CN109635200A - A kind of Collaborative Filtering Recommendation Algorithm based on MMTD and user - Google Patents
A kind of Collaborative Filtering Recommendation Algorithm based on MMTD and user Download PDFInfo
- Publication number
- CN109635200A CN109635200A CN201811548134.1A CN201811548134A CN109635200A CN 109635200 A CN109635200 A CN 109635200A CN 201811548134 A CN201811548134 A CN 201811548134A CN 109635200 A CN109635200 A CN 109635200A
- Authority
- CN
- China
- Prior art keywords
- user
- article
- scoring
- collection
- mmtd
- 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
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a kind of Collaborative Filtering Recommendation Algorithm based on MMTD and user, after calculating user interest similarity using traditional approach, the method and strategy measured again with similarity degree of the MMTD mode to the feedback behavior of user, feedback behavior measuring similarity is carried out to the higher neighbor objects of Interest Similarity, the mode for improving traditional calculating user interest similarity, is finally reached the purpose for improving recommendation results accuracy and recall rate.The present invention passes through the user interest similarity calculation by MMTD applied to the Collaborative Filtering Recommendation Algorithm based on user, existing algorithm is solved when calculating user interest similarity, lack and the science of user feedback is considered, caused by recommendation results accuracy and the lower problem of recall rate, realize the effect for improving recommendation results accuracy and recall rate.
Description
Technical field
The present invention relates to the technical field of the Collaborative Filtering Recommendation Algorithm (UserCF) based on user, in particular to a kind of bases
In the collaborative filtering recommending of intermediary's measure of truth grad (Measuring Of Medium Truth Degree, MMTD) and user
Algorithm.
Background technique
Collaborative Filtering Recommendation Algorithm based on user is exactly to find have similar tastes and interests, possess altogether using certain algorithm in simple terms
With the group of experience, the hobby of these groups is recommended to the user of same type.The response that algorithm makes information by user
The analysis of (as scored, collecting) realizes the purpose to information filtering, and then helps other users filter information.
Traditional Collaborative Filtering Recommendation Algorithm based on user specifically includes that Interest Similarity calculates and result recommendation two is big
Part.Interest Similarity calculating is one ring of most important one, and the method generally used has Jaccard formula and cosine formula
Deng.The core concept of this two methods is all to seek the intersection for the article collection that user had positive feedback, then determine divided by some
Value, Interest Similarity of the result between user, this way it is simple and it is available possess certain accuracy as a result, still
By the feedback for not having science to consider that user makes article, so that the accuracy and recall rate of recommendation results are lower, although
Accuracy and recall rate is allowed there are some promotions by improving calculating formula of similarity in recent years, but result still has the sky of promotion
Between.
In conclusion it is traditional based on the collaborative filtering of user when calculating user interest similarity, lack to
The science of family feedback considers, so that recommendation results not fully meet the interest habit of user, ultimately causes the standard of recommendation results
True property and the lower problem of recall rate.Therefore it needs to make improvement to the mode for calculating user interest similarity, it can be more
Add the interest habit for meeting user, to improve the accuracy and recall rate of recommendation results.The present invention can be well solved above-mentioned
Problem.
Summary of the invention
Goal of the invention: in order to overcome the deficiencies in the prior art, the present invention provides a kind of based on MMTD and user
Collaborative Filtering Recommendation Algorithm, solve it is traditional based on the collaborative filtering of user when calculating user interest similarity, lack
The science of user feedback is considered, so that recommendation results not fully meet the interest habit of user, the accuracy of recommendation results
The lower problem with recall rate.
