CN108399197A - Collaborative filtering recommending method based on user's confidence level and time context - Google Patents
Collaborative filtering recommending method based on user's confidence level and time context Download PDFInfo
- Publication number
- CN108399197A CN108399197A CN201810087257.3A CN201810087257A CN108399197A CN 108399197 A CN108399197 A CN 108399197A CN 201810087257 A CN201810087257 A CN 201810087257A CN 108399197 A CN108399197 A CN 108399197A
- Authority
- CN
- China
- Prior art keywords
- user
- similarity
- confidence level
- project
- time context
- 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.)
- Pending
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
-
- 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
Abstract
A kind of collaborative filtering recommending method based on user's confidence level and time context is claimed in the present invention, is related to the calculating of similarity between user in collaborative filtering recommending method.For the insufficient problem and user interest dynamic change situation of user's score data traditional method for measuring similarity in the case of extremely sparse, a kind of collaborative filtering recommending method based on user's confidence level and time context is proposed.First, the rating matrix between structuring user's and project.Secondly, what is proposed through the invention calculates the similarity between user based on user's confidence level and time context similarity calculating method.Then, according between user sequencing of similarity as a result, the optimal neighbour user of selection collects, or one similarity of setting threshold value, select more than neighbours of the user as target user of threshold value.Finally, prediction of the target user to non-scoring item is obtained using similarity as weight with obtaining target user, form Top N lists and recommend user after arest neighbors set.
Description
Technical field
The invention belongs to the technical field of personalized recommendation method, it is based particularly on user's confidence level and time context
Collaborative filtering recommending method.
Background technology
Nowadays, the resource on network is in explosive growth.On the one hand, people obtain more and more abundant letter by network
Breath, brings great convenience to life；On the other hand, while the information space of magnanimity brings user's more Multiple strategies,
But also user gets lost in the ocean of information.Although traditional search engine can with alleviating user to a certain extent information
Search Requirement, however they are presented to the same ranking results of all users, can not be directed to the hobby of different users
Personalized service is actively provided.In this context, personalized recommendation comes into being.
Personalized essence is that different services is provided for different users.In real life, personalized recommendation system
Responsibility be more like the Shopping Guide in a market, Shopping Guide needs the attributive character of each product in known market, so as to
It is that user recommends the commodity that suitable user needs in time when user visits.The research of proposed algorithm is in personalized recommendation system
Core needs outstanding suitable proposed algorithm, could be that user recommends its most possible interested new projects.Collaborative filtering pushes away
It is the most popular and widely used algorithm in commending system to recommend algorithm
But presently, personalized recommendation system has that accuracy and reliability is low.Many personalized recommendation systems
System cannot fully be recommended according to the interests change situation of user, and the especially change in the interest of user's different periods makes
It must recommend to lose accuracy and specific aim.
Invention content
The present invention proposes a kind of collaborative filtering recommending side based on user's confidence level and time context according to problem above
Method.By analyzing the behavioral data of user, modified cosine similarity computational methods are based on, in conjunction with user's confidence level and on the time
Hereafter the factor in terms of the two calculates the similarity between user more fully hereinafter, and proposes on this basis a kind of based on use
The collaborative filtering recommending method of family confidence level and time context, is as follows：
S1：Obtain user behavior data, the rating matrix between structuring user's and project.
S2：Using the similarity calculating method based on user's confidence level and time context, in modified cosine similarity
On the basis of calculate user between Interest Similarity, and choose with target user's nearest neighbor set.
S3：According to the arest neighbors set with target user obtained in S2, target user is calculated to not generating behavior project
Interest level.
S4：According to the target user obtained in S3 to not generating the interest level of behavior project, target user is carried out
Project recommendation.
User behavior data is obtained in the S1, including obtains the score data of user items, the time of scoring and project
Information.
The similarity between user is calculated in the S2, on the basis of modified cosine similarity computational methods, addition is used
Family confidence level and time context calculate the similarity between user.
The modified cosine similarity computational methods are a kind of more popular similarity calculating methods, by subtracting use
Family improves the problem of different user scoring size is not accounted in cosine similarity to the average score of project.
