1341 results about "Sequence Feature" patented technology

The invention discloses an adaptive teaching real-time feedback method based on multi-modal fusion, and the method comprises the following steps: synchronously collecting text modal information, voice modal information and image modal information generated by students in a teaching process, forming multi-modal original data information, and extracting historical student interaction behavior data; preprocessing the multi-modal original data information, and respectively generating corresponding text, voice and image sequence features; a visual feature encoder and a sequence feature encoder are adopted to encode each modal sequence feature to obtain a high-dimensional feature; inputting the modal high-dimensional features into a cross-modal fusion network for deep fusion; parameters of the feedback model are optimized through a model-independent element learning feedback regulation and control algorithm, and a personalized feedback strategy is generated; generating comprehensive feature representation according to the fusion features, and outputting personalized teaching feedback; and the interaction information is updated based on the feedback behavior data to realize closed-loop optimization.

The invention discloses an identification method and system of a short antibacterial peptide sequence, a terminal and a storage medium. Relates to the technical field of bioinformatics. A short antibacterial peptide data set is obtained in advance, a training set is obtained through division according to the short antibacterial peptide data set, and a pre-trained protein language model is finely adjusted according to the training set; obtaining sequence features of a to-be-classified sequence through the fine-tuned protein language model; obtaining physicochemical characteristics of the to-be-classified sequences through an analysis method based on pseudo amino acid components; inputting the sequence features and the physicochemical features into an attention model for feature fusion to obtain fusion features of the to-be-classified sequence; and inputting the fusion features into a classifier to obtain a prediction category of the to-be-classified sequence. According to the method, the pre-training model is finely adjusted by establishing a more perfect short antibacterial peptide data set, the sequence features and the physicochemical features of the to-be-classified sequences are fused, and the sequences are classified through the fused features, so that the recognition accuracy of the short antibacterial peptides can be effectively improved.

The invention provides a marketing path dynamic generation method and device based on behavior preferences, equipment and a medium, and the method comprises the steps: obtaining user behavior data, and constructing a user behavior graph structure according to the user behavior data, the complex association comprising similarity between users, association between contacts and interaction weights of the users to the contacts; processing short-term behavior records of the user according to a Transform network, extracting sequence features, storing statistical features of long-term behaviors of the user according to a long-term memory network, and based on a user behavior graph structure, fusing the sequence features and the statistical features to obtain fused features; by constructing the user behavior preference graph and fusing the short-term and long-term behavior characteristics of the user, accurate modeling of the user behavior preference is realized. The graph structure not only captures the global structure of the user behavior, but also can dynamically reflect the current interest of the user; according to the method, the path candidate set is dynamically generated and optimized by adopting reinforcement learning and combining real-time feedback.

The invention relates to the field of drug-target affinity prediction, in particular to a drug-target affinity prediction method based on a multi-scale mixed attention network. According to the method, global residue characteristics can be learned by protein pocket residues by utilizing a self-attention mechanism, so that the global characteristics of the protein are learned. In addition, a cross-modal feature fusion mechanism is utilized, protein and drug structural features are enhanced in sequence features, cross-modal feature fusion in entities is enhanced, meanwhile, a cross-entity interaction module is utilized to recognize key atoms of drug molecules or key residues of protein, and therefore the overall performance of the model is improved. Finally, multi-level protein feature extraction is further introduced to further extract protein features.

The invention discloses a YOLOv8 improved model training method and a ship detection method, and belongs to the technical field of target detection. According to the method, on the basis of a sample ship remote sensing image and a sample ship labeling remote sensing image, a YOLOv8 improved model comprising an OBB rotation detection model, a feature extraction network RFAnet, a dynamic attention scale sequence feature fusion network DSASS F and an SPPFLSKA module is trained. The method can effectively cope with the diversity of the target ship in scale, shape, direction and attitude, accurately captures the characteristics of the ship, and improves the ship detection precision and robustness in a complex marine environment.