Technical solution: to achieve the above object, the technical solution adopted by the present invention are as follows:
A kind of Collaborative Filtering Recommendation Algorithm based on MMTD and user, comprising the following steps:
Step 1, user interest similarity calculation:
It retrieves entire training set and records the purchased number of each article, by sorting from large to small, according to recommendation article number n choosing
F/n article forms popular article collection G ', 0 < F/n < 0.05F before selecting, and F indicates total number of items in training set, once according to user
There is the article collection of positive feedback to obtain shared article object set S,
S=N (i) ∩ N (j),
Wherein, N (i) indicates that user i once had the article collection of positive feedback, and N (j) indicates that user j once had positive feedback
Article collection;
By popular product data collection G ' introducing, obtain removing the user feedback data collection after popular article:
N (i) '=N (i)-N (i) ∩ G '
N (j) '=N (j)-N (j) ∩ G '
Establish article tabling look-up to user:
C [i] [j]=| N (i) ' ∩ N (j) ' |
If user i, j belong to the middle M that tables look-up simultaneously1The corresponding user list of a article, then enable C [i] [j]=M1,
The Interest Similarity of user i and j are calculated by improved cosine similar value formula:
Step 2, the scoring similarity degree based on MMTD calculates:
User i and j may be n to the scoring of item object1~n2Between any one positive integer value, remember predicate P (x
(i, j)) indicate that user i to be investigated is identical to item object scoring as j, ╕ P (x (i, j)) indicates that N (i) is different with N (j) ,~
P (x (i, j)) indicate i and j between it is identical it is different between, pass through and calculate distance proportion function hT(x (i, j)) obtains user i and j
To the similarity degree of item object scoring;
Relative score f (x (i, j)) is obtained to the scoring of article according to user:
F (x (i, j))=| Qig-Qjg|;
Wherein, QigScoring for user i to article g, QjgScoring for user j to article g;
For y=f (x (i, j)) using~P as symmetrical centre, left and right is respectively P He ╕ P, the value of f (x (i, j)) on number axis
It is [0, n2-n1];
The value of y=f (x (i, j)) falls in three codomain (αr+εr,αl-εl),(0,αr+εr),(αl-εl,n2-n1) in ,~P
The region of (x (i, j)) is (αr+εr,αl-εl), the region of P (x (i, j)) is (0, αr+εr) the region of , ╕ P (x (i, j)) is
(αl-εl,n2-n1), the true value of P (x (i, j)) is that the true value of 1 , ╕ P (x (i, j)) is 0;
Distance rates function relative to P (x (i, j))
Wherein,
Wherein d is the absolute value of the difference of two values, passes through distance rates function hTgThe calculating of (x (i, j)), obtains user i
The similarity degree to score with j item object g;
Shared article object set S is traversed, user i and the similarity degree of all shared article object scores of j are summed, then removes
To share the size n of article collection S, comprehensive score similarity degree h is obtainedTn(x (i, j)):
Recommendation results are according to the ascending sequencing selection neighbor objects of comprehensive score similarity degree.
It traverses training set and extracts data composition user-article collection and user-article-scoring collection, calculate user and other use
User interest similarity between family, selecting neighbor user collection M of the similarity in the user of preceding 2K as the user, K is to recommend
Number of users.
Neighbor user collection M is traversed, neighbor user is extracted from user-article-scoring concentration and concentrates corresponding user-object
Product-scoring, calculate relative score, calculate the article scoring similarity degree of candidate user, the synthesis of user is finally calculated
Article scoring similarity degree.
According to recommended user number K, the ascending sequencing selection neighbor user object of point similarity degree is judged according to synthesis.
Region representated by P (x (i, j)) accounts for region representated by 20% , ╕ P (x (i, j)) and accounts for 50% ,~P (x (i, j))
Representative region accounts for 30%.
The present invention compared with prior art, has the advantages that
The present invention by by MMTD be applied to the Collaborative Filtering Recommendation Algorithm based on user user interest similarity calculation,
For existing algorithm when calculating user interest similarity, lack and the science of user feedback considered, caused by recommendation results it is quasi-
True property and the lower problem of recall rate, realize the effect for improving recommendation results accuracy and recall rate.
Detailed description of the invention
Collaborative Filtering Recommendation Algorithm flow chart of the Fig. 1 based on MMTD and user.
The different corresponding relationship with similar value section of Fig. 2 predicate.
Fig. 3 carries out analogical object using MMTD and selects optimized flow chart.
Specific embodiment
In the following with reference to the drawings and specific embodiments, the present invention is furture elucidated, it should be understood that these examples are merely to illustrate this
It invents rather than limits the scope of the invention, after the present invention has been read, those skilled in the art are to of the invention various
The modification of equivalent form falls within the application range as defined in the appended claims.
A kind of Collaborative Filtering Recommendation Algorithm based on MMTD and user, is a kind of method of tactic, is using tradition
The side that mode calculates user interest similarity and then measured with similarity degree of the MMTD mode to the feedback behavior of user
Method and strategy carry out feedback behavior measuring similarity to the higher neighbor objects of Interest Similarity, improve traditional calculating user
The mode of Interest Similarity is finally reached the purpose for improving recommendation results accuracy and recall rate.
