339 results about "Adjacency matrix" patented technology

This invention relates to a graph node classification method based on a graph feature self-attention mechanism using ensemble learning. First, the original graph network data is input, including the node feature matrix H and the graph adjacency matrix A. The LightGBM method is used to preprocess the dataset to obtain model parameters. Finally, the expanded node feature matrix H is obtained by summarizing the tree model parameters. new ; Combine the graph adjacency matrix A and the expanded node feature matrix H new The input graph features are trained on an attention network, and the network's preference information for nodes and features is obtained after training. Finally, the trained model is used to predict the classification of the graph node dataset on the validation and test sets. This invention utilizes an ensemble learning method and incorporates node attention preferences for features, which not only efficiently leverages the high interpretability of tree models but also enriches the model's expression, allowing node features to contain more information, thus contributing to higher performance on node classification tasks.

This invention discloses a dynamic quantile filtering method for product co-occurrence networks, comprising: firstly, constructing a product co-occurrence network based on the co-occurrence relationships of products in orders; then, employing a dynamic quantile filtering strategy to denoise the network, specifically: obtaining the co-occurrence frequency weights of all product pairs and forming a co-occurrence matrix; extracting all non-zero co-occurrence frequency weights for each node; defining a local adaptive threshold based on the quantiles of these weights; retaining edges greater than or equal to the threshold; resetting the weights of edges less than the threshold to zero; and retaining isolated nodes; symmetricizing the adjacency matrix using an intersection strategy to obtain a weighted adjacency matrix; updating the network to obtain the denoised product co-occurrence network; and finally, using the Louvain algorithm to perform community partitioning on the denoised network to obtain a product community network. This invention can adaptively filter out weak connections and noisy edges in the product co-occurrence network, effectively improving the accuracy and robustness of community detection.

The application relates to an excitation system fault recording and event record analysis and diagnosis method and system, belonging to the field of excitation systems. The method comprises collecting recording files and event sequence records of the excitation system; performing multi-domain feature extraction based on the recording files to obtain a multi-domain feature tensor; performing space-time causal structure learning based on the multi-domain feature tensor and the event sequence records to output a causal adjacency matrix, a causal diagram and a time delay matrix; performing double-channel interpretable fault classification based on the multi-domain feature tensor, the causal diagram and an event time tag list E in the event sequence records to output a fault type label M and an attention space-time heat map; performing counterfactual causal tracing to obtain a root cause variable set and a causal propagation path, and outputting a diagnosis report. The application realizes intelligent diagnosis of the excitation system with signal analysis capability, causal reasoning capability and diagnosis interpretability.

This invention discloses a method for fault diagnosis of rotating machinery based on lightweight star-anchor screening and a flow-gated graph convolutional network. The method first maps vibration time-series signals into a node feature matrix and constructs a candidate adjacency matrix containing nearest neighbor edges and sequence prior edges. Star-anchor screening is used to score the node pair features of candidate edges, filtering core nodes and effective edges to generate sparse effective adjacencies. A graph convolutional network with a flow-gated mechanism is constructed, dividing the input features into propagation branches and bypass branches based on channel global statistics. Only effective channels undergo graph propagation aggregation on sparse adjacencies, and the node features are output after linear fusion with bypass branches. The fault category is then output via graph-level aggregation and a classifier. This invention achieves channel-structure dual sparsity collaborative optimization, balancing high diagnostic accuracy, noise robustness, and lightweight design. It is adaptable to edge deployment and suitable for fault diagnosis of rotating machinery such as bearings and gearboxes.

This invention discloses a method for classifying consciousness disorders in EEG signals based on dynamic graph convolution and channel attention mechanisms, relating to the field of EEG signal recognition technology. According to the method provided in the embodiments of this invention, a complete closed loop is achieved, covering uploading, preprocessing, artifact removal, segmentation, feature extraction, and model prediction. A dynamic graph convolution modeling method with a trainable adjacency matrix is ​​used to adaptively learn functional connections between EEG channels, overcoming the poor generalization problem of static adjacency matrices. Simultaneously, a joint modeling framework of explicit connectivity (PLV) + implicit connectivity (dynamic graph convolution) is used to more robustly capture cross-channel synchronization patterns under low signal-to-noise ratio conditions.

