Social network trust evaluation method and system based on multi-view graph neural network

Abstract

The application discloses a social network trust evaluation method and system based on a multi-view graph neural network, and comprises the following steps: acquiring social network behavior information, constructing different-view social behavior network graphs, and constructing a user trust relationship network graph according to trust relationships between users; based on the different-view social behavior network graphs, a multi-head attention mechanism is built to extract the influence of different types of neighbor nodes on the characteristics of a target user node and to fuse the influence, so as to output social attribute characteristics; based on the user trust relationship network graph, the attention mechanism is used to fuse neighbor information of nodes, so as to form trust characteristics of the user nodes; and the social attribute characteristics and the trust characteristics of the user nodes are combined, multi-view feature fusion is carried out, and a predicted trust relationship is output. The application solves the problems that a traditional trust evaluation method depends on a pre-defined model, has poor universality, and an existing trust relationship prediction scheme based on a graph neural network faces the problems that a trust relationship graph is difficult to acquire and feature fusion effect is poor.

Description

Technical Field

[0001] This invention belongs to the field of trust assessment technology, and specifically relates to a method and system for trust assessment of social networks based on multi-view graph neural networks. Background Technology

[0002] Trust assessment is the process of evaluating trust by quantifying attributes that influence trust. It has numerous applications in sensor networks, social networks, mobile crowdsourcing, and blockchain. Traditional trust assessment schemes are based on heuristic trust theory, typically fusing information from user attributes and user ratings using a predefined trust model to obtain a comprehensive user trust score or trust relationships between users. The effectiveness of such methods heavily relies on the predefined model, lacks universality, and cannot flexibly apply to different scenarios. To address the limitations of traditional trust assessment algorithms, some researchers have applied machine learning algorithms to trust assessment. Machine learning models automate the processing of trust-related data, extracting user trust features and predicting trust relationships between users. Traditional machine learning-based trust assessment algorithms typically rely on matrix factorization theory. They represent the trust relationships between users as a matrix, decompose the matrix to obtain node trust features and trusted features, and then perform matrix multiplication to obtain the complete trust relationship matrix. For example, Fazeli et al. explored the effectiveness of different measurement methods in matrix factorization-based trust relationship evaluation and proposed a matrix factorization-based trust evaluation algorithm based on implicit trust scores. By incorporating implicit trust scores into matrix factorization instead of explicit trust relationships, they demonstrated through experiments that the model can achieve similar performance without explicit trust scores. Similar work has also been applied in [6]. Matrix factorization-based methods have shown good effectiveness, but because they are based on shallow models, they cannot mine the deep and complex features of nodes and are difficult to effectively discover the complex nonlinear mapping relationship between user trust features and trust relationships.

[0003] To address the aforementioned problems, some scholars have introduced deep learning models for trust feature extraction and trust relationship prediction. By modeling trust relationships as graphs and processing them using graph convolutional neural networks (GCNNs), these models have achieved remarkable results. For example, Guardian uses GCNNs to process and fuse two types of user trust features separately, constructing a trust relationship prediction scheme based on GCNNs. Building upon this, KGTrust further integrates other social networks besides trust relationship networks, expanding the trust relationship graph and achieving certain performance improvements. GATTrust integrates trust features from different domains based on an attention mechanism to improve trust prediction performance. Some works further consider the dynamic nature of trust, constructing dynamic trust relationship prediction mechanisms based on dynamic graph neural networks, such as TrustGuard. These works have effectively solved the trust relationship prediction problem.

[0004] However, the above work has the following problems:

[0005] 1. Most existing trust relationship prediction schemes based on graph neural networks are geared towards trust relationship graphs. However, in practical applications, trust relationship graphs are very difficult to obtain, especially trust relationship networks with sufficient density to support accurate trust assessment.

