CN111079004B - Three-part graph random walk recommendation method based on word2vec label similarity - Google Patents
Three-part graph random walk recommendation method based on word2vec label similarity Download PDFInfo
- Publication number
- CN111079004B CN111079004B CN201911241017.5A CN201911241017A CN111079004B CN 111079004 B CN111079004 B CN 111079004B CN 201911241017 A CN201911241017 A CN 201911241017A CN 111079004 B CN111079004 B CN 111079004B
- Authority
- CN
- China
- Prior art keywords
- user
- node
- graph
- similarity
- label
- 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
- 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
- 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
- G06F16/9536—Search customisation based on social or collaborative filtering
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a method for recommending a three-part graph random walk based on word2vec label similarity, which comprises the following steps of: s1, calculating similarity between articles based on an ICF algorithm-cosRA; s2, generating a recommendation list according to the similarity of the articles and the historical behaviors of the user; s3, constructing a user-article bipartite graph according to the target user and the recommendation list, and establishing a bipartite graph model based on introduction of label nodes in the user-article bipartite graph; s4, building a word2vec model, and taking the memory vocabulary as a corpus of training word2 vec; s5, giving weights to edges among the labels of the three parts of the graph; and S6, performing random walk on the three-part graph from the user node, wherein after the random walk is performed for a plurality of times, the probability of each article node being visited converges to a number, and the visit probability is the weight of the articles in the final recommendation list.
Description
Technical Field
The invention belongs to the technical field of internet information recommendation, and particularly relates to a three-part graph random walk recommendation method based on word2vec label similarity.
Background
With the development of information technology and the internet, people gradually move from an information-deficient era to an information-overloaded era. In this age, both information consumers and information producers have met with significant challenges: as information consumers, it is very difficult to find out the information which is interested by the consumers from a large amount of information; it is very difficult for information producers to make information produced by themselves stand out, and the information producers get attention from the wide range of users. Recommendation systems are important tools to resolve this conflict. The recommendation system is used for associating users with information, on one hand, the recommendation system helps the users to find valuable information for the users, on the other hand, the information can be presented to the users interested in the information, and therefore win-win effect of information consumers and information producers is achieved.
As is well known, to solve the problem of information overload, numerous talent solutions have been proposed by countless scientists and engineers, of which the representative solution is a catalog and search engine. The search engine can enable the user to find the required information by searching the keywords. However, search engines require users to actively provide accurate keywords to find information. Therefore, many requirements of the user cannot be solved, for example, when the user cannot find the keywords which accurately describe the requirements of the user, the search engine cannot be helped. Like search engines, recommendation systems are also a tool to help users quickly find useful information. Different from a search engine, the recommendation system does not need a user to provide clear requirements, and the historical behaviors of the user are analyzed to model the interests of the user, so that information capable of meeting the interests and the requirements of the user is actively recommended to the user. Thus, the recommendation system and search engine are, in a sense, two complementary tools for the user. The search engine meets the active search requirement when the user has a definite purpose, and the recommendation system can help the user to find interesting new content when the user has no definite purpose.
A good recommender system-indiscernible recommendation algorithm, e.g. collaborative filtering [1-2] Mass diffusion [3] Thermal conduction of [4] Graph computation, etc. Among these, graph-based models are important content in recommendation systems. Google's pageank algorithm was first used for ranking web pages and was later improved to PersonalRank for recommendation systems [5] The algorithm has good performance in diversity, but the recommendation accuracy is low. The difference between the goods recommended to the user and the actual demand of the user is too large, and the experience of the user is reduced.
Reference to the literature
[1]B.Sarwar,G.Karypis,J.Konstan,J.Riedl,Item-based collaborative filtering recommendation algorithms,in:Proceedings of the 10th International Conference on World Wide Web,WWW’01,ACM,New York,NY,USA,2001,pp.285–295.
[2]D.Goldberg,D.Nichols,B.M.Oki,D.Terry,Using collaborative filtering to weave an information tapestry,Commun.ACM 35(12)(1992)61–70.
[3]Y.-C.Zhang,M.Medo,J.Ren,T.Zhou,T.Li,F.Yang,Recommendation model based on opinion diffusion,Europhys.Lett.80(6)(2007)68003.
[4]Y.-C.Zhang,M.Blattner,Y.-K.Yu,Heat conduction process on community networks as a recommendation model,Phys.Rev.Lett.99(15)(2007)154301.
