression project i ∈ I to
Amount, ru,i∈RiIt is scoring of the user u ∈ U to project i.Recommender system screens destination item i most using potential neighbor relational graph
Neighbour occupies, and then utilizes scoring of the target user u to the score in predicting user u of all nearest-neighbors of project i to project i.At this
In embodiment 1, object set is project set, and detailed process is as follows for the nearest-neighbors of screening programme:
(1) gathering is generated to close.Project i ∈ I is assigned to by certain probability according to item characteristic using Fuzzy clustering techniques
In multiple clusters, thus generates project gathering and close C.
(2) building gathering closes the corresponding potential neighbor relational graph G of CC={ VC,EC, VCIt is vertex set, ECIt is nonoriented edge collection
It closes.Scheme GCSpecific construction method is as follows: for co-occurrence each pair of in C to < i, j >, first by the corresponding vertex project i ∈ I and j ∈ I
viAnd vjIt is added to figure GCIn, then use side ei,j∈ECConnect vertex viAnd vj.During creating relational graph, if discovery two
A nonoriented edge is had existed between a project, then no longer needs to increase a line between them.Each edge indicates that one is latent in figure
Neighborhood, similarity calculation need to be passed through in the next steps, it is determined whether be nearest-neighbors.
(3) quantization figure GCEach edge weight.Relational graph GCAfter finishing, for further screening neighbor relationships, requirement figure
Middle each edge.Weighing computation method is as follows: knownIt respectively indicates the gathering that project i, j is subordinate to close, Ci,j=Ci
∩CjReferred to as project i and j shared set.When project i and j share cluster it is more (i.e. | Ci,j| value is bigger), they become each other
A possibility that nearest-neighbors, is bigger.In addition to considering | Ci,j|, side right escheat need to consider the sum of its adjoining project cluster subjected,
And the degree of the adjacent vertex on side.When the cluster that the adjoining project on some side is subordinate to is fewer, the weight Ying Yuegao on the side;When some side
Vertex degree it is smaller, the weight Ying Yuegao on the side.Side e is calculated with following formulai,j∈ECWeight
Wherein, d (*) indicates the degree of vertex *.
(4) extra comparison is cut.Minimal weight threshold value w according to settingminDelete figure GCThe weight on middle side is lower than wmin's
Side, remaining side constitute the figure G after cuttingC'。
(5) project i is chosen as target, screens the nearest-neighbors of target i.For GC'Every adjacent side of target i in figure
ei,j, compare the utility vector R of target i and j project i in user-project utility matrixiWith the utility vector R of project jjBetween
Similitude, calculate the similarity between them;Then it is screened in all of its neighbor project of target i according to neighbour's alternative condition
Its nearest-neighbors is chosen the neighbour for constituting target with the maximum preceding k project of the similarity of destination item and is collected in this example 1.
Scoring of the score in predicting target user u of each neighbour to destination item finally is concentrated to project neighbour using target user u.
[embodiment 2]
In a film recommender system, it is known that user collects U, and Item Sets I, each project is a film;FqIt indicates
The feature vector of user q, gender, age, occupation etc. for describing user;RqIndicate the utility vector of user q ∈ U, rq,m∈
RuIt is scoring of the user q to project m ∈ I.Recommender system screens the nearest-neighbors of target user q using potential neighbor relational graph,
Then the scoring using all nearest-neighbors of target user q to the score in predicting user q of project m to project m.In the present embodiment
In 2, object set is user's set, and screening the nearest-neighbors of user, detailed process is as follows:
(1) gathering is generated to close.User u ∈ U is assigned to by certain probability according to user characteristics using Fuzzy clustering techniques
In multiple clusters, thus generates user's gathering and close C.
(2) building gathering closes the corresponding potential neighbor relational graph G of CC={ VC,EC, VCIt is vertex set, ECIt is nonoriented edge collection
It closes.Scheme GCSpecific construction method is as follows: for co-occurrence each pair of in C to < q, h >, first by the corresponding vertex user q ∈ U and h ∈ U
vqWithvhIt is added to figure GCIn, then use side eq,h∈ECConnect vertex vqWithvh.During creating relational graph, if discovery two
A nonoriented edge is had existed between user, then no longer needs to increase a line between them.Each edge indicates that one is potential in figure
Neighborhood, similarity calculation need to be passed through in the next steps, it is determined whether be nearest-neighbors.