The mode of traditional calculating user interest similarity is being counted to improve the accuracy of recommendation results as far as possible
When calculating user interest similarity, the article that user had positive feedback is considered emphatically, that is, had joint act between user
Article lacks science to user feedback and considers, so that there are errors for calculating of the algorithm to the Interest Similarity of user, ultimately causes
The lower problem of the accuracy and recall rate of recommendation results.For example: having given 1 point of scoring after user A viewing film Z, used
5 points of scoring is given after family B viewing film Z, traditional algorithm can be included in using film Z as the common interest of user A and B
Interest Similarity calculates, and actually user A does not simultaneously like film Z.In order to solve this problem, the present invention is by MMTD to process
The similarity degree of the feedback for the neighbours that traditional algorithm is chosen is evaluated, and is redefined between user according to the similarity degree of feedback
Interest Similarity sequence, improve recommendation results accuracy and recall rate.
Method flow:
As shown in Figure 1, the present invention proposes a kind of UserCF algorithm based on MMTD comprising following steps:
1, user interest similarity calculation
Defining total number of items in training set is F, retrieves entire training set and records the purchased number of each article, by from greatly to
Small sequence forms popular article collection G ' according to a article of F/n (0 < F/n < 0.05F) before recommending article number n to select.
N (i) is enabled to indicate that user i once had the article collection of positive feedback, N (j) indicates that user j once had the object of positive feedback
Product collection.
Then share article object set S:
S=N (i) ∩ N (j) (1)
By popular product data collection G ' introducing, obtain removing the user feedback data collection after popular article:
N (i) '=N (i)-N (i) ∩ G ' (2)
N (j) '=N (j)-N (j) ∩ G ' (3)
For the time complexity for reducing algorithm, we establish article tabling look-up to user:
C [i] [j]=| N (i) ' ∩ N (j) ' | (4)
If user i, j belong to the middle M that tables look-up simultaneously1The corresponding user list of a article, then enable C [i] [j]=M1。
The Interest Similarity of user i and j are calculated by improved cosine similar value formula:
2, the scoring similarity degree based on MMTD calculates
User object i and j may be n to the scoring of item object1~n2Between any one positive integer value.By away from
From proportion function hT(x (i, j)) can calculate the similarity degree that i and j score to item object.
Note predicate P (x (i, j)) indicates that user i to be investigated is identical to a certain item object scoring as j, ╕ P (x (i, j))
Indicate that N (i) is different with N (j) ,~P (x (i, j)) expression i and the scoring of j between it is identical it is different between, predicate is different and similar value
The corresponding relationship in magnitude numerical value section is as shown in Figure 2.Wherein region representated by P (x (i, j)) accounts for 10% , ╕ P (x (i, j)) institute
The region of representative accounts for 60%, and region representated by~P (x (i, j)) accounts for 30%, can by calculate distance rates function h (x (i,
J) similarity degree that i and j score to item object) is obtained.
Assuming that QigScoring for user i to article g, QjgScoring for user j to article g, definition:
F (x (i, j))=| Qig-Qjg| (6)
For the y=f (x (i, j)) from number axis it is recognised that using~P as symmetrical centre on number axis, left and right is respectively P He ╕
The value of P, f (x (i, j)) are [0, n2-n1]。
The value of y=f (x (i, j)) falls in three codomain (αr+εr,αl-εl),(0,αr+εr),(αl-εl,n2-n1) in ,~P
The region of (x (i, j)) is (αr+εr,αl-εl), the region of P (x (i, j)) is (0, αr+εr) the region of , ╕ P (x (i, j)) is
(αl-εl,n2-n1), the true value of P (x (i, j)) is that the true value of 1 , ╕ P (x (i, j)) is 0.
Distance rates function relative to P (x (i, j))
Wherein
Pass through distance rates function hTgThe calculating of (x (i, j)), the phase that available user i and j scores to item object g
Like degree.
Shared article object set S is traversed, user i and the similarity degree of all shared article object scores of j are summed, then removes
To share the size n of article collection S, comprehensive score similarity degree h is obtainedTn(x (i, j)):
Recommendation results are according to the ascending sequencing selection neighbor objects of comprehensive score similarity degree.
As shown in figure 3, specific implementation step is as follows:
1, traversal training set extracts data composition user-article collection and user-article-scoring collection.Traverse user-article collection
Popular article set is obtained, it is useful then to loop through all user object removal institutes according to formula (2) and formula (3) for
The article of family positive feedback concentrates popular article.