[0006] 2. Some works that supplement the missing trust relationships based on external knowledge (such as KGTrust) fuse trust features and social attribute features through attention mechanisms. However, the two have a high degree of heterogeneity in feature distribution. Although direct fusion can enrich trust features, it cannot achieve complementary enhancement of features. It is difficult to effectively extract consistent information hidden in social attributes and trust features. Summary of the Invention

[0007] The purpose of this invention is to provide a method and system for assessing trust in social networks based on multi-view graph neural networks, in order to solve the above-mentioned problems.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] In a first aspect, the present invention provides a social network trust evaluation method based on a multi-view graph neural network, comprising:

[0010] Acquire social network behavior information, construct social behavior network graphs from different perspectives, and construct user trust relationship network graphs based on the trust relationships between users;

[0011] Based on social behavior network graphs from different perspectives, a multi-head attention mechanism is built to extract and fuse the influence of different types of neighbor nodes on the characteristics of target user nodes, and output social attribute features.

[0012] Based on the user trust relationship network graph, the attention mechanism is used to fuse the neighbor information of nodes to form the trust characteristics of user nodes;

[0013] By combining social attribute features and user node trust features, multi-view feature fusion is performed to output predicted trust relationships.

[0014] Furthermore, the acquisition of social network behavior information and the construction of social behavior network graphs from different perspectives include:

[0015] Collect social data of target users across different social behaviors / platforms;

[0016] Each social behavior / platform is processed separately to create user relationship graphs from different perspectives. Centered on the user, the interaction relationships between the user and social platform objects are collected, and these interactive objects are treated as nodes, with the interaction relationships as edges, thus constructing the social relationship graph from that perspective. , ;

[0017] Based on the interaction type, node attributes, and other information, different processing methods are used to obtain the results. Initial attribute features of nodes and edge attribute features.

[0018] Furthermore, the step of constructing a user trust relationship network graph based on the trust relationships between users includes:

[0019] Standardize the trust characteristics between users to form The system identifies discrete trust levels and models the trust relationships between users as a directed weighted graph.

[0020] Initialize out-degree trust features based on the user's trust relationships with other users and other users' trust relationships with that user. and in-degree trust features To form a complete trust relationship network diagram .

[0021] Furthermore, based on the social behavior network graph from different perspectives, a multi-head attention mechanism is constructed to extract and fuse the influence of different types of neighbor nodes on the characteristics of the target user node, outputting social attribute features, including:

[0022] Given a node From the perspective Corresponding diagram Neighbor set as ,node , and nodes The side note Features are denoted as ,but Attention parameters The following formula is used for calculation:

[0023]

[0024] Get attention score Then, given the first Layer output is Then the first The layer output is:

[0025]

[0026] Based on the attention mechanism, a multi-head attention mechanism is further designed, namely, including... Each independent attention extracts features and performs feature fusion. Assume the _th _ ... The features extracted by the attention mechanism are Then the first The layer output is:

[0027]

[0028] When all is completed After feature extraction from the layer, the final feature is output. .

[0029] Furthermore, the process of fusing neighbor information of nodes based on the user trust relationship network graph using an attention mechanism to form the trust features of user nodes includes:

[0030] Set up A graph convolutional neural network with layers of user initial in-degree trust features. Initial features Assuming the first The output of the layer is ,if There exists right Trust edge Then it is believed yes For the in-degree neighbors, the attention score is calculated as follows:

[0031]

[0032] Based on this score, the fused features are calculated:

[0033]

[0034] go through After training the layer, output its in-degree features.

[0035] Set up A graph convolutional neural network with layers of user initial in-degree trust features. Initial features Assuming the first The output of the layer is ,if There exists right Trust edge Then it is believed yes For out-degree neighbors, the attention score is calculated as follows:

[0036]

[0037] Based on this score, the fused features are calculated:

[0038]

[0039] go through After training the layer, output its in-degree features.

[0040] Trust Feature Fusion and Output: and After concatenation, input a fully connected network and an activation layer, and output... Trust characteristics .

[0041] Furthermore, the process of combining social attribute features and user node trust features to perform multi-view feature fusion and output predicted trust relationships includes:

[0042] by As a user The From one perspective, forming a user Features from a single perspective, namely In summary:

[0043]

[0044] Next, a trust feature fusion module based on multi-view learning is constructed to integrate the features. Integration into a unified feature And construct reconstruction loss Used for subsequent model training;

[0045] Trust relationship prediction based on multilayer perceptron: given a user and users and its characteristics , The input is then fed into a multilayer perceptron to predict the trust relationship. .

