Search

This article addresses the pervasive issue of fraud in financial transactions by introducing the Graph Attention Network (GAN) into graph neural networks. The article integrates Node Attention Networks and Semantic Attention Networks to construct a Dual-Head Attention Network module, enabling a comprehensive analysis of complex relationships in user transaction data. This approach adeptly handles non-linear features and intricate data interaction relationships. The article incorporates a Gradient-Boosting Decision Tree (GBDT) to enhance fraud identification to create the GBDT–Dual-channel Graph Attention Network (GBDT-DGAN). In a bid to ensure user privacy, this article introduces blockchain technology, culminating in the development of a financial anti-fraud model that fuses blockchain with the GBDT-DGAN algorithm. Experimental verification demonstrates the model’s accuracy, reaching 93.82%, a notable improvement of at least 5.76% compared to baseline algorithms such as Convolutional Neural Networks. The recall and F1 values stand at 89.5% and 81.66%, respectively. Additionally, the model exhibits superior network data transmission security, maintaining a packet loss rate below 7%. Consequently, the proposed model significantly outperforms traditional approaches in financial fraud detection accuracy and ensures excellent network data transmission security, offering an efficient and secure solution for fraud detection in the financial domain.

Recent advancements in crime prediction have increasingly focused on street networks, which offer finer granularity and a closer reflection of real-world urban dynamics. However, existing studies on street-level graph representation learning often overlook the variability in node features when aggregating information from neighboring nodes. This limitation reduces the model’s capacity to fully capture the diverse street attributes and their influence on crime patterns. To address this issue, we introduce an end-to-end deep spatio-temporal learning model that employs a graph attention mechanism (GAT) to analyze the spatio-temporal features of 110 call incidents. Experimental results show that our proposed model outperforms existing methods across multiple prediction metrics. Additionally, ablation studies confirm that the GAT’s capacity to capture spatial dependencies within the street network significantly enhances the model’s overall predictive performance.

Trajectory prediction is a key component in the development of applications such as mixed urban traffic management and public safety. Traditional models have struggled with the complexity of modeling dynamic crowd interactions, the intricacies of spatiotemporal dependencies, and environmental constraints. Addressing these challenges, this paper introduces the innovative Social Attention Graph Neural Network (SA-GAT) framework. Utilizing Long Short-Term Memory (LSTM) networks, SA-GAT encodes pedestrian trajectory data to extract temporal correlations, while Graph Attention Networks (GAT) are employed to precisely capture the subtle interactions among pedestrians. The SA-GAT framework boosts its predictive accuracy with two key innovations. First, it features a Scene Potential Module that utilizes a Scene Tensor to dynamically capture the interplay between crowds and their environment. Second, it incorporates a Transition Intention Module with a Transition Tensor, which interprets latent transfer probabilities from trajectory data to reveal pedestrians’ implicit intentions at specific locations. Based on AnyLogic modeling of the metro station on Line 10 of Chengdu Shuangliu Airport, China, numerical studies reveal that the SA-GAT model achieves a substantial reduction in ADE and FDE metrics by 34.22% and 38.04% compared to baseline models.

Drug-resistant epilepsy is frequent, persistent, and brings a heavy economic burden to patients and their families. Traditional epilepsy detection methods ignore the causal relationship of seizures and focus on a single time or spatial dimension, and the effect varies greatly in different patients. Therefore, it is necessary to research accurate automatic detection technology of epilepsy in different patients. We propose a causal-spatio-temporal graph attention network (CSTGAT), which uses transfer entropy (TE) to construct a causal graph between multiple channels, combining graph attention network (GAT) and bi-directional long short-term memory (BiLSTM) to capture temporal dynamic correlation and spatial topological structure information. The accuracy, specificity, and sensitivity of the SWEZ dataset were 97.24%, 97.92%, and 98.11%. The accuracy of the private dataset reached 98.55%. The effectiveness of each module was proven through ablation experiments and the impact of different network construction methods was compared. The experimental results indicate that the causal relationship network constructed by TE could accurately capture the information flow of epileptic seizures, and GAT and BiLSTM could capture spatiotemporal dynamic correlations. This model accurately captures causal relationships and spatiotemporal correlations on two datasets, and it overcomes the variability of epileptic seizures in different patients, which may contribute to clinical surgical planning.

Background: Cancer survival prediction is vital in improving patients’ prospects and recommending therapies. Understanding the molecular behavior of cancer can be enhanced through the integration of multi-omics data, including mRNA, miRNA, and DNA methylation data. In light of these multi-omics data, we proposed a graph attention network (GAT) model in this study to predict the survival of non-small cell lung cancer (NSCLC). Methods: The different omics data were obtained from The Cancer Genome Atlas (TCGA) and preprocessed and combined into a single dataset using the sample ID. We used the chi-square test to select the most significant features to be used in our model. We used the synthetic minority oversampling technique (SMOTE) to balance the dataset and the concordance index (C-index) to measure the performance of our model on different combinations of omics data. Results: Our model demonstrated superior performance, with the highest value of the C-index obtained when we used both mRNA and miRNA data. This demonstrates that the multi-omics approach could be effective in predicting survival. Further pathway analysis conducted with KEGG showed that our GAT model provided high weights to the features that are associated with the viral entry pathways, such as the Epstein–Barr virus and Influenza A pathways, which are involved in lung cancer development. From our findings, it can be observed that the proposed GAT model leads to a significantly improved prediction of survival by exploiting the strengths of multiple omics datasets and the findings from the enriched pathways. Our GAT model outperforms other state-of-the-art methods that are used for NSCLC prediction. Conclusions: In this study, we developed a new model for the survival prediction of NSCLC using the GAT based on multi-omics data. Our model showed outstanding predictive values, and the KEGG analysis of the selected significant features showed that they were implicated in pivotal biological processes underlying pathways such as Influenza A and the Epstein–Barr virus infection, which are linked to lung cancer progression.

Stock index fluctuations are characterized by high noise and their accurate prediction is extremely challenging. To address this challenge, this study proposes a spatial–temporal–bidirectional long short-term memory (STBL) model, incorporating spatiotemporal attention mechanisms. The model enhances the analysis of temporal dependencies between data by introducing graph attention networks with multi-hop neighbor nodes while incorporating the temporal attention mechanism of long short-term memory (LSTM) to effectively address the potential interdependencies in the data structure. In addition, by assigning different learning weights to different neighbor nodes, the model can better integrate the correlation between node features. To verify the accuracy of the proposed model, this study utilized the closing prices of the Hong Kong Hang Seng Index (HSI) from 31 December 1986 to 31 December 2023 for analysis. By comparing it with nine other forecasting models, the experimental results show that the STBL model achieves more accurate predictions of the closing prices for short-term, medium-term, and long-term forecasts of the stock index.