The application discloses a kind of breeding environment multidimensional data integration processing and early warning system.The system includes: the main control computing unit deployed in factory local, multiple sensor anchor point clusters distributed in breeding pool interior and data aggregation gateway;Each sensor anchor point cluster contains basic physicochemical probe component, environmental flow field component and biological observation component;The main control computing unit calculates physical transfer time delay according to three-dimensional flow velocity vector and aligns time series data using dynamic time warping algorithm, and the adjacency matrix between the nodes of internal model is dynamically calculated according to three-dimensional flow velocity vector to output environmental anomaly probability score;When the fluctuation rate of physicochemical parameter or environmental anomaly probability score meets the condition, data aggregation gateway sends hardware interrupt instruction to dormant biological observation component.This application can accurately compensate material propagation data time delay, accurately identify abnormalities, and reduce energy consumption using cross-modal wake-up mechanism, to ensure system space-time causal closed loop and bandwidth communication efficiency.

This invention discloses an intelligent risk assessment method and system for dam-break flood propagation in water network nodes, comprising: collecting basic data of water network engineering and dam-break scenario parameters to generate a standardized dataset; constructing a directed topological structure model of water network nodes-channels that integrates engineering semantic constraints and hydraulic propagation direction constraints, and generating a water network adjacency matrix; using a neural network model based on water network structural constraints and the physical knowledge of mass and momentum conservation to predict the propagation process of dam-break floods in the water network; converting the flood propagation prediction results into flood impact indices, calculating the risk levels of engineering facilities and population units based on the flood impact indices, and outputting risk assessment results and decision support information. This invention provides decision support information for water network scheduling and emergency management.

The application discloses a power grid power flow prediction method and system based on a physical constraint graph convolution network, and relates to the technical field of power system power flow prediction.The method comprises the following steps: acquiring the topological structure and operation data of a power grid to be analyzed, constructing an adjacency matrix, and constructing a node feature matrix and an edge label matrix; inputting the node feature matrix into a graph convolution network for topological perception learning, and acquiring node representation; mapping the node representation into initial branch power flow prediction values through a bus-branch correlation mapping layer; adjusting the initial branch power flow prediction values through a node power balance correction layer by using a correction matrix, and obtaining corrected branch power flow prediction values; and combining a root mean square error and a physical regularization loss function to iteratively update model parameters, obtaining a power flow prediction model, and performing online reasoning.The application solves the problem of node power imbalance of a traditional data-driven model, and significantly improves the physical consistency and accuracy of power flow prediction.

This invention discloses a method, apparatus, device, and storage medium for predicting transformer overload. By calculating the multi-dimensional dynamically coupled adjacency matrix of the transformer, this adjacency matrix can be used in conjunction with a spatiotemporal attention bidirectional feedback network to effectively improve the accuracy of overload prediction and reduce the false judgment rate under extreme conditions. Furthermore, both the multi-dimensional dynamically coupled adjacency matrix and the spatiotemporal attention bidirectional feedback network are lightweight parallel computing architectures, resulting in a short prediction time for the target transformer's load rate and low resource consumption, which can meet the real-time dispatching requirements of the power grid. The invention also obtains overload and overload benchmark thresholds based on the historical number of overloads in the target transformer's static equipment dataset. These thresholds are used to predict whether the target transformer is overloaded or overloaded, allowing for threshold settings based on the specific conditions of the target transformer to reflect individual differences and further improve prediction accuracy.

The application discloses a kind of based on graph structure's power grid voltage stability margin online prediction method, comprising the following steps: constructing power grid attribute graph model in graph database, and establishing event-driven mechanism to realize the incremental update of topological change;Based on parameterized graph query template, dynamically generate graph database query statement, obtain the node feature matrix and adjacency matrix of prediction time, and convert to sparse tensor format;Constructing spatio-temporal graph convolution network prediction model;Through the incremental update of topological change driven by event-driven mechanism, to drive spatio-temporal graph convolution network prediction model;The prediction result and model intermediate layer node embedding vector are written back to graph database as attribute.This application realizes the millisecond level online prediction of power grid voltage stability margin, meets the real-time, accuracy and adaptability requirements of power system online safety analysis.

The present application relates to the technical field of marine data processing and spatio-temporal prediction, and specifically discloses a sea surface core variable prediction method and system based on a wavelet-enhanced graph neural network.The method comprises the following steps: obtaining multivariate graph sequence data of a target sea surface; performing multilevel wavelet decomposition and gated fusion on each variable through a variable-level multiscale wavelet gated fusion module, and outputting enhanced multiscale time series features; inputting the enhanced multiscale time series features into a KAN-LSTM encoder module, performing recursive updating through gated fusion of a conventional convolution and a KAN convolution branch, and outputting encoder spatio-temporal features; inputting the encoder spatio-temporal features into a signed adaptive spatial graph convolution module, learning a signed sparse adaptive adjacency matrix and performing multi-order diffusion aggregation, and outputting a prediction result.The present application realizes long-term prediction of a target sea surface with high precision and high stability.

