CN108549729B - Personalized user collaborative filtering recommendation method based on coverage reduction - Google Patents
Personalized user collaborative filtering recommendation method based on coverage reduction Download PDFInfo
- Publication number
- CN108549729B CN108549729B CN201810486715.0A CN201810486715A CN108549729B CN 108549729 B CN108549729 B CN 108549729B CN 201810486715 A CN201810486715 A CN 201810486715A CN 108549729 B CN108549729 B CN 108549729B
- Authority
- CN
- China
- Prior art keywords
- user
- users
- target
- item
- target user
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
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
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a personalized user collaborative filtering recommendation method based on coverage reduction, which clearly defines the concept of a redundant user of a target user, removes the redundant user of the target user according to the function of removing redundant elements by coverage rough concentrated coverage reduction, thereby ensuring the quality of the adjacent users of the target user, and realizing the high-precision and diversified personalized recommendation for the target user by utilizing the grading information of the adjacent users with high quality.
Description
Technical Field
The invention relates to the field of recommendation systems, in particular to a coverage reduction-based personalized user collaborative filtering recommendation method capable of improving recommendation accuracy and having diversity.
Background
The recommendation system can intelligently sense the interest or the demand of the user through the personal information of the user, realize high-quality recommendation of the information and effectively solve the problem of information overload. The user collaborative filtering algorithm is one of the most widely applied and successful technologies in the field of recommendation systems, and the user collaborative filtering algorithm assumes that if the user has similar hobbies in the past, the user may also have similar hobbies in the future, and has the advantages of simple calculation, high efficiency and precision and the like. However, in the existing user collaborative filtering algorithm, the neighboring users of the target user tend to have the same hobbies, so the items with high prediction scores obtained by the neighboring users tend to be concentrated in a small number of types of items, even only popular items, and thus the diversity of recommendations is not satisfactory.
Disclosure of Invention
The invention aims to solve the technical problems in the prior art and provides a personalized user collaborative filtering recommendation method based on coverage reduction, which can improve recommendation accuracy and has diversity.
The technical solution of the invention is as follows: a personalized user collaborative filtering recommendation method based on coverage reduction is characterized by comprising the following steps in sequence:
step 1, forming a two-dimensional scoring information table by statistics:
forming a two-dimensional scoring information table according to the scoring information of the user to the articleRM={U,I,R∪{*}}(ii) a The two-dimensional scoring information tableRMIn (1),Ua set of users is represented as a set of users,Ia collection of items is represented that is,R∪{*}representing a user's set of scores for an item, wherein*Indicating that the user has not scored the item;
order useru∈UTo the articlei∈IIs scored asr u,i ∈R∪{*}And the useruIs given as an average score of;θA threshold value for scoring the user, ifr u,i ≥θIndicate the useruFavorite articlei(ii) a User' suThe scored set of items isI u ={i∈I| r u,i ≠*};For the useruAn unscored collection of items; the article attribute matrix isAM(ii) a In the user set U, if the item set favored by the user a is contained in the item set favored by the user b, the user a is called a redundant user of the target user;
and 2, reducing the redundant users by using a coverage reduction algorithm:
step 2.1 order item CollectionIAs universes of discourseIIn the universe of discourseIIn the method, each user's favorite articles form a set; in-item attribute matrixAMExtracting the favorite attributes of the target user:
in the formula (1), the reaction mixture is,mthe number of the attributes is represented and,at m a representation of one of the attributes is presented,av m representing attributesat m A value of (d);
