Features Of Nodes Research Articles

Adaptive node feature extraction in graph-based neural networks for brain diseases diagnosis using self-supervised learning

Electroencephalography (EEG) has demonstrated significant value in diagnosing brain diseases. In particular, brain networks have gained prominence as they offer additional valuable insights by establishing connections between EEG signal channels. While brain connections are typically delineated by channel signal similarity, there lacks a consistent and reliable strategy for ascertaining node characteristics. Conventional node features such as temporal and frequency domain properties of EEG signals prove inadequate for capturing the extensive EEG information. In our investigation, we introduce a novel adaptive method for extracting node features from EEG signals utilizing a distinctive task-induced self-supervised learning technique. By amalgamating these extracted node features with fundamental edge features constructed using Pearson correlation coefficients, we showed that the proposed approach can function as a plug-in module that can be integrated to many common GNN networks (e.g., GCN, GraphSAGE, GAT) as a replacement of node feature selections module. Comprehensive experiments are then conducted to demonstrate the consistently superior performance and high generality of the proposed method over other feature selection methods in various of brain disorder prediction tasks, such as depression, schizophrenia, and Parkinson’s disease. Furthermore, compared to other node features, our approach unveils profound spatial patterns through graph pooling and structural learning, shedding light on pivotal brain regions influencing various brain disorder prediction based on derived features.

Adaptive self-supervised learning for sequential recommendation

Sequential recommendation typically utilizes deep neural networks to mine rich information in interaction sequences. However, existing methods often face the issue of insufficient interaction data. To alleviate the sparsity issue, self-supervised learning is introduced into sequential recommendation. Despite its effectiveness, we argue that current self-supervised learning-based (i.e., SSL-based) sequential recommendation models have the following limitations: (1) using only a single self-supervised learning method, either contrastive self-supervised learning or generative self-supervised learning. (2) employing a simple data augmentation strategy in either the graph structure domain or the node feature domain. We believe that they have not fully utilized the capabilities of both self-supervised methods and have not sufficiently explored the advantages of combining graph augmentation schemes. As a result, they often fail to learn better item representations. In light of this, we propose a novel multi-task sequential recommendation framework named Adaptive Self-supervised Learning for sequential Recommendation (ASLRec). Specifically, our framework combines contrastive and generative self-supervised learning methods adaptively, simultaneously applying different perturbations at both the graph topology and node feature levels. This approach constructs diverse augmented graph views and employs multiple loss functions (including contrastive loss, generative loss, mask loss, and prediction loss) for joint training. By encompassing the capabilities of various methods, our model learns item representations across different augmented graph views to achieve better performance and effectively mitigate interaction noise and sparsity. In addition, we add a small proportion of random uniform noise to item representations, making the item representations more uniform and mitigating the inherent popularity bias in interaction records. We conduct extensive experiments on three publicly available benchmark datasets to evaluate our model. The results demonstrate that our approach achieves state-of-the-art performance compared to 14 other competitive methods: the hit rate (HR) improved by over 14.39%, and the normalized discounted cumulative gain (NDCG) increased by over 18.67%.

A graph convolutional network model based on regular equivalence for identifying influential nodes in complex networks

The identification of influential nodes in complex networks is an interesting problem. Traditional approaches, such as metric- and machine learning-based approaches, often fail to adequately consider the network structure or node features, limiting their applications in certain scenarios. Deep learning-based methods, particularly those using graph neural networks (GNNs), have yielded remarkable results in this field. However, there is a lack of utilization of the regular equivalence-based similarity between nodes to feature construction, which is critical to deep learning-based methods. Nevertheless, regular equivalence-based similarity is extremely suitable to the quantification of the similarity between influential nodes, as these nodes usually have similar structural properties but do not require common neighbors. Building upon graph convolutional network (GCN, a specific type of GNN) and the regular equivalence-based similarity, this study proposes a novel method for identifying influential nodes in complex networks. Particularly, the regular equivalence-based similarities between nodes are considered to construct node features; a GCN is enhanced to improve the identification of influential nodes. Experiments on several datasets indicate that the proposed method is superior to state-of-the-art methods, demonstrating the advantages of the proposed feature construction based on regular equivalence-based similarity over the identification of influential nodes.

