CN113052712B - Social data analysis method and system and storage medium - Google Patents

Social data analysis method and system and storage medium Download PDF

Info

Publication number
CN113052712B
CN113052712B CN202110246341.7A CN202110246341A CN113052712B CN 113052712 B CN113052712 B CN 113052712B CN 202110246341 A CN202110246341 A CN 202110246341A CN 113052712 B CN113052712 B CN 113052712B
Authority
CN
China
Prior art keywords
graph
data
resource
user
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.)
Active
Application number
CN202110246341.7A
Other languages
Chinese (zh)
Other versions
CN113052712A (en
Inventor
李明
张�林
黄昌勤
梁吉业
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.)
Zhejiang Normal University CJNU
Original Assignee
Zhejiang Normal University CJNU
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 Zhejiang Normal University CJNU filed Critical Zhejiang Normal University CJNU
Priority to CN202110246341.7A priority Critical patent/CN113052712B/en
Publication of CN113052712A publication Critical patent/CN113052712A/en
Application granted granted Critical
Publication of CN113052712B publication Critical patent/CN113052712B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • 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
    • 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
    • G06N3/08—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)
  • General Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Computing Systems (AREA)
  • Computational Linguistics (AREA)
  • Databases & Information Systems (AREA)
  • Evolutionary Computation (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Business, Economics & Management (AREA)
  • Molecular Biology (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The invention discloses a social data analysis method, a social data analysis system and a storage medium, and relates to the field of artificial intelligence. The social data analysis method comprises the following steps: obtaining social data, wherein the social data comprises user data and social resource data; processing the social data into graph data with semantic information; constructing a semantic graph convolution neural network model, wherein the semantic graph convolution neural network model is composed of a plurality of network residual modules; and processing the graph data by adopting the semantic graph convolutional neural network model to obtain a processing and analyzing result of the social data. Massive multi-source heterogeneous data can be processed, and the accuracy of the analysis result of the data is high.

