Multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning

Abstract

The invention relates to a multi-dimensional rumor detection method based on a graph attention mechanism and self-supervised learning. The method comprises the steps that key features of original data are extracted and subjected to preprocessing and coding expression, and the key features comprise node attribute features, graph attribute features, user features and time features; according to the message forwarding list, reconstructing a propagation graph, inputting the key features and the propagation graph into a double-layer graph attention network, and constructing an initial rumor detection classifier; processing the encoded key features by adopting a graph self-supervised learning model, and generating an embedded vector from the propagation graph; constructing an interpretation matrix based on the key features and the embedded vector, and performing attention weighting based on the interpretation matrix to obtain an adjusted feature vector; and inputting the adjusted feature vector into the initial rumor detection classifier, optimizing the initial rumor detection classifier, performing rumor detection based on the optimized initial rumor detection classifier, and outputting a rumor detection result.

Description

Technical Field

[0001] The present application relates to the technical field of rumor detection in social networks, and in particular to a multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning. Background Art

[0002] In the Internet age, the rise of social media and online platforms has greatly promoted the rapid circulation of information, but it has also led to the widespread spread of rumors and false information. These rumors not only mislead the public, but may also cause social panic, disrupt order, and even affect economic stability. Therefore, it is particularly urgent to develop effective rumor detection technology.

[0003] Most existing rumor detection methods rely mainly on the semantic information of text, so the detection effect is often limited in multilingual or cross-platform environments. Traditional models usually need to extract and train features for different languages ​​separately, resulting in weak cross-language migration capabilities. At the same time, in the early stages of rumor propagation, rumors often spread rapidly in non-textual forms through user interactions and forwarding behaviors. Traditional text analysis methods are difficult to capture these behavioral changes in a timely manner, and their response speed is relatively slow.

[0004] Secondly, many detection methods ignore the significant difference in the time distribution of user forwarding of rumors and true information. Rumors usually spread rapidly in the early stage, accompanied by a large number of users forwarding them, while the forwarding time distribution of true information is more dispersed and robust. In addition, most existing rumor detection models lack the interpretability of detection results and cannot clearly explain which features play a key role in the classification results. This not only reduces the credibility of the model, but also limits its operability in practical applications.

[0005] Therefore, in the relevant technology, there is an urgent need for a method that can effectively identify rumors and improve the speed and accuracy of rumor detection. Summary of the invention

[0006] Based on this, it is necessary to provide a multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning, which can effectively identify rumors and improve the speed and accuracy of rumor detection, in response to the above technical problems.

[0007] In the first aspect, the present application provides a multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning. The method comprises:

[0008] Extracting key features of the original data, and performing preprocessing and encoding representation, wherein the key features include node attribute features, graph attribute features, user features, and time features;

[0009] According to the message forwarding list, a propagation graph is reconstructed, and the key features and the propagation graph are input into a two-layer graph attention network to construct an initial rumor detection classifier;

[0010] A graph self-supervised learning model is used to process the encoded key features and generate embedding vectors from the propagation graph;

[0011] constructing an explanation matrix based on the key features and the embedding vector, performing attention weighting based on the explanation matrix, and obtaining an adjusted feature vector;

[0012] The adjusted feature vector is input into the initial rumor detection classifier, the initial rumor detection classifier is optimized, rumor detection is performed based on the optimized initial rumor detection classifier, and the rumor detection result is output.

[0013] Optionally, in one embodiment of the present application, the preprocessing includes data normalization processing, and the formula is:

[0014]

[0015] Among them, X scaled represents the normalized result, X min and X max Indicates the minimum and maximum values ​​of each type of feature.

[0016] Optionally, in one embodiment of the present application, the attention mechanism calculation formula of the two-layer graph attention network is:

[0017]

[0018] Among them, α ij is the attention weight of node i to node j, h i 、h j and h k is the feature vector of the node, is the domain of i, W is the shared weight matrix, a is the learnable attention weight vector, || represents the concatenation of vectors, and LeakyReLU is the activation function.

[0019] Optionally, in one embodiment of the present application, the graph self-supervised learning model adopts a DGI model to maximize the mutual information between global information and local information in the graph structure and obtain a useful embedding vector of the graph or node.

[0020] Optionally, in one embodiment of the present application, sparsity loss, orthogonality loss and regularization loss are used to optimize the loss function of the training DGI model.

