CN106991133A - It is a kind of based on any active ues group recommending method for restarting random walk model - Google Patents
It is a kind of based on any active ues group recommending method for restarting random walk model Download PDFInfo
- Publication number
- CN106991133A CN106991133A CN201710145033.9A CN201710145033A CN106991133A CN 106991133 A CN106991133 A CN 106991133A CN 201710145033 A CN201710145033 A CN 201710145033A CN 106991133 A CN106991133 A CN 106991133A
- Authority
- CN
- China
- Prior art keywords
- group
- project
- active ues
- user
- gmg
- 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
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/104—Peer-to-peer [P2P] networks
- H04L67/1044—Group management mechanisms
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of based on any active ues group recommending method for restarting random walk, mainly for the problems such as covering of fringe sport in group's recommendation process, propose a kind of group recommending method, it is main to include proposing any active ues group, user's propensity value, project coverage rate, group is inclined to deviation, each is obtained by restarting random walk model and enlivens coefficient correlation of the group on project, and the recommendation for including fringe sport is generated to groups of users, the problem of customer volume is excessive during group is recommended is reduced, while the recommendation of generation contains fringe sport.
Description
Technical field
It is particularly a kind of based on any active ues group for restarting random walk model the present invention relates to commending system technical field
Group recommendation method.
Background technology
With developing rapidly for Internet technology, quantity of service on network also therewith sharp increase however, this growth
The category that can receive, handle and effectively utilize considerably beyond personal or system.In such a case, different user can be directed to
The commending system of demand arises at the historic moment, and recommends theoretical and its correlation technique to turn into a popular research of academia and industrial quarters
Problem.
Traditional service recommendation system such as collaborative filtering, which is generally laid particular emphasis on to unique user, to be recommended, but in reality
In many daily routines of life, user is that occur in group form, such as trip tourism, it is online purchase by group.Therefore, group
Commending system needs to consider the tendency of all users simultaneously recommend on the other hand, is recommended for certain sole user
Easily tell on undesirable situation, and needs to put it into group, recommends to tend to obtain good effect by group
Really, and can effectively alleviate cold start-up problem caused by new user at present towards group commending system study by more and more
Concern, 2011 the conference of ACM commending systems (RecSys2011) with " for group of family recommend film " be the theme, held
Hereafter perceive film and recommend challenge match (CAMRa2011), promote popularization of group's recommendation in fields such as film, food and drink, tourisms
The prerequisite steps recommended as group are found with application groups, and its group division result plays an important role groups to recommendation effect
Inherent similarity determine the accuracy that group is recommended, the recommendation effect of high similarity group can meet or exceed single
The precision that user recommends, and when group size increases also with good stability.
During existing group is recommended, often drastically decline with the increase computational efficiency of group size, the overwhelming majority is calculated
Consumption comes from sluggish user, and is also urgent problem to be solved for the recommendation of fringe sport.Group recommend in
Being on the increase for group size, unique user is analyzed one by one and greatly wastes time and resource, and can not be ensured completely
Cover all items, and fringe sport cannot get appropriate recommendation.
The content of the invention
The technical problems to be solved by the invention are to overcome the deficiencies in the prior art and provide one kind and be based on restarting random trip
Any active ues group recommending method of model is walked, using any active ues as group's recommended, and mould is restarted by random walk
Type recommends suitable items for it.
The present invention uses following technical scheme to solve above-mentioned technical problem:
According to proposed by the present invention a kind of based on any active ues group recommending method for restarting random walk model, including with
Lower step:
Step 1: being inclined to deviation by calculating project coverage rate and group and obtaining any active ues group:To have simultaneously most
The subset group of large project coverage rate and minimum group tendency deviation is used as any active ues group;
Step 2: building correlation matrix SGMG, wherein, SGMGContain the degree of correlation square between any active ues group and project
Battle array SGM, the selection that each any active ues group includes to project is inclined to, and each any active ues group and the phase of each user
Guan Du, and each user equally has the selection propensity value to project;Then by correlation matrix SGMGIt is regular to turn to NGMG, make
To restart the probability transfer matrix of random walk model;
Restart random walk model Step 3: starting, the coefficient correlation of any active ues group and project is obtained, with NGMGFor
Random walk process is restarted in probability transfer matrix, the execution that iterated to each any active ues group, by defining end condition
Random walk is restarted in termination;Its steady probability matrix is calculated to each any active ues group againUntil convergence, to obtain
Any active ues group and the correlation coefficient of project, finally by K project recommendation before correlation coefficient highest to any active ues
Group.