Description

Social data analysis method and system and storage medium
Technical Field
The invention relates to the field of artificial intelligence, in particular to a social data analysis method, a social data analysis system and a storage medium.
Background
With the development of network technology, social platforms and networks are increased, and the generated social big data is also in explosive growth. The wide use of social networks generates unprecedented massive data, and the social data has the characteristics of more sources, more complex structures and the like, and is difficult to be utilized in a unified manner without processing. The general flow of social big data processing comprises data acquisition, data integration and data analysis. Data integration provides a suitable data source for a data analysis stage through a series of preprocessing such as database storage, data cleaning, conversion, dimension reduction and the like. However, the great diversity of social big data, which is usually from different sources and different types, increases the difficulty of data processing, and structured, semi-structured, and completely unstructured data make the data sources have heterogeneous, high-dimensional, and nonlinear data characteristics in different forms. The key approach to solving these problems is data integration, which aims to combine data from different sources and provide a unified view of the data to users, and how data representation is performed at the core of data integration.
In the related art, a Graph neural network (Graph neural networks) has obvious advantages in saving global structure information of a Graph and processing a complex problem of a structural relationship, so that the Graph neural network can process a semantic relationship between a complex structure of the relationship and captured data. Although the graph neural network can operate on any topological node, different neighbor nodes exist for processing each graph node, and the convolution filters share one weight matrix, so that the internal structure of the whole graph cannot be well utilized; (2) the convolutional neural network in the graph neural network can consider some local semantic information behind data, but cannot acquire semantic information which is not obviously represented in the graph, such as the semantic relation of global and local nodes; (3) although each node is used in the graph neural network, the receptive field area is too small, and when the network structure is deep, the efficiency of information exchange is influenced, and the learning of global characteristics is not facilitated. These deficiencies result in low accuracy in the analytical processing of multi-source heterogeneous social data.
Disclosure of Invention
In order to solve at least one of the above technical problems, the present invention provides a method, a system and a storage medium for analyzing social data, which can process massive multi-source heterogeneous data, and the accuracy of the analysis result of the data is high.
In a first aspect, an embodiment of the present invention provides a method for analyzing social data, including the following steps:
obtaining social data, wherein the social data comprises user data and social resource data;
processing the social data into graph data with semantic information;
constructing a semantic graph convolution neural network model, wherein the semantic graph convolution neural network model is composed of a plurality of network residual modules;
and processing the graph data by adopting the semantic graph convolutional neural network model to obtain a processing and analyzing result of the social data.
In some embodiments, the social data is in the form of a graph structure representation of:
G＝(U，I，E)；
wherein the content of the first and second substances,g represents social data; u ═ U1，u2，...，umRepresenting user nodes, wherein m represents the number of the user nodes; i ═ I1，i2，...，inRepresenting social resource nodes, and n represents the number of resource nodes;
Figure BDA0002964234100000021
representing the interaction between the user and the resource.
In some embodiments, processing the social data into graph data with semantic information comprises the steps of:
constructing a plurality of user resource bipartite graphs according to different sources of the social resources, wherein the user resource bipartite graphs are represented as follows:
G＝{(u，i，e)|u∈U，i∈I，e∈E}；
wherein, U represents a user set, I represents a resource set, E represents a set of edges between the user and the resource, and when the edge between the user and the resource is EuiWhen 1, the user u and the resource i generate social data interaction, euiWhen the value is 0, the user u and the resource i do not generate social data interaction;
and embedding the user data and the resource data into the user resource bipartite graph.
In some embodiments, said processing said social data into graph data with semantic information further comprises the steps of:
respectively inputting each user resource bipartite graph into a graph convolution neural network model to obtain a high-order characteristic matrix of each user resource bipartite graph;
