CN109033242A - A kind of collaborative filtering recommending method of combination scoring and project dependency - Google Patents
A kind of collaborative filtering recommending method of combination scoring and project dependency Download PDFInfo
- Publication number
- CN109033242A CN109033242A CN201810728955.7A CN201810728955A CN109033242A CN 109033242 A CN109033242 A CN 109033242A CN 201810728955 A CN201810728955 A CN 201810728955A CN 109033242 A CN109033242 A CN 109033242A
- Authority
- CN
- China
- Prior art keywords
- project
- scoring
- dependency
- similarity
- sim
- 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
- 238000001914 filtration Methods 0.000 title claims abstract description 28
- 238000000034 method Methods 0.000 title claims abstract description 20
- 238000004364 calculation method Methods 0.000 claims description 8
- 239000011159 matrix material Substances 0.000 claims description 5
- 238000013480 data collection Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses the collaborative filtering recommending methods of a kind of combination scoring and project dependency, comprising steps of the correlation between calculating project；In conjunction with the similarity between scoring and project dependency calculating project；Neighbours' Item Sets are found according to the similarity between project；Prediction model prediction scoring is established, personalized recommendation is provided.Data sparsity problem can be effectively relieved in the present invention, improve the forecasting accuracy of proposed algorithm.
Description
Technical field
The invention belongs to recommending learning art field, in particular to a kind of collaboration of combination scoring and project dependency
Filter recommended method.
Background technique
In recommender system, proposed algorithm is the pith that support it and operate.Currently, main recommended technology has base
Recommendation in content, the recommendation based on collaborative filtering, Knowledge based engineering recommendation and mixed recommendation etc..Wherein, collaborative filtering
The characteristics of proposed algorithm is simply easily achieved due to it, is widely used.Collaborative filtering can be divided into the association based on neighborhood
With filter algorithm and based on the collaborative filtering of model.Collaborative Filtering Recommendation Algorithm based on neighborhood includes the association based on user
With filtering recommendation algorithms and project-based collaborative filtering.Score in predicting problem, which is that one kind is important in recommender system, asks
Topic, researcher have done a large amount of related work for the collaborative filtering of the score in predicting based on user and project, obtain very
Big progress.In the collaborative filtering based on neighbour, the similarity calculation between user between project is committed step, is used
More accurate similarity calculating method can find significantly more efficient neighbours' collection, and then carry out more accurate recommendation.
Collaborative filtering based on score in predicting is a kind of important application problem, uses the collaborative filtering based on neighbour
It carries out prediction scoring and is broadly divided into three steps: 1. data collections；2. finding neighbour's collection；3. prediction scoring.It can be seen that recommending
Cheng Zhong, the selection of method for measuring similarity are the cores of entire Collaborative Filtering Recommendation Algorithm, and traditional collaborative filtering is based on user-
Project rating matrix calculates the similitude between user or between project, but the score data of user be it is very limited, lead
It causes traditional collaborative filtering to be faced with sparsity problem, and then influences to recommend quality.At the same time, in system there is also
The label information of many description users and item attribute feature is not used by, these information are dissolved into conventional recommendation algorithm,
Information deficiency problem in similarity calculation can be made up by excavating the correlation between project using the attribute tags of project, be
Alleviate Deta sparseness and is cold-started the effective ways of problem.
Summary of the invention
In order to solve the technical issues of above-mentioned background technique proposes, the present invention is intended to provide a kind of combination scoring and project phase
The collaborative filtering recommending method of closing property alleviates data sparsity problem, improves the accuracy of recommendation.
In order to achieve the above technical purposes, the technical solution of the present invention is as follows:
A kind of collaborative filtering recommending method of combination scoring and project dependency, comprising the following steps:
(1) correlation between calculating project；
(2) similarity between scoring and project dependency calculating project is combined；
(3) neighbours' Item Sets are found according to the similarity between project；
(4) prediction model prediction scoring is established, personalized recommendation is provided.
Further, in step (1), the correlation between project includes Attribute Correlation and interest correlation；The category
Property correlation is defined as follows:
Wherein, sim_attr (i, j) indicates the Attribute Correlation between project i and project j, AttriAnd AttrjIt respectively indicates
The attribute set that project i and project j possess；
The interest correlation is defined as follows:
Wherein, sim_interest (i, j) indicates the interest correlation between project i and project j, wherein UiExpression was commented on
The user of project i gathers, UjIndicate that the user for commenting on project j gathers.