[0046] Furthermore, the loss function:

[0047] View-related feature encoder : Construct a multi-layer deep encoder based on neural networks, and then... User characteristics of each view Encoding as latent features ;

[0048] users The user fused feature is obtained by fusing the features of each view. ;

[0049] View-related feature decoder A multi-layer deep decoder is constructed based on a neural network, forming an autoencoder structure with the aforementioned deep encoder, and the fused features are then processed. Decode to reconstruct the first Features of a view ;

[0050] Constructing a reconstruction loss based on an autoencoder Its definition is as follows:

[0051]

[0052] A set of edges is extracted from the trust relationships for end-to-end training, using real trust relationships. To perform supervision, the loss function is constructed as follows:

[0053]

[0054] At the same time, combining the loss function in (5), the final loss function is constructed as follows:

[0055] .

[0056] Secondly, the present invention provides a social network trust evaluation system based on a multi-view graph neural network, comprising:

[0057] The data acquisition module is used to acquire social network behavior information, construct social behavior network graphs from different perspectives, and construct user trust relationship network graphs based on the trust relationships between users.

[0058] The social attribute feature acquisition module is used to build a multi-head attention mechanism based on social behavior network graphs from different perspectives to extract the influence of different types of neighbor nodes on the features of target user nodes and fuse them to output social attribute features.

[0059] The trust feature acquisition module is used to fuse the neighbor information of nodes based on the user trust relationship network graph using an attention mechanism to form the trust features of user nodes.

[0060] The trust relationship output module is used to combine social attribute features and user node trust features to perform multi-view feature fusion and output predicted trust relationships.

[0061] Thirdly, the present invention provides a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the social network trust evaluation method based on a multi-view graph neural network.

[0062] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the social network trust evaluation method based on a multi-view graph neural network.

[0063] Compared with the prior art, the present invention has the following technical effects:

[0064] This technical solution proposes a social network trust evaluation method based on multi-view graph neural networks, aiming to solve the problems of traditional trust evaluation methods relying on predefined models and having poor universality, as well as the difficulties in obtaining trust relationship graphs and poor feature fusion effects faced by existing trust relationship prediction schemes based on graph neural networks.

[0065] By constructing user social behavior network graphs from different perspectives, this solution can integrate user behavior information from multiple social platforms and different types of users. This helps to supplement and enrich trust features through user social behavior data in scenarios where trust relationship data is scarce or insufficient, thereby improving the accuracy of trust assessment.

[0066] By utilizing multi-view learning techniques, this approach can uncover consistent trust information contained in graphs from different perspectives. This consistent information helps eliminate redundant and irrelevant features, enabling the completion and enhancement of trust features, and further improving the accuracy of trust assessment.

[0067] The heterogeneous graph attention mechanism in the proposed solution can effectively handle node and edge information, adapting to different types of graph structures. This means that the solution can be flexibly applied to various complex social network environments, including different types of social behavior graphs and trust relationship graphs, improving the model's adaptability and generalization ability.

[0068] By employing multi-perspective learning technology, this approach can integrate trust features from different perspectives to form a more comprehensive and accurate representation of user trust. This fusion strategy helps capture users' trust characteristics in different social scenarios, thereby improving the accuracy of trust relationship prediction.

[0069] The autoencoder structure in this scheme optimizes model training by constructing a reconstruction loss, ensuring that the fused features can accurately reconstruct the features of the original viewpoint. This optimization strategy helps preserve important information in the original data, further improving the accuracy of trust assessment.

[0070] In summary, this technical solution achieves effective evaluation of trust relationships in social networks by introducing multi-view graph neural networks and heterogeneous graph attention mechanisms. It not only overcomes the limitations of traditional trust evaluation methods but also improves the model's adaptability and prediction accuracy. Furthermore, this solution possesses potential advantages such as interpretability and scalability, providing new ideas and methods for future trust evaluation research. Attached Figure Description

[0071] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0072] The present invention will be further described below with reference to the accompanying drawings:

[0073] Example 1, please refer to Figure 1 Social network trust assessment methods based on multi-view graph neural networks include:

[0074] Acquire social network behavior information, construct social behavior network graphs from different perspectives, and construct user trust relationship network graphs based on the trust relationships between users;

[0075] Based on social behavior network graphs from different perspectives, a multi-head attention mechanism is built to extract and fuse the influence of different types of neighbor nodes on the characteristics of target user nodes, and output social attribute features.