aggregating the high-order characteristic matrix vertex of each user resource bipartite graph by utilizing global average pooling to obtain a representation vector of each user resource bipartite graph;
and forming graph data by using the representation vector of each user resource bipartite graph.
In some embodiments, the step of inputting each of the user resource bipartite graphs into a graph convolution neural network model to obtain a high-order feature matrix of each of the user resource bipartite graphs includes the following steps:
extracting node characteristics in the user resource bipartite graph to form an input characteristic matrix;
forming an adjacency matrix of the user resource bipartite graph according to the relationship among the nodes in the user resource bipartite graph;
and inputting the input feature matrix and the adjacency matrix into a graph convolution neural network model to obtain a high-order feature matrix.
In some embodiments, the atlas neural network model is represented as:
Z(0)＝X；
Z(j+1)＝σ(WZ(j)A)；
wherein A ∈ Rm*nRepresenting an adjacency matrix, m representing the number of user nodes, n representing the number of resource nodes, a degree matrix D representing the weight between the user nodes and the resource nodes, and based on the adjacency matrix A and the degree matrix D, the input characteristic matrix of the first layer of the graph convolution neural network model is
Figure BDA0002964234100000031
Layer j input feature matrix Z(j)∈Rm*d(j)The j-th layer outputs a high-order feature matrix z(j+1)∈Rm*d(j+1)，
Figure BDA0002964234100000032
Is a trainable weight matrix, sigma is an activation function, and d is used to describe the d-dimension characteristics of the node.
In some embodiments, the semantic graph convolutional neural network model comprises a plurality of layers of semantic graph convolutional neural networks, each layer of semantic graph convolutional neural network represented as:
Figure BDA0002964234100000033
wherein, the first and the second end of the pipe are connected with each other,
Figure BDA0002964234100000034
vertices representing the graph datai node representation after the jth layer semantic graph convolutional neural network,
Figure BDA0002964234100000035
representing a node representation of a vertex of the graph data before a level j semantic graph convolutional neural network, WzRepresenting trainable weight matrices, WzThe initial value is set to 0, and,
Figure BDA0002964234100000036
represents the affinity of vertex i and vertex l, K represents the number of nodes,
Figure BDA0002964234100000037
and (3) a representing vector of the semantic graph convolution neural network at the j-th layer of the vertex i of the graph data.
In some embodiments, the method of analyzing social data further comprises the steps of:
and visually displaying the processing and analyzing result of the social data.
In a second aspect, an embodiment of the present invention further provides a system for analyzing social data, including:
the system comprises an acquisition module, a processing module and a processing module, wherein the acquisition module is used for acquiring social data, and the social data comprises user data and social resource data;
the data processing module is used for processing the social data into graph data with semantic information;
the model construction module is used for constructing a semantic graph convolution neural network model, wherein the semantic graph convolution neural network model is composed of a plurality of network residual modules;
and the analysis module is used for processing the graph data by adopting the semantic graph convolutional neural network model to obtain a processing and analysis result of the social data.
In a third aspect, embodiments of the present invention further provide a computer storage medium, where a program executable by a processor is stored, and when the program is executed by the processor, the program is used to implement the social data analysis method as described in the foregoing first aspect.
The technical scheme of the invention at least has one of the following advantages or beneficial effects: firstly, multi-source heterogeneous social data of various network platforms are obtained, wherein the multi-source heterogeneous social data comprise user data and resource data, and then the complex multi-source heterogeneous social data are processed into graph data with semantic information in a unified mode; constructing a semantic graph convolution neural network model which is composed of a plurality of network residual modules and is connected among the plurality of network residual modules through non-local layer; and inputting the graph data into a semantic graph convolution neural network model to obtain a processing and analyzing result of the social data. According to the method, the multi-source heterogeneous social data problem is modeled into a semantic graph convolution task, and interactive social data of users and resources are processed into graph data with semantic information, so that a semantic graph convolution model obtains more accurate semantic expression, and the performance and the analysis accuracy of the model are improved.
Drawings