[5]Taher H.Haveliwala Topic-Sensitive PageRank 2002.
[6]Ling-Jiao Chen,Zi-Ke Zhang,Jin-Hu Liu,Jian Gao,Tao Zhoua,A vertex similarity index for better personalized recommendation(2017)
[7]Tomas Mikolov,Kai Chen,Greg Corrado,Jeffrey Dean Efficient Estimation of Word Representations in Vector Space 2013.
Disclosure of Invention
The invention aims to provide a three-part graph random walk recommendation method based on word2vec label similarity aiming at the defects in the prior art, so as to solve the problems that the difference between the actual demand of an article recommended to a user and the actual demand of the user is too large and the experience of the user is reduced due to the traditional algorithm.
In order to achieve the purpose, the invention adopts the technical scheme that:
a three-part graph random walk recommendation method based on word2vec label similarity comprises the following steps:
s1, calculating similarity between articles based on an ICF algorithm-cosRA;
s2, generating a recommendation list according to the similarity of the articles and the historical behaviors of the user;
s3, constructing a user-article bipartite graph according to the target user and the recommendation list, and establishing a bipartite graph model based on introduction of label nodes in the user-article bipartite graph;
s4, building a word2vec model which is trained, and taking the memory vocabulary as a corpus of the training word2 vec;
s5, giving weights to edges among the labels of the articles in the three graphs;
and S6, carrying out random walk on the three-part graph from the user node, wherein after the random walk is carried out for a plurality of times, the probability of each article node being visited converges to a number, and the visit probability is the weight of the articles in the final recommendation list.
Preferably, the ICF-based algorithm-cosRA in step S1 calculates the similarity between the articles as:
wherein the content of the first and second substances,
degree of agreement between alpha and beta of an object, K α ，K β Degree of items alpha, beta, respectively, if user i purchases item alpha, then a iα =1, otherwise a iα ＝0。
Preferably, in step S2, a recommendation list is generated according to the similarity of the items and the historical behavior of the user, wherein the length of the recommendation list is 2n and the n is 50.
Preferably, the step S3 of constructing the user-item bipartite graph according to the target user and the recommendation list, and establishing a three-part graph model based on introducing label nodes in the user-item bipartite graph includes the specific steps of:
constructing a user-article bipartite graph by a target user and 2N articles recommended by the cosrA, wherein the weight of each edge is provided by a recommendation coefficient obtained by the cosrA;
establishing a three-part graph model by introducing label nodes into a user-article two-part graph, and enabling G (u, I, T) to represent a user, an article and a label three-part graph, wherein V = u { [ U ] } V { (U { [ U ] } V } T { (U, I, T) } represents a user, an article and a label three-part graph i ∪V t 2N item sets V recommended by target user u, cosRA i And a label set V corresponding to the 2N items in the label data set t Composition of nodes V for target user u and item i There is a corresponding edge e (u, i) with a weight W u,i = cosrA (u, i), cosrA (u, i) being given by cosrA algorithmRecommending coefficients for the target user u and the item i; for each dyad (i, t) in the graph, there is a corresponding edge e (i, t) in the graph, and there is no directly connected edge between the user and the label.
Preferably, the step S5 of giving the weight to the edges between the labels of the three-part graph includes the following specific steps:
scaling the tag similarity range between 0 and 1:
wherein min = -1,max = -1,w tt’ Similarity of the tag t and the tag t' calculated for word2 vec;
building a weight S of an edge between an item and a label i,t ：
Wherein N is i Is a set of labels corresponding to the item i in the data set, n t,i Number of times item i is labeled t, n t′i Number of times an item i is labeled t', m i Is a set N i Length of (d);
calculated S i,t Needs to be scaled to between 0 and 1, i.e. S' i,t ：
Preferably, the specific steps of step S6 include:
starting from a user node, randomly walking on the three-part graph, and when walking to any node, firstly determining whether to continue walking or stop the walking according to the probability theta and walking from v u The user node starts to swim again;
if the walking is determined to continue, randomly selecting one node from the nodes pointed by the current node as the node passing the walking next time according to uniform distribution; after a plurality of random walks, the probability that each item node is visited converges to a number, and the weight of the item in the final recommendation list is the visit probability PR (v) of the item node:
wherein out (v) represents a node set with edges connected with the node v, out (v ') is a node set with edges connected with the node v', PR (v ') is the access probability of the node v' in the graph, the first N optimal item sets are taken as the final recommendation list after the wandering is finished, and S is the weight of the edges.
