Scholarlink

<p>Density peak clustering (DPC) is widely used for identifying clusters in complex datasets but heavily depends on the selection of the cut-off distance parameters, which is often chosen subjectively and can significantly impact clustering quality. To address this limitation, we propose a hybrid algorithm that integrates the secretary bird optimization algorithm (SBOA), a bio-inspired metaheuristic modeled on the hunting behavior of secretary birds, with DPC. The primary contribution of this approach is the automatic optimization of the cutoff distance parameter by efficiently exploring the parameter space and avoiding local optima. This enhancement leads to more accurate and stable cluster identification, particularly in complex medical datasets characterized by high dimensionality and noise. Comparative experiments on benchmark medical datasets against established clustering methods demonstrate that the proposed SBOA-DPC algorithm consistently outperforms alternatives in clustering accuracy and robustness, as measured by metrics such as the adjusted Rand index and adjusted mutual information. These results highlight the effectiveness of bio-inspired optimization in improving parameter selection for clustering algorithms, offering a scalable and reliable tool for medical data analysis and other applications requiring precise cluster detection.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Improving medical data clustering based on combining density peaks clustering and secretary bird optimization algorithm

  • 期刊论文

  • 发表日期：

    2026年2月11日

查看全文
  • Shaimaa Waleed Mahmood,
  • Zakariya Yahya Algamal

摘要

Density peak clustering (DPC) is widely used for identifying clusters in complex datasets but heavily depends on the selection of the cut-off distance parameters, which is often chosen subjectively and can significantly impact clustering quality. To address this limitation, we propose a hybrid algorithm that integrates the secretary bird optimization algorithm (SBOA), a bio-inspired metaheuristic modeled on the hunting behavior of secretary birds, with DPC. The primary contribution of this approach is the automatic optimization of the cutoff distance parameter by efficiently exploring the parameter space and avoiding local optima. This enhancement leads to more accurate and stable cluster identification, particularly in complex medical datasets characterized by high dimensionality and noise. Comparative experiments on benchmark medical datasets against established clustering methods demonstrate that the proposed SBOA-DPC algorithm consistently outperforms alternatives in clustering accuracy and robustness, as measured by metrics such as the adjusted Rand index and adjusted mutual information. These results highlight the effectiveness of bio-inspired optimization in improving parameter selection for clustering algorithms, offering a scalable and reliable tool for medical data analysis and other applications requiring precise cluster detection.