Further, in step (2), similarity based on project dependency is calculated separately first and based on scoring
Then similarity is calculated in conjunction with the similarity between scoring and project dependency calculating project.
Further, in step (2), the similarity calculation formula based on project dependency is as follows:
Sim_attr_interest (i, j)=sim_attr (i, j) * sim_interest (i, j)
Wherein, sim_attr_interest (i, j) indicates that the project i based on project dependency is similar between project j
Degree.
Further, in step (2), the similarity calculation formula based on scoring is as follows:
Wherein, sim_rating (i, j) indicates the similarity between the project i based on scoring and project j, Rui、RujTable respectively
Show scoring of the user u to the scoring and user u of project i to project j,It respectively indicates project i and being averaged for project j is commented
Point, UijIndicate user's collection of common comment project i and project j.
Further, as follows in conjunction with the similarity calculation formula between scoring and the project of project dependency in step (2):
Unified_sim (i, j)=r*sim_rating (i, j)+(1-r) * sim_attr_interest (i, j)
Wherein, unified_sim (i, j) indicates to combine similar between scoring and the project i and project j of project dependency
Degree, r is scale parameter.
Further, in step (3), the similarity matrix the project obtained between step (2) is ranked up, and is obtained
K neighbours project sets most like with campaign item of top, are denoted as N (i).
Further, it in step (4), predicts to score using following prediction model:
Wherein, PuiFor the scoring of prediction.
By adopting the above technical scheme bring the utility model has the advantages that
Present invention combination score information and project dependency calculate similarity, alleviate data sparsity problem, effectively mention
The accuracy that height is recommended.Method proposed by the present invention can for based on the calculated not accurate similarity of rating matrix into
Row improvement, and then improve the recommendation accuracy of recommender system.
Detailed description of the invention
Fig. 1 is overall flow figure of the invention.
Specific embodiment
Below with reference to attached drawing, technical solution of the present invention is described in detail.
As shown in Figure 1, the collaborative filtering recommending method of a kind of combination scoring and project dependency proposed by the present invention, including
Following steps:
Step 1, the correlation between calculating project；
Step 2, in conjunction with the similarity between scoring and project dependency calculating project；
Step 3, neighbours' Item Sets are found according to the similarity between project；
Step 4, prediction model prediction scoring is established, personalized recommendation is provided.
Project dependency is defined first, project dependency is divided into two aspects, first is that Attribute Correlation, but interest is related
Property.
Attribute Correlation defined formula is between project:
Wherein, sim_attr (i, j) indicates the Attribute Correlation between project i and project j, AttriAnd AttrjIt respectively indicates
The attribute set that project i and project j possess.
Interest correlation defined formula is between project:
Wherein, sim_interest (i, j) indicates the interest correlation between project i and project j, wherein UiExpression was commented on
The user of project i gathers, UjIndicate that the user for commenting on project j gathers.
Calculating formula of similarity based on project dependency is:
Sim_attr_interest (i, j)=sim_attr (i, j) * sim_interest (i, j)
Scoring similitude between calculating project i and project j is as follows:
Wherein, Rui、RujScoring of the user u to the scoring and user u of project i to project j is respectively indicated,Respectively
The average score of expression project i and project j, UijIndicate user's collection of common comment project i and project j.
Calculating formula of similarity in conjunction with scoring and project dependency is as follows:
Unified_sim (i, j)=r*sim_rating (i, j)+(1-r) * sim_attr_interest (i, j)
Wherein, the scoring similarity between sim_rating (i, j) expression project, sim_attr_interest (i, j) are indicated
The ratio of item similarity based on project dependency, the two is controlled by parameter r.
After obtaining the similarity matrix between project, we are the available top k neighbours most like with campaign item
Project set is occupied, is denoted as N (i) respectively.Neighbours' project set has been obtained, next campaign item can score pre-
It surveys.
Final prediction model is:
Embodiment is merely illustrative of the invention's technical idea, and this does not limit the scope of protection of the present invention, it is all according to
Technical idea proposed by the present invention, any changes made on the basis of the technical scheme are fallen within the scope of the present invention.