The interaction between microbes and drugs encompasses the sourcing of pharmaceutical compounds, microbial drug degradation, the development of drug resistance genes, and the impact of microbial communities on host drug metabolism and immune modulation. These interactions significantly impact drug efficacy and the evolution of drug resistance. In this study, we propose a novel predictive model, termed GCGACNN. We first collected microbe, disease, and drug association data from multiple databases and the relevant literature to construct three association matrices and generate similarity feature matrices using Gaussian similarity functions. These association and similarity feature matrices were then input into a multi-layer Graph Neural Network for feature extraction, followed by a two-dimensional Convolutional Neural Network for feature fusion, ultimately establishing an effective predictive framework. Experimental results demonstrate that GCGACNN outperforms existing methods in predictive performance.

Medium–long-term photovoltaic (PV) output forecasting is of great significance to power grid planning, power market transactions, power dispatching operations, equipment maintenance and overhaul. However, PV output fluctuates greatly due to weather changes. Furthermore, it is frequently challenging to ensure the accuracy of forecasts for medium–long-term forecasting involving a long time span. In response to the above problems, this paper proposes a medium–long-term forecasting method for PV output based on amplitude-aware permutation entropy component reconstruction and the graph attention network. Firstly, the PV output sequence data are decomposed by ensemble empirical mode decomposition (EEMD), and the decomposed intrinsic mode function (IMF) subsequences are combined and reconstructed according to the amplitude-aware permutation entropy. Secondly, the graph node feature sequence is constructed from the reconstructed subsequences, and the mutual information of the node feature sequence is calculated to obtain the graph node adjacency matrix which is applied to generate a graph sequence. Thirdly, the graph attention network is utilized to forecast the graph sequence and separate the PV output forecasting results. Finally, an actual measurement system is used to experimentally verify the proposed method, and the outcomes indicate that the proposed method, which has certain promotion value, can improve the accuracy of medium–long-term forecasting of PV output.

The accurate prediction of vessel trajectory is of crucial importance in order to improve navigational efficiency, optimize routes, enhance the effectiveness of search and rescue operations at sea, and ensure maritime safety. However, the spatial interaction among vessels can have a certain impact on the prediction accuracy of the models. To overcome such a problem in predicting the vessel trajectory, this research proposes a novel hybrid methodology incorporating the graph attention network (GAT) and long short-term memory network (LSTM). The proposed GAT-LSTM model can comprehensively consider spatio-temporal features in the prediction process, which is expected to significantly improve the accuracy and robustness of the trajectory prediction. The Automatic Identification System (AIS) data from the surrounding waters of Xiamen Port is collected and utilized as the empirical case for model validation. The experimental results demonstrate that the GAT-LSTM model outperforms the best baseline model in terms of the reduction on the average displacement error and final displacement error, which are 44.52% and 56.20%, respectively. These improvements will translate into more accurate vessel trajectories, helping to minimize route deviations and improve the accuracy of collision avoidance systems, so that this research can effectively provide support for warning about potential collisions and reducing the risk of maritime accidents.

Accurate identification of spatial domains is essential in the analysis of spatial transcriptomics data in order to elucidate tissue microenvironments and biological functions. However, existing methods only perform domain segmentation based on local or global spatial relationships between spots, resulting in an underutilization of spatial information. To this end, we propose SECE, a deep learning-based method that captures both local and global relationships among spots and aggregates their information using expression similarity and spatial similarity. We benchmarked SECE against eight state-of-the-art methods on six real spatial transcriptomics datasets spanning four different platforms. SECE consistently outperformed other methods in spatial domain identification accuracy. Moreover, SECE produced spatial embeddings that exhibited clearer patterns in low-dimensional visualizations and facilitated a more accurate trajectory inference.

In the realm of aspect-based sentiment analysis (ABSA), a paramount task is the extraction of triplets, which define aspect terms, opinion terms, and their respective sentiment orientations within text. This study introduces a novel extraction model, BiLSTM-BGAT-GCN, which seamlessly integrates graph neural networks with an enhanced biaffine attention mechanism. This model amalgamates the sophisticated capabilities of both graph attention and convolutional networks to process graph-structured data, substantially enhancing the interpretation and extraction of textual features. By optimizing the biaffine attention mechanism, the model adeptly uncovers the subtle interplay between aspect terms and emotional expressions, offering enhanced flexibility and superior contextual analysis through dynamic weight distribution. A series of comparative experiments confirm the model’s significant performance improvements across various metrics, underscoring its efficacy and refined effectiveness in ABSA tasks.

The online calibration method for current transformers is an important research direction in the field of smart grids. This article constructs a transformer error prediction model based on spatiotemporal integration. This model draws inspiration from the structure of forgetting gates in gated loop units and combines it with a graph convolutional network (GCN) that is good at capturing the spatial relationships within the graph attention network to construct an adaptive GCN. The spatial module formed by this adaptive GCN is used to model the spatial relationships in the circuit network, and the attention mechanism and gated time convolutional network are combined to form a time module to learn the temporal relationships in the circuit network. The layer that combines the time and space modules is used, which consists of a gating mechanism for spatiotemporal fusion, and a transformer error prediction model based on a spatiotemporal correlation analysis is constructed. Finally, it is verified on a real power grid operation dataset, and compared with the existing prediction methods to analyze its performance.

The construction of medical knowledge graphs (MKGs) is steadily progressing from manual to automatic methods, which inevitably introduce noise, which could impair the performance of downstream healthcare applications. Existing error detection approaches depend on the topological structure and external labels of entities in MKGs to improve their quality. Nevertheless, due to the cost of manual annotation and imperfect automatic algorithms, precise entity labels in MKGs cannot be readily obtained. To address these issues, we propose an approach named Enhancing error detection on Medical knowledge graphs via intrinsic labEL (EMKGEL). Considering the absence of hyper-view KG, we establish a hyper-view KG and a triplet-level KG for implicit label information and neighborhood information, respectively. Inspired by the success of graph attention networks (GATs), we introduce the hyper-view GAT to incorporate label messages and neighborhood information into representation learning. We leverage a confidence score that combines local and global trustworthiness to estimate the triplets. To validate the effectiveness of our approach, we conducted experiments on three publicly available MKGs, namely PharmKG-8k, DiseaseKG, and DiaKG. Compared with the baseline models, the Precision@K value improved by 0.7%, 6.1%, and 3.6%, respectively, on these datasets. Furthermore, our method empirically showed that it significantly outperformed the baseline on a general knowledge graph, Nell-995.