step 2.2. Using the obtained targetUser's favorite attributes, building a decision set of target usersDSet of decisionsDThe method comprises the following steps of (1) consisting of a collection of articles with favorite attributes:
in the formula (2), the reaction mixture is,at m (i)= av m representing an articleiIn attributeat m A value ofav m ;
Step 2.3. domain of discourseIReduction of item collections into decision sets for target usersDI.e. discourse domainD(ii) a For each useru∈UBuilding usersuIn the universe of discourseDFavorite articles set onC u :
Order toC*=D-∪C u ,C={C 1 ,C 2 …C n , C*} Form the target user's domain of discourseDA cover ofC;
Step 2.4 secondary overlay of redundant elements using overlay reduction algorithmCThe intermediate reduction is to obtain the coverage after the reduction is finishedreduct(C) And reduced usersU r :
Step 3, utilizing the reduced userU r Building target usersauCandidate neighboring users ofu;
Step 4, calculating the similarity between the target user and the candidate adjacent users, and selecting the adjacent users of the target user:
calculating target user by using Pearson similarity measurement function (5)auAnd candidate neighboring usersu∈U r The degree of similarity between the two images,
in the formula (5), the reaction mixture is,sim(au,u)representing target usersauAnd candidate neighboring usersu∈U r The degree of similarity between the two images,I au ={i ∈I|r au,i ≠*}representing target usersauThe set of items that have been evaluated,an average score value representing a target user;
then selecting the front part with high similarityKNeighbor users with candidate neighbor users as target usersN au (k);
Step 5, performing predictive scoring on the unscored goods of the target user:
proximity user based on target userN au (k)By using the adjusted weighted sum function (6) to the target userauUnscored collection of itemsCarrying out prediction scoring to obtain a prediction scoring table of the target user;
in the formula (6), the reaction mixture is,P au,i representing target usersauTo the articleiThe prediction score of (a) is determined,U i ={u∈U|r u,i ≠*}indicating the evaluated articleiA set of users of (1);λas a regularization factor:
step 6, selecting the front part with high prediction scoreNThe individual item is used as a recommendation.
The invention clearly defines the concept of the redundant user of the target user, removes the redundant user of the target user according to the function of removing redundant elements by covering rough concentrated coverage reduction, thereby ensuring the quality of the adjacent user of the target user, and realizing the high-precision and diversified personalized recommendation provided for the target user by utilizing the grading information of the adjacent user with high quality.
Drawings
FIG. 1 is a schematic flow chart of an embodiment of the present invention;
fig. 2 is a graph showing the results of the accuracy measures (MAE and RMSE) of the embodiment and the comparative example according to the present invention as the number of neighboring users of the target user varies.
Fig. 3 is a diagram illustrating the results of diversity metrics (coverages) of embodiments and comparative examples according to the present invention as the number of neighboring users of the target user changes.
Detailed Description
The invention discloses a personalized user collaborative filtering recommendation method based on coverage reduction, which is sequentially carried out according to the following steps as shown in figure 1:
step 1, forming a two-dimensional scoring information table by statistics:
forming a two-dimensional scoring information table according to the scoring information of the user to the articleRM={U,I,R∪{*}}(ii) a The two-dimensional scoring information tableRMIn (1),Ua set of users is represented as a set of users,Ia collection of items is represented that is,R∪{*}representing a user's set of scores for an item, wherein*Indicating that the user has not scored the item;
order useru∈UTo the articlei∈IIs scored asr u,i ∈R∪{*}And the useruIs given as an average score of;θA threshold value for scoring the user, ifr u,i ≥θIndicate the useruLikes and dislikesArticle (A)i(ii) a User' suThe scored set of items isI u ={i∈I| r u,i ≠*};For the useruAn unscored collection of items; the article attribute matrix isAM(ii) a In the user set U, if the item set favored by the user a is contained in the item set favored by the user b, the user a is called a redundant user of the target user;
