CN112633559A - Social relationship prediction method and system based on dynamic graph convolutional neural network - Google Patents
Social relationship prediction method and system based on dynamic graph convolutional neural network Download PDFInfo
- Publication number
- CN112633559A CN112633559A CN202011439699.3A CN202011439699A CN112633559A CN 112633559 A CN112633559 A CN 112633559A CN 202011439699 A CN202011439699 A CN 202011439699A CN 112633559 A CN112633559 A CN 112633559A
- Authority
- CN
- China
- Prior art keywords
- nodes
- dynamic graph
- node
- neural network
- social relationship
- 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
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 12
- 230000014509 gene expression Effects 0.000 claims abstract description 45
- 230000008859 change Effects 0.000 claims abstract description 26
- 238000003062 neural network model Methods 0.000 claims abstract description 25
- 238000013528 artificial neural network Methods 0.000 claims abstract description 19
- 238000013507 mapping Methods 0.000 claims abstract description 14
- 238000003860 storage Methods 0.000 claims description 6
- 230000003595 spectral effect Effects 0.000 claims description 4
- 230000007704 transition Effects 0.000 claims description 4
- 235000008694 Humulus lupulus Nutrition 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 abstract description 9
- 230000000694 effects Effects 0.000 abstract description 6
- 230000008569 process Effects 0.000 abstract description 4
- 238000012549 training Methods 0.000 description 12
- 238000010586 diagram Methods 0.000 description 9
- 238000004590 computer program Methods 0.000 description 7
- 230000006870 function Effects 0.000 description 6
- 238000002474 experimental method Methods 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 230000004931 aggregating effect Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000009191 jumping Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- General Business, Economics & Management (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Marketing (AREA)
- Molecular Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Tourism & Hospitality (AREA)
- Primary Health Care (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The utility model discloses a social relationship prediction method and system based on dynamic graph convolution neural network, including: acquiring a social relationship network dynamic graph; inputting the social relationship network dynamic graph into a trained dynamic graph convolutional neural network model, mapping people into nodes, mapping the social relationship between people into edges, and predicting the social relationship between different people which is possibly generated in the future; when the dynamic graph convolution neural network model is trained, nodes at each time point are divided into influenced nodes of a plurality of levels, the embedding change amount of each influenced node is calculated, the embedding expression of each influenced node is updated, and the dynamic graph convolution neural network model is trained through the embedding expression update of the influenced nodes at each time point. The dynamic change process is added into the node update embedded expression, so that the calculation cost is saved, and a good prediction effect is achieved.
Description
Technical Field
The invention relates to the technical field of dynamic graph expression learning, in particular to a social relationship prediction method and system based on a dynamic graph convolution neural network.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
The graph is used as the most common expression of coding relationship and structure, and is suitable for a plurality of scenes such as bioinformatics, chemistry, recommendation systems, social networks and the like, so that the graph has wide application and has more and more research works. However, unlike the feature structure of an image or language in a compact grid form, the graph is topologically formed, which is not easily handled by common machine learning methods. With the development of deep learning, the graph neural network obtains good results in the aspect of graph expression, and particularly, the node expression is updated in a mode of aggregating neighbor node expression by a method based on a graph convolution network model, so that the effect is obvious.
In a real-world scenario, graphs are usually dynamic, and the connections of their nodes and edges are constantly changing. However, most of the existing model methods are directed at training and learning under a static graph condition, and embedded parameters are learned through an operation optimization algorithm (for example, gradient descent), so when the method is applied to a dynamic graph scene, expressions are regenerated at each time step, and such an updating manner can generate huge calculation cost and is time-consuming.
After entering the internet era, the social relationship between people becomes more frequent and tight through the connection of networks, and the social circle of people is not limited within a small range simply under the background of developed social tools and multiplied information propagation speed. The social relationship is predicted in advance, so that the efficiency of person-to-person communication in various scenes can be improved, and the satisfaction degree of user interaction is improved. On the other hand, social risks can be avoided in advance, and unsafe communication behaviors in various scenes can be assisted. For example, the scholars are no longer limited to the same academic institution, more and more academic collaboration occurs among strangers, national and regional limitations are broken, and matching is performed through similar research directions and interests, so that the possibility of the academic collaboration is greatly increased. When the social relationship is dynamically predicted through the existing graph convolution neural network, the problems of high calculation cost, long consumed time and low accuracy of calculation results exist.