The invention discloses a drug target affinity prediction method and system based on a multi-scale protein attention mechanism, and belongs to the crossing field of bioinformatics and artificial intelligence. The method comprises the following steps: firstly, extracting protein sequence features through an ESM2 pre-training model, predicting that a three-dimensional structure is converted into a two-dimensional contact graph, and extracting spatial topological information in combination with a graph convolutional network; a two-dimensional attention mechanism is innovatively designed, structural features are taken as query vectors, sequence features are taken as key value pairs, and cross-modal feature fusion is realized by dynamically associating sequence semantics and spatial proximity relationships through multiple attention. Drug molecules are characterized by adopting MACCS fingerprints, are spliced with protein multi-modal features and then are optimized through a deep network, and finally an affinity value is output through a regression prediction module. According to the technology, the problem of protein heterogeneous data fusion is effectively solved, the generalization ability to unknown targets is remarkably improved, an efficient calculation tool is provided for new drug research and development and drug relocation, and the drug research and development cost can be reduced.

The invention provides a multi-modal multi-scale feature fusion-based power load prediction method and system, and relates to the technical field of power load prediction. According to the method, a cross-modal attention mechanism is adopted to fuse multi-scale load time sequence features and satellite image features; decoding the fusion features through a multi-step prediction decoder to obtain a multi-step power load prediction result; the acquisition mode of the satellite image features comprises the following steps: performing feature extraction on a satellite image based on a GRVT model to obtain the satellite image features; the GRVT model includes a number of stacked gated residual encoder network elements. According to the method, satellite image feature extraction is carried out based on the GRVT model, the global information processing advantage of Transform is utilized, the single-layer information retaining capability of a residual connection mechanism is also utilized, the two are combined to efficiently extract satellite image features, the satellite image features are fused during power load prediction, and the situation that the data type is single, and the prediction efficiency is high is avoided. And the precision of power load prediction is improved.

The invention relates to the technical field of genomics, in particular to a gene detection data cleaning method and system based on artificial intelligence. The method comprises the following steps: acquiring gene detection data, and performing preliminary filtering on the gene detection data based on removal of low-quality reads to obtain preliminary filtering data; performing linker sequence identification and removal on the preliminarily filtered data, and performing sequencing error correction to generate gene sequence data; performing GC content analysis and repetitive sequence identification on the gene sequence data to obtain sequence feature data; performing dynamic quality threshold adjustment according to the sequence feature data, and performing secondary filtering on the gene sequence data to generate high-quality sequence data; and performing sequence pattern recognition on the high-quality sequence data to obtain sequence pattern feature data. According to the method, deep analysis is carried out by utilizing sequence characteristics such as GC content and nucleotide distribution, and the recognition capability on sequence complexity and variation modes is improved.

The invention discloses a health risk assessment method and early warning system based on multi-source data analysis. The health risk assessment method comprises the following steps: S1, collecting and preprocessing multi-source health data through medical detection equipment; s2, extracting key health indexes, sequence features and statistical features based on the health data set; s3, adopting a recurrent neural network model to construct a health risk assessment model; s4, optimizing structural parameters of the health risk assessment model by adopting an improved dragonfly algorithm; s5, performing performance evaluation on the health risk evaluation model by using the optimal parameter set; s6, deploying the final health risk assessment model in a health risk assessment and early warning system; and S7, when the health risk assessment result exceeds a preset risk threshold, automatically generating early warning information. According to the method, the recurrent neural network model, the improved dragonfly algorithm and the multi-source health data fusion optimization technology are combined, and health risk assessment and early warning based on multi-source data analysis are realized.

The invention provides a drug-disease association prediction method and system based on heterogeneous node sequence representation, and belongs to the technical field of drug discovery in bioinformatics. Constructing a drug similarity network and a disease similarity network; constructing a heterogeneous biological interaction network by using the drug similarity network and the disease similarity network, and obtaining a first-order neighbor subgraph of the target node; splitting the first-order neighbor subgraph of the target node into a plurality of meta-relation bigraphs according to the meta-relation type, performing intra-domain message propagation and aggregation in each meta-relation bigraph, and updating the semantic features of the target node; fusing the semantic features of the target nodes in a cross-domain manner, updating the sequence features of the target nodes, and aggregating the features; obtaining vector representation of a target node through a multi-head self-attention mechanism by using the updated sequence features; and inputting the vector representations of the drug nodes and the disease nodes into a multi-layer perceptron to predict the drug-disease association probability. Semantic information confusion is effectively avoided, and then the accuracy of model prediction is improved.

