Community detection based on unsupervised attributed network embedding

Zhou, Xinchuang; Su, Lingtao; Li, Xiangju; Zhao, Zhongying; Li, Chao

doi:10.1016/j.eswa.2022.118937

Cited by 29 publications

(8 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…SDCN [18] constructs a dual network of DNN and GCN that proposes dual self-supervised modular optimization learning. CDBNE [19] uses modules to maximize the capture of cluster-oriented information and mine the initial community structure. Despite their advantages, most of the current methods are based on attribute graphs, as a consequence, ignore the higher-order connections between nodes and are not conducive to mining potential information.…”

Section: Related Work 21 Graph Representation Learningmentioning

confidence: 99%

See 1 more Smart Citation

Hypergraph network embedding for community detection

Xiang,

You,

Wang

et al. 2024

J Supercomput

View full text Add to dashboard Cite

Using attribute graphs for node embedding to detect community structure has become a popular research topic. However, most of the existing algorithms mainly focus on the network structure and node features, which ignore the higher-order relationships between nodes. In addition, only adopting the original graph structure will suffer from sparsity problems, and will also result in sub-optimal node clustering performance. In this paper, we propose a hypergraph network embedding (HGNE) for community detection to solve the above problems. Firstly, we construct potential connections based on the shared feature information of the nodes. By fusing the original topology with feature-based potential connections, both the explicit and implicit relationships are encoded into the node representations, thus alleviating the sparsity problem. Secondly, for integrating the higher-order relationship, we adopt hypergraph convolution to encode the higherorder correlations. To constrain the quality of the node embedding, the spectral hypergraph embedding loss is utilized. Furthermore, we design a dual-contrast mechanism, which draws similar nodes closer by comparing the representations of different views. This mechanism can efficiently prevent multi-node classes from distorting less-node classes. Finally, the dual-contrast mechanism is jointly optimized with self-training clustering to obtain more robust node representations, thus improving the clustering results. Extensive experiments on five datasets indicate the superiority and effectiveness of HGNE.

show abstract

Section: Related Work 21 Graph Representation Learningmentioning

confidence: 99%

“…The overall framework is optimized using spectral clustering loss, and structural and attribute information is integrated into a higher-order convolution kernel. CDBNE [19]: Incorporating modules to maximize module capture of mesoscopic community structure. Optimizing the learned node representation by joint learning.…”

Section: Baselinesmentioning

confidence: 99%

Hypergraph network embedding for community detection

Xiang,

You,

Wang

et al. 2024

J Supercomput

View full text Add to dashboard Cite

show abstract

“…The potential of deep learning to capture information about the attribute and structure of a graph leads to a more comprehensive representation, making it better suited for downstream community detection tasks.DAEGC [16] uses attention coefficients to represent the correlation between different nodes and capture structural information within communities. However, this overlooks the importance of node attribute information.Recently proposed CDBNE [17] uses graph attention mechanisms when encoding topological structure and node attributes. However, there are redundancy issues with edge and node information encoded using same attention mechanism.To address these issues, [18] introduces dual-view graph attention encoder which processes structural and attribute information as independent views in order to effectively learn embeddings for nodes but still faces problems related to similar information.…”

Section: Related Workmentioning

confidence: 99%

“…Inspired by the previous work of [24], we simply adopt the inner product of the GAAE encoder's output H and its transpose to predict the connection relationship between nodes in the reconstructed graph. We reconstruct the adjacency matrix as Eq.…”

Section: Decoded Of K-subgraph By Graph Attentionmentioning

confidence: 99%

Detect Local Community based on Core Node using Deep Feature Fusion

Guo,

Li,

Shi

et al. 2024

Preprint

View full text Add to dashboard Cite

Unlike global community detection, local community detection is to identify a cluster of nodes sharing similar feature information based on a given seed, which is of great significance for many real-world applications. The most popular strategies of local community detection involve either expanding local communities around seed nodes or dividing communities through subgraph clustering. However, the accuracy of many local community detection algorithms heavily relies on the quality of seed nodes. Only high-quality seed nodes can accurately detect local communities. At the same time, the inability to effectively obtain node attributes and structural information also leads to an increase in subgraph clustering error rates. In this paper, we propose a Local Community Detection based on a Core Node using deep feature fusion, named LCDCN. For the seed node, we first find the nearest node with greater significance and correlation as the core node, then construct a \(k\)-subgraph through a specific subgraph extractor based on the core node. Subsequently, two deep encoders are employed to encode and fuse the attribute and structure information of the subgraph, respectively.Finally, by optimizing the fused feature representation through a self-supervised optimization function, the local community is discovered. Extensive experiments on 10 real datasets and 4 synthetic datasets demonstrate that LCDCN outperforms its competitors in terms of performance.

show abstract

“…CDBNE [18]: presented a community detection algorithm based on unsupervised attributed network integration (CDBNE) to solve problems. They propose a framework that simultaneously learns representation based on network structure and attribute information and clustering-oriented representation.…”

Section: Related Workmentioning

confidence: 99%

PCMeans: Community Detection using LocalPagerank, Clustering, and K-Means

Louafi

Titouna

2023

Preprint

View full text Add to dashboard Cite

With the rise of social networks, the task of community detection in networks has become increasingly difficult in recent years. In this study, weintroduce a novel approach, named PCMeans, which combines PageRank, hierarchical clustering, and K-means algorithms to tackle this issue.Our technique employs local PageRank to identify the most influential nodes, followed by a hierarchical clustering strategy that determinesthe optimal number of clusters, and ultimately applies K-means learning to swiftly converge to the final community structure. PCMeansis an unsupervised method that is easy to implement, efficient, andsimple, and it addresses three common problems, including the random selection of the initial central node, specification of the number ofclasses K, and slow convergence. Our experimental findings show thatPCMeans can efficiently identify communities and outperforms otherrecent approaches on both real networks and synthetic benchmarks.

show abstract

Community detection based on unsupervised attributed network embedding

Cited by 29 publications

References 32 publications

Hypergraph network embedding for community detection

Hypergraph network embedding for community detection

Detect Local Community based on Core Node using Deep Feature Fusion

PCMeans: Community Detection using LocalPagerank, Clustering, and K-Means

Contact Info

Product

Resources

About