Disclosure of Invention
In order to solve the above problems, the present disclosure provides a social relationship prediction method and system based on a dynamic graph convolution neural network, which adds a dynamic change process to node embedded expression update, saves calculation cost, and has a good prediction effect.
In order to achieve the purpose, the following technical scheme is adopted in the disclosure:
in a first aspect, a social relationship prediction method based on a dynamic graph convolution neural network is provided, which includes,
acquiring a social relationship network dynamic graph;
inputting the social relationship network dynamic graph into a trained dynamic graph convolutional neural network model, mapping people into nodes, mapping the social relationship between people into edges, and predicting the social relationship between different people which is possibly generated in the future;
when the dynamic graph convolution neural network model is trained, nodes at each time point are divided into influenced nodes of a plurality of levels, the embedding change amount of each influenced node is calculated, the embedding expression of each influenced node is updated, and the dynamic graph convolution neural network model is trained through the embedding expression update of the influenced nodes at each time point.
In a second aspect, a social relationship prediction system based on a dynamic graph convolution neural network is provided, including:
the acquisition module is used for acquiring a social relationship network dynamic graph;
the relation prediction module is used for inputting the social relation network dynamic graph into the trained dynamic graph convolutional neural network model, mapping people into nodes, mapping the social relation among people into edges, and predicting the social relation among different people which is possibly generated in the future;
when the dynamic graph convolution neural network model is trained, nodes at each time point are divided into influenced nodes of a plurality of levels, the embedding change amount of each influenced node is calculated, the embedding expression of each influenced node is updated, and the dynamic graph convolution neural network model is trained through the embedding expression update of the influenced nodes at each time point.
In a third aspect, an electronic device is provided, which includes a memory, a processor, and computer instructions stored in the memory and executed on the processor, where the computer instructions, when executed by the processor, perform the steps of a method for predicting social relationships based on a dynamic graph convolution neural network.
In a fourth aspect, a computer-readable storage medium is provided for storing computer instructions, which when executed by a processor, perform the steps of a method for social relationship prediction based on a dynamic graph convolution neural network.
Compared with the prior art, the beneficial effect of this disclosure is:
1. when the dynamic graph is embedded, dynamic change information is added into node embedding expression updating, high-low order difference of affected nodes is considered, the characteristic change information of the nodes is transmitted to the affected nodes with different orders to complete embedding expression updating, calculation cost caused by repeated training is avoided, and meanwhile, when the dynamic graph convolution neural network model is used for predicting social relations between different people which may be generated in the future, prediction efficiency and accuracy are improved.
Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate embodiments of the application and, together with the description, serve to explain the application and are not intended to limit the application.
Fig. 1 is a schematic diagram illustrating node-embedded expression updating of a dynamic graph convolution neural network model according to embodiment 1 of the present disclosure;
fig. 2 is a display of accuracy results of dynamic graph models such as a dynamic graph convolutional neural network model and a theoretical upper bound model thereof in a node classification experiment in embodiment 1 of the present disclosure.
Fig. 3 is a graph showing accuracy results of the public model and the baseline model in the node classification experiment according to embodiment 1 of the present disclosure.
The specific implementation mode is as follows:
the present disclosure is further described with reference to the following drawings and examples.
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
In the present disclosure, terms such as "upper", "lower", "left", "right", "front", "rear", "vertical", "horizontal", "side", "bottom", and the like indicate orientations or positional relationships based on those shown in the drawings, and are only relational terms determined for convenience in describing structural relationships of the parts or elements of the present disclosure, and do not refer to any parts or elements of the present disclosure, and are not to be construed as limiting the present disclosure.
In the present disclosure, terms such as "fixedly connected", "connected", and the like are to be understood in a broad sense, and mean either a fixed connection or an integrally connected or detachable connection; may be directly connected or indirectly connected through an intermediate. The specific meanings of the above terms in the present disclosure can be determined on a case-by-case basis by persons skilled in the relevant art or technicians, and are not to be construed as limitations of the present disclosure.