As the cornerstone of intelligent transportation systems, accurate traffic prediction can reduce the pressure of urban traffic, reduce the cost of residents’ travel time, and provide a reference basis for urban construction planning. Existing traffic prediction methods focus on spatio-temporal dependence modeling, ignoring the influence of weather factors on spatio-temporal characteristics, and the prediction task has complexity and an uneven distribution in different spatio-temporal scenarios and weather changes. In view of this, we propose a weather interaction-aware spatio-temporal attention network (WST-ANet), in which we integrate feature models and dynamic graph modules in the encoder and decoder, and use a spatio-temporal weather interaction perception module for prediction. Firstly, the contextual semantics of the traffic flows are fused using a feature embedding module to improve the adaptability to weather drivers; then, an encoder–decoder is constructed by combining the Dynamic Graph Module and the WSTA Block, to extract spatio-temporal aggregated correlations in the roadway network; finally, the feature information of the encoder was weighted and aggregated using the cross-focusing mechanism, and attention was paid to the hidden state of the encoding. Traffic flow was predicted using the PeMS04 and PeMS08 datasets and compared with multiple typical baseline models. It was learned through extensive experiments that the accuracy evaluation result is the smallest in WST-ANet, which demonstrated the superiority of the proposed model. This can more accurately predict future changes in traffic in different weather conditions, providing decision makers with a basis for optimizing scenarios.

Time series anomaly detection is very important to ensure the security of industrial control systems (ICSs). Many algorithms have performed well in anomaly detection. However, the performance of most of these algorithms decreases sharply with the increase in feature dimension. This paper proposes an anomaly detection scheme based on Graph Attention Network (GAT) and Informer. GAT learns sequential characteristics effectively, and Informer performs excellently in long time series prediction. In addition, long-time forecasting loss and short-time forecasting loss are used to detect multivariate time series anomalies. Short-time forecasting is used to predict the next time value, and long-time forecasting is employed to assist the short-time prediction. We conduct a large number of experiments on industrial control system datasets SWaT and WADI. Compared with most advanced methods, we achieve competitive results, especially on higher-dimensional datasets. Moreover, the proposed method can accurately locate anomalies and realize interpretability.

In this paper, we propose a new type of vision transformer (ViT) based on graph head attention (GHA). Because the multi-head attention (MHA) of a pure ViT requires multiple parameters and tends to lose the locality of an image, we replaced MHA with GHA by applying a graph to the attention head of the transformer. Consequently, the proposed GHA maintains both the locality and globality of the input patches and guarantees the diversity of the attention. The proposed GHA-ViT commonly outperforms pure ViT-based models using small-sized CIFAR-10/100, MNIST, and MNIST-F datasets and a medium-sized ImageNet-1K dataset in scratch training. A Top-1 accuracy of 81.7% was achieved for ImageNet-1K using GHA-B, which is a base model with approximately 29 M parameters. In addition, with CIFAR-10/100, the existing ViT and parameters are reduced 17-fold and the performance increased by 0.4/4.3%, respectively. The proposed GHA-ViT shows promising results in terms of the number of parameters and operations and the level of accuracy in comparison with other state-of-the-art ViT-lightweight models.

In recent years, graph convolutional neural networks (GCNs) have become a popular research topic due to their outstanding performance in various complex network data mining tasks. However, current research on graph neural networks lacks understanding of the high-order structural features of networks, focusing mostly on node features and first-order neighbor features. This article proposes two new models, MGAT and MGATv2, by introducing high-order structure motifs that frequently appear in networks and combining them with graph attention mechanisms. By introducing a mixed information matrix based on motifs, the generation process of graph attention coefficients is improved, allowing the model to capture higher-order structural features. Compared with the latest research on various graph neural networks, both MGAT and MGATv2 achieve good results in node classification tasks. Furthermore, through various experimental studies on real datasets, we demonstrate that the introduction of network structural motifs can effectively enhance the expressive power of graph neural networks, indicating that both high-order structural features and attribute features are important components of network feature learning.

Traffic congestion and frequent traffic accidents have become the main problems affecting urban traffic. The effective location prediction of vehicle trajectory can help alleviate traffic congestion, reduce the occurrence of traffic accidents, and optimize the urban traffic system. Vehicle trajectory is closely related to the surrounding Point of Interest (POI). POI can be considered as the spatial feature and can be fused with trajectory points to improve prediction accuracy. A Local Dynamic Graph Spatiotemporal–Long Short-Term Memory (LDGST-LSTM) was proposed in this paper to extract and fuse the POI knowledge and realize next location prediction. POI semantic information was learned by constructing the traffic knowledge graph, and spatial and temporal features were extracted by combining the Graph Attention Network (GAT) and temporal attention mechanism. The effectiveness of LDGST-LSTM was verified on two datasets, including Chengdu taxi trajectory data in August 2014 and October 2018. The accuracy and robustness of the proposed model were significantly improved compared with the benchmark models. The effects of major components in the proposed model were also evaluated through an ablation experiment. Moreover, the weights of POI that influence location prediction were visualized to improve the interpretability of the proposed model.

The inability to locate device faults quickly and accurately has become prominent due to the large number of communication devices and the complex structure of secondary circuit networks in smart substations. Traditional methods are less efficient when diagnosing secondary equipment faults in smart substations, and deep learning methods have poor portability, high learning sample costs, and often require retraining a model. Therefore, a secondary equipment fault diagnosis method based on a graph attention network is proposed in this paper. All fault events are automatically represented as graph-structured data based on the K-nearest neighbors (KNNs) algorithm in terms of the feature information exhibited by the corresponding detection nodes when equipment faults occur. Then, a fault diagnosis model is established based on the graph attention network. Finally, partial intervals of a 220 kV intelligent substation are taken as an example to compare the fault localization effect of different methods. The results show that the method proposed in this paper has the advantages of higher localization accuracy, lower learning cost, and better robustness than the traditional machine learning and deep learning methods.

In recent years, with the advancement of natural language processing techniques and the release of models like ChatGPT, how language models understand questions has become a hot topic. In handling complex logical reasoning with pre-trained models, its performance still has room for improvement. Inspired by DAGN, we propose an improved DaGATN (Discourse-apperceptive Graph Attention Networks) model. By constructing a discourse information graph to learn logical clues in the text, we decompose the context, question, and answer into elementary discourse units (EDUs) and connect them with discourse relations to construct a relation graph. The text features are learned through a discourse graph attention network and applied to downstream multiple-choice tasks. Our method was evaluated on the ReClor dataset and achieved an accuracy of 74.3%, surpassing the best-known performance methods utilizing deberta-xlarge-level pre-trained models, and also performed better than ChatGPT (Zero-Shot).