[0076] Based on the user trust relationship network graph, the attention mechanism is used to fuse the neighbor information of nodes to form the trust characteristics of user nodes;

[0077] By combining social attribute features and user node trust features, multi-view feature fusion is performed to output predicted trust relationships.

[0078] First, by acquiring rich information on social network behavior, such as the frequency of user interactions, content sharing, likes, and comments, a social behavior network graph from different perspectives was constructed. This multi-dimensional data collection method ensures a comprehensive capture of users' behavioral patterns on social networks, providing a solid foundation for subsequent trust assessments.

[0079] This method can extract and fuse the influence of different types of neighbor nodes on the characteristics of target user nodes in a refined manner. This means that it can not only identify which neighbor nodes have a significant impact on the target user, but also accurately quantify these impacts, thereby generating more representative social attribute features.

[0080] In the user trust relationship network graph, this method integrates the neighbor information of nodes through an attention mechanism to form the trust features of user nodes. This mechanism can automatically focus on nodes closely related to user trust relationships and adjust the trust feature representation of user nodes accordingly, thereby improving the accuracy of trust assessment.

[0081] This method fuses social attribute features and user node trust features using multi-view features, making full use of feature information from different sources and types. This fusion strategy not only enhances the expressive power of the features but also improves the model's adaptability to complex social environments, thereby achieving more accurate predictions of trust relationships.

[0082] Because this method employs a multi-view graph neural network structure and incorporates an attention mechanism for feature extraction and fusion, it exhibits strong robustness to data noise and outliers. Furthermore, through training and learning on large-scale social network data, this method can learn more generalized social patterns and trust dynamics, thereby improving its generalization ability across different social scenarios.

[0083] Accurate trust assessment is crucial for many applications in social networks, such as recommendation systems, social search, and fraud detection. The trust assessment results provided by this method can offer strong data support for these applications, helping them better meet user needs, improve user experience, and effectively prevent potential risks and threats.

[0084] In summary, the social network trust assessment method based on multi-view graph neural networks achieves accurate assessment of trust relationships through strategies such as comprehensively capturing social behavior information, accurately extracting user features, effectively integrating trust relationships, and fusing multi-view features. This method not only improves the accuracy and robustness of trust assessment but also provides strong data support for social network applications, demonstrating broad application prospects and value.

[0085] Example 2: This invention provides a social network trust evaluation method based on multi-view graph neural networks, specifically including:

[0086] Social Behavior Information Mapping Module: By preprocessing user behavior information on social networks (such as likes, purchases, comments, etc.), this module constructs user social behavior network graphs from different perspectives for different types of user social behaviors. ;

[0087] Trust Relationship Network Graph Construction Module: Constructs a user trust relationship network graph based on the trust relationships between users. And based on a single user Construct its initial trust features ;

[0088] A user node social attribute feature extraction module based on graph attention mechanism: This module constructs a user feature extraction module based on graph neural network, builds a multi-head attention mechanism to extract and fuse the influence of different types of neighbor nodes on the target user node features, and processes them separately based on this module. and output user Corresponding feature embedding

[0089] User trust attribute feature extraction module based on graph attention mechanism: Constructs initial user features based on node trust relationships. and trust relationship network By leveraging the attention mechanism to fuse neighbor information of nodes, trust features of user nodes are formed. ;

[0090] Trust feature fusion module based on multi-view learning: As a user The From one perspective, forming a user Features from a single perspective, namely In summary:

[0091]

[0092] Next, a trust feature fusion module based on multi-view learning is constructed to integrate the features. Integration into a unified feature And construct reconstruction loss For subsequent model training

[0093] Trust relationship prediction based on multilayer perceptron: given a user and users and its characteristics , The input is then fed into a multilayer perceptron to predict the trust relationship.

