CN112507245A - Social network friend recommendation method based on graph neural network - Google Patents

Social network friend recommendation method based on graph neural network Download PDF

Info

Publication number
CN112507245A
CN112507245A CN202011414193.7A CN202011414193A CN112507245A CN 112507245 A CN112507245 A CN 112507245A CN 202011414193 A CN202011414193 A CN 202011414193A CN 112507245 A CN112507245 A CN 112507245A
Authority
CN
China
Prior art keywords
node
attribute
user
neural network
graph
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
Application number
CN202011414193.7A
Other languages
Chinese (zh)
Other versions
CN112507245B (en
Inventor
魏哲巍
陈明
丁博麟
李雅亮
袁野
杜小勇
文继荣
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Renmin University of China
Original Assignee
Renmin University of China
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Renmin University of China filed Critical Renmin University of China
Priority to CN202011414193.7A priority Critical patent/CN112507245B/en
Publication of CN112507245A publication Critical patent/CN112507245A/en
Application granted granted Critical
Publication of CN112507245B publication Critical patent/CN112507245B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G06F16/9536—Search customisation based on social or collaborative filtering
    • 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
    • G06F16/901—Indexing; Data structures therefor; Storage structures
    • G06F16/9024—Graphs; Linked lists
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/047—Probabilistic or stochastic networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The invention discloses a social network friend recommendation method based on a graph neural network, which comprises the following steps: all users and relationships among the users are converted into a graph structure. And mapping all the user attribute information into a numerical vector to obtain an attribute matrix. And executing L-layer attribute transition probability calculation to obtain attribute information of the target user and the user to be recommended with the graph structure after aggregating the domain information. And inputting the attribute information into a deep neural network for coding, thereby obtaining coded information of the host node and the other nodes to be recommended, calculating a relevance score of the host node relative to each other node to be recommended according to the coded information, and recommending friends by taking the relevance score as a recommendation measuring standard. Therefore, the social network friend recommendation method based on the graph neural network improves the efficiency of social user relevance evaluation, and improves the friend recommendation speed.