Identifying critical nodes in interdependent networks by GA-XGBoost

Once a critical node is destroyed, the interdependent network is prone to experience severe cascading failure. Due to the coupling, traditional methods are challenging to apply to interdependent networks. Here, we propose a novel comprehensive model based on machine learning. The main work of this data-driven approach is to train the model on a small set of nodes (5 % of the graph) and do the critical node identification on the rest. We collect node centrality indicators to describe the node features and provide informative input data from different dimensions. The uniform node sampling is improved to cluster oversampling, which combines K-means and Synthetic Minority Over-sampling Technique (SMOTE) to select and recreate uniformly distributed training samples. We optimize the XGBoost based on the Genetic algorithm (GA) to overcome the instability of manual parameters. Kendall's τ correlation coefficient, Jaccard similarity coefficient, R2, and RMSE are used as the model performance evaluation metrics. Experiment results confirm that the proposed GA-XGBoost model outperforms others, demonstrating higher adaptability and stability in various situations. The heuristic algorithm-optimized machine learning model offers a viable solution for identifying critical nodes in interdependent networks, which is of great significance for controlling virus propagation and preventing failures.

Multi-graph attention fusion graph neural network for remaining useful life prediction of rolling bearings

Abstract Graph neural network (GNN) has the proven ability to learn feature representations from graph data, and has been utilized for the tasks of predicting the machinery remaining useful life (RUL). However, existing methods only focus on a single graph structure and cannot integrate the correlation information contained in multi-graph structures. To address these issues, a multi-graph structure GNN prediction method with attention fusion (MGAFGNN) is proposed in this paper for GNN-based bearing RUL prediction. Specifically, a multi-channel graph attention module is designed to effectively learn the similar features of node neighbors from different graph data and capture the multi-scale latent features of nodes through the nonlinear transformation. Furthermore, a multi-graph attention fusion module (MGAFM) is proposed to extract the collaborative features from the interaction graph, thereby fusing the feature embeddings from different graph structures. The fused feature representation is sent to the long short-term memory network to further learn the temporal features and achieve RUL prediction. The experimental results on two bearing datasets demonstrate that MGAFGNN outperforms existing methods in terms of prediction performance by effectively incorporating multi-graph structural information.

DyHDGE: Dynamic heterogeneous transaction graph embedding for safety-centric fraud detection in financial scenarios

Dynamic graph fraud detection aims to distinguish fraudulent entities that deviate significantly from most benign entities within an ever-changing graph network. However, when dealing with different financial fraud scenarios, existing methods face challenges, resulting in difficulty in effectively ensuring financial security. In fraud scenarios, transaction data are generated in real time, in which a strong temporal relationship between multiple fraudulent transactions is observed. Traditional dynamic graph models struggle to effectively balance the temporal features of nodes and spatial structural features, failing to handle different types of nodes in the graph network. In this study, to extract the temporal and structural information, we proposed a dynamic heterogeneous transaction graph embedding (DyHDGE) network based on a dynamic heterogeneous transaction graph, considering both temporal and structural information while incorporating heterogeneous data. To separately extract temporal relationships between transactions and spatial structural relationships between nodes, we used a heterogeneous temporal graph representation learning module and a temporal graph structure information extraction module. Additionally, we designed two loss functions to optimize node feature representations. Extensive experiments demonstrated that the proposed DyHDGE significantly outperformed previous state-of-the-art methods on two simulated datasets of financial fraud scenarios. This capability contributes to enhancing security in financial consumption scenarios.

BrepGen: A B-rep Generative Diffusion Model with Structured Latent Geometry

This paper presents BrepGen , a diffusion-based generative approach that directly outputs a Boundary representation (B-rep) Computer-Aided Design (CAD) model. BrepGen represents a B-rep model as a novel structured latent geometry in a hierarchical tree. With the root node representing a whole CAD solid, each element of a B-rep model (i.e., a face, an edge, or a vertex) progressively turns into a child-node from top to bottom. B-rep geometry information goes into the nodes as the global bounding box of each primitive along with a latent code describing the local geometric shape. The B-rep topology information is implicitly represented by node duplication. When two faces share an edge, the edge curve will appear twice in the tree, and a T-junction vertex with three incident edges appears six times in the tree with identical node features. Starting from the root and progressing to the leaf, BrepGen employs Transformer-based diffusion models to sequentially denoise node features while duplicated nodes are detected and merged, recovering the B-Rep topology information. Extensive experiments show that BrepGen advances the task of CAD B-rep generation, surpassing existing methods on various benchmarks. Results on our newly collected furniture dataset further showcase its exceptional capability in generating complicated geometry. While previous methods were limited to generating simple prismatic shapes, BrepGen incorporates free-form and doubly-curved surfaces for the first time. Additional applications of BrepGen include CAD autocomplete and design interpolation. The code, pretrained models, and dataset are available at https://github.com/samxuxiang/BrepGen.

