CN107301583A - It is a kind of that method is recommended based on user preference and the cold start-up trusted - Google Patents
It is a kind of that method is recommended based on user preference and the cold start-up trusted Download PDFInfo
- Publication number
- CN107301583A CN107301583A CN201710384564.3A CN201710384564A CN107301583A CN 107301583 A CN107301583 A CN 107301583A CN 201710384564 A CN201710384564 A CN 201710384564A CN 107301583 A CN107301583 A CN 107301583A
- Authority
- CN
- China
- Prior art keywords
- user
- preference
- mrow
- trusted
- scoring
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
Abstract
The present invention is claimed a kind of based on user preference and the cold start-up trusted recommendation method, including step：Synthesis trust value between S1, foundation user social contact information, measure user, builds trusting relationship matrix；S2, according to user's score data, calculates the preference similarity of user, builds preference relation matrix；S3, preference relation and trusting relationship are merged using comprehensive similarity computational methods, and update the weights in comprehensive similarity using ant colony algorithm iteration, are carried out multiple-objection optimization, weights is reached adaptive optimal, build preference trusting relationship matrix；S4, based on preference trusting relationship matrix, chooses the score value that the most trusted neighbor collection of targeted customer predicts respective item for it；S5, will predict the high project recommendation of scoring to targeted customer.The present invention improves the accuracy of users to trust measurement, more accurately builds user behavior preference and improves the recommendation quality to cold start-up user.
Description
Technical field
It is more particularly to a kind of to be pushed away based on user preference and the cold start-up trusted the invention belongs to data mining collaboration field
Recommend method.
Background technology
Commending system refer to working knowledge discovery technique produce personalized recommendation so that help user substantial amounts of article,
Useful information is filtered out in product, film, music, webpage etc., has been widely used in each e-commerce platform.
Existing recommendation method be broadly divided into based on content be divided into the recommendation based on correlation rule, information filtering recommend,
Three kinds of collaborative filtering recommending etc..Especially, collaborative filtering recommending technology is to be employed to obtain a kind of most commonly used recommendation method.But
There is also the problem of user's cold start-up for collaborative filtering recommending technology.Cold start-up based on user collaborative filtered recommendation technology is recommended
The principle of method：By searching and the higher customer group of targeted customer's preference similarity, then by customer group item interested
Mesh recommends targeted customer.Therefore, it is to solve the problems, such as user's cold start-up to find with the higher other users of targeted customer's similarity
Key.At present, the method for similarity mainly utilizes user's scoring item information between calculating user, but user-project scores
Data in matrix are generally very sparse.
For colod-application family score information Sparse Problem, the most widely used at present is exactly using polytype number
According to the comprehensive similarity for calculating user.Wherein, with Web2.0 fast development, the social information of user is easier to be mined,
And reliability is higher；Therefore, the method that fusion score information and social information calculate the comprehensive similarity between user is to be employed
Obtain most widely.
Although scientific research personnel calculates the comprehensive similarity of user by using the data of both the above type, to a certain degree
On improve the personalized recommendation precision at colod-application family.But the uniqueness that two types data are individually present often is ignored, from
And there are problems that following both sides when handling two kinds of data：On the one hand, typically all only consider from single trust angle
The trusting relationship of user, i.e., only consider influence of the single aspect to users to trust, and used typically on user's scoring is similar
Similarity calculating method；On the other hand, user's comprehensive similarity calculating in, using conventional method determine user scoring it is similar,
Trust value, social similar weights distribution, causing the comprehensive similarity degree of accuracy of user has certain one-sidedness.Therefore, it is existing
User's cold start-up of some fusion trust information and score information recommends method still to fail to effectively improve the personalization at colod-application family
Recommend the degree of accuracy.