Description

Social network friend recommendation method based on graph neural network
Technical Field
The invention relates to the technical field of computers, in particular to a social network friend recommendation method based on a graph neural network.
Background
The well-spraying development of the China Mobile Internet has prompted the explosive growth of the China Mobile social market. The data of the media advisory (iMedia Research) shows that the utilization rate of the Chinese instant messaging application in the last half of 2019 reaches 96.9%, the mobile social network represented by instant messaging becomes a domestic normal state, and the scale of users is expected to break through 9 hundred million in 2020 along with the development of the domestic mobile social ecology. The huge mobile social user size also means more market possibilities, and the younger new generation mainly after 95 and 00 is gradually the mastery force of the China mobile social market. The young users prefer an easy and interesting social form, and prefer a trendy, interesting and diversified social scene.
While mobile internet provides more information and services for users, the massive data makes information processing and filtering more complicated. On the one hand, users are easily lost in a large amount of information space and stranded; on the other hand, the website loses contact with the user and cannot establish a long-term stable cooperative relationship with the user. In this context, recommendation systems have come to mind. The recommendation system is a personalized system which analyzes interests and demands of users according to existing historical data and recommends information, products, services or other users with high relevance to the users. Wherein the friend recommendation aims to recommend users with high relevance but without established friend relationship for the users. This enables users (especially new users) to quickly establish a good circle of friendships, and to merge into the information service of the social network, thereby increasing user liveness and user stickiness.
In the evaluation process of the user relevance, in order to obtain a better evaluation result, people tend to utilize user information to complete the user relevance evaluation on a social network by training a deep neural network. Among the methods of deep learning, Graph Neural Networks (Graph Neural Networks) have been widely studied and applied due to their unique design for Graph structure data and accurate correlation evaluation results. After converting the social network into an abstract graph structure, the graph neural network can learn implicit connection patterns between users. For a given user, the relevance of the remaining users to the user can be calculated from the learned patterns.
The graph neural network is an important algorithm in the field of deep learning, and the core calculation process can be summarized as the following two points:
1. and propagating the node information to the neighborhood nodes.
2. And coding the information by utilizing the deep neural network.
The time consumption of the current calculation method of the graph neural network is generally dependent on the density of the social network. For a social network with a large average relevance between users, the information propagation speed of the graph neural network is low, and the efficiency of user relevance evaluation is limited to a certain extent, so that the recommendation speed of friends is influenced.
At present, the existing technical algorithm can realize relevance evaluation on a small-scale user group, but with the arrival of a big data era, the social network scale is larger and larger, such as WeChat, Twitter (Twitter) and the like of hundreds of millions of users, and the existing technical algorithm cannot effectively calculate recommendation results on a large graph due to huge time or space overhead, which greatly affects the quality and effect of a friend recommendation function.
The information disclosed in this background section is only for enhancement of understanding of the general background of the invention and should not be taken as an acknowledgement or any form of suggestion that this information forms the prior art already known to a person skilled in the art.
Disclosure of Invention
The invention aims to provide a social network friend recommendation method based on a graph neural network, which improves the efficiency of social user relevance evaluation, thereby improving the friend recommendation speed.
In order to achieve the above object, the present invention provides a social network friend recommendation method based on a graph neural network, comprising: all users and relationships among the users are converted into a graph structure. And mapping all the user attribute information into a numerical vector to obtain an attribute matrix. And executing L-layer attribute transition probability calculation to obtain attribute information of the target user and the user to be recommended with the graph structure after aggregating the domain information. And inputting the attribute information into a deep neural network for coding, thereby obtaining coded information of the host node and the other nodes to be recommended, calculating a relevance score of the host node relative to each other node to be recommended according to the coded information, and recommending friends by taking the relevance score as a recommendation measuring standard.
In an embodiment of the present invention, the graph structure includes nodes corresponding to users and edges corresponding to relationships between users, and the target user is a sink node of the graph structure.
In one embodiment of the present invention, performing the L-layer attribute transition probability calculation includes:
for the graph structure, the sampling process is repeatedly executed w times for the sink node and the node to be recommended. And performing an update process for each column of the attribute matrix, and combining the sampling process and the update process structure.
In one embodiment of the invention, repeatedly performing the sampling process w times for the sink node and the node to be recommended comprises estimating the probability of the random walk from the node u to the node v through 1 step by using the sampling
Figure BDA0002816560860000031
Where v is any node in the graph structure.
In one embodiment of the present invention, the updating process is performed for each column of the attribute matrix, and combining the sampling process and the updating process structure comprises: and judging whether the numerical value of each item (u, f) in a certain column of the attribute matrix meets a first preset condition or not. If the first preset condition is satisfied, the method utilizesAttribute transition probability value of first updating formula to current node u
Figure BDA0002816560860000032
Updating and clearing attribute values
Figure BDA0002816560860000033