FIG. 1 is a flow chart of a method for analyzing social data provided in accordance with an embodiment of the present invention;
FIG. 2 is a schematic structural diagram of a convolutional neural network model of a semantic graph provided according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a visualization of user social data provided in accordance with an embodiment of the present invention;
fig. 4 is a schematic diagram of a result of a visualized social data analysis provided according to an embodiment of the present invention.
Detailed Description
The embodiments described in the embodiments of the present application should not be considered as limiting the present application, and all other embodiments obtained by those skilled in the art without making creative efforts shall fall within the protection scope of the present application.
In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict.
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. The terminology used herein is for the purpose of describing embodiments of the present application only and is not intended to be limiting of the application.
An embodiment of the present invention provides a method for analyzing social data, and referring to fig. 1, the method of the embodiment of the present invention includes, but is not limited to, step S110, step S120, step S130, and step S140.
Step S110, social data is obtained, wherein the social data comprises user data and social resource data.
In some embodiments, the social data is in the form of a graph structure representation of:
G＝(U，I，E)；
wherein G represents social data; u ═ U1，u2，...，umRepresenting user nodes, wherein m represents the number of the user nodes; i ═ I1，i2，...，inRepresenting social resource nodes, and n represents the number of resource nodes;
Figure BDA0002964234100000041
representing the interaction between the user and the resource.
Step S120, processing the social data into graph data with semantic information.
In some embodiments, the rules for processing social data into graph data with semantic information are: and carrying out aggregation operation on the data with the user node characteristics and the resource node characteristics by using a graph convolution neural network, continuously updating the characteristic representation of the nodes, combining the processed data from different sources, and capturing semantic information by the graph data at the moment.
Specifically, a user resource bipartite graph is respectively created according to interaction data of users and resources acquired from a plurality of social platforms and aiming at different data sources, and is used for simulating interaction relations between the users and the resources. A user resource bipartite graph is represented as:
G＝{(u，i，e)|u∈U，i∈I，e∈E}；
wherein, U represents a user set,i represents a resource set, E represents a set of edges between a user and a resource, and E represents an edge between a user and a resourceuiWhen 1, the user u interacts with the resource i to generate social data, euiIf 0, it means that the user u and the resource i do not interact, and no social data is generated. The more social data sources, the more complex the user resource bipartite graph structure, and then embedding the user data and resource data into the corresponding user resource bipartite graph. Illustratively, 4 different sources of structure data M ∈ M ═ { v, a, t, b } are used as indicators of different data sources, where v, a, t, b respectively represent multi-source heterogeneous social data generated when 4 different users and different resources perform social interaction. After the social data are represented by the user resource bipartite graph respectively, embedding the user ID and the resource ID into the corresponding user resource bipartite graph, wherein eu∈RuID embedding representing user u, ei∈RiAn ID representing resource i is embedded.
And then, inputting the constructed user resource bipartite graph into a graph convolution neural network model, and updating the embedded vector of the node according to the attribute of each neighbor node. N represents the number of user nodes, m represents the number of resource nodes, and a characteristic matrix is adopted to obtain
Figure BDA0002964234100000051
Representing the user node characteristics, the h-th row represents the characteristic vector of the user node h. Model introduction adjacency matrix A epsilon Rm*nTo represent a set of edges E, AijWeight, degree matrix D, D representing the edge between user node i and resource jij＝∑jAijBased on the adjacency matrix A and the degree matrix D, the input feature matrix of the layer 1 of the graph convolution neural network model is
Figure BDA0002964234100000052
Input feature matrix Z of j-th layer input feature matrix of graph convolution neural network model(j)∈Rm*d(j)The j layer output high-order characteristic matrix Z of the graph convolution neural network model(j+1)∈Rm*d(j+1)And follows:
Z(0)＝X；
Z(j+1)＝σ(WZ(j)A)；
wherein the content of the first and second substances,
Figure BDA0002964234100000053
is a trainable weight matrix and σ is an activation function, such as the Relu function.