[0021] Optionally, in one embodiment of the present application, the explanation matrix construction formula is:

[0022]

[0023] Among them, E * represents the explanation matrix, H * is the optimal embedding vector of the node, represents the outer product of the vectors, F is the feature vector used for transformation together with the node embedding, and ||·|| represents the modulus of the vector.

[0024] Optionally, in one embodiment of the present application, performing attention weighting based on the interpretation matrix to obtain an adjusted feature vector includes:

[0025] Averaging the features of each row and dimension of the explanation matrix to obtain the feature average weight;

[0026] The feature average weight is normalized and applied to the key feature to obtain an adjusted feature vector.

[0027] In the second aspect, the present application also provides a multi-dimensional rumor detection device based on graph attention mechanism and self-supervised learning. The device comprises:

[0028] The original feature processing module is used to extract key features of the original data, and perform preprocessing and encoding representation, wherein the key features include node attribute features, graph attribute features, user features and time features;

[0029] An initial rumor detection classifier construction module is used to reconstruct a propagation graph according to a message forwarding list, input the key features and the propagation graph into a double-layer graph attention network, and construct an initial rumor detection classifier;

[0030] An embedding vector generation module is used to process the encoded key features using a graph self-supervised learning model and generate an embedding vector from the propagation graph;

[0031] An explanation matrix construction module, used to construct an explanation matrix based on the key features and the embedding vector, and perform attention weighting based on the explanation matrix to obtain an adjusted feature vector;

[0032] The rumor detection result determination module is used to input the adjusted feature vector into the initial rumor detection classifier, optimize the initial rumor detection classifier, perform rumor detection based on the optimized initial rumor detection classifier, and output the rumor detection result.

[0033] In a third aspect, the present application further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the steps of the methods described in the above embodiments.

[0034] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method described in each of the above embodiments are implemented.

[0035] The above multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning first extracts the key features of the original data, and performs preprocessing and encoding representation, wherein the key features include node attribute features, graph attribute features, user features and time features; then, according to the message forwarding list, the propagation graph is reconstructed, and the key features and the propagation graph are input into the double-layer graph attention network to construct an initial rumor detection classifier; then, the encoded key features are processed by the graph self-supervised learning model, and an embedding vector is generated from the propagation graph; then, an explanation matrix is ​​constructed based on the key features and the embedding vector, and attention weighting is performed based on the explanation matrix to obtain an adjusted feature vector; finally, the adjusted feature vector is input into the initial rumor detection classifier, the initial rumor detection classifier is optimized, and rumor detection is performed based on the optimized initial rumor detection classifier, and the rumor detection result is output. That is to say, by integrating multiple features, including node attributes, graph attributes, user features and time features, the previous dependence on text features is broken. This multi-dimensional method not only accurately maps the interactive relationship between users and their influence in the network, but also can monitor the spread trend of rumors over time. By introducing the graph attention network, it is possible to flexibly handle complex information flows and accurately capture key nodes and patterns in the propagation chain, thereby enhancing adaptability in diverse contexts. This feature greatly improves the robustness of the model and the detection accuracy between different platforms, ensuring its wide application in different propagation environments. The accuracy and interpretability of rumor detection are further improved through self-supervised learning and the explanation matrix mechanism. The explanation matrix can intuitively display the specific contribution of each feature to the final detection result, making the detection process transparent, helping users understand the judgment basis of the model, and enhancing the credibility of the model decision. At the same time, the present invention has a highly sensitive recognition ability in the early stage of rumor propagation, and can quickly lock potential propagation paths and abnormal propagation behaviors, thereby significantly improving the response speed and accuracy of rumor detection. By quickly intervening in the early stage of rumor propagation, the model can effectively curb the spread of rumors, providing more reliable technical support and guarantee for response strategies in complex public opinion environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is an application environment diagram of a multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning in one embodiment;

[0037] Figure 21 is a flow chart of a multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning in one embodiment;

[0038] Figure 3 A schematic diagram of the structure of a multi-dimensional rumor detection method model based on graph attention mechanism and self-supervised learning in one embodiment;

[0039] Figure 4 A schematic diagram of visualization results of an explanation matrix of overall news data in one embodiment;

[0040] Figure 5 A schematic diagram of visualization results of an explanation matrix of Politics news data in one embodiment;

[0041] Figure 6 This is a diagram showing the visualization results of the explanation matrix for news data of the Viral Photos / Stories / Urban Legends category;

[0042] Figure 7 This is a diagram showing the visualization results of the explanation matrix for news data in the category War / Terrorism / Shootings.

