Clusteval

Clusteval evaluates and compares clustering methods across biomedical datasets to provide objective performance assessments using cluster validity indices.

Key Features:

• Algorithm comparison: Systematic evaluation and comparison of 13 clustering algorithms.
• Cluster validity indices: Computation and use of 13 common cluster validity indices for objective assessment.
• Diverse datasets: Benchmarking across 24 distinct datasets including gene expression profiles and protein domains.
• Large parameter sweeps: Testing up to 1,000 different parameter sets per method across all datasets.
• High-throughput evaluation: Generation of over 4 million calculated cluster validity indices.
• Benchmarking guidance: Production of tailored guidelines to aid selection of suitable clustering methods and benchmarking for method developers.

Scientific Applications:

• Method selection: Informing choice of clustering algorithms for specific biomedical data types and research needs.
• Benchmarking new methods: Providing a reference framework for developers to compare new clustering approaches against established methods.
• Analysis of gene expression and protein domains: Assessing clustering performance specifically on gene expression profiles and protein domain datasets.
• Parameter impact assessment: Evaluating how different parameter sets affect clustering performance and validity indices.

Methodology:

Evaluation of 13 clustering algorithms on 24 datasets (including gene expression profiles and protein domains), computation of 13 cluster validity indices, testing up to 1,000 parameter sets per method across datasets, resulting in over 4 million calculated cluster validity indices.