According to some specific embodiments of the present invention, after obtaining the high-order feature matrix through the graph convolution neural network model, selecting global average pooling to aggregate vertices on the bipartite graph of each user resource to obtain a representation vector of the bipartite graph of each user resource, and forming graph data by using the representation vector of the bipartite graph of each user resource. It should be noted that, during the global average pooling operation, the average operation is performed on the features of all nodes in the user resource bipartite graph, and this aggregation method is based on the assumption: different neighbor nodes have different contribution degrees to the node, namely, the user node and the resource node are also influenced by the neighbor nodes.
In particular, the graph-convolution neural network model incorporates interactive data of users and resources to enrich node representations, because such data can not only describe interests, usage habits, etc. of users, but also capture behavioral similarities between nodes of the same type. For a user node in a user resource bipartite graph, the influence of neighbor nodes is quantified by using an aggregation function f, and the output is expressed as:
hm＝f(Nu)；
wherein N isuAnd { j | (u, j) ∈ G } represents a neighbor node of the user u, j represents an interaction resource with the user u, and G represents a user resource bipartite graph.
Expressing function f by averagingavgTaking average pooling operation on node characteristics in a certain specific structure, and carrying out nonlinear transformation:
Figure BDA0002964234100000061
wherein the content of the first and second substances,
Figure BDA0002964234100000062
is a representation of a node j of the modality m,
Figure BDA0002964234100000063
is a trainable transformation matrix and Leakyrelu is a non-linear activation function.
According to some embodiments of the present invention, a new binding layer is proposed for a graph convolution neural network model, and the structural information, intrinsic information, and data source of a node are integrated into a unified representation by the following formulas:
Figure BDA0002964234100000064
wherein the content of the first and second substances,
Figure BDA0002964234100000065
is a user representation of node u when the data source is m, uidIs the embedding of a user ID, hmRepresenting structural information of the node.
U is to bemAnd uidProjection into the underlying space yields:
um＝LeakyReLU(W2，mjum)+uid；
wherein the content of the first and second substances,
Figure BDA00029642341000000611
is a trainable weight matrix, and u ismInto an ID embedding space.
The connection mode by nonlinear conversion includes:
gco(hm，um，uid)＝LeakyReLU(W3，m(hm||um))；
where, | | denotes a join operation, W3，mRepresenting trainable model parameters.
And S130, constructing a semantic graph convolution neural network model, wherein the semantic graph convolution neural network model is composed of a plurality of network residual modules.
Particularly, capturing global and long-term relationships between nodes can effectively solve the problem of small receptive field in graph volume. However, conventional graph convolution network networks compute the degree of interaction between users and resources based on the representation of the nodes of the users and resources, which limits the node feature update mechanism and cannot learn new convolution filters. Therefore, a semantic graph convolution neural network model is constructed according to the concept of the non-local mean value, and the operation of each layer of semantic graph convolution neural network of the semantic graph convolution neural network model is as follows:
Figure BDA0002964234100000066
wherein the content of the first and second substances,
Figure BDA0002964234100000067
a node representation of a vertex i representing the graph data after the level j semantic graph convolutional neural network,
Figure BDA0002964234100000068
representing a node representation of a vertex of the graph data before a level j semantic graph convolutional neural network, WzRepresenting trainable weight matrices, WzThe initial value is set to be 0, and,
Figure BDA0002964234100000069
represents the affinity of vertex i and vertex l, K represents the number of nodes,
Figure BDA00029642341000000610
and (3) a representing vector of the semantic graph convolution neural network at the j-th layer of the vertex i of the graph data. The structure of the semantic graph convolution neural network model can effectively capture the semantic relation between local nodes and global nodes. The structure of the semantic graph convolutional neural network model is shown in fig. 3, the semantic graph convolutional neural network model comprises a plurality of network residual error modules, wherein the network residual error modules comprise two SemGConv layers with 128 channels and a non-local layer, and the non-local layer is a non-local layerAnd a part layer for deepening a network structure by repeated use. At the start of the network, the SemGConv layer in one network residual module is responsible for mapping the input vectors into the potential space, while the SemGConv layer in the next adjacent network residual module is responsible for projecting the encoded features back into the output space. In addition, except for the last layer, all semantic graph convolutional neural network layers need to be subjected to batch processing standardization and ReLU function activation. The semantic graph convolution neural network model can be regarded as a message transmission system, and the information exchange efficiency is improved by processing messages in turn in stages. The two stages of information forward propagation are: the messages are updated locally and are refined by the global state of the system.