The three-part graph random walk recommendation method based on word2vec label similarity provided by the invention has the following beneficial effects:
the method is characterized in that text similarity is calculated based on natural language processing, a multi-attribute random walk model is added to update a recommendation list of a traditional collaborative filtering algorithm, so that accuracy and diversity of the algorithm are improved, similarity among article labels is obtained by constructing a word2vec model, a new edge used for random walk is generated among the labeled articles, weight of the edge is obtained through label similarity and label importance, and finally iterative walk is performed in a three-part graph through a walk algorithm to obtain the recommendation list of a user. The invention realizes more accurate and diversified recommendation, improves the recommendation efficiency and precision, solves the problem of difficult selection for users due to large information amount, and also solves the problems of overlarge difference between the actual requirements of the articles recommended to the users and reduction of the experience of the users caused by the traditional algorithm.
Drawings
FIG. 1 is a user item bipartite graph of a bipartite graph random walk recommendation method based on word2vec tag similarity.
FIG. 2 is a user item label three-part diagram of a three-part diagram random walk recommendation method based on word2vec label similarity.
FIG. 3 is a model diagram of building word2vec based on the three-part graph random walk recommendation method of word2vec tag similarity.
Detailed Description
The following description of the embodiments of the present invention is provided to facilitate the understanding of the present invention by those skilled in the art, but it should be understood that the present invention is not limited to the scope of the embodiments, and it will be apparent to those skilled in the art that various changes may be made without departing from the spirit and scope of the invention as defined and defined in the appended claims, and all matters produced by the invention using the inventive concept are protected.
According to one embodiment of the application, referring to fig. 1-3, the method for recommending the random walk of the three-part graph based on the word2vec label similarity comprises the following steps:
s1, calculating similarity between articles based on an ICF algorithm-cosRA;
s2, generating a recommendation list according to the similarity of the articles and the historical behaviors of the user;
s3, constructing a user-article bipartite graph according to the target user and the recommendation list, and establishing a three-part graph model based on introduction of label nodes in the user-article bipartite graph;
s4, building a word2vec model which is trained, and taking the memory vocabulary as a corpus of the training word2 vec;
s5, giving weights to edges among the labels of the articles in the three graphs;
and S6, carrying out random walk on the three-part graph from the user node, wherein after the random walk is carried out for a plurality of times, the probability of each article node being visited converges to a number, and the visit probability is the weight of the articles in the final recommendation list.
The invention creates a process of calculating random walk by taking natural language processing as a basis, and takes the similarity of the label on the text instead of the user behavior as a new attribute as a factor influencing recommendation. Through the attribute, the traditional recommendation system and the deep learning can be associated, and a new idea is provided for the combination of the traditional recommendation system and the deep learning. Through the model, the recommendation list provided by the conventional collaborative filtering can be optimized, and the accuracy and diversity of the recommendation system are obviously improved.
The above steps will be described in detail
S1, calculating the similarity between the articles based on an ICF algorithm-cosRA:
wherein the content of the first and second substances,
is the degree of identity between the objects alpha, beta, K α ，K β Degree of items alpha, beta, respectively, if user i purchases item alpha, then a iα =1, otherwise a iα ＝0。
And S2, generating a recommendation list for the user according to the similarity of the articles and the historical behaviors of the user, wherein the length of the recommendation list is 2N, and N is 50.
S3, constructing a user-article bipartite graph, and establishing a three-part graph model based on label nodes introduced into the user-article bipartite graph, wherein the specific steps comprise:
referring to fig. 1, a user-item bipartite graph is constructed from the target user and the 2N items recommended by cosRA, where the weight of each edge is provided by the recommendation coefficient derived from cosRA;
establishing a three-part graph model by introducing label nodes into a user-article two-part graph, and enabling G (u, I, T) to represent a user, an article and a label three-part graph, wherein V = u { [ U ] } V { (U { [ U ] } V } T { (U, I, T) } represents a user, an article and a label three-part graph i ∪V t 2N item sets V recommended by target user u, cosRA i And a label set V corresponding to the 2N items in the label data set t Composition of nodes V for target user u and item i There is a corresponding edge e (u, i) with a weight W u,i = cosRA (u, i), which is the recommendation coefficient given by the cosRA algorithm for the target user u, item i; for each dyad (i, t) in the graph, there is a corresponding edge e (i, t) in the graph, and there is no directly connected edge between the user and the label.