The invention discloses a power grid communication equipment fault identification method and system based on multi-source heterogeneous data characterization learning. The method comprises the following steps: firstly, collecting various complex multi-source heterogeneous data containing fault power grid communication equipment and preprocessing the complex multi-source heterogeneous data; extracting deep semantic features by using a pre-trained feature extraction network and obtaining a feature vector sequence; capturing time features of the feature vector sequence through a bidirectional long-short-term memory network, and obtaining complete sequence feature representation through a full connection layer; performing low-rank multi-modal fusion on the complete sequence feature representation in combination with cross-modal Transform to obtain three types of output splicing construction joint representation, and obtaining a fusion feature by using a self-attention mechanism; and finally, constructing a fault identification model by using the fusion features. Through innovative data preprocessing and data fusion technologies and in combination with an advanced deep learning model, fault-related characteristics can be automatically learned from multi-source heterogeneous data, and intelligent identification of the fault state of the power grid communication equipment is realized.

The invention discloses a nucleic acid binding protein recognition method based on a protein map and a protein language model, and belongs to the technical field of bioinformatics. According to the method, high-dimensional sequence features extracted by a ProtT5 protein language model are combined with three-dimensional structure information predicted by AlphaFold, so that the model can capture rich feature information of the sequence and the structure at the same time, and the accuracy and reliability of NABPs prediction are remarkably improved; the expression capability of sequence features is enhanced through the GAT network and the BiLSTM network, so that the constructed prediction model can more comprehensively understand important information in the protein sequence, and the prediction performance of NABPs is improved; the features output by the GAT network and the BiLSTM network are fused, and NABPs classification is performed through the MLP, so that the prediction model can make full use of feature information from different sources, and the NABPs prediction performance and robustness are improved.

The invention provides a multi-modal protein function prediction method and system, and the method comprises the following steps: carrying out the processing of a one-dimensional convolution layer and a pooling layer of protein sequence data, extracting the local dependence features in a protein sequence, and capturing the mode and key sequence features in the protein; using protein sequence data as input, and using a large-scale pre-trained protein language model to extract residue-level embedded information in a protein sequence as node features to be applied to the contact graph; the method comprises the following steps: taking protein sequence data as input, predicting the probability of contact between protein residues based on a Cb-Cb distance, and constructing a contact graph; and performing joint processing on features of different modes by using a multi-head attention mechanism and a residual structure, integrating data from different modes through feature fusion, and fusing the extracted key sequence features, the residue level embedded information and the contact graph. The prediction method is accurate in prediction and good in performance.

The invention provides a polar region sea ice thickness inversion method based on multi-source remote sensing data fusion, and the method comprises the steps: carrying out the feature extraction of SAR data, MODIS data, AMSR2 data and MIRAS data in a target region, constructing an image and sequence feature data set, carrying out the fusion of the image and sequence feature data set through employing a designed multi-modal data fusion network, and obtaining a polar region sea ice thickness inversion result. A new sequence feature data set is obtained, multiple machine learning regression models are used for training the sequence feature data set, the machine learning regression model with the optimal index is selected for sea ice thickness inversion, optical data are deleted, and a sea ice thickness inversion model is obtained through re-training; and dividing the target area into a clouded area and a cloudless area by using the cloud mask data, and applying corresponding inversion models to the clouded area and the cloudless area respectively. According to the method, the advantages of various remote sensing data in sea ice thickness inversion are fully fused, continuous large-area ice thickness information extraction of sea ice can be realized, and the method has better universality.

The invention discloses a double-flow coding landslide automatic identification method based on improved U-Net, and the method achieves the precise extraction of a landslide through the multi-dimensional features of an optical wave band, a spectral index, a terrain index, a texture index and the like. In order to make full use of sequence features after multi-dimensional feature superposition, the encoder of the invention designs two branches, one branch constructs a residual encoding module to extract spatial dimension features, and the other branch constructs a multi-scale ResNet Module (MSRNM) and a long short-term memory (LSTM) to extract sequence dimension features. Designing a fusion module to effectively fuse the spatial dimension features and the sequence dimension features to obtain deep features; and the decoder is provided with an attention gating decoding module (AGDM), so that the expression of deep layer features and shallow layer features is enhanced, and the sensitivity of the model to the landslide is remarkably improved. The constructed double-flow coding landslide identification model has good space and sequence feature learning ability, and can finely identify various landslides.