At present, there are problems such as low fault data, insufficient labeling information, and poor fault diagnosis in the field of ship engine diagnosis. To address the above problems, this paper proposes a fault diagnosis method based on probabilistic similarity and rank-order similarity of multi-head graph attention neural networks (MPGANN) models. Firstly, the ship engine dataset is used to explore the similarity between the data using the probabilistic similarity of T_SNE and the rank order similarity of Spearman’s correlation coefficient to define the neighbor relationship between the samples, and then the appropriate weights are selected for the early fusion of the two graph structures to fuse the feature information of the two scales. Finally, the graph attention neural networks (GANN) incorporating the multi-head attention mechanism are utilized to complete the fault diagnosis. In this paper, comparative experiments such as graph construction and algorithm performance are carried out based on the simulated ship engine dataset, and the experimental results show that the MPGANN outperforms the comparative methods in terms of accuracy, F1 score, and total elapsed time, with an accuracy rate of 97.58%. The experimental results show that the model proposed in this paper can still fulfill the ship engine fault diagnosis task well under unfavorable conditions such as small samples and insufficient label information, which is of practical significance in the field of intelligent ship cabins and fault diagnosis.

A growing number of studies have shown that aberrant microRNA (miRNA) expression is closely associated with the evolution and development of various complex human diseases. These key biomarkers’ identification and observation are significant for gaining a deeper understanding of disease pathogenesis and therapeutic mechanisms. Consequently, pinpointing potential miRNA–disease associations (MDA) has become a prominent bioinformatics subject, encouraging several new computational methods given the advances in graph neural networks (GNN). Nevertheless, these existing methods commonly fail to exploit the network nodes’ global feature information, leaving the generation of high-quality embedding representations using graph properties as a critical unsolved issue. Addressing these challenges, we introduce the DAEMDA, a computational method designed to optimize the current models’ efficacy. First, we construct similarity and heterogeneous networks involving miRNAs and diseases, relying on experimentally corroborated miRNA–disease association data and analogous information. Then, a newly-fashioned parallel dual-channel feature encoder, designed to better comprehend the global information within the heterogeneous network and generate varying embedding representations, follows this. Ultimately, employing a neural network classifier, we merge the dual-channel embedding representations and undertake association predictions between miRNA and disease nodes. The experimental results of five-fold cross-validation and case studies of major diseases based on the HMDD v3.2 database show that this method can generate high-quality embedded representations and effectively improve the accuracy of MDA prediction.

The agile earth observation satellite scheduling problem (AEOSSP) is a combinatorial optimization problem with time-dependent constraints. Recently, many construction heuristics and meta-heuristics have been proposed; however, existing methods cannot balance the requirements of efficiency and timeliness. In this paper, we propose a graph attention network-based decision neural network (GDNN) to solve the AEOSSP. Specifically, we first represent the task and time-dependent attitude transition constraints by a graph. We then describe the problem as a Markov decision process and perform feature engineering. On this basis, we design a GDNN to guide the construction of the solution sequence and train it with proximal policy optimization (PPO). Experimental results show that the proposed method outperforms construction heuristics at scheduling profit by at least 45%. The proposed method can also calculate the approximate profits of the state-of-the-art method with an error of less than 7% and reduce scheduling time markedly. Finally, we demonstrate the scalability of the proposed method.

Electricity load forecasting is of great significance for the overall operation of the power system and the orderly use of electricity at a later stage. However, traditional load forecasting does not consider the change in load quantity at each time point, while the information on the time difference of the load data can reflect the dynamic evolution information of the load data, which is a very important factor for load forecasting. In addition, the research topics in recent years mainly focus on the learning of the complex relationships of load sequences in time latitude by graph neural networks. The relationships between different variables of load sequences are not explicitly captured. In this paper, we propose a model that combines a differential learning network and a multidimensional feature graph attention layer, it can model the time dependence and dynamic evolution of load sequences by learning the amount of load variation at different time points, while representing the correlation of different variable features of load sequences through the graph attention layer. Comparative experiments show that the prediction errors of the proposed model have decreased by 5–26% compared to other advanced methods in the UC Irvine Machine Learning Repository Electricity Load Chart public dataset.

The prediction of drug–target interaction (DTI) is crucial to drug discovery. Although the interactions between the drug and target can be accurately verified by traditional biochemical experiments, the determination of DTI through biochemical experiments is a time-consuming, laborious, and expensive process. Therefore, we propose a learning-based framework named BG-DTI for drug–target interaction prediction. Our model combines two main approaches based on biological features and heterogeneous networks to identify interactions between drugs and targets. First, we extract original features from the sequence to encode each drug and target. Later, we further consider the relationships among various biological entities by constructing drug–drug similarity networks and target–target similarity networks. Furthermore, a graph convolutional network and a graph attention network in the graph representation learning module help us learn the features representation of drugs and targets. After obtaining the features from graph representation learning modules, these features are combined into fusion descriptors for drug–target pairs. Finally, we send the fusion descriptors and labels to a random forest classifier for predicting DTI. The evaluation results show that BG-DTI achieves an average AUC of 0.938 and an average AUPR of 0.930, which is better than those of five existing state-of-the-art methods. We believe that BG-DTI can facilitate the development of drug discovery or drug repurposing.

Agile-satellite mission planning is a crucial issue in the construction of satellite constellations. The large scale of remote sensing missions and the high complexity of constraints in agile-satellite mission planning pose challenges in the search for an optimal solution. To tackle the issue, a dynamic destroy deep-reinforcement learning (D3RL) model is designed to facilitate subsequent optimization operations via adaptive destruction to the existing solutions. Specifically, we first perform a clustering and embedding operation to reconstruct tasks into a clustering graph, thereby improving data utilization. Secondly, the D3RL model is established based on graph attention networks (GATs) to enhance the search efficiency for optimal solutions. Moreover, we present two applications of the D3RL model for intensive scenes: the deep-reinforcement learning (DRL) method and the D3RL-based large-neighborhood search method (DRL-LNS). Experimental simulation results illustrate that the D3RL-based approaches outperform the competition in terms of solutions’ quality and computational efficiency, particularly in more challenging large-scale scenarios. DRL-LNS outperforms ALNS with an average scheduling rate improvement of approximately 11% in Area instances. In contrast, the DRL approach performs better in World scenarios, with an average scheduling rate that is around 8% higher than that of ALNS.

