CN117540111A - Preference perception socialization recommendation method based on graph neural network - Google Patents
Preference perception socialization recommendation method based on graph neural network Download PDFInfo
- Publication number
- CN117540111A CN117540111A CN202410028020.3A CN202410028020A CN117540111A CN 117540111 A CN117540111 A CN 117540111A CN 202410028020 A CN202410028020 A CN 202410028020A CN 117540111 A CN117540111 A CN 117540111A
- Authority
- CN
- China
- Prior art keywords
- user
- graph
- embedding
- order
- neural network
- 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
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 44
- 238000000034 method Methods 0.000 title claims abstract description 33
- 230000008447 perception Effects 0.000 title claims abstract description 16
- 230000003993 interaction Effects 0.000 claims abstract description 69
- 238000010586 diagram Methods 0.000 claims abstract description 18
- 239000013598 vector Substances 0.000 claims description 31
- 230000004931 aggregating effect Effects 0.000 claims description 24
- 239000011159 matrix material Substances 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 9
- 238000010276 construction Methods 0.000 claims description 8
- 238000004220 aggregation Methods 0.000 claims description 6
- 230000002776 aggregation Effects 0.000 claims description 5
- 230000006399 behavior Effects 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims description 3
- 230000002452 interceptive effect Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Abstract
The invention discloses a preference perception socialization recommendation method based on a graph neural network, which belongs to the technical field of socialization recommendation methods and comprises the following steps: s1: inputting social information of a user and interaction information between the user and an article; s2: constructing a user social graph and a user article interaction graph, and initializing user and article embedding; s3: determining a final first-order user embedding; s4: dividing the user object interaction diagram into a plurality of sub-diagrams; s5: determining a second or higher order user embedding; s6: determining first-order to high-order user embedding and article embedding to carry out weighted summation to obtain final user embedding and article embedding; s7: calculating the preference degree of the user on the articles by the inner product; s8: sorting and generating a recommendation list according to the descending order of the preference degree; the method solves the problems of cold start and data sparseness of the traditional recommendation model through the socialization recommendation method, solves the problem of low precision of the traditional socialization recommendation system, and improves the recommendation precision of the recommendation model.
Description
Technical Field
The invention belongs to the technical field of social recommendation methods, and particularly relates to a preference perception social recommendation method based on a graph neural network.
Background
With the rapid development and iterative upgrade of internet technology and terminal equipment, the information transmission and communication are greatly promoted, the activities of people on the network are also more and more abundant, the information quantity in the network is also exponentially and explosively increased, when the information is in a huge amount, users are easy to lose, and the interesting and useful information of the users is difficult to directly extract from the information, the phenomenon is called information overload, and the recommendation system can solve the information overload problem by providing the user with a list of items which the users can click or purchase in a personalized way, and the recommendation system is a core information filtering tool for online services including electronic commerce, social media and the like.
Because of the limitation of the internet technology and the data acquisition mode, the traditional recommendation system can only utilize the historical interaction information between the user and the articles as modeling information to establish a preference model of the user so as to recommend other articles possibly of interest to the user, but in many practical application scenes, new users can continuously appear, and because the new users lack the historical interaction information, the traditional recommendation system cannot accurately recommend the articles possibly of interest to the users, so that the users have the problem of poor recommendation experience, namely the cold start problem, and meanwhile, the sparsity problem of the interaction data of the user and the articles is more serious along with the increasing of the number of the articles, and in order to solve the problem, the traditional recommendation system introduces the social attribute information of the user, and is called a socialization recommendation system.
The current social recommendation system recommendation method generally uses information in a social network diagram as auxiliary information, performs simple splicing with information in a user-object interaction diagram, does not well exert the value of the social network diagram, only considers the information propagation of high-order neighbor nodes in the user-object interaction diagram, ignores many useless or even negative information in the information, greatly influences the recommendation performance, and reduces the accuracy of the social recommendation method.
In view of the above, the design of the preference perception socialization recommendation method based on the graph neural network not only can solve the problems of cold start and data sparseness of the traditional recommendation model, but also can solve the problem of low precision of the traditional socialization recommendation system.
Disclosure of Invention
To solve the problems set forth in the background art. The invention provides a preference perception socialization recommendation method based on a graph neural network, which has the characteristics of solving the problems of cold start and data sparseness of a traditional recommendation model and solving the problem of low precision of the traditional socialization recommendation system.