[0094] Model training: Extract a set of edges from the trust relationships for end-to-end training, using real trust relationships. To perform supervision, the loss function is constructed as follows:

[0095]

[0096] At the same time, combining the loss function in (5), the final loss function is constructed as follows:

[0097]

[0098] The social behavior information mapping module mainly includes the following process:

[0099] Data collection: Collect social data of target users across different social behaviors / platforms;

[0100] Social Relationship Graph Construction: This involves processing each social behavior / platform separately to create user relationship graphs from various perspectives, centered around users (e.g., ...). Centered on a given object, we collect its interaction relationships with other objects (such as products, articles, etc.), treat these interacting objects as nodes, and the interaction relationships as edges, thus constructing a social relationship graph from this perspective. ;

[0101] Initial social node / edge attribute modeling: Based on interaction type, node attributes, and other information, different methods are used to obtain the model. Initial attribute features of nodes and edge attribute features, for example: The textual evaluation information of products can be processed using the BERT model to obtain edge features;

[0102] The trust relationship network graph construction module mainly includes the following process:

[0103] User trust relationship extraction: Standardize the trust characteristics between users to form... The system identifies discrete trust levels and models the trust relationships between users as a directed weighted graph.

[0104] User trust feature initialization: based on user For example, based on its trust relationships with other users and other users' trust relationships with it, out-degree trust features are initialized respectively. and in-degree trust features This forms a complete trust relationship network diagram. ;

[0105] The user node social attribute feature extraction module based on graph attention mechanism mainly includes the following process:

[0106] Attention mechanism: given a node From the perspective Corresponding diagram Neighbor set as ,node , and nodes The side note Features are denoted as ,but Attention parameters The following formula is used for calculation:

[0107]

[0108] Get attention score Then, given the first Layer output is Then the first The layer output is:

[0109]

[0110] Multi-head attention mechanism: Based on the attention mechanism, a multi-head attention mechanism is further designed, that is, it includes... Each independent attention extracts features and performs feature fusion. Assume the first... The features extracted by the attention mechanism are Then the first The layer output is:

[0111]

[0112] Social attribute feature output: When all are completed After feature extraction from the layer, the final feature is output. ;

[0113] The user trust attribute feature extraction module based on graph attention mechanism mainly includes the following process:

[0114] In-degree trust feature embedding module based on multi-layer graph convolutional network: Construction A graph convolutional neural network with layers of user initial in-degree trust features. Initial features Assuming the first The output of the layer is ,if There exists right Trust edge Then it is believed yes For the in-degree neighbors, the attention score is calculated as follows:

[0115]

[0116] Based on this score, the fused features are calculated:

[0117]

[0118] go through After training the layer, output its in-degree features.

[0119] Out-degree trust feature embedding module based on multi-layer graph convolutional network: Construction A graph convolutional neural network with layers of user initial in-degree trust features. Initial features Assuming the first The output of the layer is ,if There exists right Trust edge Then it is believed yes For out-degree neighbors, the attention score is calculated as follows:

[0120]

[0121] Based on this score, the fused features are calculated:

[0122]

[0123] go through After training the layer, output its in-degree features.

[0124] Trust Feature Fusion and Output: and After concatenation, input a fully connected network and an activation layer, and output... Trust characteristics ;

[0125] The trust feature fusion module based on multi-view learning mainly includes the following process:

[0126] View-related feature encoder : Construct a multi-layer deep encoder based on neural networks, and then... User characteristics of each view Encoding as latent features ;

[0127] Multi-view feature fusion module: integrates user... The user fused feature is obtained by fusing the features of each view. ;

[0128] View-related feature decoder A multi-layer deep decoder is constructed based on a neural network, forming an autoencoder structure with the aforementioned deep encoder, and the fused features are then processed. Decode to reconstruct the first Features of a view ;

[0129] Constructing a reconstruction loss based on an autoencoder Its definition is as follows:

[0130]

[0131] This invention addresses the trust assessment problem in social networking platforms, including ubiquitous social networks and mobile crowdsourcing. The platform primarily comprises a social networking platform and... It consists of individual users. For ease of representation, this patent uses... This refers to a social platform that primarily provides online social services such as instant messaging, information sharing, and shopping. It is also responsible for predicting trust relationships based on users' trust patterns and social behaviors to better serve them. The algorithms involved in this patent are all deployed on this platform. middle, Maintaining user social behavior graphs in different social scenarios And user trust relationship diagram The user set is represented as User use This paper proposes a social network trust evaluation method based on multi-view graph neural networks. This method models user trust relationship graphs and cross-platform, cross-scenario social relationship graphs using multi-view graphs. By comprehensively utilizing graph neural networks and multi-view representation learning techniques, it can effectively extract user trust features from graphs at different perspectives and mine consistent trust information. While supplementing the limited trust features caused by insufficient trust relationships through multiple social relationship graphs, the mining of consistent information effectively ensures the elimination of redundant and irrelevant features, thus achieving the completion and enhancement of trust features.