such as: user collectionU = { user 1, user 2, user 3, target user }, item setI = { item 1, item 2, item 3, item 4, item 5, item 6}, and the value range of the score R is [1,5 = g]. The two-dimensional score information table RM is shown in table 1:
TABLE 1
Let the threshold value of the user score be equal to 3, and the item with score greater than or equal to 3 is taken as the favorite item of the user, as shown in table 1:
the favorite items of the user 1 are { item 2, item 4, item 6 };
the favorite of user 2 is { item 4, item 6 };
the favorite of user 3 is { item 2, item 3, item 6 };
favorite items of the target user are { item 1, item 3, item 4 };
and 2, reducing the redundant users by using a coverage reduction algorithm:
step 2.1 order item CollectionIAs universes of discourseIIn the universe of discourseIIn the method, each user's favorite articles form a set; in-item attribute matrixAMExtracting the favorite attributes of the target user:
in the formula (1), the reaction mixture is,mthe number of the attributes is represented and,at m a representation of one of the attributes is presented,av m representing attributesat m A value of (d);
such as a ream of articlesI = { item 1, item 2, item 3, item 4, item 5, item 6} is used as a domain, table 2 represents an attribute matrix AM of the items, and an attribute value corresponding to the favorite item of the target user is obtained by statistics according to table 2 and the favorite item set of the target user:
comedy =3, thriller =2, action =1, drama =1, music =1,
selecting two attributes with the largest statistical value as the favorite attributes of the target user, wherein the favorite attributes of the target user are as follows:
[ comedy =1] 'Λ [ thriller =1 ]' Λ [ action =0] 'drama =0 ]' Λ [ music =0]
TABLE 2
Comedy | Thriller | Movement of | Drama (drama) | Music | |
Article 1 | 1 | 0 | 1 | 1 | 0 |
Article 2 | 1 | 1 | 0 | 1 | 0 |
Article 3 | 1 | 1 | 0 | 0 | 0 |
Article 4 | 1 | 1 | 0 | 0 | 1 |
Article 5 | 0 | 0 | 1 | 1 | 0 |
Article 6 | 1 | 1 | 1 | 0 | 1 |
Step 2.2. utilizationObtaining the favorite attributes of the target users, and constructing a decision set of the target usersDSet of decisionsDThe method comprises the following steps of (1) consisting of a collection of articles with favorite attributes:
in the formula (2), the reaction mixture is,at m (i)= av m representing an articleiIn attributeat m A value ofav m ;
If the user's favorite attributes are used: [ comedy =1]Λ thriller =1]Λ [ motion =0]Λ [ drama =0]Λ [ music =0]Building a decision set of target usersDThe decision set consists of all items that possess (comedy, thriller) properties, i.e.:
a decision set D = { item 2, item 3, item 4, item 6} may be obtained from table 2;
step 2.3. to eliminate the redundant users of the target users to the maximum extent, the domain of discourseIReduction of item collections into decision sets for target usersDI.e. discourse domainD(ii) a For each useru∈UBuilding usersuIn the universe of discourseDFavorite articles set onC u :
Order toC*=D-∪C u ,C={C 1 ,C 2 …C n , C*} Form the target user's domain of discourseDA cover ofC;
Such asC 1 = item 2, item 4, item 6;
C 2 = { substanceArticle 4, article 6 };
C 3 = item 2, item 3, item 6;
thenC = { C 1 ,C 2 ,C 3 } A set of target user decisions is formedDAn cover ofC;
Step 2.4 secondary overlay of redundant elements using overlay reduction algorithmCThe intermediate reduction is to obtain the coverage after the reduction is finishedreduct(C) (ii) a The reduction of the redundant elements means that the redundant users of the target user are all deleted, so that the reduced usersU r :
Due to the fact thatC 2 ⊂ C 1 The data is then transmitted, according to a coverage reduction algorithm,C 2 called redundant element slave overlayCIs removed therebyreduct(C)={C 1 , C 3 }; user 2 is removed as redundant user called target user so reduced userU r = user 1, user 3;
step 3, utilizing the reduced userU r Building target usersauCandidate neighboring users ofuThat is, the candidate neighboring users of the target user are { user 1, user 3 };
step 4, calculating the similarity between the target user and the candidate adjacent users, and selecting the adjacent users of the target user:
calculating target user by using Pearson similarity measurement function (5)auAnd candidate neighboring usersu∈U r The degree of similarity between the two images,
in the formula (5), the reaction mixture is,sim(au,u)representing target usersauAnd candidate neighboring usersu∈U r The degree of similarity between the two images,I au ={i ∈I|r au,i ≠*}representing target usersauThe set of items that have been evaluated,an average score value representing a target user;