[0043] Figure 8 A schematic diagram showing the comparison of the results before and after the original eigenvectors are weighted by the interpretation matrix values;

[0044] Fig. 9 A schematic diagram showing the comparison between the experimental results of the present invention and the results of other methods;

[0045] Fig.10 It is a schematic diagram of the early detection results of the present invention.

[0046] Fig.11 It is a structural block diagram of a multi-dimensional rumor detection device based on graph attention mechanism and self-supervised learning in one embodiment;

[0047] Fig.12 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0049] The multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. Among them, the terminal 102 can be but is not limited to various personal computers, laptops, smart phones, tablets, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented with an independent server or a server cluster consisting of multiple servers.

[0050] In one embodiment, Figure 2 As shown in the figure, a multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning is provided. Figure 1 The server in the example is used to illustrate the following steps:

[0051] S201: extract key features of the original data, and perform preprocessing and encoding representation, wherein the key features include node attribute features, graph attribute features, user features, and time features.

[0052] In the embodiment of the present application, first, key features are extracted from the original data, including node attribute features, graph attribute features, user features and time features, and these features are preprocessed and encoded. Among them, the node attribute features include the number of connections, clustering coefficient, standard deviation, average degree of adjacent nodes, average adjacent clustering coefficient and centrality of the node, the graph attribute features include the size, maximum depth and maximum width of the graph, the user features include the average number of users concerned, the average participation, the average number of fans, the average account age, and the time features include the total time of each cascade propagation and the total time of independent user propagation intervals. The preprocessed features are spliced ​​together as the initial feature vector for subsequent training.

[0053] In one embodiment of the present application, the preprocessing includes data normalization processing, and the formula is:

[0054]

[0055] Among them, X scaled represents the normalized result, X min and X max Indicates the minimum and maximum values ​​of each type of feature.

[0056] S203: Reconstruct the propagation graph according to the message forwarding list, input the key features and the propagation graph into the double-layer graph attention network, and build an initial rumor detection classifier.

[0057] In the embodiment of the present application, when a user forwards a piece of news or information, this forwarding behavior will form a propagation path in the user network. By analyzing the propagation path, specific cascade events can be identified, based on which the propagation graph of each piece of news is reconstructed. Afterwards, the spliced ​​initial feature vector extracted and encoded from the original data and the reconstructed propagation graph are divided into training samples and test samples, and the spliced ​​initial feature vector is used as an additional node feature of the propagation graph, and is input into a two-layer graph attention network for training to construct an initial rumor detection classifier and output preliminary detection results. Figure 3 As shown in the figure, the two-layer graph attention network module can improve the model's feature extraction and rumor recognition capabilities in the propagation network by effectively utilizing the node relationships in the graph structured data and the neighbor information on the graph.

[0058] Specifically, the two-layer graph attention network is an improved version of the Graph Attention Networks (GAT), GATv2Conv, which is an enhancement of the original GAT. In the original GAT, the attention mechanism relies on a linear combination of input features and weights, which may lead to learning information that does not fully express the graph structure under certain circumstances. GATv2 proposes a new attention calculation method, using a self-attention mechanism that can dynamically adjust the aggregation of node features according to the importance of neighboring nodes, making the attention mechanism insensitive to the order of input features, that is, solving the symmetry problem of the attention function, which enables GATv2 to flexibly capture the complex dependencies between neighboring nodes.

[0059] In one embodiment of the present application, the attention mechanism calculation formula of the two-layer graph attention network is:

[0060]

[0061] Among them, α ij is the attention weight of node i to node j, h i 、h j and h k is the feature vector of the node, is the domain of i, W is the shared weight matrix, a is the learnable attention weight vector, || represents the concatenation of vectors, and LeakyReLU is the activation function.

[0062] In one embodiment of the present application, the attention weight α ij , the feature h of each node i i It aggregates the features of its neighbor nodes and updates them into a new representation h i ′, the specific formula is as follows:

[0063]

[0064] Among them, σ is the activation function, and optionally, the SELU activation function can be used.