The application discloses a source-load joint probability prediction method and system of a physically constrained graph attention network. The method collects multi-dimensional feature data of source-load nodes in a prediction area to construct an initial node feature matrix. A similarity matrix is generated through differentiable graph structure learning. A dynamic adjacency matrix is generated through normalization and introduction of a sparse mask. Spatial feature aggregation is performed through a multi-head graph attention network to obtain node spatial encoding. The node spatial-temporal hidden state is output through an encoder. The node spatial-temporal hidden state is input into a probability prediction head to output Gaussian distribution parameters of the source-load node power. A joint loss function is constructed. The joint loss function is used for soft constraint training to output a probability prediction result. Posterior projection hard constraint correction is performed in an inference stage to obtain a corrected prediction result. The application solves the problems of lack of physical consistency and inability to quantify uncertainty in the prior art.

This invention discloses a traffic flow prediction method and system based on Transformer, belonging to the fields of intelligent transportation and deep learning technology. Addressing the technical problems of existing traffic flow prediction models, such as difficulty in simultaneously considering long-term and short-term dependencies, inability of static road network topology to characterize dynamic spatial heterogeneity, and poor modeling performance of spatiotemporal feature coupling, this invention proposes a multi-timescale adaptive graph attention Transformer model. This method first reconstructs the original traffic data at low, medium, and high time scales, and then aggregates spatiotemporal features through a temporal convolutional network and a compressed excitation network. Next, an adaptive data graph generation module learns node embedding vectors to generate an adaptive adjacency matrix that integrates static topology and dynamic associations. Finally, an encoder incorporating temporal one-dimensional convolutional multi-head attention and spatial graph attention, and a decoder integrating causal convolution and temporally gated convolution, are constructed to achieve high-precision multi-step prediction of traffic flow. This invention effectively captures the spatiotemporal dependencies of traffic flow, with prediction accuracy and generalization superior to mainstream models, and can be widely applied to urban intelligent traffic management, dynamic path planning, and traffic congestion mitigation scenarios.

The present application relates to clustering analysis technology in data mining and high-level fusion technology in information fusion, and belongs to the field of pattern recognition and intelligent information processing. It includes: 1) setting the attributes and type labels of the target; 2) representing the target data in the form of a space-time graph; 3) designing a space-time graph autoencoder with an attention mechanism, learning node representation by aggregating adjacency matrix information, and reconstructing the space-time graph network structure by calculating the inner product of node pairs; 4) constructing a self-training graph neural network model to aggregate the reduced neighbor target information; 5) designing a graph convolutional neural network and a deep neural network double self-supervised module; step 6, setting the target behavior rule label; 6) visualizing the target behavior rule. The method can solve the problems of traditional clustering methods such as strong data dependence, high computational complexity and inaccurate measurement description, and realize efficient mining and analysis of target behavior rules.

The application relates to an industrial quality prediction method and equipment based on a graph attention long short-term memory network, which comprises the following steps: acquiring a data set to be predicted in an industrial manufacturing process, performing a standardization operation on the data set to obtain a standardized data set; acquiring a low-order feature variable matrix of the standardized data set through a multilayer perception machine, and constructing an adjacency matrix; inputting the standardized data set, the adjacency matrix and the low-order feature variable matrix into a trained graph attention long short-term memory network model, splicing time feature information and space feature information output by the model through a quality estimator, and obtaining a quality prediction result; the graph attention long short-term memory network model comprises a graph attention layer and a long short-term memory layer arranged in parallel, and is used for extracting the time feature information and the space feature information, respectively. Compared with the prior art, the application has the advantages of fusing time-space features, automatically constructing an adjacency matrix and improving prediction accuracy.

