Clustering Evaluation Measure

A Clustering Evaluation Measure is an evaluation measure that measures clustering quality and clustering performance for clustering algorithms.

• AKA: Cluster Validity Measure, Clustering Quality Measure, Cluster Assessment Measure.
• Context:
  • It can typically assess cluster cohesion through similarity measurement.
  • It can typically evaluate cluster separation via dissimilarity analysis.
  • It can typically validate clustering results using quality criterion.
  • It can often guide parameter selection through optimization processes.
  • It can often support algorithm comparison by performance benchmarking.
  • It can range from being an Internal Clustering Evaluation Measure to being an External Clustering Evaluation Measure, depending on its label requirement.
  • It can range from being a Local Clustering Evaluation Measure to being a Global Clustering Evaluation Measure, depending on its evaluation scope.
  • It can range from being an Absolute Clustering Evaluation Measure to being a Relative Clustering Evaluation Measure, depending on its comparison type.
  • It can range from being a Single-Aspect Clustering Evaluation Measure to being a Multi-Aspect Clustering Evaluation Measure, depending on its quality dimensions.
  • ...
• Examples:
  • Internal Clustering Evaluation Measures, such as:
    • Silhouette Coefficient Measure for cohesion-separation balance.
    • Calinski-Harabasz Score Measure for variance ratio.
  • External Clustering Evaluation Measures, such as:
    • Adjusted Rand Index Measure for agreement measurement.
    • Normalized Mutual Information Measure for information comparison.
  • ...
• Counter-Examples:
  • Classification Accuracy Measure, which requires supervised learning.
  • Regression Error Measure, which measures continuous prediction.
• See: Evaluation Measure, Internal Clustering Evaluation Measure, External Clustering Evaluation Measure, Clustering Task, Clustering Algorithm, Cluster Validity Index, Performance Measure.

---