Integrating lexical information into Chinese character embedding is a valid method to figure out the Chinese named entity recognition (NER) issue. However, most existing methods focus only on the discovery of named entity boundaries, considering only the words matched by the Chinese characters. They ignore the association between Chinese characters and their left and right matching words. They ignore the local semantic information of the character’s neighborhood, which is crucial for Chinese NER. The Chinese language incorporates a significant number of polysemous words, meaning that a single word can possess multiple meanings. Consequently, in the absence of sufficient contextual information, individuals may encounter difficulties in comprehending the intended meaning of a text, leading to the emergence of ambiguity. We consider how to handle the issue of entity ambiguity because of polysemous words in Chinese texts in different contexts more simply and effectively. We propose in this paper the use of graph attention networks to construct relatives among matching words and neighboring characters as well as matching words and adding left- and right-matching words directly using semantic information provided by the local lexicon. Moreover, this paper proposes a short-sequence convolutional neural network (SSCNN). It utilizes the generated shorter subsequence encoded with the sliding window module to enhance the perception of local information about the character. Compared with the widely used Chinese NER models, our approach achieves 1.18%, 0.29%, 0.18%, and 1.1% improvement on the four benchmark datasets Weibo, Resume, OntoNotes, and E-commerce, respectively, and proves the effectiveness of the model.

Background : Cancer survival prediction is vital in improving patients' prospects and recommending therapies. Understanding the molecular behavior of cancer can be enhanced through the integration of multi-omics data, including mRNA, miRNA, and DNA methylation data. In light of these multi-omics data, we proposed a graph attention network (GAT) model in this study to predict the survival of non-small cell lung cancer (NSCLC). Methods : The different omics data were obtained from The Cancer Genome Atlas (TCGA) and preprocessed and combined into a single dataset using the sample ID. We used the chi-square test to select the most significant features to be used in our model. We used the synthetic minority oversampling technique (SMOTE) to balance the dataset and the concordance index (C-index) to measure the performance of our model on different combinations of omics data. Results : Our model demonstrated superior performance, with the highest value of the C-index obtained when we used both mRNA and miRNA data. This demonstrates that the multi-omics approach could be effective in predicting survival. Further pathway analysis conducted with KEGG showed that our GAT model provided high weights to the features that are associated with the viral entry pathways, such as the Epstein-Barr virus and Influenza A pathways, which are involved in lung cancer development. From our findings, it can be observed that the proposed GAT model leads to a significantly improved prediction of survival by exploiting the strengths of multiple omics datasets and the findings from the enriched pathways. Our GAT model outperforms other state-of-the-art methods that are used for NSCLC prediction. Conclusions : In this study, we developed a new model for the survival prediction of NSCLC using the GAT based on multi-omics data. Our model showed outstanding predictive values, and the KEGG analysis of the selected significant features showed that they were implicated in pivotal biological processes underlying pathways such as Influenza A and the Epstein-Barr virus infection, which are linked to lung cancer progression.

Multiple unmanned aerial vehicles (Multi-UAV) systems have recently demonstrated significant advantages in some real-world scenarios, but the limited communication range of UAVs poses great challenges to multi-UAV collaborative decision-making. By constructing the multi-UAV cooperation problem as a multi-agent system (MAS), the cooperative decision-making among UAVs can be realized by using multi-agent reinforcement learning (MARL). Following this paradigm, this work focuses on developing partially observable MARL models that capture important information from local observations in order to select effective actions. Previous related studies employ either probability distributions or weighted mean field to update the average actions of neighborhood agents. However, they do not fully consider the feature information of surrounding neighbors, resulting in a local optimum often. In this paper, we propose a novel partially multi-agent reinforcement learning algorithm to remedy this flaw, which is based on graph attention network and partially observable mean field and is named as the GPMF algorithm for short. GPMF uses a graph attention module and a mean field module to describe how an agent is influenced by the actions of other agents at each time step. The graph attention module consists of a graph attention encoder and a differentiable attention mechanism, outputting a dynamic graph to represent the effectiveness of neighborhood agents against central agents. The mean field module approximates the effect of a neighborhood agent on a central agent as the average effect of effective neighborhood agents. Aiming at the typical task scenario of large-scale multi-UAV cooperative roundup, the proposed algorithm is evaluated based on the MAgent framework. Experimental results show that GPMF outperforms baselines including state-of-the-art partially observable mean field reinforcement learning algorithms, providing technical support for large-scale multi-UAV coordination and confrontation tasks in communication-constrained environments.

In graph-structured data, the node content contains rich information. Therefore, how to effectively utilize the content is crucial to improve the performance of graph convolutional networks (GCNs) on various analytical tasks. However, current GCNs do not fully utilize the content, especially multi-order content. For example, graph attention networks (GATs) only focus on low-order content, while high-order content is completely ignored. To address this issue, we propose a novel graph attention network with adaptability that could fully utilize the features of multi-order content. Its core idea has the following novelties: First, we constructed a high-order content attention mechanism that could focus on high-order content to evaluate attention weights. Second, we propose a multi-order content attention mechanism that can fully utilize multi-order content, i.e., it combines the attention mechanisms of high- and low-order content. Furthermore, the mechanism has adaptability, i.e., it can perform a good trade-off between high- and low-order content according to the task requirements. Lastly, we applied this mechanism to constructing a graph attention network with structural symmetry. This mechanism could more reasonably evaluate the attention weights between nodes, thereby improving the convergence of the network. In addition, we conducted experiments on multiple datasets and compared the proposed model with state-of-the-art models in multiple dimensions. The results validate the feasibility and effectiveness of the proposed model.

Aspect-level sentiment analysis aims to identify the sentiment polarity of specific aspects appearing in a given sentence or review. The model based on graph structure uses a dependency tree to link the aspect word with its corresponding opinion word and achieves significant results. However, for some sentences with ambiguous syntactic structure, it is difficult for the dependency tree to accurately parse the dependencies, which introduces noise and degrades the performance of the model. Based on this, we propose a syntactic and semantic enhanced multi-layer graph attention network (SSEMGAT), which introduces constituent trees in syntactic features to compensate for dependent trees at the clause level, exploiting aspect-aware attention in semantic features to assign the attention weight of specific aspects between contexts. The enhanced syntactic and semantic features are then used to classify specific aspects of sentiment through a multi-layer graph attention network. Accuracy and Macro-F1 are used as evaluation indexes in the SemEval-2014 Task 4 Restaurant and Laptop dataset and the Twitter dataset to compare the proposed model with the baseline model and the latest model, achieving competitive results.

