.png){kind=icon}
Return
A Fast Grid-Block-Based Density Peak Clustering Algorithm
.png){kind=icon}
PRE
AI
DOI：10.1016/j.neucom.2026.133822.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Density peak clustering (DPC) algorithm has received increasing attention because of its simplicity and effectiveness in recent years. However, it still exists some drawbacks. For instance, the high computational complexity makes it difficult to deal with large-scale datasets. Furthermore, the local density needs to depend on the cutoff distance parameter, which affects the clustering performance. To address these challenges, we propose a fast Grid-Block-based density peak clustering algorithm called FastGB-DPC. It first introduces the idea of grid clustering to divide the grid space created from dataset into several grid cells. Then, it further devises a multi-level diffusion neighbor search strategy to generate Grid-Block on the grid cell, which is a collection of locally coherent and highly similar grid cells. Finally, the density peak clustering is performed by regarding the Grid-Block as the basic clustering unit instead of the original data sample, which greatly reduces the scale of data and has much less running time. Moreover, it further designs a novel double density function so as to make the clustering results independent of the cutoff distance. Extensive experiments on synthetic and real-world datasets demonstrate that FastGB-DPC not only achieves similar or even better effectiveness performance but also has much less running time than the state-of-the-arts. Meanwhile, the statistical test results indicate the statistically significant difference between algorithms. On the whole, the proposed algorithm has a better clustering effect.
Keywords:
Density peak clustering
Grid-Block
Computational complexity
Local density
Clustering algorithm
Journal
IF:
6.5
Papers:
2.5W
Citations:
6.5W
Organization
No organization information available
Cited Papers
No cited papers available