And S4, referring to FIG. 3, constructing a word2vec model, training more than 600 million Wikipedia English articles, and using the memory vocabulary as a corpus for training the word2 vec. By using the model, the text similarity between almost any English vocabulary can be calculated. If the article label is Chinese, a corpus is established by using a Chinese Wikipedia article, and the difference is that a word segmentation step needs to be added in the middle.
Step S5, the specific steps of endowing the weights of the edges among the article labels in the three-part graph comprise:
the edge e (i, t) between the article labels in the three-part graph is given weight, because the word2vec model calculates the cosine similarity between words, the range is-1 to 1, and the similarity range is scaled to be between 0 and 1 by using the following formula:
wherein, min = -1,max = -1,w tt’ Similarity of the tag t and the tag t' calculated for word2 vec;
building a weight S of an edge between an item and a label i,t ：
Wherein N is i Is a set of labels corresponding to the item i in the data set, n t,i Number of times item i is labeled t, n t′i Number of times item i is labeled t', m i Is a set N i Length of (2)
Calculated S i,t Needs to be scaled between 0 and 1, i.e. S' i,t ：
Step S6, starting from the user node, on the three-part graphRandom walk is carried out, when the walking is carried out to any node, whether the walking is continued or not is determined according to the probability theta, and the walking is stopped and is carried out from v u The user node starts to wander again;
if the walking is determined to continue, randomly selecting one node from the nodes pointed by the current node as the node passing the walking next time according to uniform distribution; after a plurality of random walks, the probability that each item node is visited converges to a number, and the weight of the item in the final recommendation list is the visit probability PR (v) of the item node:
wherein out (v) represents a node set with edges connected with the node v, out (v ') is a node set with edges connected with the node v', PR (v ') is the access probability of the node v' in the graph, the first N optimal item sets are taken as the final recommendation list after the wandering is finished, and S is the weight of the edges.
The method is characterized in that text similarity is calculated based on natural language processing, a multi-attribute random walk model is added to update a recommendation list of a traditional collaborative filtering algorithm, so that accuracy and diversity of the algorithm are improved, similarity among article labels is obtained by constructing a word2vec model, a new edge for random walk is generated among the labeled articles, weight of the edge is obtained through label similarity and label importance, and finally, iterative walk is performed in a three-part graph through a walk algorithm to obtain the recommendation list of a user. The invention realizes more accurate and diversified recommendation, improves the recommendation efficiency and precision, solves the problem of difficult selection for users due to large information amount, and also solves the problems of overlarge difference between the actual requirements of the articles recommended to the users and reduction of the experience of the users caused by the traditional algorithm.
While the embodiments of the invention have been described in detail in connection with the accompanying drawings, it is not intended to limit the scope of the invention. Various modifications and changes may be made by those skilled in the art without inventive step within the scope of the appended claims.