[0065] The SELU activation function is a self-normalizing activation function that can maintain the zero mean and unit variance of neuron output in deep neural networks, which helps solve the problem of gradient disappearance or explosion in deep networks. The formula is as follows:

[0066]

[0067] Where λ is a scalar constant used to scale the output and α is a scaling factor for the negative part.

[0068] S205: Use a graph self-supervised learning model to process the encoded key features and generate an embedding vector from the propagation graph.

[0069] In an embodiment of the present application, a graph self-supervised learning model is used to process the encoded initial feature vector, and an embedding vector is generated from the propagation graph.

[0070] In one embodiment of the present application, the graph self-supervised learning model adopts a DGI model to maximize the mutual information between global information and local information in the graph structure and obtain useful embedding vectors of the graph or node.

[0071] In one embodiment of the present application, Figure 3 As shown in the figure, the graph self-supervised learning model adopts the DGI (Deep Graph Infomax) model, which aims to obtain useful embedding vectors of graphs or nodes by maximizing the mutual information between global information and local information in the graph structure. Assume that a graph G (X, A) is given, X is the node feature matrix, the dimension is N × F, where N is the number of nodes, F is the feature dimension of each node, and the graph convolutional network GCN (Graph Convolutional Networks) is used to generate the embedding vector. The basic propagation rule of GCN is:

[0072]

[0073] in, is the adjacency matrix with self-loops added. yes The degree matrix of H (l) is the node embedding representation of the lth layer, initially H (0) =X.W (l) is a learnable weight matrix. σ is a nonlinear activation function, and here we use the RELU activation function. After that, a READOUT function is used to generate the global graph representation s.

[0074] In order to perform contrastive learning, DGI needs to generate negative samples. By randomly perturbing the node features X, a perturbed feature matrix X′ is obtained, but the adjacency matrix A remains unchanged. The node embedding for generating negative samples is:

[0075] H′=GCN(X′,A)

[0076] Afterwards, the core goal of DGI is to maximize the local embedding h of the node i The mutual information between the global graph representation s and the negative sample node embedding h is minimized i The correlation between ′ and the global representation s. This is achieved through contrastive learning, and the loss function is:

[0077]

[0078] Among them, D(h i ,s) is a discriminator used to distinguish positive samples from negative samples. It usually has the following form:

[0079]

[0080] Among them, W is the learnable weight matrix and σ is the sigmoid function.

[0081] In one embodiment of the present application, sparsity loss, orthogonality loss and regularization loss are used to optimize the loss function of the DGI model for training.

[0082] In one embodiment of the present application, in order to further optimize the DGI model and obtain the optimal embedding vector, sparsity loss, orthogonality loss and regularization loss are added to the loss function.

[0083] Sparsity loss encourages the model to generate sparse representations by controlling the sparsity of the explanation matrix, thereby improving the generalization ability of the model. Specifically, the sparsity loss focuses on making most elements in the explanation matrix close to zero, while retaining only a few important features. This helps the model focus on the most important graph features, thereby improving its interpretability and generalization ability. The calculation formula of the sparsity loss is:

[0084]

[0085] Where E is the explanation matrix, λ sparsity is the weight of the sparsity loss and ||·||1 is the L1 norm.

[0086] Orthogonality loss avoids overfitting by making the row vectors in the explanation matrix orthogonal to each other, which helps increase the discriminability of the embedding. Orthogonality loss focuses on reducing the correlation between the row vectors of the explanation matrix so that each dimension encodes different information as independently as possible. This helps improve the interpretability of the model because it ensures that each embedding dimension is defined by a different set of graph features. The calculation formula for orthogonality loss is:

[0087]

[0088] Among them, E T E is the autocorrelation matrix of the explanatory matrix, ||·|| F is the Frobenius norm, λ orthogonality is the weight of the orthogonality loss.

[0089] Regularization loss is used to limit the size of embeddings and avoid excessive weights or eigenvalues ​​to improve the stability of the model. Regularization loss focuses on controlling the norm of the embedding vector to prevent the model from being too dependent on certain specific features, thereby improving the generalization ability of the model. The calculation formula for regularization loss is:

[0090]

[0091] Among them, λ regularization is the weight of the regularization loss. i is the embedding of the i-th node. The final loss function is the weighted sum of multiple loss terms, and the formula is:

[0092]

[0093] By inputting the feature vector into the DGI model and performing training optimization, the optimal embedding vector can be obtained.