The invention provides a pre-training enhanced space-time Transform traffic flow prediction method, which belongs to the field of intelligent traffic, aims to improve the prediction performance in a complex traffic environment by designing a space-time Transform model and a pre-training strategy, and comprises the following steps of: firstly, pre-training an encoder through a mask automatic encoding strategy by utilizing long-term historical traffic flow data, and then pre-training the encoder through a mask automatic encoding strategy; extracting long sequence features by adopting time perception attention embedded with time period information; constructing a pre-training space-time sparse graph generated by a data-driven graph learning layer to capture a dynamic spatial dependency relationship; in the prediction stage, a time feature extraction module and a spatial feature extraction module are designed, an improved multi-head self-attention mechanism is utilized to model correlation between time and space, and long sequence features and spatial-temporal features are adaptively fused through a gating mechanism to obtain a prediction result. According to the method, long-term historical traffic flow sequence information can be effectively extracted, and the traffic flow can be predicted more accurately under a complex traffic condition.

The invention discloses a log anomaly detection method and device, equipment and a storage medium, and relates to the technical field of artificial intelligence, the method comprises the following steps: extracting feature data from log data, the feature data comprising log sequence features and log semantic features; inputting the log sequence features into a pre-trained strategy network to obtain a prediction result of a log sequence level; inputting the log semantic features into a fine-tuned large language model to obtain a reasoning result of a log semantic level; generating a reward signal based on the prediction result of the log sequence level and the reasoning result of the log semantic level; performing reinforcement learning fine tuning on the strategy network by using the reward signal to obtain an anomaly detection model; and performing anomaly detection on the log data based on the anomaly detection model, and outputting an anomaly detection result.

The invention discloses a multi-feature learning online video popularity prediction method based on time information perception. The method comprises the following steps: acquiring multi-feature data and playing time sequence data related to a to-be-predicted video; inputting the multi-feature data into feature sub-encoders in a preset multi-feature encoder according to categories, and determining visual, acoustic, text, social and metadata features; inputting the playing time sequence data into a hierarchical popularity sequence feature extraction network in a preset multi-feature encoder to determine popularity sequence features of the to-be-predicted video; inputting the visual, acoustic, text, social, metadata and popularity sequence features of the to-be-predicted video into a preset feature fusion network for feature fusion; and inputting the fused to-be-predicted video features into a preset popularity prediction network to obtain a prediction result. Through the technical scheme of the invention, various features related to the video popularity can be comprehensively considered, and the popularity prediction performance is improved.

The invention provides a drug target prediction method based on a 3D structure and a multilevel attention mechanism, and belongs to the technical field of drug target prediction. The technical problem to be solved is to provide a drug target prediction method based on a 3D structure and a multi-level attention mechanism. According to the technical scheme, the method comprises the following steps: collecting data information of an interaction tag among a related drug, a target sequence and a drug target protein, and processing and extracting SMILES character string characteristics of the drug; learning the numeralized protein amino acid sequence by using a convolutional neural network, completing feature coding of a target spot on the sequence through multi-layer convolution and pooling operation, and extracting target spot sequence feature data; a PDB file is read and modeled into a pocket graph, feature calculation is carried out on the pocket graph, a final protein structure representation code is generated, and structural feature data of target protein is extracted by modeling the PDB file; the method is applied to drug target prediction.

The invention discloses a fault diagnosis method and device, computer equipment and a computer readable storage medium, relates to the technical field of power systems and automation thereof, and solves the problem of fault misjudgment caused by important fault characterization of a method which is easy to lose when single modal information is used for fault diagnosis at present. The method comprises the following steps: converting a bus voltage signal into a two-dimensional time-frequency diagram based on a continuous wavelet transform model, and converting the bus voltage signal into a one-dimensional frequency spectrum sequence based on a fast Fourier transform model; performing feature extraction on the two-dimensional time-frequency graph based on a time-frequency image feature extraction model to obtain a first feature vector, and performing feature extraction on the one-dimensional frequency spectrum sequence based on a frequency spectrum sequence feature extraction model to obtain a second feature vector; performing feature fusion on the first feature vector and the second feature vector based on a feature fusion model, and obtaining a category probability vector by using a fault diagnosis model based on the fused feature vector; and determining the highest probability value in the category probability vector, and taking the corresponding fault type as a target fault type.