The content of the invention
Present invention seek to address that above problem of the prior art.Propose it is a kind of improve user between trust value metric it is accurate
Degree, while the accuracy of user's scoring Similarity Measure is improved, the cold start-up based on user preference and trust of multiple-objection optimization
Recommendation method.Technical scheme is as follows：
A kind of to recommend method based on user preference and the cold start-up trusted, it comprises the following steps：S1, foundation user social contact
Synthesis trust value between information, measure user, builds trusting relationship matrix；S2, foundation user's score data, calculate the inclined of user
Good similarity, builds preference relation matrix；S3, preference relation and step using comprehensive similarity computational methods fusion steps S2
S1 trusting relationship, and the weights in comprehensive similarity are updated using ant colony algorithm iteration, multiple-objection optimization is carried out, weights are reached
To adaptive optimal, preference trusting relationship matrix is built；S4, based on preference trusting relationship matrix, choose cold start-up user most
Trusted neighbor collection predicts that respective item is scored for it；S5, it will predict that cold start-up user is given in the high project recommendation of scoring.
Further, the step S1 builds trusting relationship matrix and comprised the following steps：
Trusting degree of the user to another user is worth to by calculating the trust between user, trust value is 0 two users
Between be not present trusting relationship, formed N × N Digraph adjacent matrix T, i.e. the trusting relationship matrix of user；Element therein
TS(ui,uj)Represent synthesis trusting degree of i-th of user to j-th of user；The N is positive integer, and i, j are no more than N
Positive integer.
Further, the method for trust value includes between calculating user in the step S1：
(1) trust of user is divided into according to the characteristic of user social contact net and directly trusts Dt and indirectly trust IDt；
Wherein, | Fu| represent user u good friend's number；N represents bang path between user and user
Sum；D represents the length of reachable path between user and user, and in order to save computing cost and avoid not
Necessary Trust transitivity, this method sets 3 >=d ＞ 1.
Further, in the step (1), the trust value between user is：
T(u,v)=Dt(u,v)+IDt(u,v)
(2) according to user, the probability that is trusted of user is divided into and local can be trusted by relative position scope in social network
Probability α and the overall situation can be trusted probability β；
(3) probability α can be trusted according to the asymmetric similar part for calculating user of trust；
Wherein, FuRepresent user u direct good friend；FvRepresent user v direct good friend；Fu∩FvRepresent user u and user
V common friend, the present invention can be trusted the contradiction that probability α is 0 in order to avoid local in the presence of directly trust between user, therefore write from memory
Recognize and set all common friend numbers up between good friend to be initially 1；α(u,v)And α(v,u)Then represent that user v is trusted by user u respectively
Probability and the probability trusted by user v of user u.
(4) probability β can be trusted by popularity mechanism calculating user's overall situation being obtained in social network according to user；
Wherein, | Iu| represent user u trust in-degree；Represent the average trust of user in social network residing for user u
In-degree；When user u trust in-degreeWhen, it is 1 that the user u overall situation, which can be trusted probability,.
(5) correlation theories knowledge in above rational formula and sociology obtains user and integrates degree of belief TS(u,v)。
TS(u,v)=α(u,v)*Dt(u,v)+βv*IDt(u,v)
Further, the preference relation matrix of the step S2 structures user is specifically included：N number of user is obtained to M item
Purpose actively scores, and the project scoring of not actively scoring is null value, forms a N × M users-project rating matrix R；It is wherein first
Plain R(ui,m)Represent scoring of i-th of user to m-th of project；Described N, M are positive integer, and i, m are no more than N successively
With M positive integer；Scoring similarity between user is obtained by Similarity Measure based on this score information so as to formed a N ×
N matrix S, i.e. user preference relation matrix；Wherein element S(ui,uj)Represent i-th of user similar to the scoring of j-th of user
Degree, the N is positive integer, and i, j are no more than N positive integer.
Further, the method for scoring similarity includes between calculating user in the step S2：
(1) the initial score similarity between user is defined using the common scoring item accountings of Jie Kaerde between user
is(u,v)：
Wherein, IuAnd IvRepresent respectively by the total number of the project of user u and user v scorings.