User's confidence level refers to that in daily life, each field has " expert ", these " experts " are in correspondence
Field input time and efforts it is more than ordinary people, therefore to place field occur thing evaluation also tend to more reference
Value.From this angle, think deeply in turn, it is believed that compared to other people making times and the more people of energy in the neck
The evaluation in domain just more has reference value.
The time context refers to that in daily life, the interest of people can vary over, very short
In time interval, user similar with target user gives approximate evaluation to identical article, indicate in this way the user with
Target user's interest is even more like, evaluates with more reference value with illustrating the user.Also, user and target user's scoring object
The proportion for the number of articles that respectively scores shared by the intersection of product is bigger, also can be from reflecting that similarity is higher between the two to a certain degree.
The similarity between user is calculated in the S2, is changed on the basis of modified cosine similarity computational methods
Into adding user's confidence level and time context to calculate the similarity between user.It is described according to user's confidence level and time
Both modes of context are completed with two level calculation between user come the method for gradually improving calculating user's similarity
Similarity calculated.The two level calculation, including the first order are the confidence level similarity calculation of user, and the second level exists
Time context is added on the basis of the first order carries out similarity calculation.
It is as follows：
1. first order user's confidence calculations：
Wherein, similarities of the sim (u, v) ' between user u and user v.RuiAnd RviIt is user u and user v to project i
Score data.WithRepresent the average score of user u and user v.N (i) represents the average score of project i.number(v)
The quantity of the project of user's v evaluations is represented, ave represents the quantity of the average ratings of project.
2. second level time context calculates：
Wherein tuiAnd tviUser u and user v is indicated respectively to scoring time of project i, N (v) and N (v) be user u and
The quantity of the scoring item of user v.μ be reflect user interest degree variation speed influence English, can be by adjustment parameter at
Reason obtains optimal solution.If user u and user v is closer to the time point of the evaluation behavior of project, the phase between user
It is bigger like spending.
3. third level user confidence level and time context calculate：
Sim (u, v) " '=α sim (u, v) '+(1- α) sim (u, v) "
The formula is the similarity calculating method based on user's confidence level and time context, carries out interest phase between user
Like the calculating of degree.Wherein α is to influence English, can be by testing acquisition optimal solution repeatedly.
The final similarity obtained between user u and user v.
Advantages of the present invention and advantageous effect：
The present invention comes more on the basis of modified cosine similarity from user's confidence level and time context these two aspects
The similarity situation between user is comprehensively calculated, it is more careful that the interest characteristics of user are considered, therefore based on use
The Collaborative Filtering Recommendation Algorithm of family confidence level and time context has higher accuracy rate and reliability.Personalized recommendation system
Recommendation results have the characteristics that more to meet the demand of user, while user's participation is also lower, reduces user and finds information
Cost.
By analyzing the behavioral data of user, from the score data of user behavior data, user's confidence level and above and below the time
Text calculates the similarity between user, it is proposed that the computational methods of the similarity based on user's confidence level and time context,
Convenient made to user is more accurately recommended, and recommendation quality can be significantly improved.
Description of the drawings
Fig. 1 is the overview flow chart of the present invention；
Fig. 2 be the present invention user between similarity calculation step schematic diagram；
Fig. 3 be the present invention user between similarity calculating method figure.
Specific implementation mode
Below in conjunction with the attached drawing in inventive embodiments, technical solution in the embodiment of the present invention carries out clear, detailed
Ground describes.Obviously, described embodiment is only a part of the embodiment of the present invention, instead of all the embodiments.
As shown in Figure 1, the collaborative filtering recommending method based on user's confidence level and time context, includes the following steps：
S1：Obtain behavioral data collection of the user to project.Data set includes the information of project, scoring of the user to project
Data, the behavioral datas such as time of scoring can embody the behavioural habits of different user.
S2：Using the similarity calculating method based on user's confidence level and time context, in modified cosine similarity
On the basis of calculate user between Interest Similarity, and choose with target user's nearest neighbor set.
S3：According to the arest neighbors set of the target user obtained in S2, target user is calculated to not generating behavior project
Interest level, to determine interest level and range of the target user to project.
