Open Access

ARTICLE

Relative-Density-Viewpoint-Based Weighted Kernel Fuzzy Clustering

Yuhan Xia¹, Xu Li¹, Ye Liu¹, Wenbo Zhou², Yiming Tang^1,3,*

1 School of Computer and Information, Hefei University of Technology, Hefei, 230601, China
2 School of Mechanical Engineering, Hefei University of Technology, Hefei, 230601, China
3 Department of Electrical and Computer Engineering, University of Alberta, Edmonton, AB T6R 2V4, Canada

* Corresponding Author: Yiming Tang. Email:

Computers, Materials & Continua 2025, 84(1), 625-651. https://doi.org/10.32604/cmc.2025.065358

Received 10 March 2025; Accepted 28 April 2025; Issue published 09 June 2025

Abstract

Applying domain knowledge in fuzzy clustering algorithms continuously promotes the development of clustering technology. The combination of domain knowledge and fuzzy clustering algorithms has some problems, such as initialization sensitivity and information granule weight optimization. Therefore, we propose a weighted kernel fuzzy clustering algorithm based on a relative density view (RDVWKFC). Compared with the traditional density-based methods, RDVWKFC can capture the intrinsic structure of the data more accurately, thus improving the initial quality of the clustering. By introducing a Relative Density based Knowledge Extraction Method (RDKM) and adaptive weight optimization mechanism, we effectively solve the limitations of view initialization and information granule weight optimization. RDKM can accurately identify high-density regions and optimize the initialization process. The adaptive weight mechanism can reduce noise and outliers’ interference in the initial cluster centre selection by dynamically allocating weights. Experimental results on 14 benchmark datasets show that the proposed algorithm is superior to the existing algorithms in terms of clustering accuracy, stability, and convergence speed. It shows adaptability and robustness, especially when dealing with different data distributions and noise interference. Moreover, RDVWKFC can also show significant advantages when dealing with data with complex structures and high-dimensional features. These advancements provide versatile tools for real-world applications such as bioinformatics, image segmentation, and anomaly detection.

---

Keywords

---