As of the present invention a kind of further based on any active ues group recommending method for restarting random walk model
The idiographic flow that any active ues group is obtained in prioritization scheme, step one is as follows:
Step 1.1, calculating scoring quantization matrix, define user's set U:U={ ui, 0≤i≤| U |, project set P:P
={ pj, 0≤j≤| P |, the interactive information between user and project is quantified with rating matrix R, then is had:
R={ rij}|U|×|P|,rij≥0
Wherein, uiFor i-th of user and i is integer, pjFor j-th of project and j is integer, rijFor the i-th of rating matrix R
The element and r of row j rowijFor interactive information score data of i-th of user to j-th of project, if rij=0 represents uiAnd pjNo
Interaction, i.e. uiActivity do not cover pj;
Step 1.2, calculating project coverage rate, the project for given the subset U', U' of user's set U cover collection PU' fixed
Justice is:
U' project coverage rate Cov (U') is P subset PU' ratio shared in P, i.e.,:
Step 1.3, calculating group tendency deviation, user's subset and all use are represented with subset error score Err (U')
Tendency deviation between family:
Wherein, avg (pj, U') and represent user's subset U' to j-th of project pjAverage score;
Step 1.4, obtain according to step 1.2 and 1.3 project coverage rate and group's tendency deviation obtains any active ues group,
Any active ues group CUG is user's set U a subset, and the size of any active ues group is k, and the subset meets following two
Condition, in the subset that all sizes of user's set U are k, has:
As of the present invention a kind of further based on any active ues group recommending method for restarting random walk model
Correlation matrix S is built in prioritization scheme, step 2GMGSpecific method is as follows:
Correlation matrix SGMGTo represent the coefficient correlation between two intermediate items:
Wherein, SGMFor the correlation matrix between any active ues group and project, SMUGBy SMUAnd SUGMultiplication is obtained, SMUFor item
Mesh and the correlation matrix of any active ues group, SUGFor user and the correlation matrix of any active ues group;
NGMGObtain by the following method:
Markov transition matrix N between any active ues group and projectGMG=col_norm (SGMG), wherein col_norm
(SGMG) represent normalized SGMG, so matrix SGMGEach row and for 1, then normalization SGMAnd SMUG, then:
Wherein, NGMGS as after normalizationGMG, NGMThe probability transfer matrix from any active ues group to project is represented,
NMUGProbability transfer matrix of the expression project to any active ues group.
As of the present invention a kind of further based on any active ues group recommending method for restarting random walk model
Prioritization scheme, the end condition is:Work as λ2Terminated during≤ε and restart walk process immediately, ε is default threshold value;
λ2Random walk process is restarted in the execution that refers to iterate to each any active ues group in step 3, resulting
The resultful variance of institute, λ2Calculated by equation below:
Wherein, G is any active ues group quantity, and M is the number of entry, and μ is statistical variable, is terminated since 1 to (G+M),Represent the v of the μ nodethIt is secondary to any active ues group or project migration,(v+1) of the μ nodethIt is secondary to
Any active ues group or project migration, node are the abstract representation of any active ues group and project;
Steady probability vectorObtained by below equation:
Wherein,The vector of first any active ues group is represented,Represent last
Item object vector, for i-th of any active ues group gi, formula (1) is performed until restraining, when formula (1) is restrained, formula
(1) result is one on gi(G+M) × 1 vector.
As of the present invention a kind of further based on any active ues group recommending method for restarting random walk model
Prioritization scheme, the ε is 0.28.
The present invention uses above technical scheme compared with prior art, with following technique effect:
(1) suitable group's recommendation results are provided the user:In group is recommended, because the long tail effect that user selects is led
Cause fringe sport to be covered by popular project, excavated by using random walk model is restarted including fringe sport, user,
Dependency relation between group and project, the group's recommendation results for including fringe sport are provided for any active ues colony.