[0132] In another embodiment of the present invention, a social network trust evaluation system based on a multi-view graph neural network is provided, which can be used to implement the above-mentioned social network trust evaluation method based on a multi-view graph neural network. Specifically, the system includes:

[0133] Acquire social network behavior information, construct social behavior network graphs from different perspectives, and construct user trust relationship network graphs based on the trust relationships between users;

[0134] Based on social behavior network graphs from different perspectives, a multi-head attention mechanism is built to extract and fuse the influence of different types of neighbor nodes on the characteristics of target user nodes, and output social attribute features.

[0135] Based on the user trust relationship network graph, the attention mechanism is used to fuse the neighbor information of nodes to form the trust characteristics of user nodes;

[0136] By combining social attribute features and user node trust features, multi-view feature fusion is performed to output predicted trust relationships.

[0137] The module division in this embodiment of the invention is illustrative and represents only one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional modules in the various embodiments of the invention can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0138] In another embodiment of the present invention, a computer device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions from the computer storage medium to achieve a corresponding method flow or function. The processor described in this embodiment can be used for operating a social network trust assessment method based on a multi-view graph neural network.

[0139] In another embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the operating system of the terminal. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the social network trust evaluation method based on multi-view graph neural networks in the above embodiments.

[0140] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can 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, etc.) containing computer-usable program code.

[0141] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can 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, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0142] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0143] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0144] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A social network trust evaluation method based on multi-view graph neural networks, characterized in that, include: Acquire social network behavior information, construct social behavior network graphs from different perspectives, and construct user trust relationship network graphs based on the trust relationships between users; Based on social behavior network graphs from different perspectives, a multi-head attention mechanism is built to extract and fuse the influence of different types of neighbor nodes on the characteristics of target user nodes, and output social attribute features. Based on the user trust relationship network graph, the attention mechanism is used to fuse the neighbor information of nodes to form the trust characteristics of user nodes; By combining social attribute features and user node trust features, multi-view feature fusion is performed, and the predicted trust relationship is output using an autoencoder and a multilayer perceptron. The acquisition of social network behavior information and the construction of social behavior network graphs from different perspectives include: Collect social data of target users across different social behaviors / platforms; Each social behavior / platform is processed separately to create user relationship graphs from different perspectives. Centered on the user, the interaction relationships between the user and social platform objects are collected, and these interactive objects are treated as nodes, with the interaction relationships as edges, thus constructing the social relationship graph from that perspective. , ; Different processing methods are used to obtain the results based on the interaction type and node attributes. Initial attribute features of mid-nodes and edge attribute features; The social behavior network graph based on different perspectives constructs a multi-head attention mechanism to extract and fuse the influence of different types of neighbor nodes on the characteristics of the target user node, outputting social attribute features, including: Given a node From the perspective Corresponding diagram Neighbor set as ,node , and nodes The side note Features are denoted as ,but Attention parameters The following formula is used for calculation: Get attention score Then, given the first Layer output is Then the first The layer output is: Based on the attention mechanism, a multi-head attention mechanism is further designed, namely, including... Each independent attention extracts features and performs feature fusion. Assume the _th _ ... The features extracted by the attention mechanism are Then the first The layer output is: When all is completed After feature extraction from the layer, the final feature is output. .

2. The social network trust evaluation method based on multi-view graph neural networks according to claim 1, characterized in that, The step of constructing a user trust relationship network diagram based on the trust relationships between users includes: Standardize the trust characteristics between users to form The system identifies discrete trust levels and models the trust relationships between users as a directed weighted graph. Initialize out-degree trust features based on the user's trust relationships with other users and other users' trust relationships with that user. and in-degree trust features To form a complete trust relationship network diagram .