S4：N projects recommend user before choosing.User can be obtained to not generating the sense of the project of behavior according to S3
Level of interest is chosen top n project therein and is recommended.
The S1 mainly divides following 2 steps:
S11：Obtain user behavior data collection.
S12：Data set is analyzed, the rating matrix between structuring user's and project.It can be according to user behavior data
Feature specifically divides the evaluation time period of project user.
As shown in Fig. 2, carrying out the similarity calculation between user in the S2.On modified cosine similarity method basis
On, add user's confidence level and time context factors.
As shown in figure 3, being the step of similarity calculation between user in the S2：
S21：Input the behavioral data collection of user u and v；
S22：Assuming that user u and user v is R respectively to the score data of project iuiAnd Rvi。WithRepresent user u and
The average score of user v.N (i) represents the average score of project i.Number (v) represents the quantity of the project of user's v evaluations,
Ave represents the quantity of the average ratings of project.
Interest Similarity sim (u, v) ' between so user u and user v can be expressed as：
The project of behavior is generated between i expressions user u and user v jointly；1/log2(1+ | N (i) |) indicate to have punished use
Influence of the popular article to their similarities in family u and user's v common interest lists.1/(1+e-(number(v)-ave)) indicate the use
Whether family is " expert ", i.e., whether than ordinary people takes more time and efforts in project.It can from formula (1)
Go out, if the project that user u and user v generate behavior jointly is more, and user v takes more times in the project,
If the project relatively more unexpected winner, then the Interest Similarity between user u and user v is higher.
S23：On the basis of S22, the calculating of the time context of user is carried out.In shorter time interval, with mesh
Mark user similar user give approximate evaluation to identical article, indicates the user and target user's interest phase in this way
Seemingly, illustrate that the evaluation of the user more has reference value.In addition, the intersection with target user's scoring item accounts for respectively scoring object
The proportion of product quantity is bigger, can also reflect that similarity between the two is higher to a certain extent.
Wherein tuiAnd tviUser u and user v is indicated respectively to scoring time of project i, N (v) and N (v) be user u and
The quantity of the scoring item of user v.μ is influence English that interest-degree changes speed, this is related with the selection of experimental subjects, can be with
By testing acquisition repeatedly.Under normal circumstances, the interest of user can change with the time.If user u and user v are to project
The scoring time of i is closer, and score value gap is smaller, then the Interest Similarity of the two is bigger.
S24：On the basis of S22 and S23, the two is combined, obtains following formula：
Sim (u, v) " '=α sim (u, v) '+(1- α) sim (u, v) " (3)
The formula is the similarity calculating method based on user's confidence level and time context, carries out interest phase between user
Like the calculating of degree.Wherein α is to influence English, can be by testing acquisition optimal solution repeatedly.
S25：On the basis of S24, it can obtain gathering with the user that target user has same interest.According to formula
(3), the similarity between target user and other users can be obtained, is then chosen a with the immediate K of target user's u interest
User carries out top-N operations and acquires.
Then according to the S3, target user u is calculated to not generating the interest level P of behavior project iui, can be expressed as：
Wherein, sim (u, v) " indicates the similitude between the immediate v user of target user's u interest,WithTable
Show the average score of the average score and user v of user u.
In above-mentioned steps S4, pass through the P in S3uiCalculate interest level of the target user to other projects, finally to
User recommends PuiThe project of N before ranking.
The present invention proposes a kind of collaboration based on user's confidence level and time context from the behavior based on user
Filter recommendation method.This method is to be established a kind of similar based on user's confidence level and time context according to user behavior data
Computational methods are spent, is then obtained according to this method and is gathered with target user the most neighbour, then calculate user to not generating behavior item
Purpose interest level, N projects before finally recommending.
It should be understood that above-mentioned specific embodiment, can make those skilled in the art and reader that this hair be more fully understood
The implementation of bright creation, it should be understood that protection scope of the present invention is not limited to such special statement and implementation
Example.Therefore, although description of the invention has been carried out detailed description with reference to drawings and examples to the invention,
It will be understood by those of skill in the art that still can be modified or replaced equivalently to the invention, in short, all are not
It is detached from technical solution and its improvement of the spirit and scope of the invention, the protection in the invention patent should all be covered
In range.
