Scholarlink

<p>A meticulous comparative analysis of 21 clustering techniques includes traditional methods such as K-means and hierarchical clustering, as well as contemporary models like spectral and density-based clustering. Performance is evaluated on benchmark datasets with varying dimensionality and noise using the Silhouette Score, Davies–Bouldin Index, and Calinski–Harabasz Index. Domain-specific applications across e-commerce, bioinformatics, geospatial analytics, finance, and recommendation systems are highlighted, along with real industrial examples (e.g., Uber, Airbnb, Amazon) and scalability strategies such as distributed processing and minibatch learning. Hyperparameter sensitivity analysis illustrates robustness across settings. Practical decision rules for dataset size, dimensionality, and noise are provided, linking algorithm choice to real-world big data, IoT, and streaming applications. The findings offer concise guidelines for selecting clustering algorithms in diverse contexts, advancing both research and practical applications in unsupervised learning.</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

Benchmarking clustering techniques: insights for comparative analysis and algorithm selection: a survey

  • 期刊论文

  • 发表日期：

    2026年3月17日

查看全文
  • Karan Jain,
  • Uddhav Pisharody,
  • Manjeet Singh,
  • Bam Bahadur Sinha

摘要

A meticulous comparative analysis of 21 clustering techniques includes traditional methods such as K-means and hierarchical clustering, as well as contemporary models like spectral and density-based clustering. Performance is evaluated on benchmark datasets with varying dimensionality and noise using the Silhouette Score, Davies–Bouldin Index, and Calinski–Harabasz Index. Domain-specific applications across e-commerce, bioinformatics, geospatial analytics, finance, and recommendation systems are highlighted, along with real industrial examples (e.g., Uber, Airbnb, Amazon) and scalability strategies such as distributed processing and minibatch learning. Hyperparameter sensitivity analysis illustrates robustness across settings. Practical decision rules for dataset size, dimensionality, and noise are provided, linking algorithm choice to real-world big data, IoT, and streaming applications. The findings offer concise guidelines for selecting clustering algorithms in diverse contexts, advancing both research and practical applications in unsupervised learning.