In order to achieve the above purpose, the present invention provides the following technical solutions: a preference perception socialization recommendation method based on a graph neural network comprises the following steps:
s1: inputting social information of a user and interaction information between the user and an article;
s2: constructing a user social graph according to the user social information, constructing a user article interaction graph according to the interaction information among the user articles, and initializing the embedding of the user and the articles;
s3: performing first-order graph convolution operation in the user social graph and the user object interaction graph by using a lightweight graph neural network, and then aggregating user embedments with interaction information and social information to obtain a final first-order user embedment;
s4: inputting the initial embedding of the user and the first-order user embedding into a sub-graph construction module, wherein the sub-graph construction module divides the user article interaction graph into a plurality of sub-graphs, and the classification basis is that users with similar preference and articles with interaction behaviors are classified;
s5: performing second-order or higher-order graph convolution operation on the user social graph and the user object interaction graph sub-graph by using a lightweight graph neural network, and then aggregating the user embedments of the two graphs to obtain second-order or higher-order user embedments;
s6: carrying out graph convolution operation on the user object interaction graph by utilizing a lightweight graph neural network to obtain first-order to high-order user embeddings and object embeddings, carrying out weighted summation on all the user embeddings to obtain final user embeddings, and carrying out weighted summation on all the object embeddings to obtain final object embeddings;
s7: calculating the preference degree of the user on the article by carrying out inner products on the end user embedding and the end article embedding;
s8: and sorting the items in descending order according to the preference degree of the user, and selecting the first N generated recommendation lists.
Further, in the step S3, the specific steps of performing the first-order graph convolution operation in the user social graph and the user object interaction graph by using the lightweight graph neural network are as follows:
s31: aggregating neighbor node information on the user social graph recursively using a lightweight graph neural network to obtain a first-order user embedding with social information：
Wherein:and->Respectively expressed as a set of neighbor nodes of users u and v having a direct social relationship in the social graph of said users, +.>An initial embedding denoted user v;
s32: aggregating neighbor node information on the user item interaction graph by utilizing the recursion of the lightweight graph neural network to obtain first-order user embedding with interaction information：
Wherein:and->Respectively expressed as a set of neighbor nodes with direct interaction with user u and item i in the user item interaction diagram, +.>Denoted as initial embedding of item i.
Further, in the step S3, the specific steps of embedding and aggregating the user with the interaction information and the social information are as follows: embedding the obtained first-order user with social informationAnd first order user embedding with interaction information +.>Aggregation to generate the final first-order user-embedded +.>The specific formula is as follows:
wherein:expressed as iteration number>、/>And->Expressed as a weight matrix>Denoted as->Function (F)>Expressed as vector concatenation operations,/->Denoted as L2 norm.
Further, in the step S4, the specific step of dividing the user object interaction diagram into a plurality of sub-diagrams by the sub-diagram construction module is:
s41: embedding the user initiallyAnd said first order user is embedded +>The fusion is as follows:
wherein:function (F)>Represented as bias term->Represented as a weight matrix;
s42: will beInputting the prediction vector into a two-layer neural network to obtain a final prediction vector:
wherein:and->Expressed as a weight matrix>And->Represented as bias term->The index of the maximum value of each row is the sequence number of the sub-graph of the user, the dimension of the predictive vector is consistent with the number of the pre-defined sub-graphs, and the unsupervised learning is used for classifying the user nodes.
Further, in the step S5, the specific steps of performing the second-order or higher-order graph rolling operation on the user social graph and the user object interaction graph sub-graph by using the lightweight graph neural network are as follows:
s51: aggregating neighbor node information on the user social graph by recursion of a lightweight graph neural network to obtain social informationStep user embedding->：
Wherein:and->Represented as a set of neighbor nodes of users u and v, respectively, in the user's social graph, +.>Expressed as +.>Embedding a layer;
s52: aggregating neighbor node information on the user item interaction graph subgraph by utilizing lightweight graph neural network recursion to obtain information with interaction informationStep user inlayEnter->：
Wherein:and->Represented as a set of neighbor nodes of user u and item i in their respective subgraphs, +.>Denoted as +.>Layer embedding.