(2) using based on user it is similar user scoring preference similar computational methods alleviate user between effort analysis to
The influence that scoring similarity between family is brought.Therefore the similarity S that scored between final user(u,v)：
S(u,v)=is(u,v)*URP(u,v)
Wherein, μuAnd μvThe average score value of user u and user v to project is represented respectively；σuAnd σvUser u is represented respectively
Standards of grading with user v are poor.
Further, the method for comprehensive similarity includes between calculating user in the step S3：
(1) comprehensive similarity between the social information and score information of user, user is merged using the method for linear weighted function
For：
CS(u,v)=a*TS(u,v)+b*S(u,v)+c*J(u,v)
Wherein, a, b, c are respectively three weight coefficients, and three meets a+b+c=1；J(u,v)Represent user u and user v
It is social similar.
(2) with the weight coefficient sequence { a, b, c } in comprehensive similarity for independent variable, it is definitely average that integrated forecasting is scored
Error and F mediations rate can obtain the object function of a multiple-objection optimization：
Wherein, S represents object function；MAE represents absolute average error；P(u,m)Represent that prediction of the user to project is scored,
It can hereafter introduce；N represents the number of times of prediction scoring；F mediations rate is formed by accuracy rate P and recall rate R fusions.
Further, F mediations rate is：
Wherein, M and N represent the item number being scored in the item number and test set of prediction scoring respectively, and H represents user
The undue item number through commenting.
(3) these targets are optimized using bee colony optimized algorithm, when object function S takes minimum value, its institute is right
The weight coefficient { a, b, c } answered is optimal solution.
Further, prediction score values of the targeted customer u to project m is calculated in the step S4：
Wherein, CS(u,v)Represent the synthesis similitude between user u and user v；R(v,m)It is scorings of the user v to project m,
TNuIt is the nearest trusted neighbor collection of targeted customer.
Further, prediction score values of the targeted customer u to project m is calculated in the step S4：
Wherein, CS(u,v)Represent the synthesis similarity measurements between user u and user v；R(v,m)It is that user v is commented project m
Point, TNuIt is the nearest trusted neighbor collection of targeted customer.
Further, for project set I to be recommended in the step S5recIt is shared | Irec| individual prediction scoring；
Descending arranges the prediction scoring of project to be recommended, top (1≤top≤M) the individual project for taking prediction scoring forward
Recommended to targeted customer.
Advantages of the present invention and have the beneficial effect that：
1. present invention refinement users to trust, is classified as directly trusting and trust, and in view of in sociology trusting indirectly
The characteristic of propagation, probability can be trusted further to improve the accuracy of users to trust measurement by being firstly introduced, and fully excavate social
Implicit information present in trust information, alleviates the Sparse Problem at colod-application family.
2. in the scoring similarity between calculating user, the present invention is tried to achieve first by Jie Kaerde similarities between user
Initial score similarity, and in view of influence of the grade scoring to user's effort analysis, further dropped using URP similarities
The error of the low scoring of the user caused by effort analysis similarity.
3. using the weight sequence in comprehensive similarity as independent variable, comprehensive absolute average error and F mediation rates obtain one
The object function of multiple-objection optimization.The present invention is iterated optimization to the two targets using ant colony algorithm first, improves algorithm
Convergence rate, make its weight coefficient adaptive.Ensure that the weight coefficient in user's comprehensive similarity is in optimum allocation, improve
The degree of accuracy that user's comprehensive similarity is calculated, so as to improve the recommendation quality recommended colod-application family.
Brief description of the drawings
Fig. 1 is that the present invention provides the schematic diagram that preferred embodiment recommends method based on user preference and the cold start-up trusted；
Fig. 2 is the trusting relationship matrix schematic diagram of user；
Fig. 3 is the trusting relationship network diagram of user；
Fig. 4 is the schematic diagram of user-project rating matrix；
Fig. 5 is the flow chart that the preferred embodiment of the present invention recommends method based on user preference and the cold start-up trusted.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, detailed
Carefully describe.Described embodiment is only a part of embodiment of the present invention.