Graph isomorphism network with weighted multi‐aggregators for building shape classification

AbstractBuilding shape cognition is essential for tasks, such as map generalization, urban modeling, and building semantics and distribution pattern recognition. Traditional geometric and statistical methods rely on human‐defined shape indicators, and spectral‐based graph neural networks (GNNs) require Laplacian eigendecomposition, resulting in high algorithmic complexity. Therefore, we proposed a low‐complexity and simple‐to‐use spatial‐domain GNN for differentiating building shapes. To examine the influence of the building vertices on their shape, we treated each building as a graph and proposed a graph isomorphic network with weighted multi‐aggregators (GIN‐WMA) by analyzing the node connectivity of a building graph. The GIN‐WMA utilizes a novel aggregator that combines the sum and max aggregators, enhancing its recognition and differentiation capabilities. This approach can effectively differentiate nodes that have identical features after aggregation by the sum aggregator. We extracted features considering both local node and global shape features, drawing inspiration from Gestalt cognitive psychology and GNN's “node–graph” differentiation strategy. In addition, we compared the performance of GIN‐WMA with existing methods, studying the effect of various node features and their combinations on classification accuracy. The results demonstrated that GIN‐WMA outperforms other methods in discriminating building shapes, demonstrating superior capabilities in shape classification and enabling end‐to‐end extraction and classification of building shapes.

Heterogenous biological network multi-task learning model for ncRNA-disease-drug association prediction

Traditional medical research is characterised by lengthy duration, significant financial investment, and substantial risk of failure. In response to these challenges, network medicine, combined with medicine and computer technology, has become an important development direction, and computational methods have been proposed to predict potential associations. However, most of the current computational methods focus on single-potential association prediction tasks, which face issues of association sparsity and weak generalisation ability. To address these challenges, we developed a heterogeneous biological network multi-task learning model (HBNMM). Unlike previous methods based on bipartite graphs, HBNMM constructs a complex heterogeneous biological network, including ncRNA-disease-drug association networks and diverse similarity networks. HBNMM applies graph attention networks to aggregate node neighbourhood information and acquire node feature embeddings, and is then trained with a multi-task learning strategy to simultaneously predict potential ncRNA-disease, ncRNA-drug, and drug-disease associations. As a result, the HBNMM achieves an excellent performance that is higher than that of the state-of-the-art models. Furthermore, five case studies supported by experiments showed powerful predictive ability for drug discovery and disease treatment.

A two-stage importance-aware subgraph convolutional network based on multi-source sensors for cross-domain fault diagnosis

Graph convolutional networks (GCNs) as the emerging neural networks have shown great success in Prognostics and Health Management because they can not only extract node features but can also mine relationship between nodes in the graph data. However, the most existing GCNs-based methods are still limited by graph quality, variable working conditions, and limited data, making them difficult to obtain remarkable performance. Therefore, it is proposed in this paper a two stage importance-aware subgraph convolutional network based on multi-source sensors named I2SGCN to address the above-mentioned limitations. In the real-world scenarios, it is found that the diagnostic performance of the most existing GCNs is commonly bounded by the graph quality because it is hard to get high quality through a single sensor. Therefore, we leveraged multi-source sensors to construct graphs that contain more fault-based information of mechanical equipment. Then, we discovered that unsupervised domain adaptation (UDA) methods only use single stage to achieve cross-domain fault diagnosis and ignore more refined feature extraction, which can make the representations contained in the features inadequate. Hence, it is proposed the two-stage fault diagnosis in the whole framework to achieve UDA. In the first stage, the multiple-instance learning is adopted to obtain the importance factor of each sensor towards preliminary fault diagnosis. In the second stage, it is proposed I2SGCN to achieve refined cross-domain fault diagnosis. Moreover, we observed that deficient and limited data may cause label bias and biased training, leading to reduced generalization capacity of the proposed method. Therefore, we constructed the feature-based graph and importance-based graph to jointly mine more effective relationship and then presented a subgraph learning strategy, which not only enriches sufficient and complementary features but also regularizes the training. Comprehensive experiments conducted on four case studies demonstrate the effectiveness and superiority of the proposed method for cross-domain fault diagnosis, which outperforms the state-of-the art methods.