[0094] S207: constructing an explanation matrix based on the key features and the embedding vector, and performing attention weighting based on the explanation matrix to obtain an adjusted feature vector.

[0095] In the embodiment of the present application, the embedding vector and the initial feature vector are combined to jointly construct an explanation matrix for evaluating the importance of each feature, and attention weighting is performed according to the importance value of each row feature of the explanation matrix to adjust the feature vector.

[0096] In one embodiment of the present application, the explanation matrix construction formula is:

[0097]

[0098] Among them, E * represents the explanation matrix, H * is the optimal embedding vector of the node, represents the outer product of the vectors, F is the feature vector used for transformation together with the node embedding, and ||·|| represents the modulus of the vector.

[0099] In one embodiment of the present application, performing attention weighting based on the interpretation matrix to obtain an adjusted feature vector includes:

[0100] Averaging the features of each row and dimension of the explanation matrix to obtain the feature average weight;

[0101] The feature average weight is normalized and applied to the key feature to obtain an adjusted feature vector.

[0102] In one embodiment of the present application, first, the features of each dimension of each row of the interpretation matrix are averaged, and the specific formula is as follows:

[0103]

[0104] Among them, w i represents the average weight of feature i, and m represents the number of columns in the explanation matrix.

[0105] After calculating the average weights of all features, these weights are normalized:

[0106]

[0107] Among them, w norm,i represents the normalized weight of feature i, w min and w max Represents the maximum and minimum values ​​among the feature average weights.

[0108] And apply the normalized weights to each feature of the original feature matrix to obtain the adjusted feature vector. The specific formula is as follows:

[0109] X adjusted =X⊙w norm

[0110] Among them, X represents the original feature matrix, X adjusted Represents the adjusted feature matrix.

[0111] S209: Inputting the adjusted feature vector into the initial rumor detection classifier, optimizing the initial rumor detection classifier, performing rumor detection based on the optimized initial rumor detection classifier, and outputting a rumor detection result.

[0112] In an embodiment of the present application, the adjusted feature matrix is ​​re-input into the constructed initial rumor detection classifier, training is performed to optimize the initial rumor detection classifier, and the final rumor detection result is output.

[0113] It should be noted that in specific applications, starting from the initial moment of rumor propagation, the rumor propagation accuracy rate and other values ​​within a fixed time interval are tested to obtain early detection results of rumors, thereby realizing rapid identification and response to rumors. For example, every 100 seconds is used as a fixed time interval to monitor the rumor propagation within the first 600 seconds.

[0114] In one embodiment of the present application, real spread data is used to verify the availability of the rumor detection model. The real spread data is all verified true and false news reports published on Twitter from 2006 to 2017 collected in the paper "The spread of true and false news online". The overall news data and the news data of the main categories (including Politics, Viral Photos / Stories / Urban Legends, War / Terrorism / Shootings) are conditionally screened, and finally 5,000 news true and false cascades are screened from the overall news data, and 1,000 news true and false cascades are screened from the specific types of news data as samples to calculate the interpretability matrix of each feature vector and embedding vector, so as to obtain the importance of the feature.

[0115] Based on the DGI model as the embedding model, the overall news data and specific types of news data (including Politics, Viral Photos / Stories / Urban Legends, War / Terrorism / Shootings) are analyzed by visualizing the explanation matrix. The results are as follows: Figure 4 , 5 , 6, and 7. It can be seen that the order of feature importance remains consistent for both overall and classified rumor data. The important features are the average degree of neighboring nodes, centrality, graph size, average number of users followed, average number of fans, average account age, total time between independent user propagation, number of node connections, maximum width of the graph, total time for each cascade propagation, and maximum depth of the graph. The contribution of other features to rumor propagation is almost zero, indicating that they have little effect on the predictive ability of the model.

[0116] In one embodiment of the present application, the importance of each feature is weighted using the explanation matrix of the overall news data and specific types of news data (including Politics, Viral Photos / Stories / Urban Legends, War / Terrorism / Shootings), and the rumor detection results before and after weighting are compared. Figure 8As shown in the figure, in the detection of the overall news data, the application of the weighted explanation matrix significantly improved the model performance. The detection results were improved from the original accuracy of 0.8384, precision of 0.9692, recall of 0.6864, and F1 score of 0.7827 to accuracy of 0.8505, precision of 0.9641, recall of 0.7210, and F1 score of 0.8050.