The present invention solve above-mentioned technical problem technical scheme be：
The invention discloses a kind of based on user preference and the cold start-up trusted recommendation method, as shown in figure 1, including following
Step：
The first step, builds the trusting relationship matrix of user.Users to trust relational matrix in obtaining as shown in Figure 2, and
The trusting relationship network between user is obtained according to it.As shown in figure 3, the trusting relationship network as between user, arrow represents two
Person is friend relation, for example, U1, which points to U2, represents that U1 and U2 is that direct good friend, i.e. U1 directly trust U2.
In the present embodiment, the data of the users to trust relational matrix got more than, by calculating between user
Trust be worth to trusting degree of the user to another user, trust value is (shows two between 0 two users in the absence of trusting relationship
Reachable path is not present in user in social networks), finally give 6 × 6 matrix T；Element T therein(ui,uj)Represent i-th
Trusting degree of the individual user to j-th of user；And i, j are no more than 6 positive integer.According between the user in above-mentioned steps S1
Synthesis trust value computing theory deduction formula and Fig. 2 in data, the present embodiment can obtain one group of related data.For example with
Family U1 and user U2 trusting relationship data：
Finally give TS(u1,u2)=0.5*0.667+0.25*1=0.5835, represents that user U1 is believed the comprehensive of user U2
It is 0.5835 to appoint degree.The trust value obtained between other users can be calculated by that analogy, so as to be formed comprising trusting degree
The trusting relationship matrix T of user.
Second step, builds the preference relation matrix of user.Commented as shown in figure 4, obtaining active of 6 users to 6 projects
Point (the user items scoring in the present invention is grade scoring, and the grade scoring is divided into 1,2,3,4,55 grades altogether；Scoring is got over
Height represents that user is higher to the preference of the project), the project scoring of not actively scoring is null value, forms 6 × 6 users-item
Mesh rating matrix R；Wherein element R(ui,im)Represent scoring of i-th of user to m-th of project；And i, m are no more than 6 just
Integer.
In the present embodiment, with the data message in user derived above-project rating matrix, similarity is passed through
Calculate and obtain the scoring similarity between user to form 6 × 6 matrix S；Wherein element S(ui,uj)Represent i-th user with
The scoring similarity degree of j-th of user, and i, j are no more than 6 positive integer.According to commenting between the user in above-mentioned steps S2
The data divided in the theory deduction formula and Fig. 4 of Similarity Measure, the present embodiment can obtain one group of related data.Such as user U1
With user U2 preference relation data：
Finally give S(u1,u2)=is(u1,u2)·URP(u1,u2)=0.25*0.421=0.105, represents user U1 and user
U2 scoring similarity be 0.105, i.e. preference it is similar be 0.105.The scoring obtained between other users can be calculated by that analogy
Similarity, so as to form the preference relation matrix S of the user comprising preference relation.
3rd step, builds the preference trusting relationship matrix of user.In the present embodiment, the trust obtained according to the first step
The preference relation that relation and second step are obtained, calculates the comprehensive similarity obtained between user, so as to obtain by comprehensive similarity
One 6 × 6 Matrix C S, i.e. user preference trusting relationship matrix；Wherein Elements C S(ui,uj)Represent i-th of user and j-th of use
The comprehensive similarity at family, and i, j are no more than 6 positive integer.According to the comprehensive similarity between the user in above-mentioned steps S3
The theory deduction formula of calculating and obtained preference relation and trusting relationship data, the present embodiment can obtain one group of dependency number
According to.Such as user U1 is social similar to user U2's：
Comprehensive similarity between user containing weight coefficient can be obtained with this present embodiment：
CS(u1,u2)=a*0.75+b*0.105+c*0.333
With the weight coefficient sequence { a, b, c } in comprehensive similarity for independent variable, the absolute average mistake of integrated forecasting scoring
Difference and F mediation rates obtain the object function S of a multiple-objection optimization, and excellent to the progress of these targets using bee colony optimized algorithm
Change, when object function S values are minimum so as to find optimal weight coefficient set { a, b, c }, it is assumed that weight coefficient now is most
The figure of merit is { 0.4,0.4,0.2 }.Then
CS(u1,u2)=0.4 × 0.75+0.4 × 0.105+0.2 × 0.333=0.4086
The comprehensive similarity obtained between other users can be calculated by that analogy, so as to obtain the preference trusting relationship of user
Matrix.
