Robust Graph Clustering Network for Multiple Missing Data

2026-09-25

AI Summary

This study addresses the challenge of simultaneous missing node attributes and graph structure, which causes error propagation and blurred cluster boundaries in graph clustering. To this end, we propose a robust graph clustering framework based on Relational Graph Convolutional Networks (RGCNs). Methodologically, a decoupled dual-branch imputation mechanism is designed to mitigate single-view bias, while a multi-hyperspherical mixture prior is introduced for variational inference to optimize latent space geometry. Furthermore, a boundary-aware contrastive augmentation objective is constructed to enhance inter-class separability. By integrating graph neural networks, hyperspherical modeling, and contrastive learning, the proposed framework significantly outperforms existing state-of-the-art methods across various missing-data scenarios, achieving highly accurate and robust clustering under multi-source data incompleteness.

Abstract

Clustering on graphs where both node attributes and structural links are partially missing remains a challenging task. Existing methods typically rely on imputation-then-clustering on single-view missingness incomplete graphs, which are vulnerable to cross-view error propagation and cluster-boundary blurring under simultaneous attribute and structure missingness. To address these limitations, we propose a Robust Graph Clustering Network for Multiple Missing Data (RGCN), which is designed to handle simultaneous node attribute and graph structure incompleteness. RGCN introduces three key innovations: First, we design a view-decoupled dual-branch imputation to mitigate interference and enable mutual enhancement in recovering missing data. Second, we employ a multi-hyperspherical mixture prior to enhance intra-cluster compactness and inter-cluster separability on a directional latent manifold. Third, a boundary-aware contrastive enhancement objective mitigates the blurring of clusters caused by imputation bias. Extensive experiments on real-world datasets demonstrate that RGCN consistently outperforms state-of-the-art baselines under various missing patterns.

Problem

• Graph Clustering
• Missing Data
• Incomplete Graphs
• Cross-view Error Propagation
• Cluster-boundary Blurring

Innovation

• Graph Clustering
• Missing Data Imputation
• View-Decoupled Dual-Branch
• Multi-Hyperspherical Mixture Prior
• Boundary-Aware Contrastive Learning