The invention provides a small target detection method and system based on an improved YOLOv8 structure, and the method comprises the following steps: carrying out the preprocessing of a to-be-detected image, so as to obtain an input standardized image; inputting the standardized image into a YOLOv8 backbone network, and extracting feature maps of different scales; the feature map is input to a YOLOv8 neck network for feature fusion, the neck network introduces an ASF mechanism, a scale sequence feature fusion module and a three-feature encoder module are integrated, cross-scale enhancement and integration are performed on features of different resolutions, and a multi-scale fusion feature map is generated; constructing P2, P3, P4 and P5 detection layers based on the fused feature map; the P2-P5 detection layer predicts candidate target frames and category confidence of the candidate target frames respectively; soft non-maximum suppression is adopted to process the candidate target frame, a linear or Gaussian attenuation function is utilized to adjust confidence, a redundant frame is suppressed, and a final detection result is output. According to the method, the recognition capability of small targets in dense and complex scenes is remarkably improved.

The invention relates to the technical field of network security of a power monitoring system, and discloses an attack event prediction method and system based on low-intensity abnormal mode data, and the method comprises the steps: intercepting abnormal data fragments from historical multi-source heterogeneous data based on the low-intensity abnormal mode data selected from the historical multi-source heterogeneous data; analyzing to obtain a target abnormal data fragment; training the gating loop network model based on target key time sequence abnormal features obtained by performing feature extraction on the target abnormal data fragments to obtain a future time sequence feature prediction model; performing structured processing on the low-intensity abnormal event to obtain a spatial correlation feature topological structure; obtaining future time sequence features based on the future time sequence feature prediction model; and obtaining an attack event prediction result based on an abnormal event evolution graph constructed by fusing the future time sequence feature and the spatial correlation feature topological structure. According to the method provided by the invention, accurate identification can be carried out in the preparation stage of the attack event.

The invention relates to the technical field of computer processing, and provides a data prediction method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring historical business data of a target domain, wherein the historical business data comprises time sequence characteristics and non-time sequence characteristics; decomposing the time sequence characteristics to obtain a season term, a trend term and a residual term; performing prediction based on the trend term and the residual term to obtain trend term change and residual term change, and fusing the trend term change, the residual term change and the seasonal term to obtain an initial data prediction result; and constructing a decision tree based on the initial data prediction result and the non-time-sequence features, and obtaining a final data prediction result based on the decision tree. The problem that complex features cannot be well learned in the prior art is solved.

The invention discloses a text abstract generation method and system based on sparse attention acceleration, and the method comprises the steps: reading long text data, carrying out word segmentation and embedded coding processing, extracting a sequence feature vector, and mapping the sequence feature vector into a query matrix Q, a key matrix K and a value matrix V; constructing an abstract generation network, wherein the abstract generation network comprises a sparse attention calculation module, a feedforward calculation module, a prediction head module and a key value cache module; inputting the query matrix Q, the key matrix K and the value matrix V into an abstract generation network, passing through a backbone network formed by stacking a sparse attention calculation module and a feedforward calculation module for multiple times, processing by a prediction header module, and caching a historical decoding state in real time through a key value caching module to obtain an initial text feature vector; and based on the initial text feature vector, executing an autoregressive decoding process through an abstract generation network, and outputting a final abstract result. According to the method, the decoding process can be accelerated, and the semantic integrity and coherence of the generated abstract are ensured.

The invention provides a property collection monitoring and early warning method and system based on artificial intelligence. The method comprises the following steps: acquiring historical time series data of a user, preprocessing the historical time series data of the user, and extracting features from the preprocessed data; establishing a time weighted attention mechanism model based on features extracted from the divided training set; measuring the difference between the prediction probability of the time-weighted attention mechanism and the actual risk category by using a cross entropy loss function, and adjusting a time-weighted attention mechanism model; the method comprises the following steps: acquiring user time sequence feature data, inputting the user time sequence feature data into an adjusted time weighted attention mechanism model, and adopting a collection strategy according to a model prediction result. The time weighted attention mechanism and the weighted cross entropy loss function can be combined to improve the prediction accuracy of the model for the customer payment risk, so that the collection strategy is adopted in a targeted manner, the collection efficiency is improved, and the resource allocation is optimized.