4th step, finds the most trusted neighbor collection of targeted customer and obtains the prediction scoring of project.Obtained according to the 3rd step
User preference trusting relationship matrix data, descending arrange targeted customer comprehensive similarity, selection determine targeted customer
Trusted neighbor collection (taken in present embodiment K be 10)；The pre- of the trusted neighbor set pair project is calculated by score in predicting formula
Test and appraisal point, i.e. prediction score value of the targeted customer to the project.In the present embodiment, it is assumed that U2 is credible for U1 only one
Neighbours, then user U1 be to project I3 prediction score value：
5th step, will predict the higher project recommendation of scoring to colod-application family.The targeted customer that the 4th step of foundation is obtained is to item
Purpose predicts score information, and descending arranges the prediction scoring of project to be recommended, selection prediction scoring item to be recommended in the top
Mesh is recommended targeted customer.
This recommends method based on user preference and the cold start-up trusted, and can not only excavate the credible good of more colod-application families
Friend solves the Sparse Problem at colod-application family；And the method for multiple-objection optimization improves the essence that the comprehensive similarity between user is calculated
Exactness, the preference trusting relationship between user is shown from diversiform data angle；Improve the personalized recommendation quality to colod-application family.
The above embodiment is interpreted as being merely to illustrate the present invention rather than limited the scope of the invention.
After the content for the record for having read the present invention, technical staff can make various changes or modifications to the present invention, these equivalent changes
Change and modification equally falls into the scope of the claims in the present invention.
Claims (8)
1. a kind of recommend method based on user preference and the cold start-up trusted, it is characterised in that comprises the following steps：S1, foundation
Synthesis trust value between user social contact information, measure user, builds trusting relationship matrix；S2, foundation user's score data, are calculated
The preference similarity of user, builds preference relation matrix；S3, the preference pass using comprehensive similarity computational methods fusion steps S2
System and step S1 trusting relationship, and the weights in comprehensive similarity are updated using ant colony algorithm iteration, multiple-objection optimization is carried out,
Weights is reached adaptive optimal, build preference trusting relationship matrix；S4, based on preference trusting relationship matrix, choose cold start-up
The most trusted neighbor collection of user predicts that respective item is scored for it；S5, the high project recommendation of scoring will be predicted to cold start-up
User.
2. according to claim 1 recommend method based on user preference and the cold start-up trusted, it is characterised in that the step
Rapid S1 builds trusting relationship matrix and comprised the following steps：
It is worth to by calculating the trust between user between trusting degree of the user to another user, two users that trust value is 0 not
There is trusting relationship, form N × N Digraph adjacent matrix T, i.e. the trusting relationship matrix of user；Element T S therein(ui,uj)
Represent synthesis trusting degree of i-th of user to j-th of user；The N is positive integer, and i, j are the just whole of no more than N
Number.