[0117] For certain types of news data, the weighted explanation matrix also shows an optimization effect:

[0118] In the Politics news data, the detection results improved from 0.9475 accuracy, 0.9786 precision, 0.9150 recall, and 0.9457 F1 score to 0.9651 accuracy, 0.9947 precision, 0.9350 recall, and 0.9639 F1 score. In the Viral Photos / Stories / Urban Legends news data, the detection results improved from 0.8321 accuracy, 0.9879 precision, 0.6643 recall, and 0.7983 F1 score to 0.8429 accuracy, 0.9983 precision, 0.6857 recall, and 0.8136 F1 score. In the news data of the War / Terrorism / Shootings category, the detection results are improved from accuracy 0.8679, precision 0.9893, recall 0.7357, and F1 score 0.8477 to accuracy 0.9036, precision 0.9982, recall 0.8071, and F1 score 0.8933. It can be seen that by introducing the weighting mechanism of the explanation matrix, the rumor detection model has achieved performance improvement on news data of various categories, which shows the important role of the explanation matrix in improving the prediction accuracy and interpretability of the model.

[0119] In one embodiment of the present application, a DTC (Decision Tree Classifier) ​​model, an SVM (Support Vector Machine) model, a GraphSAGE model, a GCN model, a GAE (Graph Auto-Encoder) model, and a GRU (Gated Recurrent Unit) model are used as comparative examples to perform rumor detection under the same data. Fig. 9As shown in the figure, when comparing the performance of different models, the DTC model achieved an accuracy of 0.7543, a precision of 0.7292, a recall of 0.7202, and an F1 score of 0.7247; the SVM model achieved an accuracy of 0.7854, a precision of 0.7632, a recall of 0.7378, and an F1 score of 0.7253; the GraphSAGE model achieved an accuracy of 0.8088, a precision of 0.9120, a recall of 0.6356, and an F1 score of 0. .7491 F1 score; GCN model achieved 0.7939 accuracy, 0.8528 precision, 0.6540 recall, and 0.7403 F1 score; GAE model achieved 0.8248 accuracy, 0.9642 precision, 0.6172 recall, and 0.7598 F1 score; GRU model achieved 0.8307 accuracy, 0.9541 precision, 0.6421 recall, and 0.7853 F1 score. In contrast, the multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning achieved excellent performance of 0.8505 accuracy, 0.9641 precision, 0.7210 recall, and 0.8050 F1 score, outperforming other compared models in all evaluation indicators. This result further verifies the effectiveness of the method of the present invention and its adaptability to rumor identification tasks.

[0120] In one embodiment of the present application, Fig.10As shown in Figure 1, the first 600 seconds of rumor propagation were monitored regularly to evaluate the effectiveness of the method in early detection of rumors. In the first 100 seconds, the accuracy of early detection reached 0.8087, precision 0.9598, recall 0.5968, and F1 score 0.7360. When the time reached 200 seconds, the accuracy of early detection increased to 0.8272, precision 0.9676, recall 0.6343, and F1 score 0.7663. At 300 seconds, the accuracy of early detection further increased to 0.8380, precision 0.9781, recall 0.6496, and F1 score 0.7807. At 400 seconds, the accuracy of early detection slightly increased to 0.8423, precision 0.9355, recall 0.6945, and F1 score 0.7972. By 500 seconds, the accuracy of early detection increased to 0.8495, the precision was 0.9893, the recall was 0.6698, and the F1 score was 0.7988. At 600 seconds, the accuracy of early detection stabilized at 0.8498, the precision was 0.9609, the recall was 0.6917, and the F1 score was 0.8044. It can be seen that this method can quickly identify rumors in a short period of time, and its detection performance gradually increases and stabilizes over time. This early detection capability is crucial for quickly responding to and controlling the spread of rumors, and helps take action before rumors have a widespread impact.