then selecting the front part with high similarityKNeighbor users with candidate neighbor users as target usersN au (k);
Namely, the similarity of the target user and the user 1, and the similarity of the target user and the user 3 are respectively calculated by using a Pearson similarity measurement function:
sim (target user, user 1) = -0.76
sim (target user, user 3) = -0.53
If two candidate neighbor users with the highest similarity are selected as the neighbor users of the target user, the neighbor user N of the target userTarget user(2) = user 3, user 1;
step 5, performing predictive scoring on the unscored goods of the target user:
proximity user based on target userN au (k)By using the adjusted weighted sum function (6) to the target userauUnscored collection of itemsCarrying out prediction scoring to obtain a prediction scoring table of the target user;
in the formula (6), the reaction mixture is,P au,i representing target usersauTo the articleiThe prediction score of (a) is determined,U i ={u∈U|r u,i ≠*}indicating the evaluated articleiA set of users of (1);λas a regularization factor:
According to the scoring information of the adjacent users of the target user, the objects 5 and 6 which are not scored by the target user are predicted and scored by utilizing the adjusting weighting function, and the result is as follows:
Ptarget user, item 5= 2.16
PTarget user, item 6= 4.93
And 6, selecting an article with the highest predicted score value as a recommendation result, and recommending the article 6 to the target user.
Experiment:
(1) using public data sets
The field of using recommendation systems is often used to test the public data set MovieLens of the performance of recommendation systems. The data set contains 943 users, 1682 movies and 100000 scores, the scores are distributed as {1,2,3,4,5}, and each user scores at least 20 movies.
(2) Evaluation metrics
The invention adopts the average absolute error MAE and the root mean square error RMSE to measure the accuracy of the algorithm, and the MAE and the RMSE measure the accuracy of the recommendation result by calculating the deviation between the actual score and the predicted score of the user, therefore, the smaller the MAE and the RMSE are, the higher the recommendation accuracy is:
coverage is used to measure the diversity of the algorithm. The Coverage indicates a ratio of the item types that can be recommended to the target user to all the target users as the evaluation item types, and thus, the higher the Coverage, the more diverse the recommendations.
In the formula (10), the compound represented by the formula (10),in which S isu,iRepresenting the neighboring users of user u who have evaluated item i.
(3) Parameter setting
The method adopts the Pearson similarity measurement function to calculate the similarity of the users, and uses the adjusting weighting function to predict and grade the unscored goods of the target user. And selecting the top two favorite object attributes of the target user as the favorite attributes of the target user. For clearly comparing the collaborative filtering algorithm of the present invention with the conventional user, the number of neighboring users of the target user, K ∈ {20,25,30, …,60 }. The number of recommended items is set to {2,4,6,8,10,12 }.
(4) Comparison and analysis of the results
The personalized user collaborative filtering recommendation method based on coverage reduction is represented by CBCF, the traditional user collaborative filtering algorithm is represented by UBCF, and the results of accuracy measurement MAE and RMSE are shown in FIG. 2. As can be seen from the data in fig. 2, as the number of neighboring users of the target user increases, the MAE and RMSE results of the CBCF algorithm are always smaller than those of the UBCF algorithm. Since the smaller the MAE and RMSE, the higher the recommendation accuracy, CBCF is able to recommend items with higher accuracy than UBCF. Fig. 3 shows the result of the diversity metric Coverage, and it can be seen from fig. 3 that as the number of neighboring users of the target user increases, the Coverage of the CBCF algorithm is significantly greater than that of the UBCF. Since the higher the Coverage, the more diverse the recommendations, CBCF can recommend more diverse items than UBCF. According to the comprehensive experiment results, the method and the device can provide high-precision and diversified recommendation results at the same time, so that the target user can be personally recommended.