(2) computation complexity is recommended by reduction group:Because number of users is sharply increased in network, user during group is recommended
Scale also increases therewith, instead of original group, effectively shields what inactive user brought by defining any active ues group
Consumption is calculated, in the case where ensureing project coverage rate, computational efficiency is improved and has saved calculating time and resource.
Brief description of the drawings
Fig. 1 is the process for excavating any active ues group.
Fig. 2 is the process for restarting random walk.
Fig. 3 is overall flow figure of the present invention.
Embodiment
Technical scheme is described in further detail below in conjunction with the accompanying drawings:
In order to illustrate group recommending method of the present invention, we provide following preferred example, and detailed elaborates
A kind of implementation process based on any active ues group recommending method for restarting random walk model.
Be given below a kind of related notion based in any active ues group recommending method for restarting random walk model and
Specifically describe:
(1) groups of users:The set that user with similar selection tendency is formed.In such as online film comment community
Film circles.
(2) propensity value is selected:The quantized value that user is inclined to a certain purpose, different application scene quantification manner is not
Together.
(3) project coverage rate:Commending system is between a kind of utilization user and the content information and user items of project
Interactive information, recommend the information filtering system of suitable project to suitable user.
(4) any active ues group:Due to it is expected that the tendency preference of total user can fully be reflected in any active ues group, use
Subset error score Err (U') represents the tendency deviation between user's subset and total user:
Wherein avg (pj, U') and represent user's subset U' to project pjAverage score, it is clear that Err (U') is smaller, subset U'
It more can fully reflect the preference tendency of total user group.But if Cov (U') is smaller, user's subset U' can not fully be covered
All projects.Therefore, any active ues group (Active User Group, AUG) is user's set U a subset, and size is
K, it meets following two conditions simultaneously, in the subset that all sizes of user's set U are k, has:
(5) random walk model is restarted:Random walk (random walk) is a kind of mathematical statistical model, earliest by
Pearson is proposed.Random walk is made up of a series of track, and the motion of each step is all random, and this random process can
Represented with Markov Chain, the transition probability for being moved to another point from a point is unrelated with the time.Restart random walk
(random walk with restart, RWR) model is proposed by Grady, and image segmentation is used for earliest.Restarting random walk is
A kind of random walk of specific type, there is two kinds of selections when that will carry out next moved further:One kind be with certain probability according to
State-transition matrix is randomly chosen next state, and another is to start random walk with certain probability selection arbitrfary point,
It is mainly used in calculating figure the structure sexual intercourse between any two points.
RWR process is defined as:
Wherein c (0≤c≤1) is return probability,The unit vector for being 1 for i-th dimension,For the node of graph after normalization
Weighted adjacent matrix, j, [r when initiali,j] it is coefficient correlation of the node i relative to j.Then:
(6) scoring fusion:The prediction scoring or recommended project list of scoring fusion method fusion user obtains the pre- of group
Test and appraisal point or group's recommendation list.According to user u during the fusion process that scoresxIn group giIn relative weight w (ux,gi) and use
Family uxTo project itemjPrediction scoring pred (ux,itemj) calculate group giTo project itemjPrediction scoring pred
(gi,itemj):
Specific steps are expressed as follows:
Fig. 1 is the process for excavating any active ues group, step 1) it is inclined to deviation by calculating project coverage rate and is lived
Jump groups of users, and idiographic flow is as follows:
Step 1.1) calculating scoring quantization matrix, general, define user's set U:U={ ui, 0≤i≤| U |, wherein ui
For unique user and i is integer, project set P:P={ pj, 0≤j≤| P |, wherein pjFor single project and j is integer;User
Interactive information between project is quantified with rating matrix R, then is had:
R={ rij}|U|×|P|,rij≥0 (6)
Wherein, uiFor i-th of user and i is integer, pjFor j-th of project and j is integer, rijFor the i-th of rating matrix R
The element and r of row j rowijFor interactive information score data of i-th of user to j-th of project, if rij=0 represents uiAnd pjNo
Interaction, i.e. uiActivity do not cover pj;