And step S140, processing the graph data by adopting the semantic graph convolutional neural network model to obtain a processing and analyzing result of the social data.
The graph data after the multi-graph aggregation can be used in a semantic graph convolution neural network to obtain a processing and analyzing result of the social data. It should be noted that, on the basis of the graph convolution neural network, a learnable weight matrix M e R is addedk*kThen the formula Z in the graph convolution before(j+1)＝σ(WZ(j)A) The evolution becomes:
Z(j+1)＝σ(WZ(j)ρi(M⊙A))；
wherein ρiThe Softmax nonlinear transformation of all input feature matrices for vertex i, an operation is a matrix element operation.
After training of the semantic graph neural network model, the above formula can be further expanded by learning the weight matrix M, and different weight matrices are applied to the output node characteristics:
Figure BDA0002964234100000071
where | | | denotes a join operation, WdIs the d-th row of the parameter matrix.
According to some embodiments of the invention, the method for analyzing social data further comprises:
and S150, visually displaying the processing and analyzing result of the social data.
Specifically, the accuracy of the social data analysis method of the scheme can be visually checked conveniently by visually displaying the processing and analysis results of the social data, and illustratively, the social preference condition of each user is analyzed according to social big data generated by each user by using different social resources. First, 5 users were randomly drawn and their used social resources were collected. The social data representation of the user is then visualized in a two-dimensional graph using t-distributed random neighbor embedding. And finally, inputting the social data of 5 users into a semantic graph convolution neural network model and then sequentially outputting the classification results.
Referring to FIG. 3, the left square block diagram represents structured data, the right square block diagram represents unstructured data, each point in the diagram represents data generated by user and resource interaction, and the different grayscales of the points represent different users. The data of the oval box represents data sources, and the data sources comprise video social resources and text social resources. The classification result is shown in fig. 4, some users prefer to use the video social resource, and some users prefer to use the text social resource. The visualization can reflect the source and the structure of data more intuitively, and the aggregation of different interactive data can analyze the preference of a user on a certain social resource.
An embodiment of the present invention further provides a system for analyzing social data, including:
the acquisition module is used for acquiring social data, wherein the social data comprises user data and social resource data.
And the data processing module is used for processing the social data into graph data with semantic information.
And the model construction module is used for constructing a semantic graph convolution neural network model, wherein the semantic graph convolution neural network model is composed of a plurality of network residual modules.
And the analysis module is used for processing the graph data by adopting the semantic graph convolutional neural network model to obtain a processing and analysis result of the social data.
An embodiment of the present invention also provides a computer-readable storage medium storing computer-executable instructions for execution by one or more control processors, e.g., to perform the steps described in the above embodiments.
One of ordinary skill in the art will appreciate that all or some of the steps, systems, and methods disclosed above may be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on computer readable media, which may include computer storage media (or non-transitory media) and communication media (or transitory media). The term computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data, as is well known to those of ordinary skill in the art. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, Digital Versatile Disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can accessed by a computer. In addition, communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media as known to those skilled in the art.
While the preferred embodiments of the present invention have been described in detail, it will be understood by those skilled in the art that the foregoing and various other changes, omissions and deviations in the form and detail thereof may be made without departing from the scope of this invention.