Example 1
The utility model discloses a social relation prediction method based on dynamic graph convolution neural network, which directly considers the influence of dynamic change on the embedded expression of the related nodes, thereby dynamically updating the expression of partial nodes, in order to accurately predict the social relation between different people which may be generated in the future through the existing social relation network, and comprises the following steps:
acquiring a social relationship network dynamic graph;
inputting the social relationship network dynamic graph into a trained dynamic graph convolutional neural network model, mapping people into nodes, mapping the social relationship between people into edges, and predicting the social relationship between different people which is possibly generated in the future;
when the dynamic graph convolution neural network model is trained, nodes at each time point are divided into influenced nodes of a plurality of levels, the embedding change amount of each influenced node is calculated, the embedding expression of each influenced node is updated, and the dynamic graph convolution neural network model is trained through the embedding expression update of the influenced nodes at each time point.
Further, the obtained time-varying social relationship network dynamic graph is divided into two sections according to the time relationship, the social relationship dynamic graph obtained in the first half time is used as a training set to train the dynamic graph convolution neural network model, and the social relationship dynamic graph obtained in the second half time is used as a test set to test the dynamic graph convolution neural network model.
Furthermore, the nodes at each time point are divided into multi-order affected nodes according to the influence distance, when the edge between the two nodes changes, the two nodes belong to the first-order affected node, and the node which is k-1 times of jumping away from the first-order affected node is the k-order affected node.
Further, the characteristic change at each time point is calculated, and the characteristic change is transmitted to each corresponding step of affected nodes according to the hierarchical sequence, so that the embedding change amount of each step of affected nodes is obtained.
Further, the feature change at a certain time point is calculated by subtracting the embedding sum of the disappeared neighbor nodes from the embedding sum of the newly appeared neighbor nodes.
Further, the updating mode of the embedded expression of the first-order affected node is as follows:
wherein
For the node embedding expression at time step t,
for node-embedded expression at time step t +1, W0Learnable parameters, W, representing the node's own state transitions1A state transition learning parameter representing the impact of an affected node on a certain neighbor node for a first order.
Furthermore, a spectral domain convolution method is adopted to update the embedded expression of the affected nodes of the second order and above.
And further, training a dynamic graph convolution neural network model through a graph reconstruction error according to the embedded expression of each order of affected nodes at each time point.
A social relationship prediction method based on a dynamic graph convolution neural network disclosed in this embodiment is described by taking a high-energy physical (HEP-TH) data set as an example.
This data set contains the partnership of paper authors at the high energy physics theory conference between 1993 and 2003.
And predicting the cooperation relationship of the future student thesis by adopting a dynamic graph convolution neural network model through the known cooperation relationship of the student thesis in the data set.
According to the known student paper cooperative relationship, the concrete steps of predicting the possible student paper cooperative relationship in the future by adopting the dynamic graph convolution neural network model are as follows:
step S1, a network of student paper collaboration relations is created for the papers published each month. Similarly, the corresponding scholars paper partnership network for 60 consecutive months is selected as the whole data in the data set.
In step S2, the first half of the time step is used to form the training set, and the second half of the time step is used as the test set, and the report result is the average of the predicted performance of the relationship at all time points in the test set.
Step S3, using the traditional graph convolution neural network as a static basic model, setting the node embedding dimension as 100, and setting the maximum order of the default updating as 2.
Step S4, using the embedded expression learned at time step t to evaluate the ability to predict the probability of scholars 'partnership that may be generated in the future at time step t +1, AUC (area enclosed by the subject' S working characteristic curve and coordinate axis) and F1-score (harmonic mean of accuracy and recall) are selected as indicators.
The accuracy of the prediction method disclosed by the embodiment and the accuracy of the baseline model in the node classification experiment are compared and analyzed, the accuracy result is shown in fig. 2, and the training time, AUC and F1-score index results are shown in fig. 3.
The embedding of the graph convolution neural network at each time step is calculated by the following procedure.
And (1) acquiring connection structure states and initial node characteristics of the graph at different time points.
Step (2) according to the change condition of the edge, dividing the nodes into different hierarchies according to the influence distance. For example, node v1And v2The connection relation of the edges between the two nodes is changed, then the two nodes belong to the first-order affected node, v1And v2The first-order neighbor node of (1) is a second-order affected node, and so on, the k-order affected node is a node which is k-1 hops away from the first-order affected node, as shown in fig. 1.
And (3) calculating the feature changes at different time points, and then transmitting the change information to each-step affected node according to the hierarchical sequence, thereby updating the features at different time points.