Claims (8)
1. a kind of collaborative filtering recommending method of combination scoring and project dependency, which comprises the following steps:
(1) correlation between calculating project；
(2) similarity between scoring and project dependency calculating project is combined；
(3) neighbours' Item Sets are found according to the similarity between project；
(4) prediction model prediction scoring is established, personalized recommendation is provided.
2. combining the collaborative filtering recommending method of scoring and project dependency according to claim 1, which is characterized in that in step
Suddenly in (1), the correlation between project includes Attribute Correlation and interest correlation；The Attribute Correlation is defined as follows:
Wherein, sim_attr (i, j) indicates the Attribute Correlation between project i and project j, AttriAnd AttrjRespectively indicate project i
The attribute set possessed with project j.
The interest correlation is defined as follows:
Wherein, sim_interest (i, j) indicates the interest correlation between project i and project j, wherein UiProject i was commented in expression
User set, UjIndicate that the user for commenting on project j gathers.
3. combining the collaborative filtering recommending method of scoring and project dependency according to claim 2, which is characterized in that in step
Suddenly in (2), the similarity based on project dependency and the similarity based on scoring are calculated separately first, then calculates combination
Similarity between scoring and project dependency calculating project.
4. combining the collaborative filtering recommending method of scoring and project dependency according to claim 3, which is characterized in that in step
Suddenly in (2), the similarity calculation formula based on project dependency is as follows:
Sim_attr_interest (i, j)=sim_attr (i, j) * sim_interest (i, j)
Wherein, sim_attr_interest (i, j) indicates the similarity between the project i based on project dependency and project j.
5. combining the collaborative filtering recommending method of scoring and project dependency according to claim 4, which is characterized in that in step
Suddenly in (2), the similarity calculation formula based on scoring is as follows:
Wherein, sim_rating (i, j) indicates the similarity between the project i based on scoring and project j, Rui、RujRespectively indicate use
Scoring of the family u to the scoring and user u of project i to project j,Respectively indicate the average score of project i and project j, Uij
Indicate user's collection of common comment project i and project j.
6. combining the collaborative filtering recommending method of scoring and project dependency according to claim 5, which is characterized in that in step
Suddenly as follows in conjunction with the similarity calculation formula between scoring and the project of project dependency in (2):
Unified_sim (i, j)=r*sim_rating (i, j)+(1-r) * sim_attr_interest (i, j)
Wherein, unified_sim (i, j) indicates to combine the similarity between scoring and the project i and project j of project dependency, and r is
Scale parameter.
7. combining the collaborative filtering recommending method of scoring and project dependency according to claim 6, which is characterized in that in step
Suddenly in (3), the similarity matrix the project obtained between step (2) is ranked up, and it is a most like with campaign item to obtain top k
Neighbours' project set, be denoted as N (i).
8. combining the collaborative filtering recommending method of scoring and project dependency according to claim 7, which is characterized in that in step
Suddenly it in (4), predicts to score using following prediction model:
Wherein, PuiFor the scoring of prediction.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810728955.7A CN109033242A (en) | 2018-07-05 | 2018-07-05 | A kind of collaborative filtering recommending method of combination scoring and project dependency |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810728955.7A CN109033242A (en) | 2018-07-05 | 2018-07-05 | A kind of collaborative filtering recommending method of combination scoring and project dependency |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109033242A true CN109033242A (en) | 2018-12-18 |
Family
ID=65521821
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810728955.7A Pending CN109033242A (en) | 2018-07-05 | 2018-07-05 | A kind of collaborative filtering recommending method of combination scoring and project dependency |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109033242A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110879855A (en) * | 2019-10-28 | 2020-03-13 | 世纪保众(北京)网络科技有限公司 | Insurance item comparison method, device, equipment and storage medium |