Further, in the step S5, the specific steps of aggregating the user embedments of the two graphs are as follows:
the obtained social information-carrying informationStep user embedding->And said +.>Step user embeddingAggregation to generate the final second or higher order user-embedded +.>The specific formula is as follows:
wherein:expressed as iteration number>、/>And->Expressed as a weight matrix>Denoted as->Function (F)>Expressed as vector concatenation operations,/->Denoted as L2 norm.
Further, the specific steps of the step S6 are as follows:
s61: aggregating usersN users and object sets of (a)>The feature vectors of the M items in (a) are all initialized to one +.>Vector of dimension,/->The feature vectors of the individual users are represented as a matrix +.>User->Is expressed as +.>，/>The eigenvectors of the individual items are represented as a matrix +.>Article->Is expressed as +.>；
S62: the feature vectors of the user and the object passing through the first picture scroll layer are calculated, and the specific formula is as follows:
wherein:and->Represented as a set of neighbor nodes of users u and v in the social graph, respectively, +.>Expressed as user +.>Is->Layer feature vectors;
s63: through K-round propagation, user u embedding is obtained asAnd->And carrying out weighted summation on each layer of user embedding to obtain final user embedding, wherein the specific formula is as follows:
wherein:an embedded weight factor denoted as layer i, < >>A final embedding denoted user u;
s64: the article i is obtained by K-round propagation, isAnd carrying out weighted summation on each layer of object embedding to obtain a final object embedding, wherein the specific formula is as follows:
wherein:an embedded weight factor denoted as layer i, < >>Denoted as the final embedding of item i.
Further, in the step S7, the formula of the embedded inner product is calculated as follows:
wherein:represented as predicted preference of user u for item i, < >>Represented as the final embedding of user u within a preset time period,/->Denoted as the final embedding of item i for a preset period of time.
Compared with the prior art, the invention has the beneficial effects that:
according to the social recommendation method, the feature vectors of the users are enriched by acquiring the social information of the users in the social network diagram, and meanwhile, the user object interaction diagram is split into a plurality of sub-diagrams, so that the propagation of negative information can be effectively prevented, the problems of cold start and data sparseness of the traditional recommendation model are solved through the social recommendation method, the problem of low precision of the traditional social recommendation system is solved, and the recommendation precision of the recommendation model is improved.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, the present invention provides the following technical solutions: a preference perception socialization recommendation method based on a graph neural network comprises the following steps:
s1: inputting social information of a user and interaction information between the user and an article;
obtaining social information of a user: acquiring a binding relation between a user identifier and a login identifier in a social network application, wherein the user identifier is an identifier currently used in the social network application and comprises, but not limited to, a mobile terminal and a computer terminal identifier, sending a social information acquisition request of the user identifier to the social network application through the binding relation, and returning the social information of the user identifier comprising a name, a mailbox, a mobile phone, position information and time information to be requested by the social network application after the binding relation is verified successfully, so as to realize related acquisition;
and acquiring interaction information between the user and the article: acquiring interaction information between a user and an article based on the same method;
s2: constructing a user social graph according to the user social information, constructing a user article interaction graph according to the interaction information among the user articles, and initializing the embedding of the user and the articles;
building a user social graph: constructing a social network diagram according to the user social information，/>Represented asIf->Then indicate user +.>And user->With direct social connection, if->Then indicate user +.>And user->There is no direct social connection;
building a user article interaction diagram: constructing a user-article interaction diagram according to the interaction information between the user and the article，/>Denoted as->Wherein->And->Respectively as a set of users and items, if +.>Then indicate user +.>And articles->No interactive information is provided, if->Then indicate user +.>And articles->Interaction information is arranged between the two devices;
s3: performing first-order graph convolution operation in a user social graph and a user object interactive graph by using a lightweight graph neural network, and then aggregating user embedments with interactive information and social information to obtain a final first-order user embedment;
s31: aggregating neighbor node information on a user social graph using lightweight graph neural network recursion to obtain first-order user embedding with social information：
Wherein:and->Respectively expressed as a set of neighbor nodes of users u and v having a direct social relationship in the user's social graph, +.>An initial embedding denoted user v;
s32: aggregating neighbor node information on user item interaction graph by utilizing lightweight graph neural network recursion to obtain first-order user embedding with interaction information：
Wherein:and->Respectively expressed as a set of neighbor nodes with direct interaction with user u and item i in the user item interaction diagram,/or->Denoted as initial embedding of item i;
embedding the obtained first-order user with social informationAnd first order user embedding with interaction information +.>Aggregation to generate the final first-order user-embedded +.>The specific formula is as follows:
wherein:expressed as iteration number>、/>And->Expressed as a weight matrix>Denoted as->Function (F)>Expressed as vector concatenation operations,/->Expressed as L2 norm;