DMFVAE: miRNA-disease associations prediction based on deep matrix factorization method with variational autoencoder

MicroRNAs (miRNAs) are closely related to numerous complex human diseases, therefore, exploring miRNA-disease associations (MDAs) can help people gain a better understanding of complex disease mechanism. An increasing number of computational methods have been developed to predict MDAs. However, the sparsity of the MDAs may hinder the performance of many methods. In addition, many methods fail to capture the nonlinear relationships of miRNA-disease network and inadequately leverage the features of network and neighbor nodes. In this study, we propose a deep matrix factorization model with variational autoencoder (DMFVAE) to predict potential MDAs. DMFVAE first decomposes the original association matrix and the enhanced association matrix, in which the enhanced association matrix is enhanced by self-adjusting the nearest neighbor method, to obtain sparse vectors and dense vectors, respectively. Then, the variational encoder is employed to obtain the nonlinear latent vectors of miRNA and disease for the sparse vectors, and meanwhile, node2vec is used to obtain the network structure embedding vectors of miRNA and disease for the dense vectors. Finally, sample features are acquired by combining the latent vectors and network structure embedding vectors, and the final prediction is implemented by convolutional neural network with channel attention. To evaluate the performance of DMFVAE, we conduct five-fold cross validation on the HMDD v2.0 and HMDD v3.2 datasets and the results show that DMFVAE performs well. Furthermore, case studies on lung neoplasms, colon neoplasms, and esophageal neoplasms confirm the ability of DMFVAE in identifying potential miRNAs for human diseases.

Prediction method of pharmacokinetic parameters of small molecule drugs based on GCN network model.

Accurately predicting plasma protein binding rate (PPBR) and oral bioavailability (OBA) helps to better reveal the absorption and distribution of drugs in the human body and subsequent drug design. Although machine learning models have achieved good results in prediction accuracy, they often suffer from insufficient accuracy when dealing with data with irregular topological structures. In view of this, this study proposes a pharmacokinetic parameter prediction framework based on graph convolutional networks (GCN), which predicts the PPBR and OBA of small molecule drugs. In the framework, GCN is first used to extract spatial feature information on the topological structure of drug molecules, in order to better learn node features and association information between nodes. Then, based on the principle of drug similarity, this study calculates the similarity between small molecule drugs, selects different thresholds to construct datasets, and establishes a prediction model centered on the GCN algorithm. The experimental results show that compared with traditional machine learning prediction models, the prediction model constructed based on the GCN method performs best on PPBR and OBA datasets with an inter-molecular similarity threshold of 0.25, with MAE of 0.155 and 0.167, respectively. In addition, in order to further improve the accuracy of the prediction model, GCN is combined with other algorithms. Compared to using a single GCN method, the distribution of the predicted values obtained by the combined model is highly consistent with the true values. In summary, this work provides a new method for improving the rate of early drug screening in the future.

Towards multimodal graph neural networks for surgical instrument anticipation

PurposeDecision support systems and context-aware assistance in the operating room have emerged as the key clinical applications supporting surgeons in their daily work and are generally based on single modalities. The model- and knowledge-based integration of multimodal data as a basis for decision support systems that can dynamically adapt to the surgical workflow has not yet been established. Therefore, we propose a knowledge-enhanced method for fusing multimodal data for anticipation tasks.MethodsWe developed a holistic, multimodal graph-based approach combining imaging and non-imaging information in a knowledge graph representing the intraoperative scene of a surgery. Node and edge features of the knowledge graph are extracted from suitable data sources in the operating room using machine learning. A spatiotemporal graph neural network architecture subsequently allows for interpretation of relational and temporal patterns within the knowledge graph. We apply our approach to the downstream task of instrument anticipation while presenting a suitable modeling and evaluation strategy for this task.ResultsOur approach achieves an F1 score of 66.86% in terms of instrument anticipation, allowing for a seamless surgical workflow and adding a valuable impact for surgical decision support systems. A resting recall of 63.33% indicates the non-prematurity of the anticipations.ConclusionThis work shows how multimodal data can be combined with the topological properties of an operating room in a graph-based approach. Our multimodal graph architecture serves as a basis for context-sensitive decision support systems in laparoscopic surgery considering a comprehensive intraoperative operating scene.