Claims (4)
1. A three-part graph random walk recommendation method based on word2vec label similarity is characterized by comprising the following steps:
s1, calculating similarity between articles based on an ICF algorithm-cosRA;
s2, generating a recommendation list according to the similarity of the articles and the historical behaviors of the user;
s3, constructing a user-article bipartite graph according to the target user and the recommendation list, and establishing a bipartite graph model based on introduction of label nodes in the user-article bipartite graph, wherein the bipartite graph model specifically comprises the following steps:
constructing a user-article bipartite graph by the target user and 2N articles recommended by the cosRA, wherein the weight of each edge is provided by a recommendation coefficient obtained by the cosRA;
establishing a three-part graph model by introducing label nodes into a user-article two-part graph, and enabling G (u, I, T) to represent a user, an article and a label three-part graph, wherein V = u { [ U ] } V { (U { [ U ] } V } T { (U, I, T) } represents a user, an article and a label three-part graph i ∪V t 2N item sets V recommended by target user u, cosRA i And a label set V corresponding to the 2N items in the label data set t Composition of nodes V for target user u and item i There is a corresponding edge e (u, i) with a weight W u,i = cosRA (u, i), which is the recommendation coefficient given by the cosRA algorithm for the target user u, item i; for each binary group (i, t) in the graph, the graph has a corresponding edge e (i, t), and there is no edge directly connected between the user and the label;
s4, constructing a word2vec model which is trained, and taking a memory vocabulary as a corpus of the trained word2 vec;
s5, giving weights to edges among the labels of the three parts of the graph;
s6, random walk is carried out on the three-part graph from the user node, after the random walk is carried out for a plurality of times, the probability of each article node being visited converges to a number, the visit probability is the weight of the articles in the final recommendation list, and the method specifically comprises the following steps:
starting from the user node, random walk is carried out on the three-part graph, when the user node walks to any node, whether the walk is continued or stopped is determined according to the probability theta, and the walk is started from v u The user node starts to swim again;
if the walking is determined to continue, randomly selecting one node from the nodes pointed by the current node as the node passing the walking next time according to uniform distribution; after a plurality of random walks, the probability that each item node is visited converges to a number, and the weight of the item in the final recommendation list is the visit probability PR (v) of the item node:
wherein, out (v) represents a node set with edges connected with the node v, out (v ') is a node set with edges connected with the node v', PR (v ') is the visit probability of the node v' in the graph, the first N optimal item sets are taken as the final recommendation lists after the walking is finished, and S is the weight of the edges.
2. The word2vec tag similarity-based bipartite graph random walk recommendation method according to claim 1, wherein the ICF-based algorithm-cosRA in step S1 calculates similarity between articles as:
wherein the content of the first and second substances,
degree of agreement between alpha and beta of an object, K α ，K β Degree of items alpha, beta, respectively, if user i purchases item alpha, then a iα =1, otherwise a iα ＝0。
3. The word2vec tag similarity-based three-part graph random walk recommendation method according to claim 1, wherein a recommendation list is generated in the step S2 according to similarity of the articles and historical behaviors of the user, the length of the recommendation list is 2n, and n is 50.
4. The word2vec label similarity-based random walk recommendation method for the three-part graph according to claim 1, wherein the step S5 of giving weights to edges among the labels of the items in the three-part graph comprises the specific steps of:
scaling the tag similarity range between 0 and 1:
wherein min = -1,max = -1,w tt’ Similarity of the tag t and the tag t' calculated for word2 vec;
building a weight S of an edge between an item and a label i，t ：
Wherein N is i Is a set of labels corresponding to the item i in the data set, n t,i Number of times item i is labeled t, n t′i Number of times item i is labeled t', m i Is a set N i Length of (d);
calculated S i，t Needs to be scaled to between 0 and 1, i.e. S' i，t ：
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911241017.5A CN111079004B (en) | 2019-12-06 | 2019-12-06 | Three-part graph random walk recommendation method based on word2vec label similarity |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911241017.5A CN111079004B (en) | 2019-12-06 | 2019-12-06 | Three-part graph random walk recommendation method based on word2vec label similarity |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111079004A CN111079004A (en) | 2020-04-28 |
| CN111079004B true CN111079004B (en) | 2023-03-31 |
Family
ID=70312962