Wherein the characteristics at different time points vary
Subtracting the total of the embedded expressions of the influence caused by the disappeared neighbor nodes from the total of the embedded expressions of the influence caused by the newly appeared neighbor nodes, and calculating by the following formula:
where u is a neighbor node of v, Nt(v) And Nt+1(v) Respectively representing the neighbor nodes of the node v at time steps t and t +1,
representing the vector representation of node u at time t.
For the embedding of the first-order affected nodes, the updating mode is as follows:
wherein
For node embedding at time step t,
and embedding the nodes at the time step t + 1. W0Learnable parameters, W, representing the node's own state transitions1A state transition learning parameter representing the impact of an affected node on a certain neighbor node for a first order.
Two methods based on a space domain and a frequency domain are adopted for the high-order affected nodes, and the high-order affected nodes are affected nodes of the second order and above.
The updating mode based on the spatial domain is similar to that of the first-order affected node, and the embedding of the k-order affected node needs to be updated according to the embedding change quantity of the k-1-order node:
for ease of calculation, the update may be performed in the form of a matrix, as follows:
wherein
Showing the updating of the affected node of order k,
represents the amount of change of the embedded expression caused by the update of the influenced node of k order, W0Learnable parameters, W, representing the node's own state transitionskA state transition learning parameter representing an influence of the influenced node on a certain node for the k-th order.
The embedded quantity of the high-order affected nodes is updated by adopting a spectral domain convolution method, and considering that an updating mode based on a spatial domain cannot be popularized to all nodes, the overall updating mode is designed, and according to the approximate principle of the spectral domain convolution and the Chebyshev polynomial, the updating mode is designed as follows:
wherein Z istFor the feature expression of each node, WsTo approximate the parameters of the frequency domain convolution, INIs the identity matrix, D is the degree matrix of the graph, and A is the adjacency matrix of the graph.
Step S4: and carrying out dynamic graph convolution neural network model training through graph reconstruction errors according to finally obtained embedding amount expressions of each order of affected nodes at different time points so as to achieve the optimal effect.
The dynamic graph convolution neural network model is trained in an unsupervised training mode, and mainly aims at the loss function of structure maintenance as follows:
a neighbor node where u is v, u-Is a random point and σ is a sigmoid function.
Of course, supervision training can be performed according to task needs, and only corresponding loss needs to be added.
The method is suitable for embedding the dynamic learning graph on the basis of the graph convolution network model. The characteristic change information of the nodes is transmitted to the affected nodes with different orders to complete embedding and updating, so that the calculation cost caused by repeated training is avoided, the effect is obviously improved, and experiments show that the method is superior to the existing dynamic graph learning method.
Meanwhile, when the embedded expressions of the affected nodes of different orders are updated, the high-low order difference of the affected range is considered, and the embedded change amounts are sequentially distributed to the affected nodes of different orders according to the high-low difference, so that the node feature expression is more accurate, the accuracy of model training is further improved, and the effect of model prediction is improved.
The training process of the dynamic graph convolution neural network model provided by the disclosure is not only used for predicting social relations, but also used for predicting other tasks with unknown dynamic graph structures. The method can be applied to various downstream tasks, and the calculation time is shortened and the accuracy is improved.
The present disclosure also provides two different methods for updating the embedding amount of the high-order affected nodes, so that the updating method for the embedding amount of the high-order affected nodes can be popularized to all nodes.
Example 2
In this embodiment, a social relationship prediction system based on a dynamic graph convolution neural network is disclosed, which includes:
the acquisition module is used for acquiring a social relationship network dynamic graph;
the relation prediction module is used for inputting the social relation network dynamic graph into the trained dynamic graph convolutional neural network model, mapping people as nodes and mapping the social relation between people as edges, and predicting the social relation of different people which is possibly generated in the future;
when the dynamic graph convolution neural network model is trained, nodes at each time point are divided into influenced nodes of a plurality of levels, the embedding change amount of each influenced node is calculated, the embedding expression of each influenced node is updated, and the dynamic graph convolution neural network model is trained through the embedding expression update of the influenced nodes at each time point.
Example 3
In this embodiment, an electronic device is disclosed, which includes a memory, a processor, and computer instructions stored in the memory and executed on the processor, where the computer instructions, when executed by the processor, perform the steps of the method for predicting social relationships based on a dynamic graph convolution neural network disclosed in embodiment 1.