Step 1.2) project coverage rate is calculated, the project for given the subset U', U' of user's set U covers collection PU' fixed
Justice is:
U' project coverage rate Cov (U') is P subset PU' ratio shared in P, i.e.,:
Step 1.3) group's tendency deviation is calculated, due to it is expected that inclining for total user can fully be reflected in any active ues group
To preference, the tendency deviation between user's subset and total user is represented with subset error score Err (U'):
Wherein, avg (pj, U') and represent user's subset U' to j-th of project pjAverage score;
Step 1.4) any active ues group is obtained, project coverage rate is obtained according to step 1.2 and 1.3 and group is inclined to partially
Difference, any active ues group (Active User Group) is user's set U a subset, and the size of any active ues group is
K, the subset meets following two conditions, in the subset that all sizes of user's set U are k, has:
Fig. 2 restarts the process of random walk, step 2) build the correlation probabilities matrix for enlivening group and project, specific side
Method is as follows:
Correlation matrix S is built firstGMG, the correlation matrix to represent two intermediate items between coefficient correlation:
Wherein, SGMFor the correlation matrix between any active ues group and project, SMUGBy SMUAnd SUGMultiplication is obtained, SMUFor item
Mesh and the correlation matrix of any active ues group, SUGFor user and the correlation matrix of group;
Markov transition matrix N between any active ues group and projectGMG=col_norm (SGMG), wherein col_norm
(SGMG) represent normalized SGMG, so matrix SGMGEach row and for 1, then normalization SGMAnd SMUG, then:
Wherein, NGMGS as after normalizationGMG, NGMThe probability transfer matrix from any active ues group to project is represented,
NMUGProbability transfer matrix of the expression project to any active ues group.
Step 3) start and restart random walk model, the coefficient correlation for enlivening group and project is obtained, specific method is as follows:
It is general to restart random walk process and iterate execution, until reaching a stable state, Wo Menke
To restart random walk by defining end condition and terminating, to obtain the accurate estimation to node of graph coefficient correlation.Restart with
The resultful variance of machine migration institute is calculated by equation below:
Wherein, G is any active ues group quantity, and M is the number of entry, and μ is statistical variable, is terminated since 1 to (G+M),Represent the v of the μ nodethIt is secondary to any active ues group or project migration,(v+1) of the μ nodethIt is secondary to
Any active ues group or project migration, node are the abstract representation of any active ues group and project;
The end condition is:Work as λ2Terminated during≤ε and restart walk process immediately, ε is default threshold value;
In order to obtain coefficient correlation between any active ues group and project by restarting random walk model, we shift in application
Matrix NGMG, therefore steady probability vectorObtained by below equation:
Cost and memory space are calculated in order to save, amended equation is:
Wherein, g andRespectively represent first group's any active ues to last project to
Amount.For each any active ues group gi, formula (15) is performed until restraining, when formula (15) is restrained, the knot of equation (15)
Fruit is one on gi(G+M) × 1 vector.Any active ues group giWith project mjBetween coefficient correlation it is higher, then group pair
The preference of the project is higher.
Fig. 3 gives a kind of overall flow based on any active ues group recommending method for restarting random walk model.It is false
Provided with a film score data system, comprising 6020 users, and 5763 films, wherein containing group's letter of user
Breath, score information, scoring quantity, film and film quantity.Comprise the following steps that:
The first step:Scoring according to user to film project, builds the scoring quantization matrix of user and project.
Second step:Calculating project coverage rate, passes through the scoring item involved by the subset of groups of users and project set
The project coverage rate of ratio calculation user's subset.
3rd step:Group's tendency deviation is calculated, by calculating average score and group of the subset groups of users to projects
Group's tendency deviation is used as to the ratio of the average score of respective item.
4th step:The project coverage rate obtained according to second step and the 3rd step and group's tendency deviation obtain any active ues group
Group, while the group's subset met is any active ues group.
5th step:The correlation matrix between any active ues group and project is built, is deposited to improve computational efficiency and saving
Cost is stored up, normalization is carried out to correlation matrix.