Utilizing a second updating formula to carry out attribute value of a neighbor node v of the current node u at the next layer
Figure BDA0002816560860000034
And (6) updating. And after the updating is finished, combining the sampling process and the result obtained in the sampling process by using a third updating formula.
In an embodiment of the present invention, the first preset condition is:
Figure BDA0002816560860000035
the first update formula is:
Figure BDA0002816560860000036
the second update formula is:
Figure BDA0002816560860000037
wherein the content of the first and second substances,
Figure BDA0002816560860000038
the attribute transition probability of the current node u at the current layer l is 0 initially. Wherein the content of the first and second substances,
Figure BDA0002816560860000039
and the attribute value of the current node u at the current layer l is user information at the beginning. Wherein the content of the first and second substances,
Figure BDA00028165608600000310
and (5) transferring the probability for the updated attribute of the node v at the 1+1 st layer. And theta is an error parameter determined by a user according to the actual situation. Wherein d (u) is the number of neighbors of the node u. If the value of each item (u, f) does not satisfy the first preset condition, the updating process is ended.
Compared with the prior art, according to the social network friend recommendation method based on the graph neural network, the attribute transition probability of the user can be obtained under the sub-linear time complexity by combining the sampling process and the updating process, the dependency relationship between the transition probability calculation time and the social network consistency is eliminated, the social user correlation evaluation efficiency is improved, and the friend recommendation speed is improved.
Drawings
FIG. 1 is a flow chart of a social network friend recommendation method based on a graph neural network according to an embodiment of the present invention;
fig. 2 is a schematic flowchart of step S3 in fig. 1.
Detailed Description
The following detailed description of the present invention is provided in conjunction with the accompanying drawings, but it should be understood that the scope of the present invention is not limited to the specific embodiments.
Throughout the specification and claims, unless explicitly stated otherwise, the word "comprise", or variations such as "comprises" or "comprising", will be understood to imply the inclusion of a stated element or component but not the exclusion of any other element or component.
Fig. 1 is a flowchart illustrating a social network friend recommendation method based on a graph neural network according to an embodiment of the present invention. As shown in fig. 1, a social network friend recommendation method based on a graph neural network according to a preferred embodiment of the present invention includes:
s1, converting users, users and relationships among the users on the social network platform into a graph structure G, wherein the graph structure G comprises nodes corresponding to the users and edges corresponding to the relationships among the users, and a target user is a host node of the graph structure;
firstly, converting the user, the user and the relationship among the users into a graph structure G, wherein the graph structure G comprises all social users, namely a target user and other users, the social users correspond to nodes on the graph structure, the concern relationship among the social users corresponds to edges on the graph structure, the target user is a host node u of the graph structure, and the graph structure G comprises n nodes.
In the embodiment of the invention, the users refer to all registered users on the social platform, and the relationship among the users can be specifically the attention relationship among the users. For example, all registered buddies on Facebook and the buddy relationship network.
Specifically, for social networks with concern relationships, such as microblogs, facebooks, instagrams, and the like, users of the social networks correspond to nodes on the graph structure, and concern relationships among the users correspond to edges on the graph structure. In particular, if an a-user is interested in a B-user or a B-user is interested in an a-user, a bidirectional edge from the a-user node to the B-user node is established on the graph structure. The number of edges that a node has is referred to as "degree".
For the social networks with friend relationships, such as WeChat and QQ, users on the social networks correspond to graph nodes, and the friend relationships correspond to edges on graph structures. Specifically, if a buddy relationship exists between the a-user and the B-user (i.e., A, B are buddies of each other), a bidirectional edge is established from the a-user node to the B-user node on the graph structure.
When the graph structure is stored, an adjacency list is constructed for each node on the graph and is used for storing all neighbors of the node. For any node v on the graph, the length of the corresponding adjacent entry table is d (v), and the length of the corresponding adjacent entry table represents the degree of departure of the node v.
S2, mapping the user attribute information into a numerical vector to obtain an attribute matrix R(0)；
The method comprises the steps of transversely splicing various different types of information of a user into a vector with the length of F, wherein the user information comprises numerical value type or character type information, the numerical value type information can be directly used, and the character type information is mapped into the numerical value type information by using a word embedding technology. Using a two-dimensional matrix R of n rows and F columns(0)Information of all users is stored.
S3, executing L-layer attribute transition probability calculation to obtain attribute information P after neighborhood information is aggregated between a graph structure target user and a user to be recommended, wherein the user corresponds to a node on the graph structure;
and executing L-layer attribute transition probability calculation, and updating L according to the following formula until L is greater than a preset value L.
The update formula is: l + 1.
Wherein the preset value of 1 is 0.
Before step S3 is executed, each node on the graph structure maintains 3L vectors corresponding to the sampling probability, the residual value, and the attribute transition probability value of the node at each layer. For example, for any node v on the graph, it needs to maintain 3L vectors:
Figure BDA0002816560860000061
the attribute transition probability value of the corresponding node v at the 0 th, 1 st, … th and L-th layers,
Figure BDA0002816560860000062
the sampling probability value of the corresponding node v at the 0 th, 1 st, … th and L level,
Figure BDA0002816560860000063