Claims (1)
1. A personalized user collaborative filtering recommendation method based on coverage reduction is characterized by comprising the following steps in sequence:
step 1, forming a two-dimensional scoring information table by statistics:
forming a two-dimensional scoring information table according to the scoring information of the user to the articleRM={U,I,R∪{*}}(ii) a The two-dimensional scoring information tableRMIn (1),Ua set of users is represented as a set of users,Ia collection of items is represented that is,R∪{*}representing a user's set of scores for an item, wherein*Indicating that the user has not scored the item;
order useru∈UTo the articlei∈IIs scored asr u,i ∈R∪{*}And the useruIs given as an average score of;θA threshold value for scoring the user, ifr u,i ≥θIndicate the useruFavorite articlei(ii) a User' suThe scored set of items isI u ={i∈I|r u,i ≠*};For the useruAn unscored collection of items; the article attribute matrix isAM(ii) a In the user set U, if the item set favored by the user a is contained in the item set favored by the user b, the user a is called a redundant user of the target user;
and 2, reducing the redundant users by using a coverage reduction algorithm:
step 2.1 order item CollectionIAs universes of discourseIIn the universe of discourseIIn the method, each user's favorite articles form a set; in-item attribute matrixAMExtracting the favorite attributes of the target user:
in the formula (1), the reaction mixture is,mthe number of the attributes is represented and,at m represents aThe number of the attributes is one,av m representing attributesat m A value of (d);
step 2.2, utilizing the obtained favorite attributes of the target users to construct a decision set of the target usersDSet of decisionsDThe method comprises the following steps of (1) consisting of a collection of articles with favorite attributes:
D={i∈I|at 1(i)= av 1 ,at 2 (i)=av 2 ,…,at m (i)=av m } (2)
in the formula (2), the reaction mixture is,at m (i)= av m representing an articleiIn attributeat m A value ofav m ;
Step 2.3. domain of discourseIReduction of item collections into decision sets for target usersDI.e. discourse domainD(ii) a For each useru ∈UBuilding usersuIn the universe of discourseDFavorite articles set onC u :
Order toC*=D-∪C u ,C={C 1 ,C 2 …C n , C*} Form the target user's domain of discourseDA cover ofC;
Step 2.4, redundant elements are covered by using an covering reduction algorithmCThe intermediate reduction is to obtain the coverage after the reduction is finishedreduct(C) And reduced usersU r :
Step 3, utilizing the reduced userU r Building target usersauCandidate neighboring users ofu;
Step 4, calculating the similarity between the target user and the candidate adjacent users, and selecting the adjacent users of the target user:
calculating target user by using Pearson similarity measurement function (5)auAnd candidate neighboring usersu∈U r The degree of similarity between the two images,
in the formula (5), the reaction mixture is,sim(au,u)representing target usersauAnd candidate neighboring usersu∈U r The degree of similarity between the two images,I au ={i∈I| r au,i ≠*}representing target usersauThe set of items that have been evaluated,an average score value representing a target user;
then selecting the front part with high similarityKNeighbor users with candidate neighbor users as target usersN au (k);
Step 5, performing predictive scoring on the unscored goods of the target user:
proximity user based on target userN au (k)By using the adjusted weighted sum function (6) to the target userauUnscored collection of itemsCarrying out prediction scoring to obtain a prediction scoring table of the target user;
in the formula (6), the reaction mixture is,P au,i representing target usersauTo the articleiThe prediction score of (a) is determined,U i ={u∈U|r u,i ≠*}indicating the evaluated articleiA set of users of (1);λas a regularization factor:
step six, selecting the front part with high prediction scoreNThe individual item is used as a recommendation.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810486715.0A CN108549729B (en) | 2018-05-21 | 2018-05-21 | Personalized user collaborative filtering recommendation method based on coverage reduction |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810486715.0A CN108549729B (en) | 2018-05-21 | 2018-05-21 | Personalized user collaborative filtering recommendation method based on coverage reduction |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108549729A CN108549729A (en) | 2018-09-18 |
CN108549729B true CN108549729B (en) | 2021-06-11 |
Family
ID=63495385