3. according to claim 2 recommend method based on user preference and the cold start-up trusted, it is characterised in that the step
The method of trust value includes between calculating user in rapid S1：
(1) trust of user is divided into according to the characteristic of user social contact net and directly trusts Dt and indirectly trust IDt；
(2) according to user, the probability that is trusted of user is divided into and local can be trusted probability by relative position scope in social network
α and the overall situation can be trusted probability β；
(3) probability α can be trusted according to the asymmetric similar part for calculating user of trust；
(4) probability β can be trusted by popularity mechanism calculating user's overall situation being obtained in social network according to user；
(5) correlation theories knowledge in above rational formula and sociology obtains user and integrates degree of belief TS(u,v),
TS(u,v)=α(u,v)*Dt(u,v)+βv*IDt(u,v)。
4. according to claim 2 recommend method based on user preference and the cold start-up trusted, it is characterised in that the step
The preference relation matrix that rapid S2 builds user is specifically included：Obtain active of N number of user to M project to score, do not score actively
Project scoring be null value, formed a N × M users-project rating matrix R；Wherein element R(ui,m)Represent i-th of user couple
The scoring of m-th of project；Described N, M are positive integer, and i, m are no more than N and M positive integer successively；Based on this scoring
Information is obtained the scoring similarity between user by Similarity Measure and closed so as to form the preference of a N N matrix S, i.e. user
It is matrix；Wherein element S(ui,uj)The scoring similarity degree of i-th of user and j-th of user are represented, the N is positive integer, and
I, j are no more than N positive integer.
5. according to claim 4 recommend method based on user preference and the cold start-up trusted, it is characterised in that the step
The method of scoring similarity includes between calculating user in rapid S2：
(1) the common scoring item accountings of Jie Kaerde between user are defined as to the initial score similarity between user；
(2) using effort analysis between alleviating user based on the similar similar computational methods of user's scoring preference of user user
The influence that scoring similarity is brought.
6. according to claim 1 recommend method based on user preference and the cold start-up trusted, it is characterised in that the step
The method of comprehensive similarity includes between calculating user in rapid S3：
(1) using the social information and score information of comprehensive similarity computational methods fusion user；
(2) using the weight sequence in comprehensive similarity as independent variable, comprehensive absolute average error and F mediations rate are available more than one
The object function of objective optimization；
(3) multiple objective function is optimized using bee colony optimized algorithm so as to find optimal weight coefficient collection a, b,
C }, it is finally reached weight coefficient adaptive.
7. according to claim 6 recommend method based on user preference and the cold start-up trusted, it is characterised in that the step
Prediction score values of the targeted customer u to project m is calculated in rapid S4：
<mrow>
<msub>
<mi>P</mi>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>m</mi>
<mo>)</mo>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mrow>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>v</mi>
<mo>&Element;</mo>
<msub>
<mi>TN</mi>
<mi>u</mi>
</msub>
</mrow>
</munder>
<msub>
<mi>CS</mi>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
</msub>
<mo>*</mo>
<msub>
<mi>R</mi>
<mrow>
<mo>(</mo>
<mi>v</mi>
<mo>,</mo>
<mi>m</mi>
<mo>)</mo>
</mrow>
</msub>
</mrow>
<mrow>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mi>v</mi>
<mo>&Element;</mo>
<msub>
<mi>TN</mi>
<mi>u</mi>
</msub>
</mrow>
</munder>
<mrow>
<mo>|</mo>
<mrow>
<msub>
<mi>CS</mi>
<mrow>
<mo>(</mo>
<mi>u</mi>
<mo>,</mo>
<mi>v</mi>
<mo>)</mo>
</mrow>
</msub>
</mrow>
<mo>|</mo>
</mrow>
</mrow>
</mfrac>
</mrow>
Wherein, CS(u,v)Represent the synthesis similarity measurements between user u and user v；R(v,m)It is scorings of the user v to project m, TNu
It is the nearest trusted neighbor collection of targeted customer.
