.png){kind=icon}
.png){kind=icon}
Return
Radial search-based graph clustering method
.png){kind=icon}
PRE
AI
DOI：10.1016/j.neucom.2025.131421.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Graph-based clustering methods represent data samples as nodes and their relationships as edges, which effectively capture the complex structures within similarity graphs. However, many graph-based clustering methods rely on traditional spectral clustering to partition the graph, which may overlook crucial local structural information and affect clustering accuracy. To address this, we propose a Radial Search-Based Graph Clustering (RSGC) method, which can detect clusters with arbitrary shapes and densities, even in complex scenarios such as high-dimensional or multi-peak distributions. We propose a Radial Search Allocation (RSA) method for initial partitioning of the similarity graph, which constructs well-structured single-peak sub-graphs by fully considering the local structure. Additionally, we propose a method for calculating sub-graph similarity based on the importance of cross-cluster edges in the similarity graph, and obtain the final clustering result by merging highly similar sub-graphs. Experimental validations on synthetic and real datasets demonstrate the effectiveness of the RSGC method.
Journal
IF:
6.5
Papers:
2.5W
Citations:
6.5W