6th step:Random walk is restarted in startup.Each any active ues group nodes are iterated computing straight by given threshold
To convergence.
7th step:Any active ues group obtained by restarting after random walk and the correlation coefficient of project, by correlation
Any active ues group g is given in K film project recommendation before degree coefficient highest.
Above content is to combine specific preferred embodiment further description made for the present invention, it is impossible to assert
The specific implementation of the present invention is confined to these explanations.For general technical staff of the technical field of the invention,
On the premise of not departing from present inventive concept, some simple deductions or replacement can also be made, should all be considered as belonging to the present invention's
Protection domain.
Claims (5)
1. it is a kind of based on any active ues group recommending method for restarting random walk model, it is characterised in that to comprise the following steps:
Step 1: being inclined to deviation by calculating project coverage rate and group and obtaining any active ues group:There to be maximal term simultaneously
The subset group of mesh coverage rate and minimum group tendency deviation is used as any active ues group;
Step 2: building correlation matrix SGMG, wherein, SGMGContain the correlation matrix between any active ues group and project
SGM, the selection that each any active ues group includes to project is inclined to, and each any active ues group is related to each user's
Degree, and each user equally has the selection propensity value to project;Then by correlation matrix SGMGIt is regular to turn to NGMG, as
Restart the probability transfer matrix of random walk model;
Restart random walk model Step 3: starting, the coefficient correlation of any active ues group and project is obtained, with NGMGFor probability
Random walk process is restarted in transfer matrix, the execution that iterated to each any active ues group, is terminated by defining end condition
Restart random walk;Its steady probability matrix is calculated to each any active ues group againUntil convergence, to be enlivened
The correlation coefficient of groups of users and project, finally gives any active ues group by K project recommendation before correlation coefficient highest.
2. according to claim 1 a kind of based on any active ues group recommending method for restarting random walk model, it is special
Levy and be, the idiographic flow that any active ues group is obtained in step one is as follows:
Step 1.1, calculating scoring quantization matrix, define user's set U:U={ ui, 0≤i≤| U |, project set P:P=
{pj, 0≤j≤| P |, the interactive information between user and project is quantified with rating matrix R, then is had:
R={ rij}|U|×|P|,rij≥0
Wherein, uiFor i-th of user and i is integer, pjFor j-th of project and j is integer, rijArranged for rating matrix R the i-th row j
Element and rijFor interactive information score data of i-th of user to j-th of project, if rij=0 represents uiAnd pjWithout interaction,
That is uiActivity do not cover pj;
Step 1.2, calculating project coverage rate, the project for given the subset U', U' of user's set U cover collection PU'It is defined as:
U' project coverage rate Cov (U') is P subset PU'The shared ratio in P, i.e.,:
Step 1.3, calculate group's tendency deviation, represented with subset error score Err (U') user's subset and total user it
Between tendency deviation:
Wherein, avg (pj, U') and represent user's subset U' to j-th of project pjAverage score;
Step 1.4, obtain according to step 1.2 and 1.3 project coverage rate and group's tendency deviation obtains any active ues group, it is active
Groups of users CUG is user's set U a subset, and the size of any active ues group is k, and the subset meets following two bars
Part, in the subset that all sizes of user's set U are k, has:
3. according to claim 1 a kind of based on any active ues group recommending method for restarting random walk model, it is special
Levy and be, correlation matrix S is built in step 2GMGSpecific method is as follows:
Correlation matrix SGMGTo represent the coefficient correlation between two intermediate items:
Wherein, SGMFor the correlation matrix between any active ues group and project, SMUGBy SMUAnd SUGMultiplication is obtained, SMUFor project with
The correlation matrix of any active ues group, SUGFor user and the correlation matrix of any active ues group;
NGMGObtain by the following method:
Markov transition matrix N between any active ues group and projectGMG=col_norm (SGMG), wherein col_norm
(SGMG) represent normalized SGMG, so matrix SGMGEach row and for 1, then normalization SGMAnd SMUG, then:
Wherein, NGMGS as after normalizationGMG, NGMRepresent the probability transfer matrix from any active ues group to project, NMUGTable
Probability transfer matrix of the aspect mesh to any active ues group.