[0121] In the above multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning, first, the key features of the original data are extracted, and preprocessed and encoded, wherein the key features include node attribute features, graph attribute features, user features and time features; then, according to the message forwarding list, the propagation graph is reconstructed, and the key features and the propagation graph are input into the double-layer graph attention network to construct an initial rumor detection classifier; then, the encoded key features are processed by the graph self-supervised learning model, and an embedding vector is generated from the propagation graph; then, an explanation matrix is ​​constructed based on the key features and the embedding vector, and attention weighting is performed based on the explanation matrix to obtain an adjusted feature vector; finally, the adjusted feature vector is input into the initial rumor detection classifier, the initial rumor detection classifier is optimized, and rumor detection is performed based on the optimized initial rumor detection classifier, and the rumor detection result is output. That is to say, by fusing multiple features, including node attributes, graph attributes, user features and time features, the previous dependence on text features is broken. This multi-dimensional method not only accurately maps the interactive relationship between users and their influence in the network, but also can monitor the spread trend of rumors over time. By introducing the graph attention network, it is possible to flexibly handle complex information flows and accurately capture key nodes and patterns in the propagation chain, thereby enhancing adaptability in diverse contexts. This feature greatly improves the robustness of the model and the detection accuracy between different platforms, ensuring its wide application in different propagation environments. The accuracy and interpretability of rumor detection are further improved through self-supervised learning and the explanation matrix mechanism. The explanation matrix can intuitively display the specific contribution of each feature to the final detection result, making the detection process transparent, helping users understand the judgment basis of the model, and enhancing the credibility of the model decision. At the same time, the present invention has a highly sensitive recognition ability in the early stage of rumor propagation, and can quickly lock potential propagation paths and abnormal propagation behaviors, thereby significantly improving the response speed and accuracy of rumor detection. By quickly intervening in the early stage of rumor propagation, the model can effectively curb the spread of rumors, providing more reliable technical support and guarantee for response strategies in complex public opinion environments.

[0122] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0123] Based on the same inventive concept, the embodiment of the present application also provides a multi-dimensional rumor detection device based on graph attention mechanism and self-supervised learning for implementing the multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more embodiments of the multi-dimensional rumor detection device based on graph attention mechanism and self-supervised learning provided below can be found in the above limitations on the multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning, which will not be repeated here.

[0124] In one embodiment, Fig.11 As shown, a multi-dimensional rumor detection device 1100 based on graph attention mechanism and self-supervised learning is provided, including: an original feature processing module 1101, an initial rumor detection classifier construction module 1103, an embedding vector generation module 1105, an explanation matrix construction module 1107 and a rumor detection result determination module 1109, wherein:

[0125] The original feature processing module 1101 is used to extract key features of the original data, and perform preprocessing and encoding representation, wherein the key features include node attribute features, graph attribute features, user features and time features.

[0126] The initial rumor detection classifier construction module 1103 is used to reconstruct the propagation graph according to the message forwarding list, input the key features and the propagation graph into the double-layer graph attention network, and construct the initial rumor detection classifier.

[0127] The embedding vector generation module 1105 is used to process the encoded key features using a graph self-supervised learning model and generate an embedding vector from the propagation graph.

[0128] The explanation matrix construction module 1107 is used to construct an explanation matrix based on the key features and the embedding vector, and perform attention weighting based on the explanation matrix to obtain an adjusted feature vector.

[0129] The rumor detection result determination module 1109 is used to input the adjusted feature vector into the initial rumor detection classifier, optimize the initial rumor detection classifier, perform rumor detection based on the optimized initial rumor detection classifier, and output the rumor detection result.

[0130] In one embodiment of the present application, the preprocessing includes data normalization processing, and the formula is:

[0131]

[0132] Among them, X scaled represents the normalized result, X min and X max Indicates the minimum and maximum values ​​of each type of feature.

[0133] In one embodiment of the present application, the attention mechanism calculation formula of the two-layer graph attention network is:

[0134]

[0135] Among them, α ij is the attention weight of node i to node j, h i 、h j and h k is the feature vector of the node, is the domain of i, W is the shared weight matrix, a is the learnable attention weight vector, || represents the concatenation of vectors, and LeakyReLU is the activation function.

[0136] In one embodiment of the present application, the graph self-supervised learning model adopts a DGI model to maximize the mutual information between global information and local information in the graph structure and obtain useful embedding vectors of the graph or node.

[0137] In one embodiment of the present application, sparsity loss, orthogonality loss and regularization loss are used to optimize the loss function of the DGI model for training.