8. according to claim 7 recommend method based on user preference and the cold start-up trusted, it is characterised in that the step
For project set I to be recommended in rapid S5recIt is shared | Irec| individual prediction scoring；Descending arranges the pre- test and appraisal of project to be recommended
Point, top (1≤top≤M) the individual project for taking prediction scoring forward is recommended to targeted customer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710384564.3A CN107301583B (en) | 2017-05-26 | 2017-05-26 | Cold start recommendation method based on user preference and trust |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710384564.3A CN107301583B (en) | 2017-05-26 | 2017-05-26 | Cold start recommendation method based on user preference and trust |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107301583A true CN107301583A (en) | 2017-10-27 |
| CN107301583B CN107301583B (en) | 2020-11-17 |
Family
ID=60137410
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710384564.3A Active CN107301583B (en) | 2017-05-26 | 2017-05-26 | Cold start recommendation method based on user preference and trust |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107301583B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108198045A (en) * | 2018-01-30 | 2018-06-22 | 东华大学 | The design method of mixing commending system based on e-commerce website data mining |
| CN109101667A (en) * | 2018-09-29 | 2018-12-28 | 新乡学院 | A kind of personalized recommendation method based on explicit trust and implicit trust |
| CN109697628A (en) * | 2018-11-12 | 2019-04-30 | 平安科技（深圳）有限公司 | Product data method for pushing and device, storage medium, computer equipment |
| CN109948067A (en) * | 2019-02-22 | 2019-06-28 | 哈尔滨工业大学（深圳） | The information-pushing method and system of user's enigmatic language justice LR model are trusted in a kind of fusion |
| CN111241418A (en) * | 2020-01-07 | 2020-06-05 | 北京邮电大学 | Information recommendation method and device based on local weighted centrality trust inference |
| CN111796992A (en) * | 2019-04-09 | 2020-10-20 | Oppo广东移动通信有限公司 | Behavior preference determination method and apparatus, storage medium and electronic device |
| CN112347368A (en) * | 2019-08-07 | 2021-02-09 | 青岛海大新星软件咨询有限公司 | Recommendation method and device |
| CN112883289A (en) * | 2021-04-16 | 2021-06-01 | 河北工程大学 | PMF recommendation method based on social trust and tag semantic similarity |
| CN113344671A (en) * | 2021-06-23 | 2021-09-03 | 昆明理工大学 | Trust factor fused personalized recommendation model and construction method |
| CN109697628B (en) * | 2018-11-12 | 2024-05-10 | 平安科技（深圳）有限公司 | Product data pushing method and device, storage medium and computer equipment |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103995823A (en) * | 2014-03-25 | 2014-08-20 | 南京邮电大学 | Information recommending method based on social network |
| US8903822B2 (en) * | 2010-04-13 | 2014-12-02 | Konkuk University Industrial Cooperation Corp. | Apparatus and method for measuring contents similarity based on feedback information of ranked user and computer readable recording medium storing program thereof |
-
2017
- 2017-05-26 CN CN201710384564.3A patent/CN107301583B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8903822B2 (en) * | 2010-04-13 | 2014-12-02 | Konkuk University Industrial Cooperation Corp. | Apparatus and method for measuring contents similarity based on feedback information of ranked user and computer readable recording medium storing program thereof |
| CN103995823A (en) * | 2014-03-25 | 2014-08-20 | 南京邮电大学 | Information recommending method based on social network |
Non-Patent Citations (2)
| Title |
|---|
| 卢竹兵: "基于信任关系的协同过滤推荐策略研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 范永全 等: "一种融合用户偏好与信任度的增强协同过滤推荐方法", 《西华大学学报( 自然科学版)》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108198045A (en) * | 2018-01-30 | 2018-06-22 | 东华大学 | The design method of mixing commending system based on e-commerce website data mining |
| CN109101667B (en) * | 2018-09-29 | 2021-07-09 | 新乡学院 | Personalized recommendation method based on explicit trust and implicit trust |
| CN109101667A (en) * | 2018-09-29 | 2018-12-28 | 新乡学院 | A kind of personalized recommendation method based on explicit trust and implicit trust |
| CN109697628A (en) * | 2018-11-12 | 2019-04-30 | 平安科技（深圳）有限公司 | Product data method for pushing and device, storage medium, computer equipment |