Click-through rate prediction model based on graph networks and feature squeeze-and-excitation mechanism

PurposePredicting a user’s click-through rate on an advertisement or item often uses deep learning methods to mine hidden information in data features, which can provide users with more accurate personalized recommendations. However, existing works usually ignore the problem that the drift of user interests may lead to the generation of new features when they compute feature interactions. Based on this, this paper aims to design a model to address this issue.Design/methodology/approachFirst, the authors use graph neural networks to model users’ interest relationships, using the existing user features as the node features of the graph neural networks. Second, through the squeeze-and-excitation network mechanism, the user features and item features are subjected to squeeze operation and excitation operation, respectively, and the importance of the features is adaptively adjusted by learning the channel weights of the features. Finally, the feature space is divided into multiple subspaces to allocate features to different models, which can improve the performance of the model.FindingsThe authors conduct experiments on two real-world data sets, and the results show that the model can effectively improve the prediction accuracy of advertisement or item click events.Originality/valueIn the study, the authors propose graph network and feature squeeze-and-excitation model for click-through rate prediction, which is used to dynamically learn the importance of features. The results indicate the effectiveness of the model.

Improving the Risk Prediction of the 2015 ATA Recurrence Risk Stratification in Papillary Thyroid Cancer.

Various prognostic factors are expected to refine the American Thyroid Association (ATA) recurrence risk stratification for patients with papillary thyroid cancer (PTC). However, it remains unclear to what extent integrating these factors improves patient treatment decision-making. We developed two predictive models for structural incomplete response (SIR) at the one-year follow-up visit, based on comprehensive clinical data from a retrospective cohort of 2539 patients. Model 1 included the recurrence risk stratification and lymph node features (i.e., number and ratio of metastatic lymph nodes, N stage). Model 2 further incorporated preablation stimulated thyroglobulin (s-Tg). An independent cohort of 746 patients was used for validation analysis. We assessed the models' predictive performance compared to the recurrence risk stratification using the integrated discrimination improvement (IDI) and the continuous net reclassification improvement (NRI). The clinical utility of the models was evaluated using decision curve analysis. Both Model 1 and Model 2 outperformed the recurrence risk stratification in predicting SIR, with improved correct classification rates (Model 1: IDI=0.02, event NRI=42.31%; Model 2: IDI=0.07, event NRI=53.54%). The decision curves indicated that both models provided greater benefits over the risk stratification system in clinical decision-making. In the validation set, Model 2 maintained similar performance while Model 1 did not significantly improve correct reclassification. The inclusion of lymph node features and s-Tg showed potential to enhance the predictive accuracy and clinical utility of the existing risk stratification system for PTC patients.

Open-world structured sequence learning via dense target encoding

Structured sequences are popularly used to describe graph data with time-evolving node features and edges. A typical real-world scenario of structured sequences is that unknown class labels continuously arrive and thus the training and testing often across different class spaces. This scenario is also referred to as the open-world learning problem on structured sequences. In this paper, we present a new Dense Open-world Structured Sequence Learning model (DOSSL for short) to learn graph streams in the open-world learning setting. To capture both structural and temporal information, DOSSL uses a GNN-based stochastic recurrent neural network for learning node representation in graph streams, then a truncated Laplacian distribution to describe the latent distribution of graph nodes, and a sampling function is used to generate node representations. Further, DOSSL learns dense target embeddings for the known classes to improve the compactness of known class distribution and reserve enough space for open-world unknown classes. The ultimate open-world classifier is optimized to detect the samples from unknown classes under the constraints of DVAE loss, label loss, class uncertainty loss, and dense target loss. Through empirical analysis conducted on real-world datasets, it has been demonstrated that the advanced technique known as DOSSL exhibits the ability to acquire precise node classifiers by harnessing the power of graph streams.

Social Network Forensics Analysis Model Based on Network Representation Learning.

The rapid evolution of computer technology and social networks has led to massive data generation through interpersonal communications, necessitating improved methods for information mining and relational analysis in areas such as criminal activity. This paper introduces a Social Network Forensic Analysis model that employs network representation learning to identify and analyze key figures within criminal networks, including leadership structures. The model incorporates traditional web forensics and community algorithms, utilizing concepts such as centrality and similarity measures and integrating the Deepwalk, Line, and Node2vec algorithms to map criminal networks into vector spaces. This maintains node features and structural information that are crucial for the relational analysis. The model refines node relationships through modified random walk sampling, using BFS and DFS, and employs a Continuous Bag-of-Words with Hierarchical Softmax for node vectorization, optimizing the value distribution via the Huffman tree. Hierarchical clustering and distance measures (cosine and Euclidean) were used to identify the key nodes and establish a hierarchy of influence. The findings demonstrate the effectiveness of the model in accurately vectorizing nodes, enhancing inter-node relationship precision, and optimizing clustering, thereby advancing the tools for combating complex criminal networks.