2, according to formula (4), traverse user-article collection establishes article-user and tables look-up C [i] [j].
3, it is tabled look-up C [i] [j] according to article-user, excludes not generate behavior to same article with active user
User.User-article collection after traversal exclusion calculates the interest phase between the user for possessing shared article according to formula (5)
Like value, finished until all users calculate between any two.
4, the descending sequence of the value for the Interest Similarity that each user obtains according to step 3, the user of 2K, builds before selecting
It waits and selects neighbor objects collection.
5, according to formula (1), formula (6)-(8), user-article-scoring collection traverse candidate neighbor object set, calculate and use
The comprehensive score similarity degree at family.
6, the comprehensive score similarity degree between the user obtained according to step 5, preferential selection are located at the use in P (x (i, j))
Then the ascending sequencing selection neighbor user pair of point similarity degree is judged according to synthesis in source of the family as recommendation results
As.
The present invention improves the accurate of algorithm while improving the recall rate of the Collaborative Filtering Recommendation Algorithm based on user
Property, the related problem of effective solution belongs to the research field of proposed algorithm.Traditional is calculated based on user collaborative filtered recommendation
Method lacks the evaluation method of science to user feedback, causes due to considering the similarity between user's positive feedback article emphatically
The lower problem of the recall rate and accuracy of recommendation results.For such problems, method of the invention is in conventional method meter
After calculating user interest similarity, using MMTD as the evaluation method to user feedback, to calculate the comprehensive score similarity of user,
According to the comprehensive score similarity of user, more preferably neighbor objects are selected, is finally reached and is improved recommendation results accuracy and recall
The effect of rate, the present invention can optimize the evaluation method of user interest similarity, improve the accuracy and recall rate of recommendation results.
The above is only a preferred embodiment of the present invention, it should be pointed out that: for the ordinary skill people of the art
For member, various improvements and modifications may be made without departing from the principle of the present invention, these improvements and modifications are also answered
It is considered as protection scope of the present invention.
Claims (5)
1. a kind of Collaborative Filtering Recommendation Algorithm based on MMTD and user, which comprises the following steps:
Step 1, user interest similarity calculation:
It retrieves entire training set and records the purchased number of each article, by sorting from large to small, before recommending article number n selection
F/n article forms popular article collection G ', 0 < F/n < 0.05F, and F indicates total number of items in training set, once had according to user
The article collection of positive feedback obtains shared article object set S,
S=N (i) ∩ N (j),
Wherein, N (i) indicates that user i once had the article collection of positive feedback, and N (j) indicates that user j once had the object of positive feedback
Product collection;
By popular product data collection G ' introducing, obtain removing the user feedback data collection after popular article:
N (i) '=N (i)-N (i) ∩ G '
N (j) '=N (j)-N (j) ∩ G '
Establish article tabling look-up to user:
C [i] [j]=| N (i) ' ∩ N (j) ' |
If user i, j belong to the middle M that tables look-up simultaneously1The corresponding user list of a article, then enable C [i] [j]=M1,
The Interest Similarity of user i and j are calculated by improved cosine similar value formula:
Step 2, the scoring similarity degree based on MMTD calculates:
User i and j may be n to the scoring of item object1~n2Between any one positive integer value, remember predicate P (x (i, j))
Indicating that user i to be investigated is identical to item object scoring as j, ╕ P (x (i, j)) indicates that N (i) is different with N (j) ,~P (x (i,
J)) indicate i and j between it is identical it is different between, pass through and calculate distance proportion function hT(x (i, j)) obtains user i and j to article
The similarity degree of object score;
Relative score f (x (i, j)) is obtained to the scoring of article according to user:
F (x (i, j))=| Qig-Qjg|;
Wherein, QigScoring for user i to article g, QjgScoring for user j to article g;
Y=f (x (i, j)) is using~P as symmetrical centre on number axis, and left and right is respectively P He ╕ P, the value of f (x (i, j)) be [0,
n2-n1];
The value of y=f (x (i, j)) falls in three codomain (αr+εr,αl-εl),(0,αr+εr),(αl-εl,n2-n1) in ,~P (x (i,
J) region) is (αr+εr,αi-εl), the region of P (x (i, j)) is (0, αr+εr) the region of , ╕ P (x (i, j)) is (αl-εl,
n2-n1), the true value of P (x (i, j)) is that the true value of 1 , ╕ P (x (i, j)) is 0;
Distance rates function relative to P (x (i, j))
Wherein,
Wherein d is the absolute value of the difference of two values, passes through distance rates function hTgThe calculating of (x (i, j)), obtains user i and j couples
The similarity degree of item object g scoring;
Shared article object set S is traversed, user i and the similarity degree of all shared article object scores of j are summed, then divided by altogether
There is the size n of article collection S, obtains comprehensive score similarity degree hTn(x (i, j)):
Recommendation results are according to the ascending sequencing selection neighbor objects of comprehensive score similarity degree.