4. according to claim 1 a kind of based on any active ues group recommending method for restarting random walk model, it is special
Levy and be, the end condition is:Work as λ2Terminated during≤ε and restart walk process immediately, ε is default threshold value;
λ2Random walk process is restarted in the execution that refers to iterate to each any active ues group in step 3, and resulting is all
As a result variance, λ2Calculated by equation below:
Wherein, G is any active ues group quantity, and M is the number of entry, and μ is statistical variable, is terminated since 1 to (G+M),Represent
The v of the μ nodethIt is secondary to any active ues group or project migration,(v+1) of the μ nodethIt is secondary to be used to active
Family group or project migration, node are the abstract representation of any active ues group and project;
Steady probability vector cgiMObtained by below equation:
Wherein, cgiM(1:G the vector of first any active ues group, c) are representedgiM(G+1:G+M last project) is represented
Vector, for i-th of any active ues group gi, formula (1) is performed until restraining, when formula (1) is restrained, the knot of formula (1)
Fruit is one on gi(G+M) × 1 vector.
5. according to claim 4 a kind of based on any active ues group recommending method for restarting random walk model, it is special
Levy and be, the ε is 0.28.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710145033.9A CN106991133B (en) | 2017-03-13 | 2017-03-13 | It is a kind of based on any active ues group recommending method for restarting random walk model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710145033.9A CN106991133B (en) | 2017-03-13 | 2017-03-13 | It is a kind of based on any active ues group recommending method for restarting random walk model |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106991133A true CN106991133A (en) | 2017-07-28 |
CN106991133B CN106991133B (en) | 2019-08-06 |
Family
ID=59413212
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710145033.9A Active CN106991133B (en) | 2017-03-13 | 2017-03-13 | It is a kind of based on any active ues group recommending method for restarting random walk model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106991133B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107590243A (en) * | 2017-09-14 | 2018-01-16 | 中国人民解放军信息工程大学 | The personalized service recommendation method to be sorted based on random walk and diversity figure |
CN108628967A (en) * | 2018-04-23 | 2018-10-09 | 西安交通大学 | A kind of e-learning group partition method generating network similarity based on study |
WO2019072040A1 (en) * | 2017-10-10 | 2019-04-18 | 阿里巴巴集团控股有限公司 | Random walking and cluster-based random walking method, apparatus and device |
CN111062800A (en) * | 2019-11-27 | 2020-04-24 | 同盾控股有限公司 | Data processing method and device, electronic equipment and computer readable medium |
CN111209745A (en) * | 2018-11-02 | 2020-05-29 | 北京好啦科技有限公司 | Information reliability evaluation method, equipment and storage medium |
US10776334B2 (en) | 2017-10-10 | 2020-09-15 | Alibaba Group Holding Limited | Random walking and cluster-based random walking method, apparatus and device |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101437220A (en) * | 2008-09-18 | 2009-05-20 | 广州五度信息技术有限公司 | System and method for implementing mutual comment and color bell recommendation between users |
CN102663128A (en) * | 2012-04-24 | 2012-09-12 | 南京师范大学 | Recommending system of large-scale collaborative filtering |
CN105677647A (en) * | 2014-11-17 | 2016-06-15 | 中国移动通信集团广东有限公司 | Individual recommend method and system |
-
2017
- 2017-03-13 CN CN201710145033.9A patent/CN106991133B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101437220A (en) * | 2008-09-18 | 2009-05-20 | 广州五度信息技术有限公司 | System and method for implementing mutual comment and color bell recommendation between users |
CN102663128A (en) * | 2012-04-24 | 2012-09-12 | 南京师范大学 | Recommending system of large-scale collaborative filtering |
CN105677647A (en) * | 2014-11-17 | 2016-06-15 | 中国移动通信集团广东有限公司 | Individual recommend method and system |
Non-Patent Citations (1)
Title |
---|
原福永等: "一种在信任网络中随机游走的推荐算法", 《现代图书情报技术》 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107590243A (en) * | 2017-09-14 | 2018-01-16 | 中国人民解放军信息工程大学 | The personalized service recommendation method to be sorted based on random walk and diversity figure |