[0138] In one embodiment of the present application, the explanation matrix construction formula is:

[0139]

[0140] Among them, E * represents the explanation matrix, H * is the optimal embedding vector of the node, represents the outer product of the vectors, F is the feature vector used for transformation together with the node embedding, and ||·|| represents the modulus of the vector.

[0141] In one embodiment of the present application, performing attention weighting based on the interpretation matrix to obtain an adjusted feature vector includes:

[0142] Averaging the features of each row and dimension of the explanation matrix to obtain the feature average weight;

[0143] The feature average weight is normalized and applied to the key feature to obtain an adjusted feature vector.

[0144] Each module in the above-mentioned multi-dimensional rumor detection device based on graph attention mechanism and self-supervised learning can be implemented in whole or in part by software, hardware and their combination. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0145] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig.12 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a multi-dimensional rumor detection method based on a graph attention mechanism and self-supervised learning is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse.

[0146] Those skilled in the art will understand that Fig.12 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0147] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0148] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0149] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0150] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0151] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0152] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0153] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning, characterized in that: The method comprises: Extract key features of the original data, and perform preprocessing and encoding representation, wherein the key features include node attribute features, graph attribute features, user features, and time features; According to the message forwarding list, a propagation graph is reconstructed, and the key features and the propagation graph are input into a two-layer graph attention network to construct an initial rumor detection classifier; A graph self-supervised learning model is used to process the encoded key features and generate embedding vectors from the propagation graph; constructing an explanation matrix based on the key features and the embedding vector, performing attention weighting based on the explanation matrix, and obtaining an adjusted feature vector; The adjusted feature vector is input into the initial rumor detection classifier, the initial rumor detection classifier is optimized, rumor detection is performed based on the optimized initial rumor detection classifier, and the rumor detection result is output.

2. According to claim 1, the multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning is characterized in that: The preprocessing includes data normalization processing, and the formula is: Among them, X scaled represents the normalized result, X min and X max Indicates the minimum and maximum values ​​of each type of feature.

3. The multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning according to claim 1 is characterized in that: The attention mechanism calculation formula of the two-layer graph attention network is: Among them, α ij is the attention weight of node i to node j, h i 、h j and h k is the feature vector of the node, is the domain of i, W is the shared weight matrix, α is the learnable attention weight vector, || represents the concatenation of vectors, and LeakyReLU is the activation function.

4. The multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning according to claim 1 is characterized in that: The graph self-supervised learning model adopts the DGI model to maximize the mutual information between global information and local information in the graph structure and obtain useful embedding vectors of the graph or nodes.

5. The multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning according to claim 4 is characterized in that: Sparsity loss, orthogonality loss and regularization loss are used to optimize the loss function of the DGI model for training.

6. The multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning according to claim 1 is characterized in that: The explanation matrix construction formula is: Among them, E * represents the explanation matrix, H * is the optimal embedding vector of the node, represents the outer product of the vectors, F is the feature vector used for transformation together with the node embedding, and ||·|| represents the modulus of the vector.

7. The multi-dimensional rumor detection method based on graph attention mechanism and self-supervised learning according to claim 1 is characterized in that: The attention weighting is performed based on the explanation matrix to obtain the adjusted feature vector, which includes: Averaging the features of each row and dimension of the explanation matrix to obtain the feature average weight; The feature average weight is normalized and applied to the key feature to obtain an adjusted feature vector.

8. A multi-dimensional rumor detection device based on graph attention mechanism and self-supervised learning, characterized in that: The device comprises: The original feature processing module is used to extract key features of the original data, and perform preprocessing and encoding representation, wherein the key features include node attribute features, graph attribute features, user features and time features; An initial rumor detection classifier construction module is used to reconstruct a propagation graph according to a message forwarding list, input the key features and the propagation graph into a double-layer graph attention network, and construct an initial rumor detection classifier; An embedding vector generation module is used to process the encoded key features using a graph self-supervised learning model and generate an embedding vector from the propagation graph; An explanation matrix construction module, used to construct an explanation matrix based on the key features and the embedding vector, and perform attention weighting based on the explanation matrix to obtain an adjusted feature vector; The rumor detection result determination module is used to input the adjusted feature vector into the initial rumor detection classifier, optimize the initial rumor detection classifier, perform rumor detection based on the optimized initial rumor detection classifier, and output the rumor detection result.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images