s4: inputting initial embedding and first-order user embedding of a user into a sub-graph construction module, wherein the sub-graph construction module divides a user article interaction graph into a plurality of sub-graphs, and the classification basis is users with similar preference and articles with interaction behaviors;
s41: embedding a user initiallyAnd first-order user embedding->The fusion is as follows:
wherein:denoted as->Function (F)>Represented as bias term->Represented as a weight matrix;
s42: will beInputting the prediction vector into a two-layer neural network to obtain a final prediction vector:
wherein:and->Expressed as a weight matrix>And->Represented as bias term->The index of the maximum value of each row is the sequence number of the subgraph of the user, the dimension of the predictive vector is consistent with the number of the predefined subgraphs, and the unsupervised learning is used for classifying the user nodes;
s5: performing second-order or higher-order graph convolution operation on the user social graph and the user object interaction graph subgraph by using a lightweight graph neural network, and then aggregating the user embedments of the two graphs to obtain second-order or higher-order user embedments;
s51: aggregating neighbor node information on a user social graph by utilizing lightweight graph neural network recursion to obtain social informationStep user embedding->：
Wherein:and->Represented as a set of neighbor nodes of users u and v in the user's social graph, respectively, +.>Expressed as +.>Embedding a layer;
s52: aggregating neighbor node information on user object interaction graph subgraphs by utilizing lightweight graph neural network recursion to obtain information with interaction informationStep user embedding->：
Wherein:and->Represented as a set of neighbor nodes of user u and item i in their respective subgraphs, +.>Denoted as +.>Embedding a layer;
the obtained social information-carrying deviceStep user embedding->And +.>Step user embedding->Aggregation to generate the final second or higher order user-embedded +.>The specific formula is as follows:
wherein:expressed as iteration number>、/>And->Expressed as a weight matrix>Denoted as->Function (F)>Expressed as vector concatenation operations,/->Expressed as L2 norm;
s6: carrying out graph convolution operation on the user object interaction graph by utilizing a lightweight graph neural network to obtain first-order to high-order user embeddings and object embeddings, carrying out weighted summation on all the user embeddings to obtain final user embeddings, and carrying out weighted summation on all the object embeddings to obtain final object embeddings;
s61: initializing feature vectors of N users and M objects as oneVector of dimension,/->The feature vectors of the individual users are represented as a matrix +.>User->Is expressed as +.>，/>The eigenvectors of the individual articles are represented as a matrixArticle->Is expressed as +.>；
S62: the feature vectors of the user and the object passing through the first picture scroll layer are calculated, and the specific formula is as follows:
wherein:and->Represented as a set of neighbor nodes of users u and v in the social graph, respectively, +.>Expressed as user +.>Is->Layer feature vectors;
s63: through K-round propagation, user u embedding is obtained asAnd->And carrying out weighted summation on each layer of user embedding to obtain final user embedding, wherein the specific formula is as follows:
wherein:an embedded weight factor denoted as layer i, < >>A final embedding denoted user u;
s64: the article i is obtained by K-round propagation, isAnd carrying out weighted summation on each layer of object embedding to obtain a final object embedding, wherein the specific formula is as follows:
wherein:an embedded weight factor denoted as layer i, < >>Denoted final inlay of item i;
s7: calculating the preference degree of the user on the article by carrying out inner products on the end user embedding and the end article embedding;
the formula for calculating the embedded inner product is as follows:
wherein:represented as predicted preference of user u for item i, < >>Represented as the final embedding of user u within a preset time period,/->Representing the final embedding of the object i within a preset time period;
s8: and sorting the items in descending order according to the preference degree of the items by the user, and selecting the first N generated recommendation lists.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (8)
1. The preference perception socialization recommendation method based on the graph neural network is characterized by comprising the following steps of:
s1: inputting social information of a user and interaction information between the user and an article;
s2: constructing a user social graph according to the user social information, constructing a user article interaction graph according to the interaction information among the user articles, and initializing the embedding of the user and the articles;
s3: performing first-order graph convolution operation in the user social graph and the user object interaction graph by using a lightweight graph neural network, and then aggregating user embedments with interaction information and social information to obtain a final first-order user embedment;
s4: inputting the initial embedding of the user and the first-order user embedding into a sub-graph construction module, wherein the sub-graph construction module divides the user article interaction graph into a plurality of sub-graphs, and the classification basis is that users with similar preference and articles with interaction behaviors are classified;
s5: performing second-order or higher-order graph convolution operation on the user social graph and the user object interaction graph sub-graph by using a lightweight graph neural network, and then aggregating the user embedments of the two graphs to obtain second-order or higher-order user embedments;
s6: carrying out graph convolution operation on the user object interaction graph by utilizing a lightweight graph neural network to obtain first-order to high-order user embeddings and object embeddings, carrying out weighted summation on all the user embeddings to obtain final user embeddings, and carrying out weighted summation on all the object embeddings to obtain final object embeddings;
s7: calculating the preference degree of the user on the article by carrying out inner products on the end user embedding and the end article embedding;
s8: and sorting the items in descending order according to the preference degree of the user, and selecting the first N generated recommendation lists.