Cross-modality person re-identification (ReID) aims at searching a pedestrian image of RGB modality from infrared (IR) pedestrian images and vice versa. Recently, some approaches have constructed a graph to learn the relevance of pedestrian images of distinct modalities to narrow the gap between IR modality and RGB modality, but they omit the correlation between IR image and RGB image pairs. In this paper, we propose a novel graph model called Local Paired Graph Attention Network (LPGAT). It uses the paired local features of pedestrian images from different modalities to build the nodes of the graph. For accurate propagation of information among the nodes of the graph, we propose a contextual attention coefficient that leverages distance information to regulate the process of updating the nodes of the graph. Furthermore, we put forward Cross-Center Contrastive Learning (C3L) to constrain how far local features are from their heterogeneous centers, which is beneficial for learning the completed distance metric. We conduct experiments on the RegDB and SYSU-MM01 datasets to validate the feasibility of the proposed approach.

Aspect-based sentiment analysis is a fine-grained sentiment analysis that focuses on the sentiment polarity of different aspects of text, and most current research methods use a combination of dependent syntactic analysis and graphical neural networks. In this paper, a graph attention network aspect-based sentiment analysis model based on the weighting of dependencies (WGAT) is designed to address the problem in that traditional models do not sufficiently analyse the types of syntactic dependencies; in the proposed model, graph attention networks can be weighted and averaged according to the importance of different nodes when aggregating information. The model first transforms the input text into a low-dimensional word vector through pretraining, while generating a dependency syntax graph by analysing the dependency syntax of the input text and constructing a dependency weighted adjacency matrix according to the importance of different dependencies in the graph. The word vector and the dependency weighted adjacency matrix are then fed into a graph attention network for feature extraction, and sentiment polarity is predicted through the classification layer. The model can focus on syntactic dependencies that are more important for sentiment classification during training, and the results of the comparison experiments on the Semeval-2014 laptop and restaurant datasets and the ACL-14 Twitter social comment dataset show that the WGAT model has significantly improved accuracy and F1 values compared to other baseline models, validating its effectiveness in aspect-level sentiment analysis tasks.

Most popular graph attention networks treat pixels of a feature map as individual nodes, which makes the feature embedding extracted by the graph convolution lack the integrity of the object. Moreover, matching between a template graph and a search graph using only part-level information usually causes tracking errors, especially in occlusion and similarity situations. To address these problems, we propose a novel end-to-end graph attention tracking framework that has high symmetry, combining traditional cross-correlation operations directly. By utilizing cross-correlation operations, we effectively compensate for the dispersion of graph nodes and enhance the representation of features. Additionally, our graph attention fusion model performs both part-to-part matching and global matching, allowing for more accurate information embedding in the template and search regions. Furthermore, we optimize the information embedding between the template and search branches to achieve better single-object tracking results, particularly in occlusion and similarity scenarios. The flexibility of graph nodes and the comprehensiveness of information embedding have brought significant performance improvements in our framework. Extensive experiments on three challenging public datasets (LaSOT, GOT-10k, and VOT2016) show that our tracker outperforms other state-of-the-art trackers.

Traffic forecasting has been regarded as the basis for many intelligent transportation system (ITS) applications, including but not limited to trip planning, road traffic control, and vehicle routing. Various forecasting methods have been proposed in the literature, including statistical models, shallow machine learning models, and deep learning models. Recently, graph neural networks (GNNs) have emerged as state-of-the-art traffic forecasting solutions because they are well suited for traffic systems with graph structures. This survey aims to introduce the research progress on graph neural networks for traffic forecasting and the research trends observed from the most recent studies. Furthermore, this survey summarizes the latest open-source datasets and code resources for sharing with the research community. Finally, research challenges and opportunities are proposed to inspire follow-up research.

Movie scene event extraction is a practical task in media analysis, which aims at extracting structured events from unstructured movie scripts. However, although there have been many studies regarding open domain event extraction, there have only been a few studies focusing on movie scene event extraction. Specifically aimed at instances where different argument roles have the same characteristics in a movie scene, we propose the utilization of the correlation between different argument roles, which is beneficial for both movie scene trigger extraction (trigger identification and classification) and movie scene argument extraction (argument identification and classification) in event extraction. To model the correlation between different argument roles, we propose the superior role concept (SRC), a high-level role concept based upon the ordinary argument role. In this paper, we introduce a new movie scene event extraction model with two main features: (1) an attentive high-level argument role module to capture SRC information and (2) an SRC-based graph attention network (GAT) to fuse the argument role correlation information into semantic embeddings. To evaluate the performance of our model, we constructed a movie scene event extraction dataset named MovieSceneEvent and also conducted experiments on a widely used dataset to compare the results with other models. The experimental results show that our model outperforms competitive models, and the correlation information of argument roles helps to improve the performance of movie scene event extraction.

In ambient-assisted living facilitated by smart home systems, the recognition of daily human activities is of great importance. It aims to infer the household’s daily activities from the triggered sensor observation sequences with varying time intervals among successive readouts. This paper introduces a novel deep learning framework based on embedding technology and graph attention networks, namely the time-oriented and location-oriented graph attention (TLGAT) networks. The embedding technology converts sensor observations into corresponding feature vectors. Afterward, TLGAT provides a sensor observation sequence as a fully connected graph to the model’s temporal correlation as well as the sensor’s location correlation among sensor observations and facilitates the feature representation of each sensor observation through receiving other sensor observations and weighting operations. The experiments were conducted on two public datasets, based on the diverse setups of sensor event sequence length. The experimental results revealed that the proposed method achieved favorable performance under diverse setups.

Graph neural networks have been successfully applied to sleep stage classification, but there are still challenges: (1) How to effectively utilize epoch information of EEG-adjacent channels owing to their different interaction effects. (2) How to extract the most representative features according to confused transitional information in confused stages. (3) How to improve classification accuracy of sleep stages compared with existing models. To address these shortcomings, we propose a multi-layer graph attention network (MGANet). Node-level attention prompts the graph attention convolution and GRU to focus on and differentiate the interaction between channels in the time-frequency domain and the spatial domain, respectively. The multi-head spatial-temporal mechanism balances the channel weights and dynamically adjusts channel features, and a multi-layer graph attention network accurately expresses the spatial sleep information. Moreover, stage-level attention is applied to easily confused sleep stages, which effectively improves the limitations of a graph convolutional network in large-scale graph sleep stages. The experimental results demonstrated classification accuracy; MF1 and Kappa reached 0.825, 0.814, and 0.775 and 0.873, 0.801, and 0.827 for the ISRUC and SHHS datasets, respectively, which showed that MGANet outperformed the state-of-the-art baselines.