Claims (5)
1. the collaborative filtering recommending method based on user's confidence level and time context, which is characterized in that include the following steps：
S1：Obtain user behavior data, the rating matrix between structuring user's and project；
S2：Using the similarity calculating method based on user's confidence level and time context, on modified cosine similarity basis
The upper Interest Similarity calculated between user, and choose and target user's nearest neighbor set；
S3：According to the arest neighbors set with target user obtained in S2, target user is calculated to not generating the sense of behavior project
Level of interest；
S4：According to the target user obtained in S3 to not generating the interest level of behavior project, project is carried out to target user
Recommend.
2. the collaborative filtering recommending method according to claim 1 based on user's confidence level and time context, feature
It is：User behavior data is obtained in the S1, including obtains the score information that user generates project, and user comments project
Between timesharing and the essential information of project.
3. the collaborative filtering recommending method according to claim 1 based on user's confidence level and time context, feature
It is：The similarity between user is calculated based on user's confidence level and the method for time context according to proposition in the S2,
From user's confidence level, the similarity degree between user is calculated in terms of the punishment of popular article and time context, is then selected
Take the nearest neighbor set with target user.
4. the collaborative filtering recommending method according to claim 3 based on user's confidence level and time context, feature
It is：User's confidence level refers to putting into more time and efforts, this kind of user than other people in some field user
Appraisal result it is more convincing；The time context refer in shorter time interval, it is similar with target user
User gives approximate evaluation to identical project.
5. the collaborative filtering recommending method according to claim 1 based on user's confidence level and time context, feature
It is：The similarity between user is calculated in the S2, is similar in modified cosine to the rating matrix of project according to user
On the basis of degree, confidence level and the time context of user is considered to calculate the similarity between user more fully hereinafter.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810087257.3A CN108399197A (en) | 2018-01-30 | 2018-01-30 | Collaborative filtering recommending method based on user's confidence level and time context |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810087257.3A CN108399197A (en) | 2018-01-30 | 2018-01-30 | Collaborative filtering recommending method based on user's confidence level and time context |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108399197A true CN108399197A (en) | 2018-08-14 |
Family
ID=63095212
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810087257.3A Pending CN108399197A (en) | 2018-01-30 | 2018-01-30 | Collaborative filtering recommending method based on user's confidence level and time context |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108399197A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109857935A (en) * | 2019-01-24 | 2019-06-07 | 腾讯科技（深圳）有限公司 | A kind of information recommendation method and device |
| CN111222054A (en) * | 2020-01-03 | 2020-06-02 | 中国计量大学 | Session social contact recommendation method based on context neighbor relation modeling |
| CN112487782A (en) * | 2020-12-11 | 2021-03-12 | 厦门市美亚柏科信息股份有限公司 | Article popularity calculation method based on article similarity quantity |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102479202A (en) * | 2010-11-26 | 2012-05-30 | 卓望数码技术（深圳）有限公司 | Recommendation system based on domain expert |
| CN103745100A (en) * | 2013-12-27 | 2014-04-23 | 浙江大学 | Item-based explicit and implicit feedback mixing collaborative filtering recommendation algorithm |
| CN105574216A (en) * | 2016-03-07 | 2016-05-11 | 达而观信息科技（上海）有限公司 | Personalized recommendation method and system based on probability model and user behavior analysis |
| CN105975496A (en) * | 2016-04-26 | 2016-09-28 | 清华大学 | Music recommendation method and device based on context sensing |
| CN106682121A (en) * | 2016-12-09 | 2017-05-17 | 广东工业大学 | Time utility recommendation method based on interest change of user |
| CN107507073A (en) * | 2017-09-14 | 2017-12-22 | 中国人民解放军信息工程大学 | Based on the service recommendation method for trusting extension and the sequence study of list level |