CN107590243B (en) * | 2017-09-14 | 2019-09-06 | 中国人民解放军信息工程大学 | The personalized service recommendation method to be sorted based on random walk and diversity figure |
WO2019072040A1 (en) * | 2017-10-10 | 2019-04-18 | 阿里巴巴集团控股有限公司 | Random walking and cluster-based random walking method, apparatus and device |
US10776334B2 (en) | 2017-10-10 | 2020-09-15 | Alibaba Group Holding Limited | Random walking and cluster-based random walking method, apparatus and device |
US10901971B2 (en) | 2017-10-10 | 2021-01-26 | Advanced New Technologies Co., Ltd. | Random walking and cluster-based random walking method, apparatus and device |
CN108628967A (en) * | 2018-04-23 | 2018-10-09 | 西安交通大学 | A kind of e-learning group partition method generating network similarity based on study |
CN108628967B (en) * | 2018-04-23 | 2020-07-28 | 西安交通大学 | Network learning group division method based on learning generated network similarity |
CN111209745A (en) * | 2018-11-02 | 2020-05-29 | 北京好啦科技有限公司 | Information reliability evaluation method, equipment and storage medium |
CN111209745B (en) * | 2018-11-02 | 2022-04-22 | 北京好啦科技有限公司 | Information reliability evaluation method, equipment and storage medium |
CN111062800A (en) * | 2019-11-27 | 2020-04-24 | 同盾控股有限公司 | Data processing method and device, electronic equipment and computer readable medium |
CN111062800B (en) * | 2019-11-27 | 2023-09-08 | 同盾控股有限公司 | Data processing method, device, electronic equipment and computer readable medium |
Also Published As
Publication number | Publication date |
---|---|
CN106991133B (en) | 2019-08-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106991133A (en) | It is a kind of based on any active ues group recommending method for restarting random walk model | |
CN100504856C (en) | Method and apparatus for characterizing documents based on clusters of related words | |
Li et al. | Clustering approach for hybrid recommender system | |
Chung | Graph theory in the information age | |
Marlin et al. | Recommender systems, missing data and statistical model estimation | |
CN107657034A (en) | A kind of event social networks proposed algorithm of social information enhancing | |
CN111144933A (en) | Commodity recommendation method and device, electronic equipment and storage medium | |
Chen et al. | Collaborative filtering based on demographic attribute vector | |
CN113094558A (en) | Network node influence sequencing method based on local structure | |
CN114817508A (en) | Sparse graph and multi-hop attention fused session recommendation system | |
CN104008301B (en) | A kind of field concept hierarchical structure method for auto constructing | |
Shams et al. | Item-based collaborative ranking | |
CN112989211B (en) | Method and system for determining information similarity | |
Yang et al. | A nearest neighbor based personal rank algorithm for collaborator recommendation | |
Sitkrongwong et al. | Latent probabilistic model for context-aware recommendations | |
Hafiene et al. | A new structural and semantic approach for identifying influential nodes in social networks | |
CN101833565B (en) | Method for actively selecting related feedbacks of representative image | |
CN108334645A (en) | One kind feeding back newer activity recommendation method based on graph model | |
CN112307343B (en) | Cross-E-book city user alignment method based on double-layer iterative compensation and full-face representation | |
Singh et al. | Recommendations using absorbing random walks | |
CN113362034A (en) | Position recommendation method | |
CN110175296B (en) | Node recommendation method and server in network graph and storage medium | |
Wang et al. | A Tri-Attention Neural Network Model-BasedRecommendation | |
CN114596109A (en) | Method and device for determining recommendation information, electronic equipment and storage medium | |
Bei et al. | Trial and error in influential social networks |
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 | ||
EE01 | Entry into force of recordation of patent licensing contract | ||
EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20170728 Assignee: NUPT INSTITUTE OF BIG DATA RESEARCH AT YANCHENG Assignor: NANJING University OF POSTS AND TELECOMMUNICATIONS Contract record no.: X2020980007071 Denomination of invention: An active user group recommendation method based on restart random walk model Granted publication date: 20190806 License type: Common License Record date: 20201026 |