The invention discloses an antibacterial peptide recognition method and system based on a sequence-structure dual-channel neural network, and the method comprises the steps: splicing amino acid features and amino acid-level manual features extracted by ProtT5 to obtain peptide embedding, and transmitting the peptide embedding to a sequence channel composed of a plurality of Transform blocks to extract the sequence features of the peptide; predicting a three-dimensional structure of the peptide by using ESM-Fold to construct an adjacency graph, taking amino acid features obtained by ESM-2 as node features of the adjacency graph, and performing layer-by-layer extraction and enhancement by fusing structural channels of multi-head graph attention, a residual network, layer normalization and a feedforward neural network; and carrying out maximum pooling and splicing on the sequence features and the structural features, and then, carrying out antibacterial peptide prediction. According to the method, a multi-feature fusion strategy is adopted, meanwhile, the three-dimensional structure information of the antibacterial peptide is introduced, and the sequence and the structural features are fused through a two-channel architecture, so that the recognition accuracy of the antibacterial peptide is effectively improved.

The invention relates to the technical field of pharmaceutical knowledge maps, in particular to a pharmaceutical knowledge map construction method and system based on artificial intelligence, and the method comprises the following steps: querying and collecting a molecular structure of a drug and a corresponding target protein sequence through a database, and carrying out the numerical coding of the molecular structure data of the drug, molecular fingerprints and protein structural domain features are extracted, and a drug and protein feature set is formed by combining drug chemical attributes and protein sequence features. According to the invention, through accurate analysis of the molecular structure of the drug and the target protein sequence thereof, the innovative scheme significantly enhances the understanding of the interaction of the drug and the protein, so that researchers can directly extract key features from data and monitor the dynamic change of the drug effect, thereby not only accelerating the development process of the drug, but also improving the development efficiency of the drug. By dynamically tracking the interaction between the side effect of the medicine and the pathological characteristics, the scheme provides powerful data support for personalized medical treatment.

The invention provides a multi-modal characterization molecular property prediction method based on hierarchical bidirectional cross attention, and relates to the technical field of machine learning assisted organic chemistry, and the method comprises the following steps: S10, generating same-molecule multiple sequences for data enhancement; s20, coding the sequence features through a pre-trained molecular language model MolBERT; s30, performing multi-modal feature fusion through a layered bidirectional cross attention mechanism; s40, establishing a prediction head; s50, in the reasoning stage, only the feature extraction and fusion steps are executed, and a molecular property prediction result is output through the trained prediction head. According to the method, the molecular sequence, the topological graph structure and the fingerprint features are effectively integrated, so that the prediction precision of the model on a plurality of MoleculeNet (molecular network benchmark) public data sets is superior to that of an existing method.

The invention discloses a radar signal feature representation method based on self-supervised contrast learning. The method comprises the steps of obtaining radar signal time sequence data and performing preprocessing; performing data enhancement to obtain a strong enhancement view and a weak enhancement view, inputting the strong enhancement view and the weak enhancement view into an encoder to obtain a potential embedding representation, inputting the potential embedding representation into a time comparison module to perform a task of cross-view time step prediction, outputting a corresponding context vector, and calculating time comparison loss; inputting context vectors corresponding to the strong enhancement view and the weak enhancement view into a context comparison module to carry out self-supervised comparison learning and calculate context comparison loss; a total loss function is obtained through integration, a model formed by the time comparison module and the context comparison module is trained, a radar signal time sequence feature representation model is obtained, and time sequence dependence, context relation and cross-view consistency in a radar signal time sequence can be efficiently and accurately captured; therefore, the ability to understand radar signals is enhanced, and a complex radar application scene can be dealt with more flexibly.

The invention provides a distribution network traveling wave fault point intelligent positioning method and system based on deep time sequence feature learning, and belongs to the technical field of intelligent power distribution detection based on deep learning. Firstly, a high-speed traveling wave sensor is arranged on a distribution line, multi-dimensional three-phase voltage and current signals are collected, and a data set of fault types, phases, grades and positions is constructed; then, fault state features are extracted through topology perception normalization, symmetric component mapping and two-channel time sequence modeling, and accurate recognition of fault types, related phases and grades is achieved through an attention mechanism; furthermore, a fault intelligent positioning model based on a topological graph is constructed, a tower sensing weight and a line information bearing weight are introduced, and space-time embedding is extracted through a graph attention network and an echo state network, so that line fault classification and accurate positioning are realized. Finally, through combination of off-line model training and on-line system deployment, real-time identification and positioning of power distribution network faults are realized, and accuracy, robustness and response speed of fault diagnosis are effectively improved.