Claims (7)

1. A method for analyzing social data is characterized by comprising the following steps:
obtaining social data, wherein the social data comprises user data and social resource data;
constructing a plurality of user resource bipartite graphs according to different sources of the social resources, wherein the user resource bipartite graphs are represented as follows:
G＝{(u，i，eui)|u∈U，i∈I，eui∈E}；
wherein G represents a user resource bipartite graph, U represents a user set, I represents a resource set, E represents a set of edges between a user and a resource, and when an edge E between the user and the resource is a set of edgesuiWhen 1, the user u and the resource i generate social data interaction, euiWhen the value is 0, the user u and the resource i do not generate social data interaction;
embedding the user data and the resource data into the user resource bipartite graph;
respectively inputting each user resource bipartite graph into a graph convolution neural network model to obtain a high-order characteristic matrix of each user resource bipartite graph;
aggregating the high-order characteristic matrix vertex of each user resource bipartite graph by utilizing global average pooling to obtain a representation vector of each user resource bipartite graph;
forming graph data using the representation vector of each of the user resource bipartite graphs;
constructing a semantic graph convolution neural network model, wherein the semantic graph convolution neural network model is composed of a plurality of network residual modules;
processing the graph data by adopting the semantic graph convolutional neural network model to obtain a processing and analyzing result of the social data;
the operation definition of the semantic graph convolutional neural network of each layer of the semantic graph convolutional neural network model is as follows:
Figure FDA0003609803570000011
wherein the content of the first and second substances,
Figure FDA0003609803570000012
a node representation of a vertex i representing the graph data after the level j semantic graph convolutional neural network,
Figure FDA0003609803570000013
representing a node representation of a vertex of the graph data before a level j semantic graph convolutional neural network, WzRepresenting trainable weight matrices, WzThe initial value is set to be 0, and,
Figure FDA0003609803570000014
represents the affinity of vertex i and vertex l, K represents the number of nodes,
Figure FDA0003609803570000015
and (3) representing the vector representation of the semantic graph convolution neural network of the vertex l of the graph data at the j-th layer.
2. The method for analyzing social data according to claim 1, wherein the social data is represented in a graph structure, and the representation of the social data is as follows:
G′＝(U，I，E)；
wherein G' represents the social data; u ═ U1，u2，...，um}，umRepresenting user nodes, wherein m represents the number of the user nodes; i ═ I1，i2，...，in}，inRepresenting social resource nodes, wherein n represents the number of the resource nodes;
Figure FDA0003609803570000016
representing the interaction between the user and the resource.
3. The method for analyzing social data according to claim 2, wherein the step of inputting each of the user resource bipartite graphs into a graph convolution neural network model to obtain a high-order feature matrix of each of the user resource bipartite graphs comprises the steps of:
extracting node characteristics in the user resource bipartite graph to form an input characteristic matrix;
forming an adjacency matrix of the user resource bipartite graph according to the relationship among the nodes in the user resource bipartite graph;
and inputting the input feature matrix and the adjacency matrix into a graph convolution neural network model to obtain a high-order feature matrix.
4. The method of analyzing social data of claim 3, wherein the graph convolutional neural network model is represented as:
Z(0)＝X；
Z(j+1)＝σ(WZ(j)A)；
wherein A ∈ Rm*nRepresenting an adjacency matrix, wherein m represents the number of user nodes, n represents the number of resource nodes, a degree matrix D represents a weight between the user nodes and the resource nodes, and based on the adjacency matrix A and the degree matrix D, an input characteristic matrix of a layer 1 of the graph convolutional neural network model is
Figure FDA0003609803570000021
The jth layer input feature matrix Z of the graph convolution neural network model(j)∈Rm*d(j)The j layer output high-order characteristic matrix Z of the graph convolution neural network model(j+1)∈Rm*d(j+1)，W∈Rd(j)*d(j+1)Is a trainable weight matrix, sigma is an activation function, and d is used for describing the dimensional characteristics of the nodes.
5. The method for analyzing social data according to claim 4, further comprising the steps of:
and visually displaying the processing and analyzing result of the social data.
6. A system for analyzing social data, comprising:
the system comprises an acquisition module, a processing module and a display module, wherein the acquisition module is used for acquiring social data, and the social data comprises user data and social resource data;
the data processing module is used for constructing a plurality of user resource bipartite graphs according to different sources of the social resources, embedding the user data and the resource data into the user resource bipartite graphs, respectively inputting each user resource bipartite graph into a graph convolution neural network model to obtain a high-order characteristic matrix of each user resource bipartite graph, aggregating the high-order characteristic matrix vertex of each user resource bipartite graph by utilizing global average pooling to obtain a representation vector of each user resource bipartite graph, and forming graph data by utilizing the representation vector of each user resource bipartite graph;
the model construction module is used for constructing a semantic graph convolution neural network model, wherein the semantic graph convolution neural network model is composed of a plurality of network residual modules;
the analysis module is used for processing the graph data by adopting the semantic graph convolutional neural network model to obtain a processing and analysis result of the social data;
wherein, the user resource bipartite graph is represented as:
G＝{(u，i，eui)|u∈U，i∈I，eui∈E}；
wherein G represents a user resource bipartite graph, U represents a user set, I represents a resource set, E represents a set of edges between a user and a resource, and when an edge E between the user and the resource is a set of edgesuiWhen 1, the user u and the resource i generate social data interaction, euiWhen the value is 0, the user u and the resource i do not generate social data interaction;
the operation definition of the semantic graph convolutional neural network of each layer of the semantic graph convolutional neural network model is as follows:
Figure FDA0003609803570000031
wherein the content of the first and second substances,
Figure FDA0003609803570000032
a node representation of a vertex i representing the graph data after the jth layer semantic graph convolutional neural network,
Figure FDA0003609803570000033
representing a node representation of a vertex of the graph data before a level j semantic graph convolutional neural network, WzRepresenting trainable weight matrices, WzThe initial value is set to be 0, and,
Figure FDA0003609803570000034
represents the affinity of vertex i and vertex l, K represents the number of nodes,
Figure FDA0003609803570000035
and (3) representing the vector representation of the semantic graph convolution neural network of the vertex l of the graph data at the j-th layer.
7. A computer storage medium in which a processor-executable program is stored, wherein the processor-executable program, when executed by the processor, is for implementing a method of social data analysis as claimed in any one of claims 1 to 5.
CN202110246341.7A 2021-03-05 2021-03-05 Social data analysis method and system and storage medium Active CN113052712B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110246341.7A CN113052712B (en) 2021-03-05 2021-03-05 Social data analysis method and system and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110246341.7A CN113052712B (en) 2021-03-05 2021-03-05 Social data analysis method and system and storage medium