This invention discloses a fault diagnosis method for aero-engines based on an adaptive spatiotemporal decoupled graph convolutional network, comprising: acquiring and preprocessing multi-channel sensor signals to construct time-series samples; adaptively constructing an adjacency matrix to form a fault signal graph structure representing the spatial correlation of multi-source signals; inputting the fault signal graph structure into a graph convolutional network to extract spatial features; inputting the extracted spatial feature sequence into a bidirectional gated recurrent unit network to obtain temporal features; introducing a global attention mechanism to the temporal features to assign adaptive weights to features at different time steps; decoupling the spatiotemporal features using a variational autoencoder decoupling layer based on a Gaussian mixture distribution; and outputting the fault category of the aero-engine to complete the fault diagnosis. This invention significantly improves the accuracy, robustness, and engineering applicability of aero-engine fault diagnosis under complex operating conditions by jointly modeling the spatiotemporal characteristics of fault signals and effectively decoupling and fusing features.

The present disclosure provides biomarker screening model training method and device, method and device, network, equipment and medium, and relates to the technical field of image processing. The implementation scheme of the present disclosure is: a heterogeneous graph construction module, configured to convert an obtained target data set into a heterogeneous graph structure; a double-flow graph convolution network module, wherein a protein interaction flow network is configured to output a first graph-level feature based on a first adjacency matrix; a gene regulation flow network is configured to output a second graph-level feature based on a second adjacency matrix; an attention fusion classification module is configured to weight and fuse the first graph-level feature and the second graph-level feature, and input the fused feature into a classifier to obtain a prediction probability value for a target disease category; and an explainable attribution module is configured to integrate a gradient along a path from a baseline input to an actual input, quantify a contribution score of each protein node feature in the heterogeneous graph structure to the prediction probability value, and output a candidate biomarker combination according to the contribution score.

This invention provides a network attack and defense adversarial testing method and system based on deep reinforcement learning, belonging to the field of network security technology. The method includes constructing a network topology adjacency matrix and performing spectral decomposition; calculating structural distances based on node spectral coordinates to divide attack clusters and configuring attack agents; the attack agents executing attack actions to obtain rewards; recording node state changes, calculating attack propagation speed, and identifying dominant paths, encoding these paths as feature vectors and inputting them into defense agents; the defense agents executing defense actions to obtain rewards; and updating the agents' network parameters through reinforcement learning using the reward feedback from both sides. This invention can automatically identify critical attack propagation paths, achieving dynamic and adaptive attack and defense adversarial testing, improving the realism and automation level of the test.

This invention provides a method, system, and medium for predicting waterway vessel congestion based on multi-agent collaboration. The prediction method includes: at each time step, dynamically calculating the connectivity weights between nodes based on the vessel traffic flow characteristics, waterway environmental characteristics, and historical congestion characteristics of the current node, generating an adjacency matrix for the current time step, and constructing a dynamic waterway topology graph; inputting the dynamic waterway topology graph and corresponding node features into a two-branch temporal graph convolutional network, outputting short-term and long-term feature vectors; using a deep reinforcement learning-driven adaptive gating fusion module to generate a fusion weight vector in real time, and weighting and fusing the short-term and long-term feature vectors according to the fusion weight vector to generate multi-scale spatiotemporal fusion features; inputting the multi-scale spatiotemporal fusion features into a full-cycle prediction output head, and outputting short-term, medium-term, and long-term waterway congestion prediction results in parallel. This invention significantly improves prediction accuracy and adaptability.

The application provides a driving fatigue recognition method and system based on dynamic graph neural network and multi-scale convolution, relates to the technical field of driving fatigue recognition, and comprises the following steps: S1, collecting multi-channel electroencephalogram signals, and dividing the signals into sample segments composed of time sequence data of multiple channels after pretreatment; S2, inputting the sample segments into a multi-scale convolution neural network, extracting and fusing time sequence fusion features; S3, inputting the time sequence fusion features into a mixed attention mechanism for processing, and then performing dynamic adjustment through a convolution guided attention mechanism to construct key time sequence dependency relationships and output time sequence enhancement features; S4, taking the time sequence enhancement features as node features, performing spatial feature propagation and fusion through a graph convolution network based on a trainable adaptive adjacency matrix, and outputting spatial embedding features corresponding to each channel; and S5, inputting the spatial embedding features into a classifier and outputting corresponding fatigue state recognition results.