It has become a tendency to use a combination of autoencoders and graph neural networks for attribute graph clustering to solve the community detection problem. However, the existing methods do not consider the influence differences between node neighborhood information and high-order neighborhood information, and the fusion of structural and attribute features is insufficient. In order to make better use of structural information and attribute information, we propose a model named community detection fusing graph attention network (CDFG). Specifically, we firstly use an autoencoder to learn attribute features. Then the graph attention network not only calculates the influence weight of the neighborhood node on the target node but also adds the high-order neighborhood information to learn the structural features. After that, the two features are initially fused by the balance parameter. The feature fusion module extracts the hidden layer representation of the graph attention layer to calculate the self-correlation matrix, which is multiplied by the node representation obtained by the preliminary fusion to achieve secondary fusion. Finally, the self-supervision mechanism makes it face the community detection task. Experiments are conducted on six real datasets. Using four evaluation metrics, the CDFG model performs better on most datasets, especially for the networks with longer average paths and diameters and smaller clustering coefficients.

Network alignment, which is also known as user identity linkage, is a kind of network analysis task that predicts overlapping users between two different social networks. This research direction has attracted much attention from the research community, and it is considered to be one of the most important research directions in the field of social network analysis. There are many different models for finding users that overlap between two networks, but most of these models use separate and different techniques to solve prediction problems, with very little work that has combined them. In this paper, we propose a method that combines different embedding techniques to solve the network alignment problem. Each association network alignment technique has its advantages and disadvantages, so combining them together will take full advantage and can overcome those disadvantages. Our model combines three-level embedding techniques of text-based user attributes, a graph attention network, a graph-drawing embedding technique, and fuzzy c-mean clustering to embed each piece of network information into a low-dimensional representation. We then project them into a common space by using canonical correlation analysis and compute the similarity matrix between them to make predictions. We tested our network alignment model on two real-life datasets, and the experimental results showed that our method can considerably improve the accuracy by about 10–15% compared to the baseline models. In addition, when experimenting with different ratios of training data, our proposed model could also handle the over-fitting problem effectively.

In recent years, pre-trained language models, represented by the bidirectional encoder representations from transformers (BERT) model, have achieved remarkable success in machine reading comprehension (MRC). However, limited by the structure of BERT-based MRC models (for example, restrictions on word count), such models cannot effectively integrate significant features, such as syntax relations, semantic connections, and long-distance semantics between sentences, leading to the inability of the available models to better understand the intrinsic connections between text and questions to be answered based on it. In this paper, a multi-granularity syntax guidance (MgSG) module that consists of a “graph with dependence” module and a “graph with entity” module is proposed. MgSG selects both sentence and word granularities to guide the text model to decipher the text. In particular, syntactic constraints are used to guide the text model while exploiting the global nature of graph neural networks to enhance the model’s ability to construct long-range semantics. Simultaneously, named entities play an important role in text and answers and focusing on entities can improve the model’s understanding of the text’s major idea. Ultimately, fusing multiple embedding representations to form a representation yields the semantics of the context and the questions. Experiments demonstrate that the performance of the proposed method on the Stanford Question Answering Dataset is better when compared with the traditional BERT baseline model. The experimental results illustrate that our proposed “MgSG” module effectively utilizes the graph structure to learn the internal features of sentences, solve the problem of long-distance semantics, while effectively improving the performance of PrLM in machine reading comprehension.

Many web platforms now include recommender systems. Network representation learning has been a successful approach for building these efficient recommender systems. However, learning the mutual influence of nodes in the network is challenging. Indeed, it carries collaborative signals accounting for complex user-item interactions on user decisions. For this purpose, in this paper, we develop a Mutual Interaction Graph Attention Network “MIGAN”, a new algorithm based on self-supervised representation learning on a large-scale bipartite graph (BGNN). Experimental investigation with real-world data demonstrates that MIGAN compares favorably with the baselines in terms of prediction accuracy and recommendation efficiency.

Computational prediction of ligand–target interactions is a crucial part of modern drug discovery as it helps to bypass high costs and labor demands of in vitro and in vivo screening. As the wealth of bioactivity data accumulates, it provides opportunities for the development of deep learning (DL) models with increasing predictive powers. Conventionally, such models were either limited to the use of very simplified representations of proteins or ineffective voxelization of their 3D structures. Herein, we present the development of the PSG-BAR (Protein Structure Graph-Binding Affinity Regression) approach that utilizes 3D structural information of the proteins along with 2D graph representations of ligands. The method also introduces attention scores to selectively weight protein regions that are most important for ligand binding. Results: The developed approach demonstrates the state-of-the-art performance on several binding affinity benchmarking datasets. The attention-based pooling of protein graphs enables identification of surface residues as critical residues for protein–ligand binding. Finally, we validate our model predictions against an experimental assay on a viral main protease (Mpro)—the hallmark target of SARS-CoV-2 coronavirus.

Point cloud registration is a fundamental task in many applications such as localization, mapping, tracking, and reconstruction. Successful registration relies on extracting robust and discriminative geometric features. Though existing learning-based methods require high computing capacity for processing a large number of raw points at the same time, computational capacity limitation is not an issue thanks to powerful parallel computing process using GPU. In this paper, we introduce a framework that efficiently and economically extracts dense features using a graph attention network for point cloud matching and registration (DFGAT). The detector of the DFGAT is responsible for finding highly reliable key points in large raw data sets. The descriptor of the DFGAT takes these keypoints combined with their neighbors to extract invariant density features in preparation for the matching. The graph attention network (GAT) uses the attention mechanism that enriches the relationships between point clouds. Finally, we consider this as an optimal transport problem and use the Sinkhorn algorithm to find positive and negative matches. We perform thorough tests on the KITTI dataset and evaluate the effectiveness of this approach. The results show that this method with the efficiently compact keypoint selection and description can achieve the best performance matching metrics and reach the highest success ratio of 99.88% registration in comparison with other state-of-the-art approaches.