-
2018
- 2018-01-30 CN CN201810087257.3A patent/CN108399197A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102479202A (en) * | 2010-11-26 | 2012-05-30 | 卓望数码技术（深圳）有限公司 | Recommendation system based on domain expert |
| CN103745100A (en) * | 2013-12-27 | 2014-04-23 | 浙江大学 | Item-based explicit and implicit feedback mixing collaborative filtering recommendation algorithm |
| CN105574216A (en) * | 2016-03-07 | 2016-05-11 | 达而观信息科技（上海）有限公司 | Personalized recommendation method and system based on probability model and user behavior analysis |
| CN105975496A (en) * | 2016-04-26 | 2016-09-28 | 清华大学 | Music recommendation method and device based on context sensing |
| CN106682121A (en) * | 2016-12-09 | 2017-05-17 | 广东工业大学 | Time utility recommendation method based on interest change of user |
| CN107507073A (en) * | 2017-09-14 | 2017-12-22 | 中国人民解放军信息工程大学 | Based on the service recommendation method for trusting extension and the sequence study of list level |
Non-Patent Citations (1)
| Title |
|---|
| XINHUA WANG: "Social recommendation algorithm based on the context of time and tags", 《2015 THIRD INTERNATIONAL CONFERENCE ON ADVANCED CLOUD AND BIGDATA》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109857935A (en) * | 2019-01-24 | 2019-06-07 | 腾讯科技（深圳）有限公司 | A kind of information recommendation method and device |
| CN109857935B (en) * | 2019-01-24 | 2021-08-24 | 腾讯科技（深圳）有限公司 | Information recommendation method and device |
| CN111222054A (en) * | 2020-01-03 | 2020-06-02 | 中国计量大学 | Session social contact recommendation method based on context neighbor relation modeling |
| CN111222054B (en) * | 2020-01-03 | 2020-12-11 | 中国计量大学 | Session social contact recommendation method based on context neighbor relation modeling |
| CN112487782A (en) * | 2020-12-11 | 2021-03-12 | 厦门市美亚柏科信息股份有限公司 | Article popularity calculation method based on article similarity quantity |
| CN112487782B (en) * | 2020-12-11 | 2024-04-09 | 厦门市美亚柏科信息股份有限公司 | Article popularity calculation method based on similar quantity of articles |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103377250B (en) | Top k based on neighborhood recommend method | |
| CN105574216A (en) | Personalized recommendation method and system based on probability model and user behavior analysis | |
| CN108665323B (en) | Integration method for financial product recommendation system | |
| CN109064285A (en) | A kind of acquisition commercial product recommending sequence and Method of Commodity Recommendation | |
| WO2019056661A1 (en) | Search term pushing method and device, and terminal | |
| CN102902744A (en) | Book recommendation method | |
| CN107526850A (en) | Social networks friend recommendation method based on multiple personality feature mixed architecture | |
| CN109902235A (en) | User preference based on bat optimization clusters Collaborative Filtering Recommendation Algorithm | |
| CN108334592A (en) | A kind of personalized recommendation method being combined with collaborative filtering based on content | |
| CN108399197A (en) | Collaborative filtering recommending method based on user's confidence level and time context | |
| CN103678323A (en) | Friend recommendation method and system in SNS network | |
| CN107402961B (en) | Recommendation method and device and electronic equipment | |
| CN105468628B (en) | A kind of sort method and device | |
| CN103870972A (en) | Data recommendation method and data recommendation system | |
| CN108280124A (en) | Product classification method and device, ranking list generation method and device, electronic equipment | |
| CN108563690A (en) | A kind of collaborative filtering recommending method based on object-oriented cluster | |
| CN108733834A (en) | The user oriented recommendation method, apparatus of one kind and storage medium | |
| CN110910207A (en) | Method and system for improving commodity recommendation diversity | |
| CN108241619A (en) | A kind of recommendation method based on the more interest of user | |
| CN109299426A (en) | A kind of recommended method and device of accurate top information | |
| CN111915409B (en) | Item recommending method, device, equipment and storage medium based on item | |
| JP5586530B2 (en) | Information presentation device | |
| CN110020152B (en) | Application recommendation method and device | |
| CN109446402B (en) | Searching method and device | |
| CN108415928A (en) | A kind of book recommendation method and system based on weighted blend k- nearest neighbor algorithms |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180814 |
|
| WD01 | Invention patent application deemed withdrawn after publication |