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810486715.0A Active CN108549729B (en) | 2018-05-21 | 2018-05-21 | Personalized user collaborative filtering recommendation method based on coverage reduction |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108549729B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110990689B (en) * | 2019-11-11 | 2023-04-18 | 辽宁师范大学 | Collaborative filtering recommendation algorithm based on rough set rule extraction |
CN112612967B (en) * | 2020-09-11 | 2023-07-18 | 辽宁师范大学 | Collaborative filtering recommendation method based on link clustering and reduction |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102662975A (en) * | 2012-03-12 | 2012-09-12 | 浙江大学 | Bidirectional and clustering mixed friend recommendation method |
CN102982107A (en) * | 2012-11-08 | 2013-03-20 | 北京航空航天大学 | Recommendation system optimization method with information of user and item and context attribute integrated |
CN106934071A (en) * | 2017-04-27 | 2017-07-07 | 北京大学 | Recommendation method and device based on Heterogeneous Information network and Bayes's personalized ordering |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8260787B2 (en) * | 2007-06-29 | 2012-09-04 | Amazon Technologies, Inc. | Recommendation system with multiple integrated recommenders |
-
2018
- 2018-05-21 CN CN201810486715.0A patent/CN108549729B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102662975A (en) * | 2012-03-12 | 2012-09-12 | 浙江大学 | Bidirectional and clustering mixed friend recommendation method |
CN102982107A (en) * | 2012-11-08 | 2013-03-20 | 北京航空航天大学 | Recommendation system optimization method with information of user and item and context attribute integrated |
CN106934071A (en) * | 2017-04-27 | 2017-07-07 | 北京大学 | Recommendation method and device based on Heterogeneous Information network and Bayes's personalized ordering |
Non-Patent Citations (1)
Title |
---|
Neighbor selection for user-based collaborative filtering using covering-based rough sets;Zhipeng Zhang et al.;《Annals of Operations Research》;20171231;全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN108549729A (en) | 2018-09-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106802956B (en) | Movie recommendation method based on weighted heterogeneous information network | |
CN108665323A (en) | A kind of integrated approach for finance product commending system | |
WO2010037286A1 (en) | Collaborative filtering-based recommendation method and system | |
Gogna et al. | DiABlO: Optimization based design for improving diversity in recommender system | |
Li et al. | Content-based filtering recommendation algorithm using HMM | |
CN105354260B (en) | The Mobile solution of a kind of mosaic society's network and item characteristic recommends method | |
CN112948625B (en) | Film recommendation method based on attribute heterogeneous information network embedding | |
CN106127506B (en) | recommendation method for solving cold start problem of commodity based on active learning | |
US9147009B2 (en) | Method of temporal bipartite projection | |
CN107229876B (en) | Collaborative filtering recommendation method meeting difference privacy | |
CN107633444A (en) | Commending system noise filtering methods based on comentropy and fuzzy C-means clustering | |
CN108549729B (en) | Personalized user collaborative filtering recommendation method based on coverage reduction | |
CN106095974B (en) | Recommendation system score prediction and recommendation algorithm based on network structure similarity | |
CN111143699B (en) | Recommendation system based on similarity and confidence coefficient clustering | |
CN107025311A (en) | A kind of Bayes's personalized recommendation method and device based on k nearest neighbor | |
CN109544231A (en) | Logistics distribution service personalization recommended method based on Logistics Information Platform | |
CN106846029B (en) | Collaborative filtering recommendation algorithm based on genetic algorithm and novel similarity calculation strategy | |
CN103942298B (en) | Recommendation method and system based on linear regression | |
CN106202151A (en) | One is used for improving the multifarious method of personalized recommendation system | |
CN104008193A (en) | Information recommending method based on typical user group finding technique | |
CN112149734A (en) | Cross-domain recommendation method based on stacked self-encoder | |
CN106649733B (en) | Online video recommendation method based on wireless access point context classification and perception | |
CN108415926B (en) | Collaborative filtering recommendation method for eliminating scoring noise of original scoring data | |
CN106651461A (en) | Film personalized recommendation method based on gray theory | |
WO2021135192A1 (en) | User curiosity-based bayesian personalized ranking recommendation 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 |