| CN114547279A (en) * | 2022-02-21 | 2022-05-27 | 电子科技大学 | Judicial recommendation method based on mixed filtering |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104317900A (en) * | 2014-10-24 | 2015-01-28 | 重庆邮电大学 | Multiattribute collaborative filtering recommendation method oriented to social network |
| CN104751353A (en) * | 2015-04-10 | 2015-07-01 | 中国石油大学(华东) | Cluster and Slope One prediction based collaborative filtering method |
| CN104794635A (en) * | 2015-04-17 | 2015-07-22 | 南京大学 | Cellphone model recommendation system based on collaborative filtering recommendation algorithm |
| CN107341204A (en) * | 2017-06-22 | 2017-11-10 | 电子科技大学 | A kind of collaborative filtering recommending method and system for merging article label information |
-
2018
- 2018-07-05 CN CN201810728955.7A patent/CN109033242A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104317900A (en) * | 2014-10-24 | 2015-01-28 | 重庆邮电大学 | Multiattribute collaborative filtering recommendation method oriented to social network |
| CN104751353A (en) * | 2015-04-10 | 2015-07-01 | 中国石油大学(华东) | Cluster and Slope One prediction based collaborative filtering method |
| CN104794635A (en) * | 2015-04-17 | 2015-07-22 | 南京大学 | Cellphone model recommendation system based on collaborative filtering recommendation algorithm |
| CN107341204A (en) * | 2017-06-22 | 2017-11-10 | 电子科技大学 | A kind of collaborative filtering recommending method and system for merging article label information |
Non-Patent Citations (2)
| Title |
|---|
| 张晓琳，付英姿，褚培肖: "杰卡德相似系数在推荐系统中的应用", 《计算机技术与发展》 * |
| 高娜，杨明: "一种改进的结合标签和评分的协同过滤推荐算法", 《南京师大学报( 自然科学版)》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110879855A (en) * | 2019-10-28 | 2020-03-13 | 世纪保众(北京)网络科技有限公司 | Insurance item comparison method, device, equipment and storage medium |
| CN114547279A (en) * | 2022-02-21 | 2022-05-27 | 电子科技大学 | Judicial recommendation method based on mixed filtering |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Cavusgil et al. | Middle-class consumers in emerging markets: Conceptualization, propositions, and implications for international marketers | |
| Dennis Wei | Beyond the Sunan model: trajectory and underlying factors of development in Kunshan, China | |
| CN103886067B (en) | Method for recommending books through label implied topic | |
| CN104462560B (en) | A kind of recommendation method of personalized recommendation system | |
| López-Morales et al. | Policies and technologies for a sustainable use of water in Mexico: A scenario analysis | |
| CN103177093B (en) | A kind of general recommendations method and system based on object tag | |
| CN104090936B (en) | News recommendation method based on hypergraph sequencing | |
| CN102841929A (en) | Recommending method integrating user and project rating and characteristic factors | |
| CN103886001A (en) | Personalized commodity recommendation system | |
| CN103150667B (en) | A kind of personalized recommendation method based on body construction | |
| CN103345699A (en) | Personalized food recommendation method based on commodity forest system | |
| CN107391670A (en) | A kind of mixing recommendation method for merging collaborative filtering and user property filtering | |
| CN109947987A (en) | A kind of intersection collaborative filtering recommending method | |
| Kashnitsky et al. | Unequally ageing regions of Europe: Exploring the role of urbanization | |
| Heikkila et al. | Seven prototypical Chinese cities | |
| Peilin et al. | Make efforts to develop an olive-shaped distribution pattern: an analysis based on data from the Chinese Social Survey for 2006-2013 | |
| CN109033242A (en) | A kind of collaborative filtering recommending method of combination scoring and project dependency | |
| CN107357845A (en) | A kind of tour interest commending system and recommendation method based on Spark | |
| CN102542409A (en) | Recruitment information display method based on socialized network system | |
| CN104978406A (en) | User behavior analysis method of Internet platform | |
| CN104731887B (en) | A kind of user method for measuring similarity in collaborative filtering | |
| CN109522487A (en) | A kind of dining room personalized recommendation method based on comment | |
| CN107194492A (en) | The optimization method that a kind of businessman based on position social networks is recommended | |
| CN104462597B (en) | A kind of positive negativity of synthetic user scores and the collaborative filtering method of scoring preference heterogeneity | |
| CN106204153A (en) | A kind of two-staged prediction Top N proposed algorithm based on attribute proportion similarity |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181218 |
|
| RJ01 | Rejection of invention patent application after publication |