2. the Collaborative Filtering Recommendation Algorithm based on MMTD and user according to claim 1, it is characterised in that: traversal training set
Data composition user-article collection and user-article-scoring collection are extracted, the user interest phase between user and other users is calculated
Like degree, selecting similarity in the user of preceding 2K is recommended user's number as neighbor user the collection M, K of the user.
3. the Collaborative Filtering Recommendation Algorithm based on MMTD and user according to claim 2, it is characterised in that: traversal neighbours use
Family collection M extracts neighbor user from user-article-scoring concentration and concentrates corresponding user-article-scoring, calculates opposite comment
Point, the article scoring similarity degree of candidate user is calculated, the synthesis that user is finally calculated judges a point similarity degree.
4. the Collaborative Filtering Recommendation Algorithm based on MMTD and user according to claim 3, it is characterised in that: used according to recommendation
Amount K judges the ascending sequencing selection neighbor user object of point similarity degree according to synthesis.
5. the Collaborative Filtering Recommendation Algorithm based on MMTD and user according to claim 1, it is characterised in that: P (x (i, j))
Representative region accounts for region representated by 20% , ╕ P (x (i, j)) and accounts for 50%, and region representated by~P (x (i, j)) accounts for
30%.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811548134.1A CN109635200B (en) | 2018-12-18 | 2018-12-18 | Collaborative filtering recommendation method based on intermediary truth degree measurement and user |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811548134.1A CN109635200B (en) | 2018-12-18 | 2018-12-18 | Collaborative filtering recommendation method based on intermediary truth degree measurement and user |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109635200A true CN109635200A (en) | 2019-04-16 |
CN109635200B CN109635200B (en) | 2022-02-01 |
Family
ID=66074945
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811548134.1A Active CN109635200B (en) | 2018-12-18 | 2018-12-18 | Collaborative filtering recommendation method based on intermediary truth degree measurement and user |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109635200B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110427565A (en) * | 2019-06-10 | 2019-11-08 | 深圳大学 | A kind of item recommendation method based on collaborative filtering, intelligent terminal and storage medium |
CN111915409A (en) * | 2020-08-11 | 2020-11-10 | 深圳墨世科技有限公司 | Article recommendation method, device and equipment based on article and storage medium |
CN113254789A (en) * | 2021-06-30 | 2021-08-13 | 中国气象局公共气象服务中心(国家预警信息发布中心) | Method and device for pushing meteorological service content |
CN116561603A (en) * | 2023-07-10 | 2023-08-08 | 深圳益普睿达市场咨询有限责任公司 | User matching method and device based on data analysis |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103544708A (en) * | 2013-10-31 | 2014-01-29 | 南京邮电大学 | Image quality objective evaluation method based on MMTD |
CN103678432A (en) * | 2013-04-07 | 2014-03-26 | 南京邮电大学 | Webpage main body extraction method based on webpage main body features and intermediate true values |
US8738467B2 (en) * | 2006-03-16 | 2014-05-27 | Microsoft Corporation | Cluster-based scalable collaborative filtering |
CN104123554A (en) * | 2013-04-23 | 2014-10-29 | 南京邮电大学 | SIFT image characteristic extraction method based on MMTD |
CN105677638A (en) * | 2016-01-05 | 2016-06-15 | 北京工业大学 | Web information extraction method |
-
2018
- 2018-12-18 CN CN201811548134.1A patent/CN109635200B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8738467B2 (en) * | 2006-03-16 | 2014-05-27 | Microsoft Corporation | Cluster-based scalable collaborative filtering |
CN103678432A (en) * | 2013-04-07 | 2014-03-26 | 南京邮电大学 | Webpage main body extraction method based on webpage main body features and intermediate true values |
CN104123554A (en) * | 2013-04-23 | 2014-10-29 | 南京邮电大学 | SIFT image characteristic extraction method based on MMTD |
CN103544708A (en) * | 2013-10-31 | 2014-01-29 | 南京邮电大学 | Image quality objective evaluation method based on MMTD |
CN105677638A (en) * | 2016-01-05 | 2016-06-15 | 北京工业大学 | Web information extraction method |
