.png){kind=icon}
.png){kind=icon}
Return
Structure-aware granular ball clustering
.png){kind=icon}
PRE
AI
DOI：10.1016/j.ins.2026.123224.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In the context of the growing prevalence of large-scale data, traditional clustering algorithms face significant bottlenecks in terms of computational efficiency and structural representation capability. This paper proposes a Structure-Aware Granular Ball Clustering method (SAGBC). The SAGBC framework is the first to be developed based on the soft affiliation graph and the dual-constraint connectivity criterion for clustering. Specifically, SAGBC adopts granular balls as the fundamental modeling units and constructs a soft affiliation graph between data points and granular balls to achieve structure-aware representations of complex cluster formations. A dual-constraint connectivity criterion, which integrates spatial proximity with structural similarity, is developed to form connected components of granular balls and effectively capture underlying cluster structures. Extensive experiments on 17 synthetic and real-world datasets demonstrate that SAGBC achieves higher clustering accuracy and computational efficiency than 10 baseline methods, validating its effectiveness for large-scale data analysis. The source code is publicly available at https://github.com/Du-Team/SAGBC.
Keywords:
Granular ball
Clustering
Granular computing
Structure awareness
Large-scale data
Shared nearest neighbors