Example 4
In this embodiment, a computer readable storage medium is disclosed for storing computer instructions, which when executed by a processor, perform the steps of a method for social relationship prediction based on dynamic graph convolution neural network disclosed in embodiment 1.
The above description is only a preferred embodiment of the present application and is not intended to limit the present application, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present application shall be included in the protection scope of the present application.
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.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting the same, and although the present invention is described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications and equivalents may be made to the embodiments of the invention without departing from the spirit and scope of the invention, which is to be covered by the claims.
Claims (10)
1. A social relationship prediction method based on a dynamic graph convolutional neural network is characterized by comprising the following steps:
acquiring a social relationship network dynamic graph;
inputting the social relationship network dynamic graph into a trained dynamic graph convolutional neural network model, mapping people as nodes, mapping the person-to-person social relationship as edges, and predicting the social relationship between different people which is possibly generated in the future;
when the dynamic graph convolution neural network model is trained, nodes at each time point are divided into influenced nodes of a plurality of levels, the embedding change amount of each influenced node is calculated, the embedding expression of each influenced node is updated, and the dynamic graph convolution neural network model is trained through the embedding expression update of the influenced nodes at each time point.
2. The method as claimed in claim 1, wherein the nodes at each time point are divided into multi-order affected nodes according to the influence distance, when the edge between two nodes changes, the two nodes belong to the first-order affected node, and the node which is k-1 hops away from the first-order affected node is the k-order affected node.
3. The method as claimed in claim 1, wherein the feature change at each time point is calculated, and the feature change is transmitted to each corresponding affected node in a hierarchical order, so as to obtain the embedded change amount of each affected node.
4. The method of claim 1, wherein the feature change at a certain time point is calculated by subtracting the embedding sum of disappeared neighbor nodes from the embedding sum of newly appeared neighbor nodes.
5. The method for predicting social relationships based on the dynamic graph convolution neural network as claimed in claim 1, wherein the embedded expression of the first-order affected nodes is updated in a manner that:
wherein
For the node embedding expression at time step t,
for node-embedded expression at time step t +1, W0Learnable parameters, W, representing the node's own state transitions1A state transition learning parameter representing the impact of an affected node on a certain neighbor node for a first order.
6. The method as claimed in claim 1, wherein the spectral domain convolution method is used to update the embedded expression of the affected nodes of the second or higher order.
7. The method of claim 1, wherein the dynamic graph convolution neural network model is trained through graph reconstruction errors according to the embedded expression of each order of affected nodes at each time point.
8. A social relationship prediction system based on a dynamic graph convolutional neural network is characterized by comprising:
the acquisition module is used for acquiring a social relationship network dynamic graph;
the relation prediction module is used for inputting the social relation network dynamic graph into the trained dynamic graph convolutional neural network model, mapping people as nodes and mapping the social relation between people as edges, and predicting the social relation between different people which is possibly generated in the future;
when the dynamic graph convolution neural network model is trained, nodes at each time point are divided into influenced nodes of a plurality of levels, the embedding change amount of each influenced node is calculated, the embedding expression of each influenced node is updated, and the dynamic graph convolution neural network model is trained through the embedding expression update of the influenced nodes at each time point.
9. An electronic device comprising a memory and a processor, and computer instructions stored on the memory and executed on the processor, wherein the computer instructions, when executed by the processor, perform the steps of a method for social relationship prediction based on dynamic graph convolution neural network as claimed in any one of claims 1 to 7.