Non-Patent Citations (1)
Title |
---|
周宁宁 等: "基于中介真值程度度量处理图像的应用理论研究", 《南京邮电大学学报》 * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110427565A (en) * | 2019-06-10 | 2019-11-08 | 深圳大学 | A kind of item recommendation method based on collaborative filtering, intelligent terminal and storage medium |
CN110427565B (en) * | 2019-06-10 | 2023-02-24 | 深圳大学 | Article recommendation method based on collaborative filtering, intelligent terminal and storage medium |
CN111915409A (en) * | 2020-08-11 | 2020-11-10 | 深圳墨世科技有限公司 | Article recommendation method, device and equipment based on article and storage medium |
CN111915409B (en) * | 2020-08-11 | 2023-08-04 | 深圳墨世科技有限公司 | Item recommending method, device, equipment and storage medium based on item |
CN113254789A (en) * | 2021-06-30 | 2021-08-13 | 中国气象局公共气象服务中心(国家预警信息发布中心) | Method and device for pushing meteorological service content |
CN116561603A (en) * | 2023-07-10 | 2023-08-08 | 深圳益普睿达市场咨询有限责任公司 | User matching method and device based on data analysis |
CN116561603B (en) * | 2023-07-10 | 2023-09-01 | 深圳益普睿达市场咨询有限责任公司 | User matching method and device based on data analysis |
Also Published As
Publication number | Publication date |
---|---|
CN109635200B (en) | 2022-02-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109635200A (en) | A kind of Collaborative Filtering Recommendation Algorithm based on MMTD and user | |
CN110110094A (en) | Across a network personage's correlating method based on social networks knowledge mapping | |
CN109299380B (en) | Exercise personalized recommendation method based on multi-dimensional features in online education platform | |
CN102609523B (en) | The collaborative filtering recommending method classified based on taxonomy of goods and user | |
CN106021298B (en) | A kind of collaborative filtering recommending method and system based on asymmetric Weighted Similarity | |
CN104899246B (en) | Collaborative filtering recommending method based on blurring mechanism user scoring neighborhood information | |
CN109902235A (en) | User preference based on bat optimization clusters Collaborative Filtering Recommendation Algorithm | |
CN105787068B (en) | The academic recommended method and system analyzed based on citation network and user's proficiency | |
CN105574216A (en) | Personalized recommendation method and system based on probability model and user behavior analysis | |
CN106484876A (en) | A kind of based on typical degree and the collaborative filtering recommending method of trust network | |
CN108550068A (en) | A kind of individual commodity recommendation method and system based on user behavior analysis | |
CN109472286A (en) | Books in University Library recommended method based on interest-degree model Yu the type factor | |
CN108132964A (en) | A kind of collaborative filtering method to be scored based on user item class | |
CN108446350B (en) | Recommendation method based on topic model analysis and long and short interests of user | |
CN106980646A (en) | The method applied based on popularity to the influencing mechanism analysis of user interest and its in proposed algorithm | |
CN108228867A (en) | A kind of theme collaborative filtering recommending method based on viewpoint enhancing | |
CN109241203A (en) | A kind of user preference and distance weighted clustering method of time of fusion factor | |
CN107562947A (en) | A kind of Mobile Space-time perceives the lower dynamic method for establishing model of recommendation service immediately | |
CN108664564A (en) | A kind of improvement collaborative filtering recommending method based on item contents feature | |
CN107391670A (en) | A kind of mixing recommendation method for merging collaborative filtering and user property filtering | |
CN110083764A (en) | A kind of collaborative filtering cold start-up way to solve the problem | |
CN106846088A (en) | A kind of Method of Commodity Recommendation of the product electric business website that disappears soon | |
CN111681084A (en) | E-commerce platform recommendation method based on social relationship influence factors | |
CN110362755A (en) | A kind of recommended method of the hybrid algorithm based on article collaborative filtering and correlation rule | |
CN108694234A (en) | A kind of service recommendation model based on improvement collaborative filtering |
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 |