2. The preference perception socialization recommendation method based on the graph neural network according to claim 1, wherein the method comprises the following steps: in the step S3, the specific steps of performing the first-order graph convolution operation in the user social graph and the user object interaction graph by using the lightweight graph neural network are as follows:
s31: aggregating neighbor node information on the user social graph recursively using a lightweight graph neural network to obtain a first-order user embedding with social information：
Wherein:and->Respectively expressed as a set of neighbor nodes of users u and v having a direct social relationship in the social graph of said users, +.>An initial embedding denoted user v;
s32: aggregating neighbor node information on the user item interaction graph by utilizing the recursion of the lightweight graph neural network to obtain first-order user embedding with interaction information：
Wherein:and->Respectively expressed as a set of neighbor nodes with direct interaction with user u and item i in the user item interaction diagram, +.>Denoted as initial embedding of item i.
3. The preference perception socialization recommendation method based on the graph neural network according to claim 1, wherein the method comprises the following steps: in the step S3, the specific steps of embedding and aggregating the user with the interaction information and the social information are as follows: embedding the obtained first-order user with social informationAnd first order user embedding with interaction information +.>Aggregation to generate the final first-order user-embedded +.>The specific formula is as follows:
wherein:expressed as iteration number>、/>And->Expressed as a weight matrix>Denoted as->Function (F)>Expressed as vector concatenation operations,/->Denoted as L2 norm.
4. The preference perception socialization recommendation method based on the graph neural network according to claim 1, wherein the method comprises the following steps: in the step S4, the specific steps of the sub-graph construction module dividing the user object interaction graph into a plurality of sub-graphs are as follows:
s41: will beInitial embedding of the userAnd said first order user is embedded +>The fusion is as follows:
wherein:denoted as->Function (F)>Represented as bias term->Represented as a weight matrix;
s42: will beInputting the prediction vector into a two-layer neural network to obtain a final prediction vector:
wherein:and->Expressed as a weight matrix>And->Represented as bias term->The index of the maximum value of each row is the sequence number of the sub-graph of the user, the dimension of the predictive vector is consistent with the number of the pre-defined sub-graphs, and the unsupervised learning is used for classifying the user nodes.
5. The preference perception socialization recommendation method based on the graph neural network according to claim 1, wherein the method comprises the following steps: in the step S5, the specific steps of performing the second-order or higher-order graph convolution operation on the user social graph and the user object interaction graph subgraph by using the lightweight graph neural network are as follows:
s51: aggregating neighbor node information on the user social graph by recursion of a lightweight graph neural network to obtain social informationStep user embedding->：
Wherein:and->Represented as a set of neighbor nodes of users u and v, respectively, in the user's social graph, +.>Expressed as +.>Embedding a layer;
s52: aggregating neighbor node information on the user item interaction graph subgraph by utilizing lightweight graph neural network recursion to obtain information with interaction informationStep user embedding->：
Wherein:and->Represented as a set of neighbor nodes of user u and item i in their respective subgraphs, +.>Denoted as +.>Layer embedding.