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911241017.5A Active CN111079004B (en) | 2019-12-06 | 2019-12-06 | Three-part graph random walk recommendation method based on word2vec label similarity |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111079004B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111723578B (en) * | 2020-06-09 | 2023-11-17 | 平安科技（深圳）有限公司 | Hot spot prediction method and device based on random walk model and computer equipment |
| CN112015987B (en) * | 2020-08-28 | 2022-10-04 | 青岛格兰德信用管理咨询有限公司 | Potential customer recommendation system and method based on enterprise tags |
| CN112214793A (en) * | 2020-09-30 | 2021-01-12 | 南京邮电大学 | Random walk model recommendation method based on fusion of differential privacy |
| CN112380417A (en) * | 2020-12-01 | 2021-02-19 | 厦门市美亚柏科信息股份有限公司 | Webpage recommendation method based on labels and graphs, terminal equipment and storage medium |
| CN112463989A (en) * | 2020-12-11 | 2021-03-09 | 交控科技股份有限公司 | Knowledge graph-based information acquisition method and system |
| CN112948238B (en) * | 2021-02-01 | 2023-05-02 | 成都信息工程大学 | Quantification method for diversity of recommendation system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101853470A (en) * | 2010-05-28 | 2010-10-06 | 浙江大学 | Collaborative filtering method based on socialized label |
| US8832091B1 (en) * | 2012-10-08 | 2014-09-09 | Amazon Technologies, Inc. | Graph-based semantic analysis of items |
| CN107369069A (en) * | 2017-07-07 | 2017-11-21 | 成都理工大学 | A kind of Method of Commodity Recommendation based on triangle area computation schema |
| CN108765042A (en) * | 2018-03-22 | 2018-11-06 | 浙江大学城市学院 | A kind of retail shop's proposed algorithm based on user Yu retail shop's network-physical space interbehavior |
| CN110020228A (en) * | 2019-04-08 | 2019-07-16 | 浙江大学城市学院 | A kind of relevance evaluation method for Internet of Things Item Information searching order |
-
2019
- 2019-12-06 CN CN201911241017.5A patent/CN111079004B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101853470A (en) * | 2010-05-28 | 2010-10-06 | 浙江大学 | Collaborative filtering method based on socialized label |
| US8832091B1 (en) * | 2012-10-08 | 2014-09-09 | Amazon Technologies, Inc. | Graph-based semantic analysis of items |
| CN107369069A (en) * | 2017-07-07 | 2017-11-21 | 成都理工大学 | A kind of Method of Commodity Recommendation based on triangle area computation schema |
| CN108765042A (en) * | 2018-03-22 | 2018-11-06 | 浙江大学城市学院 | A kind of retail shop's proposed algorithm based on user Yu retail shop's network-physical space interbehavior |
| CN110020228A (en) * | 2019-04-08 | 2019-07-16 | 浙江大学城市学院 | A kind of relevance evaluation method for Internet of Things Item Information searching order |
Non-Patent Citations (3)
| Title |
|---|
| Joseph Johnson;Yiu-Kai Ng.Using tripartite graphs to make long tail recommendations.《2017 8th International Conference on Information, Intelligence, Systems & Applications (IISA)》.2017,全文. * |
| 张艳梅;王璐;曹怀虎;毛国君.基于用户-兴趣-项目三部图的推荐算法.《模式识别与人工智能》.2015,全文. * |
| 雷鹏程;苟云倩;何强.基于关联规则的体检数据挖掘.《内蒙古科技与经济》.2018,全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111079004A (en) | 2020-04-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111079004B (en) | Three-part graph random walk recommendation method based on word2vec label similarity | |
| Wang et al. | Opportunity model for e-commerce recommendation: right product; right time | |
| CN104199896B (en) | The video similarity of feature based classification is determined and video recommendation method | |
| Liu et al. | Personalized travel package recommendation | |
| Gorla et al. | Probabilistic group recommendation via information matching | |
| Anwaar et al. | HRS-CE: A hybrid framework to integrate content embeddings in recommender systems for cold start items | |
| CN103824213A (en) | Individualized recommendation method based on user preferences and commodity properties | |
| CN110083764A (en) | A kind of collaborative filtering cold start-up way to solve the problem | |
| CN110851651A (en) | Personalized video recommendation method and system | |
| You et al. | QoS evaluation for web service recommendation | |
| Zhou et al. | Rank2vec: learning node embeddings with local structure and global ranking | |
| Hawashin et al. | Improving recommender systems using co-appearing and semantically correlated user interests | |
| Ai et al. | Online-rating prediction based on an improved opinion spreading approach | |
| Aliannejadi et al. | Venue suggestion using social-centric scores | |
| CN114358807A (en) | User portrayal method and system based on predictable user characteristic attributes | |
| CN110968675B (en) | Recommendation method and system based on multi-field semantic fusion | |
| Rupasingha et al. | Calculating Web service similarity using ontology learning with machine learning | |
| Wang et al. | Recommendation algorithm based on graph-model considering user background information | |
| Liao et al. | Jam: Joint attention model for next event recommendation in event-based social networks | |
| Sakai et al. | Topic set size design with the evaluation measures for short text conversation | |
| Zeng et al. | A recommendation algorithm for point of interest using time-based collaborative filtering | |
| Movahedian et al. | A tag-based recommender system using rule-based collaborative profile enrichment | |
| Nguyen et al. | Applying hidden topics in ranking social update streams on twitter | |
| Shang et al. | An improved video recommendations based on the hyperlink-graph model | |
| Basile et al. | Augmenting a content-based recommender system with tags for cultural heritage personalization |
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 |