Publications (2)

Publication Number Publication Date
CN113052712A CN113052712A (en) 2021-06-29
CN113052712B true CN113052712B (en) 2022-05-31

Family

ID=76510344

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110246341.7A Active CN113052712B (en) 2021-03-05 2021-03-05 Social data analysis method and system and storage medium

Country Status (1)

Country Link
CN (1) CN113052712B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113609306B (en) * 2021-08-04 2024-04-23 北京邮电大学 Social network link prediction method and system for anti-residual diagram variation self-encoder

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110851491A (en) * 2019-10-17 2020-02-28 天津大学 Network link prediction method based on multiple semantic influences of multiple neighbor nodes

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CA3061667C (en) * 2018-11-15 2024-03-12 Element Ai Inc. Systems and methods for assessing item compatibility
CN110659723B (en) * 2019-09-03 2023-09-19 腾讯科技（深圳）有限公司 Data processing method and device based on artificial intelligence, medium and electronic equipment
CN111784081B (en) * 2020-07-30 2022-03-01 南昌航空大学 Social network link prediction method adopting knowledge graph embedding and time convolution network

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110851491A (en) * 2019-10-17 2020-02-28 天津大学 Network link prediction method based on multiple semantic influences of multiple neighbor nodes

Also Published As

Publication number Publication date
CN113052712A (en) 2021-06-29

Similar Documents

Publication Publication Date Title
CN109948029B (en) Neural network self-adaptive depth Hash image searching method
CN112529168B (en) GCN-based attribute multilayer network representation learning method
Li et al. Consensus, polarization and clustering of opinions in social networks
CN111967972B (en) Financial product recommendation method and device
CN111784081A (en) Social network link prediction method adopting knowledge graph embedding and time convolution network
CN111563192B (en) Entity alignment method, device, electronic equipment and storage medium
CN111651671B (en) User object recommendation method, device, computer equipment and storage medium
CN113962358B (en) Information diffusion prediction method based on time sequence hypergraph attention neural network
CN115688913A (en) Cloud-side collaborative personalized federal learning method, system, equipment and medium
CN114265986A (en) Information pushing method and system fusing knowledge graph structure and path semantics
CN115293919A (en) Graph neural network prediction method and system oriented to social network distribution generalization
CN113052712B (en) Social data analysis method and system and storage medium
US11922336B2 (en) Architecture and method for providing insights in networks domain
US20240078473A1 (en) Systems and methods for end-to-end machine learning with automated machine learning explainable artificial intelligence
CN114463596A (en) Small sample image identification method, device and equipment of hypergraph neural network
US20240037133A1 (en) Method and apparatus for recommending cold start object, computer device, and storage medium
Zhang et al. Knowledge graph driven recommendation model of graph neural network
CN109977194B (en) Text similarity calculation method, system, device and medium based on unsupervised learning
CN116955846A (en) Cascade information propagation prediction method integrating theme characteristics and cross attention
CN116484016A (en) Time sequence knowledge graph reasoning method and system based on automatic maintenance of time sequence path
US11947503B2 (en) Autoregressive graph generation machine learning models
Le Masson et al. Designing in an innovative design regime—introduction to CK design theory
CN110442690B (en) Query optimization method, system and medium based on probabilistic reasoning
Li et al. Learning roles with emergent social value orientations
Chen et al. Social recommendation using quantified social tie strength

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