| CN109697628B (en) * | 2018-11-12 | 2024-05-10 | 平安科技（深圳）有限公司 | Product data pushing method and device, storage medium and computer equipment |
| CN109948067A (en) * | 2019-02-22 | 2019-06-28 | 哈尔滨工业大学（深圳） | The information-pushing method and system of user's enigmatic language justice LR model are trusted in a kind of fusion |
| CN111796992A (en) * | 2019-04-09 | 2020-10-20 | Oppo广东移动通信有限公司 | Behavior preference determination method and apparatus, storage medium and electronic device |
| CN111796992B (en) * | 2019-04-09 | 2023-03-21 | Oppo广东移动通信有限公司 | Behavior preference determination method and apparatus, storage medium and electronic device |
| CN112347368A (en) * | 2019-08-07 | 2021-02-09 | 青岛海大新星软件咨询有限公司 | Recommendation method and device |
| CN111241418B (en) * | 2020-01-07 | 2023-04-18 | 北京邮电大学 | Information recommendation method and device based on local weighted centrality trust inference |
| CN111241418A (en) * | 2020-01-07 | 2020-06-05 | 北京邮电大学 | Information recommendation method and device based on local weighted centrality trust inference |
| CN112883289B (en) * | 2021-04-16 | 2022-05-06 | 河北工程大学 | PMF recommendation method based on social trust and tag semantic similarity |
| CN112883289A (en) * | 2021-04-16 | 2021-06-01 | 河北工程大学 | PMF recommendation method based on social trust and tag semantic similarity |
| CN113344671A (en) * | 2021-06-23 | 2021-09-03 | 昆明理工大学 | Trust factor fused personalized recommendation model and construction method |
| CN113344671B (en) * | 2021-06-23 | 2023-04-07 | 昆明理工大学 | Trust factor fused personalized recommendation model and construction method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107301583B (en) | 2020-11-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107301583A (en) | It is a kind of that method is recommended based on user preference and the cold start-up trusted | |
| CN109508428B (en) | Interest point recommendation method based on interest point true popularity and implicit trust mining | |
| CN109271582B (en) | Personalized information recommendation method based on path with attribute element | |
| CN109408734A (en) | A kind of collaborative filtering recommending method of fuse information Entropy conformability degree and dynamic trust | |
| CN106503028B (en) | Recommendation method and system | |
| CN106528643B (en) | Multi-dimensional comprehensive recommendation method based on social network | |
| CN106682114A (en) | Personalized recommending method fused with user trust relationships and comment information | |
| CN107220365A (en) | Accurate commending system and method based on collaborative filtering and correlation rule parallel processing | |
| CN104935963A (en) | Video recommendation method based on timing sequence data mining | |
| Jiao et al. | A novel learning rate function and its application on the SVD++ recommendation algorithm | |
| Kommineni et al. | Machine learning based efficient recommendation system for book selection using user based collaborative filtering algorithm | |
| Eliyas et al. | Recommendation systems: Content-based filtering vs collaborative filtering | |
| CN105354260B (en) | The Mobile solution of a kind of mosaic society's network and item characteristic recommends method | |
| CN104166732B (en) | Project collaboration filtering recommendation method based on global scoring information | |
| CN109918562A (en) | A kind of recommended method based on communities of users and scoring joint community | |
| CN106991173A (en) | Collaborative filtering recommending method based on user preference | |
| CN106384259A (en) | Recommend system solution method for fusing social information | |
| CN106651461A (en) | Film personalized recommendation method based on gray theory | |
| CN109857928A (en) | User preference prediction technique based on polynary credit evaluation | |
| CN113065918A (en) | Sparse trust recommendation method based on semi-supervised learning | |
| CN110825978B (en) | Multitask collaborative filtering method based on neighbor user feature sharing | |
| Zhou | Shrinkage estimation of log-odds ratios for comparing mobility tables | |
| Zhang et al. | Multi-view dynamic heterogeneous information network embedding | |
| Cui et al. | DMFA-SR: Deeper membership and friendship awareness for social recommendation | |
| CN102073709A (en) | Degree-reduced nearest neighbor (DRNN)-based similarity computing method |
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 |