6. The preference perception socialization recommendation method based on the graph neural network according to claim 1, wherein the method comprises the following steps: in the step S5, the specific steps of embedding and aggregating the users of the two graphs are as follows:
will getTo with social informationStep user embedding->And +.>Step user embedding->Aggregation to generate the final second or higher order user-embedded +.>The specific formula is as follows:
wherein:expressed as iteration number>、/>And->Expressed as a weight matrix>Denoted as->Function (F)>Expressed as vector concatenation operations,/->Denoted as L2 norm.
7. The preference perception socialization recommendation method based on the graph neural network according to claim 1, wherein the method comprises the following steps: the specific steps of the step S6 are as follows:
s61: aggregating usersN users and object sets of (a)>The feature vectors of the M items in (a) are all initialized to one +.>Vector of dimension,/->The feature vectors of the individual users are represented as a matrix +.>User->Is expressed as +.>，/>The eigenvectors of the individual items are represented as a matrix +.>Article->Is expressed as +.>；
S62: the feature vectors of the user and the object passing through the first picture scroll layer are calculated, and the specific formula is as follows:
wherein:and->Represented as a set of neighbor nodes of users u and v in the social graph, respectively, +.>Expressed as user +.>A kind of electronic deviceLayer feature vectors;
s63: through K-round propagation, user u embedding is obtained asAnd->And carrying out weighted summation on each layer of user embedding to obtain final user embedding, wherein the specific formula is as follows:
wherein:an embedded weight factor denoted as layer i, < >>A final embedding denoted user u;
s64: the article i is obtained by K-round propagation, isAnd carrying out weighted summation on each layer of object embedding to obtain a final object embedding, wherein the specific formula is as follows:
wherein:an embedded weight factor denoted as layer i, < >>Denoted as the final embedding of item i.
8. The preference perception socialization recommendation method based on the graph neural network according to claim 1, wherein the method comprises the following steps: in the step S7, the formula of the embedded inner product is calculated as follows:
wherein:represented as predicted preference of user u for item i, < >>Denoted as the final embedding of user u within a preset period of time,denoted as the final embedding of item i for a preset period of time.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410028020.3A CN117540111B (en) | 2024-01-09 | 2024-01-09 | Preference perception socialization recommendation method based on graph neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410028020.3A CN117540111B (en) | 2024-01-09 | 2024-01-09 | Preference perception socialization recommendation method based on graph neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117540111A true CN117540111A (en) | 2024-02-09 |
| CN117540111B CN117540111B (en) | 2024-03-26 |
Family
ID=89792286
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410028020.3A Active CN117540111B (en) | 2024-01-09 | 2024-01-09 | Preference perception socialization recommendation method based on graph neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117540111B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111428147A (en) * | 2020-03-25 | 2020-07-17 | 合肥工业大学 | Social recommendation method of heterogeneous graph volume network combining social and interest information |
| US20210110436A1 (en) * | 2019-10-14 | 2021-04-15 | Visa International Service Association | Group item recommendations for ephemeral groups based on mutual information maximization |
| CN113407861A (en) * | 2021-05-19 | 2021-09-17 | 中国科学技术大学 | Social recommendation method based on decentralized graph neural network |
| CN113918833A (en) * | 2021-10-22 | 2022-01-11 | 重庆理工大学 | Product recommendation method realized through graph convolution collaborative filtering of social network relationship |
| CN114036406A (en) * | 2021-11-04 | 2022-02-11 | 南京大学 | Recommendation method and system based on graph contrast learning and social network enhancement |
| CN115203550A (en) * | 2022-07-14 | 2022-10-18 | 齐鲁工业大学 | Social recommendation method and system for enhancing neighbor relation |
| WO2022222037A1 (en) * | 2021-04-20 | 2022-10-27 | 中国科学院深圳先进技术研究院 | Interpretable recommendation method based on graph neural network inference |
| US20220414792A1 (en) * | 2021-06-24 | 2022-12-29 | Jinan University | Social recommendation method based on multi-feature heterogeneous graph neural networks |