(3) quantization figure GCEach edge weight.Relational graph GCAfter finishing, for further screening neighbor relationships, requirement figure
Middle each edge.Weighing computation method is as follows: knownIt respectively indicates the gathering that user q, h are subordinate to close, Cq,h=Cq
∩ChReferred to as user q and h shared set.When user q and h share cluster it is more (i.e. | Cq,h| value is bigger), they become each other
A possibility that nearest-neighbors, is bigger.In addition to considering | Cq,h|, side right escheat need to consider that it abuts the sum of user's cluster subjected,
And the degree of the adjacent vertex on side.When the cluster that the adjoining user on some side is subordinate to is fewer, the weight Ying Yuegao on the side;When some side
Vertex degree it is smaller, the weight Ying Yuegao on the side.Therefore, side e can be calculated with following formulaq,hWeight:
Wherein, d (*) indicates the degree of vertex *.
(4) extra comparison is cut.Minimal weight threshold value w according to settingminDelete figure GCThe weight on middle side is lower than wmin's
Side, remaining side constitute the figure G after cuttingC'。
(5) user q is chosen as target, screens the nearest-neighbors of target user q.For GC' in figure target q every neighbour
Edge fit eq,h, compare the utility vector R of user q in user-project utility matrix of target q and hqWith the utility vector of user h
RhBetween similitude, calculate the similarity between them;Then all of its neighbor user of the foundation neighbour's alternative condition in target q
Middle its nearest-neighbors of screening are chosen in this example 2 and constitute target with the maximum preceding k user of the similarity of target user q
Neighbour's collection.Finally concentrate each neighbour to the score in predicting target user q of project to destination item using the neighbour of target user q
Scoring.
Specific embodiments described above has carried out further the purpose of the present invention, technical scheme and beneficial effects
Detailed description, it should be understood that being not limited to this hair the foregoing is merely specific embodiments of the present invention
Bright range, any those skilled in the art, that is made under the premise of not departing from design and the principle of the present invention is equal
Variation and modification, should belong to the scope of protection of the invention.
Claims (4)
1. utilizing the method for potential neighbor relational graph screening nearest-neighbors in recommender system, which is characterized in that including following step
It is rapid:
Step 1, to set i ∈ O, O be the object set for needing to screen nearest-neighbors, and i is object, using Fuzzy clustering techniques according to object
Feature vector, object i is assigned in multiple clusters by preset probability, the gathering containing K object cluster is thus generated and closes
C;
Step 2, building gathering close the corresponding potential neighbor relational graph G of CC={ VC,EC, wherein VCIt is vertex set, ECIt is undirected
Line set;It is specific as follows:
If object i and object j appears in gathering simultaneously and closes in the same cluster c of C, object i and j be referred to as co-occurrence to and be denoted as <
I, j >;For each pair of co-occurrence in gathering conjunction C to < i, j >, first by the corresponding vertex v of object i and jiAnd vjIt is added to figure GCIn,
If not there is nonoriented edge between two objects i and j, side e is usedi,jConnect vertex viAnd vj;Wherein, scheme GCMiddle each edge ei,j
Indicate a potential neighborhood, side ei,jTwo adjacent vertex vsiAnd vjCorresponding object i and j is known as contiguous object, j ∈ O,
C ∈ C, vi∈VC, vj∈VC, ei,j∈EC;
Step 3, quantization figure GCThe weight of middle each edge;
Side e is calculated using following formula in step 3i,jWeight ei,j.weight:
Wherein, CiIndicate that the gathering that object i is subordinate to is closed, CjIndicate that the gathering that object j is subordinate to is closed, Ci,jThe collection shared for object i and j
It closes,Ci,j=Ci∩Cj, | * | for membership in set *, d (*) indicates the degree of vertex *;
Step 4, to figure GCIt is cut, deletes potential neighbor relational graph GCThe weight on middle side is lower than wminSide, remaining side is constituted
One new figure GC';Wherein, wminFor the minimal weight threshold value of setting;
Step 5 chooses object i as target, utilizes the potential neighbor relational graph G after cuttingC'Screen the nearest-neighbors of target, needle
To GC'Every adjacent side e of target i in figurei,j, relatively and calculate utility vector RiWith RjBetween similarity, then according to close
Adjacent alternative condition screens its nearest-neighbors in all of its neighbor object of target i;Wherein, RiIndicate the utility vector of object i, Rj
Indicate the utility vector of object j.
2. the method according to claim 1 for screening nearest-neighbors using potential neighbor relational graph in recommender system,
It is characterized in that, neighbour's alternative condition refers to that the maximum preceding k object of the similarity of selection and target constitutes target in the step 5
Neighbour collection.
3. the method according to claim 1 for screening nearest-neighbors using potential neighbor relational graph in recommender system,
It is characterized in that, K >=1.