and corresponding to the residual values of the node v at the 0 th, 1 st, … th and L-th layers. Removing device
Figure BDA0002816560860000064
All other values are initialized to 0, in addition to being initialized to encoded information.
In the process of calculating the L-layer attribute transition probability, the sampling probability, the attribute transition probability and the residual value of each node in the 1 layer are correspondingly updated. 1 is an intermediate variable used for marking the number of updating layers of the current numerical value and determining the stop time of the whole calculation process, and the weighted sum of the sampling probability, the attribute transition probability and the residual value of the previous L layers is used as the estimated value of the final attribute transition probability P after the stop
Figure BDA0002816560860000065
It can be shown that P and
Figure BDA0002816560860000066
the error between the two does not exceed theta, wherein theta is determined by a user according to actual conditionsAn error parameter.
And S4, inputting the attribute transition probability P obtained in the S3 into a deep neural network for coding, obtaining coded information of the host node and the other nodes to be recommended, calculating a relevance score of the host node relative to each other node to be recommended according to the coded information, and performing friend recommendation by taking the relevance score as a recommendation measuring standard.
Coding by using a deep neural network with the attribute transition probability P as input information to obtain coding vectors h of a host node u and the rest nodes v to be recommendeduAnd hvThe correlation pi of u and v is calculated according to the following formula(u，v)。
Dependence pi(u，v)The calculation formula of (2) is as follows: pi(u，v)＝hu·hv。
And finding the first t nodes with the highest correlation value with the host node u from all the nodes to perform friend recommendation, wherein t is a preset value and can be specified by a user and refers to the number of users recommending a target user.
In addition, for the calculation of the attribute transition probability, when w is set to
Figure BDA0002816560860000071
Theta is set as
Figure BDA0002816560860000072
Embodiments of the invention may be in
Figure BDA0002816560860000073
The calculation of the attribute transfer probability of the host node and the node to be recommended under the absolute error epsilon constraint threshold is completed within the time complexity, | C | is the total number of the host node and the node to be recommended in the corresponding graph structure, and d is the node average degree on the graph structure.
The existing attribute transition probability calculation method can only be used in
Figure BDA0002816560860000074
In the time of (2) to obtain a calculation result under the absolute error epsilon, or in O (Ln)dF) to obtain accurate results.
Fig. 2 is a schematic specific flowchart of step S3 in fig. 1, and as shown in fig. 2, step S3 specifically includes:
s31, for the graph structure, repeatedly executing w times of sampling process on the host node and the node to be recommended;
specifically, the sampling process is as follows: if the random walk starting from the node u reaches the node v through 1 step, the following formula is used
Figure BDA0002816560860000075
And (6) updating.
Figure BDA0002816560860000081
The update formula is:
Figure BDA0002816560860000082
and v is any node in the graph structure, and the random walk is a neighbor node which randomly walks to the current node with equal probability.
And S32, executing an updating process for each column of the attribute matrix, and combining the results of the sampling process and the updating process to obtain the attribute information of each user for subsequent calculation.
In an embodiment of the present invention, step S32 specifically includes:
s321, for the attribute matrix R of the certain layer 1(l)Judging whether the numerical value of each item (u, f) in the column meets a first preset condition one by one, if so, utilizing a first updating formula to transfer the probability numerical value of the attribute of the current node u
Figure BDA0002816560860000083
Update and purge
Figure BDA0002816560860000084
Using a second updating formula to determine the attribute value of the neighboring node v of u in the next layer
Figure BDA0002816560860000085
And (6) updating.
The first preset condition is as follows:
Figure BDA0002816560860000086
the first update formula is:
Figure BDA0002816560860000087
the second update formula is:
Figure BDA0002816560860000088
wherein the content of the first and second substances,
Figure BDA0002816560860000089
the attribute transition probability of the current node u at the current layer 1 is 0 initially,
Figure BDA00028165608600000810
the attribute value of the current node u at the current layer l is initially user information,
Figure BDA00028165608600000811
representing the attribute transition probability of the updated node v on the 1+1 st layer, theta is an error parameter determined by a user according to the actual situation, d (u) is the number of neighbors of the node u, and if the numerical values of all the items (u, f) do not meet the first preset condition, the updating process is ended;
s322, after the updating process is finished, combining the results obtained in the sampling process and the updating process by using a third updating formula;
the third update formula is:
Figure BDA0002816560860000091
wherein gamma islThe weight parameters determined by the user according to the actual conditions meet
Figure BDA0002816560860000092
Figure BDA0002816560860000093
Is an estimate of the L-layer attribute transition probability P.
In a word, the social network friend recommendation method based on the graph neural network can obtain the attribute transition probability of the user under the sub-linear time complexity by combining the sampling process and the updating process, gets rid of the dependency relationship between the transition probability calculation time and the social network consistency, improves the efficiency of social user correlation evaluation, and accordingly improves the friend recommendation speed.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The foregoing descriptions of specific exemplary embodiments of the present invention have been presented for purposes of illustration and description. It is not intended to limit the invention to the precise form disclosed, and obviously many modifications and variations are possible in light of the above teaching. The exemplary embodiments were chosen and described in order to explain certain principles of the invention and its practical application to enable one skilled in the art to make and use various exemplary embodiments of the invention and various alternatives and modifications as are suited to the particular use contemplated. It is intended that the scope of the invention be defined by the claims and their equivalents.