| CN115659019A (en) * | 2022-10-21 | 2023-01-31 | 南京邮电大学 | Collaborative filtering recommendation method, system, device and medium based on lightweight graph neural network |
-
2024
- 2024-01-09 CN CN202410028020.3A patent/CN117540111B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20210110436A1 (en) * | 2019-10-14 | 2021-04-15 | Visa International Service Association | Group item recommendations for ephemeral groups based on mutual information maximization |
| CN111428147A (en) * | 2020-03-25 | 2020-07-17 | 合肥工业大学 | Social recommendation method of heterogeneous graph volume network combining social and interest information |
| WO2022222037A1 (en) * | 2021-04-20 | 2022-10-27 | 中国科学院深圳先进技术研究院 | Interpretable recommendation method based on graph neural network inference |
| CN113407861A (en) * | 2021-05-19 | 2021-09-17 | 中国科学技术大学 | Social recommendation method based on decentralized graph neural network |
| US20220414792A1 (en) * | 2021-06-24 | 2022-12-29 | Jinan University | Social recommendation method based on multi-feature heterogeneous graph neural networks |
| CN113918833A (en) * | 2021-10-22 | 2022-01-11 | 重庆理工大学 | Product recommendation method realized through graph convolution collaborative filtering of social network relationship |
| CN114036406A (en) * | 2021-11-04 | 2022-02-11 | 南京大学 | Recommendation method and system based on graph contrast learning and social network enhancement |
| CN115203550A (en) * | 2022-07-14 | 2022-10-18 | 齐鲁工业大学 | Social recommendation method and system for enhancing neighbor relation |
| CN115659019A (en) * | 2022-10-21 | 2023-01-31 | 南京邮电大学 | Collaborative filtering recommendation method, system, device and medium based on lightweight graph neural network |
Non-Patent Citations (4)
| Title |
|---|
| SHIWEN WU, FEI SUN, WENTAO ZHANG, XU XIE, BIN CUI: "Graph Neural Networks in Recommender Systems: A Survey", 《ACM COMPUTING SURVEYS》, 3 December 2022 (2022-12-03) * |
| 何昊晨;张丹红;: "基于多维社交关系嵌入的深层图神经网络推荐方法", 计算机应用, no. 10, 28 July 2020 (2020-07-28) * |
| 吴国栋;查志康;涂立静;陶鸿;宋福根: "图神经网络推荐研究进展", 《智能系统学报》, 20 March 2020 (2020-03-20) * |
| 李诗文;潘善亮;: "基于注意力机制的神经网络贝叶斯群组推荐算法", 计算机应用与软件, no. 05, 12 May 2020 (2020-05-12) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117540111B (en) | 2024-03-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110309427B (en) | Object recommendation method and device and storage medium | |
| CN111382309B (en) | Short video recommendation method based on graph model, intelligent terminal and storage medium | |
| CN103020116B (en) | The method of the powerful user of automatic screening on social media network | |
| CN112650929B (en) | Graph neural network recommendation method integrating comment information | |
| CN105069172A (en) | Interest label generating method | |
| CN110162711B (en) | Resource intelligent recommendation method and system based on network embedding method | |
| CN110069713B (en) | Personalized recommendation method based on user context perception | |
| CN113761359A (en) | Data packet recommendation method and device, electronic equipment and storage medium | |
| CN112579822A (en) | Video data pushing method and device, computer equipment and storage medium | |
| Gong et al. | Hybrid deep neural networks for friend recommendations in edge computing environment | |
| CN115964568A (en) | Personalized recommendation method based on edge cache | |
| Sun et al. | Graph force learning | |
| WO2020147259A1 (en) | User portait method and apparatus, readable storage medium, and terminal device | |
| CN111988668B (en) | Video recommendation method and device, computer equipment and storage medium | |
| CN112765490A (en) | Information recommendation method and system based on knowledge graph and graph convolution network | |
| CN117540111B (en) | Preference perception socialization recommendation method based on graph neural network | |
| CN112734519B (en) | Commodity recommendation method based on convolution self-encoder network | |
| CN116012161A (en) | Risk analysis method, device and equipment for user group | |
| CN113327154B (en) | E-commerce user message pushing method and system based on big data | |
| CN117033754A (en) | Model processing method, device, equipment and storage medium for pushing resources | |
| CN114547310A (en) | False news early detection method, system, equipment and medium | |
| CN114610967A (en) | Data augmentation method applied to user portrait field | |
| CN114461928A (en) | Session recommendation method based on graph neural network | |
| Tang | Artificial intelligence-based e-commerce platform based on SaaS and neural networks | |
| CN112699302A (en) | Real-time recommendation method based on geodesic distance |
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 |