This invention relates to the field of intelligent brain disease diagnosis technology, specifically providing an auxiliary diagnostic method and system based on multimodal decoupled dynamic graph learning. The method includes: acquiring and preprocessing multimodal data (such as neuroimaging, genetic markers, etc.) of the subject; extracting common pathological information and modality-specific features through a shared encoder and modality-specific encoders respectively, and optimizing the separation process using a decoupling loss function; furthermore, fusing all modality embeddings using a multi-head self-attention mechanism with a masked matrix to generate initial node representations, where the mask is used to suppress modality self-attention; subsequently, performing hierarchical dynamic graph convolution based on the node representations: in each layer, dynamically updating the graph adjacency matrix by combining the current node representation with the original features, and iteratively optimizing the node representations through message passing; finally, inputting the optimized representations into a classifier to obtain disease prediction results. This invention improves the automation performance and reliability of diagnosis.

The present application relates to the technical field of electric power metering, and provides a voltage transformer error prediction model construction method, an error evaluation method and a device, the construction method comprising: collecting voltage and current phasors of a monitoring node and obtaining three-phase admittance parameters; constructing a node feature matrix based on the voltage and current phasors, establishing a graph adjacency matrix based on the three-phase admittance parameters, and using the node feature matrix and the graph adjacency matrix to perform error prediction by using a graph neural network to obtain voltage and current prediction errors; obtaining voltage estimated true values based on the voltage phasors and the voltage prediction errors, obtaining current estimated true values based on the current phasors and the current prediction errors, and determining voltage drop and current balance constraint losses of a three-phase transmission line based on the voltage estimated true values and the current estimated true values; and performing parameter iteration on the graph neural network based on the voltage drop and the current balance constraint losses to obtain an error prediction model. The method and device provided by the present application can maintain stable performance under different operating conditions and network topologies.

The application provides a driving intention recognition method based on functional connectivity and a graph neural network, comprising the following steps: S1, collecting electroencephalogram signals of a driver during driving, and preprocessing original electroencephalogram data; S2, calculating power spectral densities of each frequency band of the preprocessed electroencephalogram signals as electroencephalogram signal frequency domain features; S3, constructing an adjacency matrix as an initial graph structure based on electrode spatial proximity and functional connectivity; and S4, inputting the frequency domain features in S2 and the adjacency matrix obtained in S3 into a graph attention network for feature aggregation, inputting the extracted feature expression into a classifier to classify driving intentions and outputting results. The method can solve the problems of single electroencephalogram feature extraction and poor expression ability of the classification model in the existing driving intention prediction method, improve the accuracy of driving intention classification, improve the interpretability of the classification results, and better apply the method to driving state perception and auxiliary decision-making of a human-machine co-driving system.

This invention discloses a broadband oscillation disturbance source localization method based on compressed sensing and graph convolutional neural networks (GCNs), comprising two stages: offline training and online localization. In the offline training stage, electrical quantities under subsynchronous / supersynchronous broadband oscillation modes are obtained using measurement data from actual systems or simulation examples, constructing a broadband oscillation offline sample library. Compressed sensing is used to encode and compress the electrical quantities in the offline sample library, obtaining a training set and a test set. The constructed GCN localization model is trained using the training set until the test set reaches the target localization accuracy, resulting in a broadband oscillation localization model. In the online localization stage, a substation collects electrical quantity data and compresses and encodes it using compressed sensing technology; the substation data is uploaded to the master station; at the master station, the feature matrix and system adjacency matrix are input into the trained GCN localization model, and the oscillation source location is output. This invention's method is applicable to broadband oscillation disturbance source localization under various operating conditions.

This invention discloses an ecological management zoning method and system based on static and dynamic supply and demand matching, belonging to the field of ecological management technology. The invention constructs a node feature matrix by calculating a comprehensive static supply and demand matching index and a comprehensive dynamic change trend index, then generates a dynamic spatiotemporal adjacency matrix to construct a spatiotemporal physical connectivity graph. This graph is then input into a spatiotemporal graph neural network for processing, outputting predicted static supply and demand matching indices and predicted dynamic change trend indices. Based on the zero-value boundaries of these indices, the study area is divided into initial four-level basic management zones. A Markov decision process is then constructed, and a Pareto optimal solution set is obtained through reinforcement learning algorithms to fine-tune the spatial boundaries of the initial four-level basic management zones. The resulting refined ecological management zoning map and management priority sequence are output, making the zoning results more targeted for management. This solves the problem that existing ecological management zoning results lag behind actual changes and are difficult to effectively support adaptive management.