Global–local spatio-temporal graph convolutional networks for video summarization

Video summarization aims to create concise and accurate summary to enable users to quickly grasp the key content of the original video for facilitating efficient video browsing. Most existing video summarization methods mainly employ recurrent neural networks to capture long-term dependencies in videos, yielding remarkable results. Nevertheless, these methods overlook the potential spatial features inside the video when modeling the video. To tackle this issue, we introduce a global–local spatio-temporal graph convolutional networks for video summarization (GL-STGCN). Initially inspired by the concept of 3-D convolution, we segment the video into non-overlapping segments to capture localized spatial features of sequential frames. Subsequently, a spatial graph is constructed for each segment, with a fixed time interval between neighboring spatial graphs. Then we use the pooling method to randomly delete the redundant nodes in the graph. We then leverage a temporal gating convolutional network to extract the global temporal relationships within the video. Employing the spatial features, a spatial graph convolutional network is utilized to capture the spatial connections among frames. As the graph node information evolves, the node features furnish a more precise depiction of the video content. Consequently, we employ the temporal gating convolutional network once more to refine the global temporal relations within the video. Extensive experiments on two public datasets are conducted in this paper, showing that our proposed method outperforms most state-of-the-art video summarization methods in terms of performance. Experimental results demonstrate the effectiveness of integrating global temporal and local spatial relationships.

Graph Adaptive Attention Network with Cross-Entropy.

Non-Euclidean data, such as social networks and citation relationships between documents, have node and structural information. The Graph Convolutional Network (GCN) can automatically learn node features and association information between nodes. The core ideology of the Graph Convolutional Network is to aggregate node information by using edge information, thereby generating a new node feature. In updating node features, there are two core influencing factors. One is the number of neighboring nodes of the central node; the other is the contribution of the neighboring nodes to the central node. Due to the previous GCN methods not simultaneously considering the numbers and different contributions of neighboring nodes to the central node, we design the adaptive attention mechanism (AAM). To further enhance the representational capability of the model, we utilize Multi-Head Graph Convolution (MHGC). Finally, we adopt the cross-entropy (CE) loss function to describe the difference between the predicted results of node categories and the ground truth (GT). Combined with backpropagation, this ultimately achieves accurate node classification. Based on the AAM, MHGC, and CE, we contrive the novel Graph Adaptive Attention Network (GAAN). The experiments show that classification accuracy achieves outstanding performances on Cora, Citeseer, and Pubmed datasets.

Dynamic community detection algorithm based on hyperbolic graph convolution

PurposeCommunity detection is a key factor in analyzing the structural features of complex networks. However, traditional dynamic community detection methods often fail to effectively solve the problems of deep network information loss and computational complexity in hyperbolic space. To address this challenge, a hyperbolic space-based dynamic graph neural network community detection model (HSDCDM) is proposed.Design/methodology/approachHSDCDM first projects the node features into the hyperbolic space and then utilizes the hyperbolic graph convolution module on the Poincaré and Lorentz models to realize feature fusion and information transfer. In addition, the parallel optimized temporal memory module ensures fast and accurate capture of time domain information over extended periods. Finally, the community clustering module divides the community structure by combining the node characteristics of the space domain and the time domain. To evaluate the performance of HSDCDM, experiments are conducted on both artificial and real datasets.FindingsExperimental results on complex networks demonstrate that HSDCDM significantly enhances the quality of community detection in hierarchical networks. It shows an average improvement of 7.29% in NMI and a 9.07% increase in ARI across datasets compared to traditional methods. For complex networks with non-Euclidean geometric structures, the HSDCDM model incorporating hyperbolic geometry can better handle the discontinuity of the metric space, provides a more compact embedding that preserves the data structure, and offers advantages over methods based on Euclidean geometry methods.Originality/valueThis model aggregates the potential information of nodes in space through manifold-preserving distribution mapping and hyperbolic graph topology modules. Moreover, it optimizes the Simple Recurrent Unit (SRU) on the hyperbolic space Lorentz model to effectively extract time series data in hyperbolic space, thereby enhancing computing efficiency by eliminating the reliance on tangent space.