Claims (7)

1. A social network friend recommendation method based on a graph neural network is characterized by comprising the following steps:
converting all users and the relations among the users into a graph structure;
mapping all user attribute information into numerical vectors to obtain an attribute matrix;
executing L-layer attribute transition probability calculation to obtain attribute information of the target user and the user to be recommended after aggregating the domain information; and
and inputting the attribute information into a deep neural network for coding, thereby obtaining coded information of the host node and the other nodes to be recommended, calculating a relevance score of the host node relative to each other node to be recommended according to the coded information, and recommending friends by taking the relevance score as a recommendation measuring standard.
2. The graph neural network-based social network friend recommendation method of claim 1, wherein the graph structure comprises nodes corresponding to the users and edges corresponding to relationships between the users, and the target user is a sink node of the graph structure.
3. The graph neural network-based social network friend recommendation method of claim 2, wherein performing the L-layer attribute transition probability calculation comprises:
for the graph structure, repeatedly executing a sampling process for w times on the sink node and the node to be recommended; and
an update process is performed for each column of the attribute matrix, and the sampling process and update process structures are combined.
4. The graph neural network-based social network friend recommendation method of claim 3, wherein repeatedly performing w sampling processes on the sink node and the to-be-recommended node comprises estimating probability of a random walk starting from node u to reach node v through l steps by using sampling
Figure FDA0002816560850000011
Where v is any node in the graph structure.
5. The graph neural network-based social network friend recommendation method of claim 3, wherein an update process is performed for each column of the attribute matrix, and combining a sampling process and an update process structure comprises:
judging whether the numerical value of each item (u, f) in a certain column of the attribute matrix meets a first preset condition or not;
if the first preset condition is met, utilizing a first updating formula to transfer the probability value of the attribute of the current node u
Figure FDA0002816560850000021
Updating and clearing attribute values
Figure FDA0002816560850000022
Utilizing a second updating formula to carry out attribute value of a neighbor node v of the current node u at the next layer
Figure FDA0002816560850000023
Updating is carried out; and
and after the updating is finished, combining the sampling process and the result obtained in the sampling process by using a third updating formula.
6. The social network friend recommendation method based on graph neural network of claim 5, wherein the first preset condition is:
Figure FDA0002816560850000024
the first update formula is:
Figure FDA0002816560850000025
the second update formula is:
Figure FDA0002816560850000026
wherein the content of the first and second substances,
Figure FDA0002816560850000027
the attribute transition probability of the current node u at the current layer l is 0 initially;
wherein the content of the first and second substances,
Figure FDA0002816560850000028
the attribute value of the current node u in the current layer l is initially user information;
wherein the content of the first and second substances,
Figure FDA0002816560850000029
the attribute transition probability of the updated node v at the l +1 th layer is obtained;
wherein theta is an error parameter determined by a user according to an actual condition;
wherein d (u) is the number of neighbors of the node u;
if the value of each item (u, f) does not satisfy the first preset condition, the updating process is ended.
7. The graph neural network-based social network friend recommendation method of claim 5, wherein the third update formula is:
Figure FDA0002816560850000031
wherein, γlThe weight parameters determined by the user according to the actual conditions meet
Figure FDA0002816560850000032
Figure FDA0002816560850000033
Is an estimate of the L-layer attribute transition probability P.
CN202011414193.7A 2020-12-03 2020-12-03 Social network friend recommendation method based on graph neural network Active CN112507245B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011414193.7A CN112507245B (en) 2020-12-03 2020-12-03 Social network friend recommendation method based on graph neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011414193.7A CN112507245B (en) 2020-12-03 2020-12-03 Social network friend recommendation method based on graph neural network