Circular RNAs (circRNAs) are covalently closed single-stranded RNA molecules, which have many biological functions. Previous experiments have shown that circRNAs are involved in numerous biological processes, especially regulatory functions. It has also been found that circRNAs are associated with complex diseases of human beings. Therefore, predicting the associations of circRNA with disease (called circRNA-disease associations) is useful for disease prevention, diagnosis and treatment. In this work, we propose a novel computational approach called GGCDA based on the Graph Attention Network (GAT) and Graph Convolutional Network (GCN) to predict circRNA-disease associations. Firstly, GGCDA combines circRNA sequence similarity, disease semantic similarity and corresponding Gaussian interaction profile kernel similarity, and then a random walk with restart algorithm (RWR) is used to obtain the preliminary features of circRNA and disease. Secondly, a heterogeneous graph is constructed from the known circRNA-disease association network and the calculated similarity of circRNAs and diseases. Thirdly, the multi-head Graph Attention Network (GAT) is adopted to obtain different weights of circRNA and disease features, and then GCN is employed to aggregate the features of adjacent nodes in the network and the features of the nodes themselves, so as to obtain multi-view circRNA and disease features. Finally, we combined a multi-layer fully connected neural network to predict the associations of circRNAs with diseases. In comparison with state-of-the-art methods, GGCDA can achieve AUC values of 0.9625 and 0.9485 under the results of fivefold cross-validation on two datasets, and AUC of 0.8227 on the independent test set. Case studies further demonstrate that our approach is promising for discovering potential circRNA-disease associations.

Aspect-level sentiment classification aims to predict the sentiment polarities towards the target aspects given in sentences. To address the issues of insufficient semantic information extraction and high computational complexity of attention mechanisms in existing aspect-level sentiment classification models based on deep learning, a contextual graph attention network (CGAT) is proposed. The proposed model adopts two graph attention networks to aggregate syntactic structure information into target aspects and employs a contextual attention network to extract semantic information in sentence-aspect sequences, aiming to generate aspect-sensitive text features. In addition, a syntactic attention mechanism based on syntactic relative distance is proposed, and the Gaussian function is cleverly introduced as a syntactic weight function, which can reduce computational complexities and effectively highlight the words related to aspects in syntax. Experiments on three public sentiment datasets show that the proposed model can make better use of semantic information and syntactic structure information to improve the accuracy of sentiment classification.

In observations of Earth, the existence of clouds affects the quality and usability of optical remote sensing images in practical applications. Many cloud removal methods have been proposed to solve this issue. Among these methods, synthetic aperture radar (SAR)-based methods have more potential than others because SAR imaging is hardly affected by clouds, and can reflect ground information differences and changes. While SAR images used as auxiliary information for cloud removal may be blurred and noisy, the similar non-local information of spectral and electromagnetic features cannot be effectively utilized by traditional cloud removal methods. To overcome these weaknesses, we propose a novel cloud removal method using SAR-optical data fusion and a graph-based feature aggregation network (G-FAN). First, cloudy optical images and contemporary SAR images are concatenated and transformed into hyper-feature maps by pre-convolution. Second, the hyper-feature maps are inputted into the G-FAN to reconstruct the missing data of the cloud-covered area by aggregating the electromagnetic backscattering information of the SAR image, and the spectral information of neighborhood and non-neighborhood pixels in the optical image. Finally, post-convolution and a long skip connection are adopted to reconstruct the final predicted cloud-free images. Both the qualitative and quantitative experimental results from the simulated data and real data experiments show that our proposed method outperforms traditional deep learning methods for cloud removal.

In the COVID-19 epidemic the mildly symptomatic and asymptomatic infections generate a substantial portion of virus spread; these undetected individuals make it difficult to assess the effectiveness of preventive measures as most epidemic prevention strategies are based on the detected data. Effectively identifying the undetected infections in local transmission will be of great help in COVID-19 control. In this work, we propose an RNA virus transmission network representation model based on graph attention networks (RVTR); this model is constructed using the principle of natural language processing to learn the information of gene sequence and using a graph attention network to catch the topological character of COVID-19 transmission networks. Since SARS-CoV-2 will mutate when it spreads, our approach makes use of graph context loss function, which can reflect that the genetic sequence of infections with close spreading relation will be more similar than those with a long distance, to train our model. Our approach shows its ability to find asymptomatic spreaders both on simulated and real COVID-19 datasets and performs better when compared with other network representation and feature extraction methods.

With the rapid development of Earth observation technology, how to effectively and efficiently detect changes in multi-temporal images has become an important but challenging problem. Relying on the advantages of high performance and robustness, object-based change detection (CD) has become increasingly popular. By analyzing the similarity of local pixels, object-based CD aggregates similar pixels into one object and takes it as the basic processing unit. However, object-based approaches often have difficulty capturing discriminative features, as irregular objects make processing difficult. To address this problem, in this paper, we propose a novel superpixel-based multi-scale Siamese graph attention network (MSGATN) which can process unstructured data natively and extract valuable features. First, a difference image (DI) is generated by Euclidean distance between bitemporal images. Second, superpixel segmentation is employed based on DI to divide each image into many homogeneous regions. Then, these superpixels are used to model the problem by graph theory to construct a series of nodes with the adjacency between them. Subsequently, the multi-scale neighborhood features of the nodes are extracted through applying a graph convolutional network and concatenated by an attention mechanism. Finally, the binary change map can be obtained by classifying each node by some fully connected layers. The novel features of MSGATN can be summarized as follows: (1) Training in multi-scale constructed graphs improves the recognition over changed land cover of varied sizes and shapes. (2) Spectral and spatial self-attention mechanisms are exploited for a better change detection performance. The experimental results on several real datasets show the effectiveness and superiority of the proposed method. In addition, compared to other recent methods, the proposed can demonstrate very high processing efficiency and greatly reduce the dependence on labeled training samples in a semisupervised training fashion.

The treatment of complex diseases by using multiple drugs has become popular. However, drug-drug interactions (DDI) may give rise to the risk of unanticipated adverse effects and even unknown toxicity. Therefore, for polypharmacy safety it is crucial to identify DDIs and explore their underlying mechanisms. The detection of DDI in the wet lab is expensive and time-consuming, due to the need for experimental research over a large volume of drug combinations. Although many computational methods have been developed to predict DDIs, most of these are incapable of predicting potential DDIs between drugs within the DDI network and new drugs from outside the DDI network. In addition, they are not designed to explore the underlying mechanisms of DDIs and lack interpretative capacity. Thus, here we propose a novel method of GNN-DDI to predict potential DDIs by constructing a five-layer graph attention network to identify k-hops low-dimensional feature representations for each drug from its chemical molecular graph, concatenating all identified features of each drug pair, and inputting them into a MLP predictor to obtain the final DDI prediction score. The experimental results demonstrate that our GNN-DDI is suitable for each of two DDI predicting scenarios, namely the potential DDIs among known drugs in the DDI network and those between drugs within the DDI network and new drugs from outside DDI network. The case study indicates that our method can explore the specific drug substructures that lead to the potential DDIs, which helps to improve interpretability and discover the underlying interaction mechanisms of drug pairs.