Separation
How distinctly the resulting groups, labels, or outcomes are separated.
Features
Evaluation
availableAssess runs using standardized metrics and higher-level analytical pillars.
Overview
Ounias evaluates analytical runs using metrics generated by the relevant pipeline and presents selected measures through a common evaluation structure.
The goal is not to collapse analysis into a single score, but to make important analytical qualities easier to inspect and compare across runs.
Evaluation pillars
Relevant metrics are organized into higher-level dimensions so runs produced by different methods can still be compared coherently.
Explanatory power
How clearly the resulting structure can be associated with meaningful features or explanations.
Stability
How consistently the analytical result persists under controlled variation.
Orderliness
How structured or coherent the resulting analytical organization is where the concept applies.
Method fit
How well the selected method appears suited to the characteristics of the dataset and analysis.
Comparison
- Compare metrics across runs
- Inspect standardized evaluation pillars
- Compare different algorithms
- Compare alternative parameter configurations
- Compare feature-space and embedded-space analyses
- Use radar and bar views for compact multi-run comparison
Interpretation
- Evaluation metrics describe properties of an analytical result; they do not establish whether that result is scientifically meaningful.
- No individual metric or aggregate pillar should be interpreted independently of the underlying method and research context.
- Comparability does not imply that every metric has identical meaning across every analytical pipeline.