The application belongs to the technical field of oil well fault detection, and particularly relates to an oil well sensor spatial relationship modeling method based on centrality guided graph convolution, which comprises the following steps: firstly, valid nodes are reserved through node integrity maintenance, and state and trend features are extracted by using a double-view feature generation method; secondly, a weighted hybrid adjacency matrix is constructed by fusing prior physical connection and data-driven correlation; then, based on the weighted hybrid adjacency matrix and the node centrality score, a weight parameter for guiding spatial information aggregation is generated; finally, based on the weight parameter, centrality guided graph convolution operation is performed on a two-dimensional feature vector of the sensor node to update the node feature, and spatial relationship modeling is completed. The application introduces and optimizes a physical graph topology in oil well fault detection for the first time, effectively models the complex spatial dependence between sensors, significantly improves the interpretability, accuracy and robustness of a subsequent fault detection task, and is particularly suitable for identifying long-time evolving faults.

This invention relates to the field of environmental anomaly early warning technology, and discloses an automatic early warning system for environmental anomalies in urban ecologically sensitive areas based on spatiotemporal data mining. The system includes: extracting intermediate center scalars; calculating rainfall concentration coefficients; generating bottleneck boundary retention and collapse coefficients; generating a dynamic adjacency matrix; outputting predicted future anomaly intensity values; generating cumulative anomaly energy; and outputting automatic early warning values ​​for the entire area. This invention constructs structural segments by spatially overlapping the catchment area of ​​the total urban ecologically sensitive area with riverside or boundary strip areas. It uses intermediate center scalars to identify topological bottleneck segments, employs a dynamic adjacency matrix to allow the propagation weights between structural segments to change in real time with the boundary hydraulic conditions, and integrates the bottleneck boundary retention and collapse coefficients into the gating mechanism of the spatiotemporal graph convolutional network to achieve adaptive adjustment of the model's time memory. It outputs standardized early warning values ​​for the entire area, adapting to practical application scenarios for environmental management in urban ecologically sensitive areas.

The application provides an article classification method and device based on graph attention diffusion, the article classification method comprising: obtaining article information of an article to be classified, and constructing graph structure data of the article to be classified; determining an attention coefficient between two adjacent article nodes in the graph structure data based on a latent representation vector of each article node and using a graph attention mechanism; performing iterative attention diffusion on an attention matrix based on an adjacency matrix of the graph structure data and a jump back probability coefficient, to obtain a deep-level attention matrix; and performing feature aggregation on the latent representation vector based on the deep-level attention matrix, to obtain an attention aggregation feature of each article node, so as to determine a classification result corresponding to each article to be classified. Through the above method, the information from a high-order neighborhood is quantitatively aggregated, and an over-smoothing phenomenon is avoided, thereby improving the accuracy and stability of processing an article classification task.

A multi-agent deep reinforcement learning method for optimizing the photovoltaic (PV) carrying capacity of distribution networks based on graph attention mechanism includes: acquiring information on the access nodes and access capacity parameters of various controllable devices; establishing a distribution network PV carrying capacity optimization model considering various controllable devices; constructing a MADRL model based on the PV carrying capacity of the distribution network, and defining the state space, action space, reward function, state transition function, and discount factor of the MADRL model's Markov decision process; acquiring the distribution network topology, constructing an adjacency matrix, and constructing a graph attention network based on the state space information and adjacency matrix, embedding deep reinforcement learning (DRL) Actor and Critic networks, and training based on typical daily load and normalized irradiance data; after training, saving the agent weight results to achieve optimal control of the distribution network carrying capacity. This method can explicitly utilize the distribution network graph structure information, enhance the collaborative control capability of multiple devices, and improve the stability and robustness of the strategy in different scenarios.

This invention proposes a collaborative evolutionary classification method based on cross-view structural attribute dual-domain semantic mining, belonging to the field of multi-view semi-supervised classification. This invention acquires multi-view features from training samples, constructs K-nearest neighbor graphs for each view, and adaptively prunes based on the shared support relationships of the graph structure to obtain a purified adjacency matrix. Subsequently, structural and attribute representations are extracted through a shared graph convolutional network, and the consistency between the two in a unified space is constrained using a structure-attribute consistency loss. Simultaneously, learnable view position encoding and masked self-attention mechanisms are introduced for progressive fusion, and cross-view evolutionary patterns are learned through adjacent view mapping to form an evolutionary loss. Finally, the evolutionary representations of each view are concatenated and mapped to a sample-level fused representation, which is then input into the classifier for prediction. The model is jointly optimized by the classification loss, consistency loss, and evolutionary loss. Compared with other methods, this invention significantly improves the accuracy of classification results with a small amount of labeled data.