4. the method according to claim 1 for screening nearest-neighbors using potential neighbor relational graph in recommender system,
It is characterized in that, side ei,jWeight and relational graph GCMiddle side ei,jThe object i and j abutted shared cluster is related.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610909600.9A CN106354886B (en) | 2016-10-18 | 2016-10-18 | The method of potential neighbor relational graph screening nearest-neighbors is utilized in recommender system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610909600.9A CN106354886B (en) | 2016-10-18 | 2016-10-18 | The method of potential neighbor relational graph screening nearest-neighbors is utilized in recommender system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106354886A CN106354886A (en) | 2017-01-25 |
CN106354886B true CN106354886B (en) | 2019-05-28 |
Family
ID=57863321
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610909600.9A Active CN106354886B (en) | 2016-10-18 | 2016-10-18 | The method of potential neighbor relational graph screening nearest-neighbors is utilized in recommender system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106354886B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108848152B (en) * | 2018-06-05 | 2021-09-21 | 腾讯科技(深圳)有限公司 | Object recommendation method and server |
CN110097893B (en) * | 2019-05-23 | 2021-04-20 | 北京搜狐新媒体信息技术有限公司 | Audio signal conversion method and device |
CN115277156B (en) * | 2022-07-22 | 2023-05-23 | 福建师范大学 | User identity privacy protection method for resisting neighbor attack in social network |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102135999A (en) * | 2011-03-25 | 2011-07-27 | 南京财经大学 | User credibility and item nearest neighbor combination Internet recommendation method |
CN102663010A (en) * | 2012-03-20 | 2012-09-12 | 复旦大学 | Personalized image browsing and recommending method based on labelling semantics and system thereof |
CN103761237A (en) * | 2013-12-04 | 2014-04-30 | 南京邮电大学 | Collaborative filtering recommending method based on characteristics and credibility of users |
CN104202211A (en) * | 2014-08-25 | 2014-12-10 | 电子科技大学 | Autonomous system level network topology identification method combining active and passive measurement |
-
2016
- 2016-10-18 CN CN201610909600.9A patent/CN106354886B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102135999A (en) * | 2011-03-25 | 2011-07-27 | 南京财经大学 | User credibility and item nearest neighbor combination Internet recommendation method |
CN102663010A (en) * | 2012-03-20 | 2012-09-12 | 复旦大学 | Personalized image browsing and recommending method based on labelling semantics and system thereof |
CN103761237A (en) * | 2013-12-04 | 2014-04-30 | 南京邮电大学 | Collaborative filtering recommending method based on characteristics and credibility of users |
CN104202211A (en) * | 2014-08-25 | 2014-12-10 | 电子科技大学 | Autonomous system level network topology identification method combining active and passive measurement |
Non-Patent Citations (1)
Title |
---|
"基于社会网络分析的协同推荐方法改进";冯勇等;《计算机应用》;20130331;841-844页 |
Also Published As
Publication number | Publication date |
---|---|
CN106354886A (en) | 2017-01-25 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103544216B (en) | The information recommendation method and system of a kind of combination picture material and keyword | |
CN106844416B (en) | A kind of sub-topic method for digging | |
Xiaomei et al. | Microblog sentiment analysis with weak dependency connections | |
Kundu et al. | Fuzzy-rough community in social networks | |
CN107038184B (en) | A kind of news recommended method based on layering latent variable model | |
CN106354886B (en) | The method of potential neighbor relational graph screening nearest-neighbors is utilized in recommender system | |
CN109902235A (en) | User preference based on bat optimization clusters Collaborative Filtering Recommendation Algorithm | |
CN104778237A (en) | Individual recommending method and system based on key users | |
CN109947987A (en) | A kind of intersection collaborative filtering recommending method | |
Botta et al. | Finding network communities using modularity density | |
CN104077723A (en) | Social network recommending system and social network recommending method | |
CN109447261A (en) | A method of the network representation study based on multistage neighbouring similarity | |
CN109213926A (en) | A kind of location recommendation method divided based on community with Multi-source Information Fusion | |
CN107341199A (en) | A kind of recommendation method based on documentation & info general model | |
Yigit et al. | Extended topology based recommendation system for unidirectional social networks | |
CN106127260A (en) | A kind of multi-source data fuzzy clustering algorithm of novelty | |
CN115329215A (en) | Recommendation method and system based on self-adaptive dynamic knowledge graph in heterogeneous network | |
CN106649380A (en) | Hot spot recommendation method and system based on tag | |
CN110390058A (en) | Consider the credible mixed recommendation method of Web service of timeliness | |
CN114461879A (en) | Semantic social network multi-view community discovery method based on text feature integration | |
CN105468669B (en) | A kind of adaptive microblog topic method for tracing merging customer relationship | |
CN106919647B (en) | Clustering-based network structure similarity recommendation method | |
Ren et al. | Deep structural enhanced network for document clustering | |
CN109299849B (en) | Group demand level calculation method in social network | |
Liang et al. | A utility-based recommendation approach for academic literatures |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | 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 |