Publications (2)

Publication Number Publication Date
CN112507245A true CN112507245A (en) 2021-03-16
CN112507245B CN112507245B (en) 2023-07-18

Family

ID=74970480

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011414193.7A Active CN112507245B (en) 2020-12-03 2020-12-03 Social network friend recommendation method based on graph neural network

Country Status (1)

Country Link
CN (1) CN112507245B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113407862A (en) * 2021-05-20 2021-09-17 桂林电子科技大学 Sequential social contact recommendation method, system and storage medium based on door mechanism
CN113901333A (en) * 2021-10-11 2022-01-07 东方财富信息股份有限公司 Friend recommendation method integrating graph structure and text features
CN114265988A (en) * 2021-12-20 2022-04-01 人工智能与数字经济广东省实验室(广州) Friend recommendation method based on importance score
CN114579877A (en) * 2022-03-09 2022-06-03 北京睿芯高通量科技有限公司 Graph neural network-based recommendation method and system with subtraction mechanism

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1498791A1 (en) * 2003-07-16 2005-01-19 United Technologies Corporation Model predictive control
CN104731962A (en) * 2015-04-03 2015-06-24 重庆邮电大学 Method and system for friend recommendation based on similar associations in social network
US20180232459A1 (en) * 2017-02-10 2018-08-16 Johnson Controls Technology Company Building management system with timeseries processing
CN109492770A (en) * 2018-10-08 2019-03-19 浙江大学 A kind of net with attributes embedding grammar based on the sequence of personalized relationship
CN110705304A (en) * 2019-08-09 2020-01-17 华南师范大学 Attribute word extraction method
CN111159483A (en) * 2019-12-26 2020-05-15 华中科技大学 Social network diagram abstract generation method based on incremental calculation
CN111372255A (en) * 2020-02-13 2020-07-03 北京联合大学 Neighborhood relation prediction method and system based on graph convolution neural network
CN111506833A (en) * 2020-06-12 2020-08-07 中国人民大学 Friend recommendation method based on single-source SimRank accurate solution
CN111581532A (en) * 2020-05-08 2020-08-25 深圳市万佳安人工智能数据技术有限公司 Social network friend-making recommendation method and system based on random block

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1498791A1 (en) * 2003-07-16 2005-01-19 United Technologies Corporation Model predictive control
CN104731962A (en) * 2015-04-03 2015-06-24 重庆邮电大学 Method and system for friend recommendation based on similar associations in social network
US20180232459A1 (en) * 2017-02-10 2018-08-16 Johnson Controls Technology Company Building management system with timeseries processing
CN109492770A (en) * 2018-10-08 2019-03-19 浙江大学 A kind of net with attributes embedding grammar based on the sequence of personalized relationship
CN110705304A (en) * 2019-08-09 2020-01-17 华南师范大学 Attribute word extraction method
CN111159483A (en) * 2019-12-26 2020-05-15 华中科技大学 Social network diagram abstract generation method based on incremental calculation
CN111372255A (en) * 2020-02-13 2020-07-03 北京联合大学 Neighborhood relation prediction method and system based on graph convolution neural network
CN111581532A (en) * 2020-05-08 2020-08-25 深圳市万佳安人工智能数据技术有限公司 Social network friend-making recommendation method and system based on random block
CN111506833A (en) * 2020-06-12 2020-08-07 中国人民大学 Friend recommendation method based on single-source SimRank accurate solution