10. A computer-readable storage medium storing computer instructions which, when executed by a processor, perform the steps of a method for social relationship prediction based on dynamic graph convolution neural network as claimed in any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011439699.3A CN112633559B (en) | 2020-12-10 | 2020-12-10 | Social relationship prediction method and system based on dynamic graph convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011439699.3A CN112633559B (en) | 2020-12-10 | 2020-12-10 | Social relationship prediction method and system based on dynamic graph convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112633559A true CN112633559A (en) | 2021-04-09 |
| CN112633559B CN112633559B (en) | 2022-08-09 |
Family
ID=75309272
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011439699.3A Active CN112633559B (en) | 2020-12-10 | 2020-12-10 | Social relationship prediction method and system based on dynamic graph convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112633559B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200152289A1 (en) * | 2018-11-09 | 2020-05-14 | The Broad Institute, Inc. | Compressed sensing for screening and tissue imaging |
| CN111222054A (en) * | 2020-01-03 | 2020-06-02 | 中国计量大学 | Session social contact recommendation method based on context neighbor relation modeling |
| CN111462088A (en) * | 2020-04-01 | 2020-07-28 | 深圳前海微众银行股份有限公司 | Data processing method, device, equipment and medium based on graph convolution neural network |
| CN111582538A (en) * | 2020-03-25 | 2020-08-25 | 清华大学 | Community value prediction method and system based on graph neural network |
-
2020
- 2020-12-10 CN CN202011439699.3A patent/CN112633559B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200152289A1 (en) * | 2018-11-09 | 2020-05-14 | The Broad Institute, Inc. | Compressed sensing for screening and tissue imaging |
| CN111222054A (en) * | 2020-01-03 | 2020-06-02 | 中国计量大学 | Session social contact recommendation method based on context neighbor relation modeling |
| CN111582538A (en) * | 2020-03-25 | 2020-08-25 | 清华大学 | Community value prediction method and system based on graph neural network |
| CN111462088A (en) * | 2020-04-01 | 2020-07-28 | 深圳前海微众银行股份有限公司 | Data processing method, device, equipment and medium based on graph convolution neural network |
Non-Patent Citations (6)
| Title |
|---|
| JINGXIN LIU 等: "K-core based temporal graph convolutional network for dynamic graph", 《IEEE》 * |
| L CHEN 等: "Friend Recommendation Based on Multi-Social Graph Convolutional Network", 《IEEE》 * |
| W LIN 等: "Guardian: Evaluating Trust in Online Social Networks with Graph Convolutional Networks", 《IEEE INFOCOM 2020 - IEEE CONFERENCE ON COMPUTER COMMUNICATIONS》 * |
| YUYANG YE 等: "Identifying High Potential Talent: A Neural Network Based Dynamic Social Profiling Approach", 《IEEE》 * |
| 林锦发 等: "基于一维卷积神经网络的多特征社交网络流行度预测研究", 《现代计算机(专业版)》 * |
| 陈可佳 等: "条件变分时序图自编码器", 《计算机研究与发展》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112633559B (en) | 2022-08-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Stanton et al. | Inside the integrated assessment models: Four issues in climate economics | |
| CA3080050A1 (en) | Training tree-based machine-learning modeling algorithms for predicting outputs and generating explanatory data | |
| CN107391512B (en) | Method and device for predicting knowledge graph | |
| CN110414679A (en) | Model training method, device, electronic equipment and computer readable storage medium | |
| CN112187554B (en) | Operation and maintenance system fault positioning method and system based on Monte Carlo tree search | |
| CN104200045A (en) | Parallel computing method for distributed hydrodynamic model of large-scale watershed system | |
| CN112566093B (en) | Terminal relation identification method and device, computer equipment and storage medium | |
| CN113807520A (en) | Knowledge graph alignment model training method based on graph neural network | |
| CN109344969B (en) | Neural network system, training method thereof, and computer-readable medium | |
| CN111768096A (en) | Rating method and device based on algorithm model, electronic equipment and storage medium | |
| Liao et al. | Adaptive configuration for heterogeneous participants in decentralized federated learning | |
| CN112100450A (en) | Graph calculation data segmentation method, terminal device and storage medium | |
| CN112182306B (en) | Uncertain graph-based community discovery method | |
| CN116681104B (en) | Model building and realizing method of distributed space diagram neural network | |
| CN113194493B (en) | Wireless network data missing attribute recovery method and device based on graph neural network | |
| Couckuyt et al. | Towards efficient multiobjective optimization: multiobjective statistical criterions | |
| CN112633559B (en) | Social relationship prediction method and system based on dynamic graph convolutional neural network | |
| CN112396166A (en) | Graph convolution neural network training method and device based on mixed granularity aggregator | |
| CN115878096A (en) | Deep learning model unified application method, device, server and storage medium | |
| CN116341634A (en) | Training method and device for neural structure search model and electronic equipment | |
| CN115599990A (en) | Knowledge perception and deep reinforcement learning combined cross-domain recommendation method and system | |
| CN113722554A (en) | Data classification method and device and computing equipment | |
| CN112783747A (en) | Execution time prediction method and device for application program | |
| Gao et al. | Environmental protection of rural ecotourism using PSR and MDP models | |
| CN116527411B (en) | Data security intelligent protection model construction method and device and collaboration platform |
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 |