3. The social network trust evaluation method based on multi-view graph neural networks according to claim 1, characterized in that, The user trust relationship network graph, which uses an attention mechanism to fuse neighbor information of nodes to form trust features of user nodes, includes: Set up A graph convolutional neural network with layers of user initial in-degree trust features. Initial features Assuming the first The output of the layer is ,if There exists right Trust edge Then it is believed yes For the in-degree neighbors, the attention score is calculated as follows: Based on this score, the fused features are calculated: go through After training the layer, output its in-degree features. Set up A graph convolutional neural network with layers of user initial in-degree trust features. Initial features Assuming the first The output of the layer is ,if There exists right Trust edge Then it is believed yes For out-degree neighbors, the attention score is calculated as follows: Based on this score, the fused features are calculated: go through After training the layer, output its in-degree features. Trust Feature Fusion and Output: and After concatenation, input a fully connected network and an activation layer, and output... Trust characteristics .

4. The social network trust evaluation method based on multi-view graph neural networks according to claim 1, characterized in that, The process of combining social attribute features and user node trust features to perform multi-view feature fusion and output predicted trust relationships includes: by As a user The From one perspective, forming a user Features from a single perspective, namely : Next, a trust feature fusion module based on multi-view learning is constructed to integrate the features. Integration into a unified feature And construct reconstruction loss Used for subsequent model training; Trust relationship prediction based on multilayer perceptron: given a user and users and its characteristics , The input is then fed into a multilayer perceptron to predict the trust relationship. .

5. The social network trust evaluation method based on multi-view graph neural networks according to claim 4, characterized in that, View-related feature encoder : Construct a multi-layer deep encoder based on neural networks, and then... User characteristics of each view Encoding as latent features ; users The user fused feature is obtained by fusing the features of each view. ; View-related feature decoder A multi-layer deep decoder is constructed based on a neural network, forming an autoencoder structure with the aforementioned deep encoder, and the fused features are then processed. Decode to reconstruct the first Features of a view ; Constructing a reconstruction loss based on an autoencoder Its definition is as follows: A set of edges is extracted from the trust relationships for end-to-end training, using real trust relationships. To perform supervision, the loss function is constructed as follows: The final loss function is constructed as follows: 。 6. A social network trust evaluation system based on multi-view graph neural networks, characterized in that, include: The data acquisition module is used to acquire social network behavior information, construct social behavior network graphs from different perspectives, and construct user trust relationship network graphs based on the trust relationships between users. The social attribute feature acquisition module is used to build a multi-head attention mechanism based on social behavior network graphs from different perspectives to extract the influence of different types of neighbor nodes on the features of target user nodes and fuse them to output social attribute features. The trust feature acquisition module is used to fuse the neighbor information of nodes based on the user trust relationship network graph using an attention mechanism to form the trust features of user nodes. The trust relationship output module is used to combine social attribute features and user node trust features to perform multi-view feature fusion and output predicted trust relationships. The acquisition of social network behavior information and the construction of social behavior network graphs from different perspectives include: Collect social data of target users across different social behaviors / platforms; Each social behavior / platform is processed separately to create user relationship graphs from different perspectives. Centered on the user, the interaction relationships between the user and social platform objects are collected, and these interactive objects are treated as nodes, with the interaction relationships as edges, thus constructing the social relationship graph from that perspective. , ; Different processing methods are used to obtain the results based on the interaction type and node attributes. Initial attribute features of mid-nodes and edge attribute features; The social behavior network graph based on different perspectives constructs a multi-head attention mechanism to extract and fuse the influence of different types of neighbor nodes on the characteristics of the target user node, outputting social attribute features, including: Given a node From the perspective Corresponding diagram Neighbor set as ,node , and nodes The side note Features are denoted as ,but Attention parameters The following formula is used for calculation: Get attention score Then, given the first Layer output is Then the first The layer output is: Based on the attention mechanism, a multi-head attention mechanism is further designed, namely, including... Each independent attention extracts features and performs feature fusion. Assume the _th _ ... The features extracted by the attention mechanism are Then the first The layer output is: When all is completed After feature extraction from the layer, the final feature is output. .

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the social network trust evaluation method based on a multi-view graph neural network as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the social network trust assessment method based on a multi-view graph neural network as described in any one of claims 1 to 5.

Images