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
C. GKANTSIDIS 等: "Random walks in peer-to-peer networks", 《IEEE INFOCOM 2004》, pages 130 *
李瑞雪: "基于多信息源图嵌入的好友推荐算法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》, no. 2019, pages 138 - 1317 *
王慧雪: "基于node2vec 的社区检测方法", 《计算机与数字工程》, vol. 48, no. 2, pages 403 - 408 *
陈婉杰 等: "基于网络表示学习的非单一维度的社区发现算法", 《计算机应用》, vol. 39, no. 12, pages 3467 - 3475 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113407862A (en) * 2021-05-20 2021-09-17 桂林电子科技大学 Sequential social contact recommendation method, system and storage medium based on door mechanism
CN113407862B (en) * 2021-05-20 2022-06-17 桂林电子科技大学 Sequential social contact recommendation method, system and storage medium based on door mechanism
CN113901333A (en) * 2021-10-11 2022-01-07 东方财富信息股份有限公司 Friend recommendation method integrating graph structure and text features
CN113901333B (en) * 2021-10-11 2024-05-28 东方财富信息股份有限公司 Friend recommendation method integrating graph structure and text characteristics
CN114265988A (en) * 2021-12-20 2022-04-01 人工智能与数字经济广东省实验室(广州) Friend recommendation method based on importance score
CN114579877A (en) * 2022-03-09 2022-06-03 北京睿芯高通量科技有限公司 Graph neural network-based recommendation method and system with subtraction mechanism

Also Published As

Publication number Publication date
CN112507245B (en) 2023-07-18

Similar Documents

Publication Publication Date Title
CN112507245A (en) Social network friend recommendation method based on graph neural network
CN109508419B (en) Knowledge learning-based recommendation method and system
CN107330798B (en) Method for identifying user identity between social networks based on seed node propagation
CN113792937B (en) Social network influence prediction method and device based on graph neural network
CN105956093B (en) A kind of personalized recommendation method based on multiple view anchor point figure Hash technology
CN112418525B (en) Method and device for predicting social topic group behaviors and computer storage medium
CN108647800B (en) Online social network user missing attribute prediction method based on node embedding
CN111428127B (en) Personalized event recommendation method and system integrating theme matching and bidirectional preference
CN113094593B (en) Social network event recommendation method, system, device and storage medium
CN112989169B (en) Target object identification method, information recommendation method, device, equipment and medium
CN112182424A (en) Social recommendation method based on integration of heterogeneous information and isomorphic information networks
CN111767472A (en) Method and system for detecting abnormal account of social network
CN109740106A (en) Large-scale network betweenness approximation method based on graph convolution neural network, storage device and storage medium
CN112667920A (en) Text perception-based social influence prediction method, device and equipment
CN110659394A (en) Recommendation method based on two-way proximity
CN110008411B (en) Deep learning interest point recommendation method based on user sign-in sparse matrix
CN110287424B (en) Collaborative filtering recommendation method based on single-source SimRank
CN109857928B (en) User preference prediction method based on multivariate credit evaluation
WO2018077301A1 (en) Account screening method and apparatus
CN117216376A (en) Fair perception recommendation system and recommendation method based on depth map neural network
CN110348469A (en) A kind of user's method for measuring similarity based on DeepWalk internet startup disk model
CN114493674A (en) Advertisement click rate prediction model and method
CN116932873A (en) Video account recommending method, device, equipment, storage medium and program product
CN113032685B (en) Object pushing method, device, equipment and storage medium based on social relationship
CN116089722B (en) Implementation method, device, computing equipment and storage medium based on graph yield label

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